• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1<!DOCTYPE html>
2<html lang="en">
3<head>
4<meta charset="UTF-8">
5<!--[if IE]><meta http-equiv="X-UA-Compatible" content="IE=edge"><![endif]-->
6<meta name="viewport" content="width=device-width, initial-scale=1.0">
7<meta name="generator" content="Asciidoctor 1.5.7.1">
8<meta name="author" content="Robert J. Simpson, Qualcomm (Editor), John Kessenich, Google (Editor), Dave Baldwin and Randi Rost (Version 1.1 Authors)">
9<title>The OpenGL ES&#174; Shading Language, Version 3.20.6</title>
10<link rel="stylesheet" href="https://fonts.googleapis.com/css?family=Open+Sans:300,300italic,400,400italic,600,600italic%7CNoto+Serif:400,400italic,700,700italic%7CDroid+Sans+Mono:400,700">
11<style>
12/* Asciidoctor default stylesheet | MIT License | http://asciidoctor.org */
13/* Uncomment @import statement below to use as custom stylesheet */
14/*@import "https://fonts.googleapis.com/css?family=Open+Sans:300,300italic,400,400italic,600,600italic%7CNoto+Serif:400,400italic,700,700italic%7CDroid+Sans+Mono:400,700";*/
15article,aside,details,figcaption,figure,footer,header,hgroup,main,nav,section,summary{display:block}
16audio,canvas,video{display:inline-block}
17audio:not([controls]){display:none;height:0}
18script{display:none!important}
19html{font-family:sans-serif;-ms-text-size-adjust:100%;-webkit-text-size-adjust:100%}
20a{background:transparent}
21a:focus{outline:thin dotted}
22a:active,a:hover{outline:0}
23h1{font-size:2em;margin:.67em 0}
24abbr[title]{border-bottom:1px dotted}
25b,strong{font-weight:bold}
26dfn{font-style:italic}
27hr{-moz-box-sizing:content-box;box-sizing:content-box;height:0}
28mark{background:#ff0;color:#000}
29code,kbd,pre,samp{font-family:monospace;font-size:1em}
30pre{white-space:pre-wrap}
31q{quotes:"\201C" "\201D" "\2018" "\2019"}
32small{font-size:80%}
33sub,sup{font-size:75%;line-height:0;position:relative;vertical-align:baseline}
34sup{top:-.5em}
35sub{bottom:-.25em}
36img{border:0}
37svg:not(:root){overflow:hidden}
38figure{margin:0}
39fieldset{border:1px solid silver;margin:0 2px;padding:.35em .625em .75em}
40legend{border:0;padding:0}
41button,input,select,textarea{font-family:inherit;font-size:100%;margin:0}
42button,input{line-height:normal}
43button,select{text-transform:none}
44button,html input[type="button"],input[type="reset"],input[type="submit"]{-webkit-appearance:button;cursor:pointer}
45button[disabled],html input[disabled]{cursor:default}
46input[type="checkbox"],input[type="radio"]{box-sizing:border-box;padding:0}
47button::-moz-focus-inner,input::-moz-focus-inner{border:0;padding:0}
48textarea{overflow:auto;vertical-align:top}
49table{border-collapse:collapse;border-spacing:0}
50*,*::before,*::after{-moz-box-sizing:border-box;-webkit-box-sizing:border-box;box-sizing:border-box}
51html,body{font-size:100%}
52body{background:#fff;color:rgba(0,0,0,.8);padding:0;margin:0;font-family:"Noto Serif","DejaVu Serif",serif;font-weight:400;font-style:normal;line-height:1;position:relative;cursor:auto;tab-size:4;-moz-osx-font-smoothing:grayscale;-webkit-font-smoothing:antialiased}
53a:hover{cursor:pointer}
54img,object,embed{max-width:100%;height:auto}
55object,embed{height:100%}
56img{-ms-interpolation-mode:bicubic}
57.left{float:left!important}
58.right{float:right!important}
59.text-left{text-align:left!important}
60.text-right{text-align:right!important}
61.text-center{text-align:center!important}
62.text-justify{text-align:justify!important}
63.hide{display:none}
64img,object,svg{display:inline-block;vertical-align:middle}
65textarea{height:auto;min-height:50px}
66select{width:100%}
67.center{margin-left:auto;margin-right:auto}
68.stretch{width:100%}
69.subheader,.admonitionblock td.content>.title,.audioblock>.title,.exampleblock>.title,.imageblock>.title,.listingblock>.title,.literalblock>.title,.stemblock>.title,.openblock>.title,.paragraph>.title,.quoteblock>.title,table.tableblock>.title,.verseblock>.title,.videoblock>.title,.dlist>.title,.olist>.title,.ulist>.title,.qlist>.title,.hdlist>.title{line-height:1.45;color:#7a2518;font-weight:400;margin-top:0;margin-bottom:.25em}
70div,dl,dt,dd,ul,ol,li,h1,h2,h3,#toctitle,.sidebarblock>.content>.title,h4,h5,h6,pre,form,p,blockquote,th,td{margin:0;padding:0;direction:ltr}
71a{color:#2156a5;text-decoration:underline;line-height:inherit}
72a:hover,a:focus{color:#1d4b8f}
73a img{border:none}
74p{font-family:inherit;font-weight:400;font-size:1em;line-height:1.6;margin-bottom:1.25em;text-rendering:optimizeLegibility}
75p aside{font-size:.875em;line-height:1.35;font-style:italic}
76h1,h2,h3,#toctitle,.sidebarblock>.content>.title,h4,h5,h6{font-family:"Open Sans","DejaVu Sans",sans-serif;font-weight:300;font-style:normal;color:#ba3925;text-rendering:optimizeLegibility;margin-top:1em;margin-bottom:.5em;line-height:1.0125em}
77h1 small,h2 small,h3 small,#toctitle small,.sidebarblock>.content>.title small,h4 small,h5 small,h6 small{font-size:60%;color:#e99b8f;line-height:0}
78h1{font-size:2.125em}
79h2{font-size:1.6875em}
80h3,#toctitle,.sidebarblock>.content>.title{font-size:1.375em}
81h4,h5{font-size:1.125em}
82h6{font-size:1em}
83hr{border:solid #ddddd8;border-width:1px 0 0;clear:both;margin:1.25em 0 1.1875em;height:0}
84em,i{font-style:italic;line-height:inherit}
85strong,b{font-weight:bold;line-height:inherit}
86small{font-size:60%;line-height:inherit}
87code{font-family:"Droid Sans Mono","DejaVu Sans Mono",monospace;font-weight:400;color:rgba(0,0,0,.9)}
88ul,ol,dl{font-size:1em;line-height:1.6;margin-bottom:1.25em;list-style-position:outside;font-family:inherit}
89ul,ol{margin-left:1.5em}
90ul li ul,ul li ol{margin-left:1.25em;margin-bottom:0;font-size:1em}
91ul.square li ul,ul.circle li ul,ul.disc li ul{list-style:inherit}
92ul.square{list-style-type:square}
93ul.circle{list-style-type:circle}
94ul.disc{list-style-type:disc}
95ol li ul,ol li ol{margin-left:1.25em;margin-bottom:0}
96dl dt{margin-bottom:.3125em;font-weight:bold}
97dl dd{margin-bottom:1.25em}
98abbr,acronym{text-transform:uppercase;font-size:90%;color:rgba(0,0,0,.8);border-bottom:1px dotted #ddd;cursor:help}
99abbr{text-transform:none}
100blockquote{margin:0 0 1.25em;padding:.5625em 1.25em 0 1.1875em;border-left:1px solid #ddd}
101blockquote cite{display:block;font-size:.9375em;color:rgba(0,0,0,.6)}
102blockquote cite::before{content:"\2014 \0020"}
103blockquote cite a,blockquote cite a:visited{color:rgba(0,0,0,.6)}
104blockquote,blockquote p{line-height:1.6;color:rgba(0,0,0,.85)}
105@media screen and (min-width:768px){h1,h2,h3,#toctitle,.sidebarblock>.content>.title,h4,h5,h6{line-height:1.2}
106h1{font-size:2.75em}
107h2{font-size:2.3125em}
108h3,#toctitle,.sidebarblock>.content>.title{font-size:1.6875em}
109h4{font-size:1.4375em}}
110table{background:#fff;margin-bottom:1.25em;border:solid 1px #dedede}
111table thead,table tfoot{background:#f7f8f7}
112table thead tr th,table thead tr td,table tfoot tr th,table tfoot tr td{padding:.5em .625em .625em;font-size:inherit;color:rgba(0,0,0,.8);text-align:left}
113table tr th,table tr td{padding:.5625em .625em;font-size:inherit;color:rgba(0,0,0,.8)}
114table tr.even,table tr.alt,table tr:nth-of-type(even){background:#f8f8f7}
115table thead tr th,table tfoot tr th,table tbody tr td,table tr td,table tfoot tr td{display:table-cell;line-height:1.6}
116h1,h2,h3,#toctitle,.sidebarblock>.content>.title,h4,h5,h6{line-height:1.2;word-spacing:-.05em}
117h1 strong,h2 strong,h3 strong,#toctitle strong,.sidebarblock>.content>.title strong,h4 strong,h5 strong,h6 strong{font-weight:400}
118.clearfix::before,.clearfix::after,.float-group::before,.float-group::after{content:" ";display:table}
119.clearfix::after,.float-group::after{clear:both}
120*:not(pre)>code{font-size:.9375em;font-style:normal!important;letter-spacing:0;padding:.1em .5ex;word-spacing:-.15em;background-color:#f7f7f8;-webkit-border-radius:4px;border-radius:4px;line-height:1.45;text-rendering:optimizeSpeed;word-wrap:break-word}
121*:not(pre)>code.nobreak{word-wrap:normal}
122*:not(pre)>code.nowrap{white-space:nowrap}
123pre,pre>code{line-height:1.45;color:rgba(0,0,0,.9);font-family:"Droid Sans Mono","DejaVu Sans Mono",monospace;font-weight:400;text-rendering:optimizeSpeed}
124em em{font-style:normal}
125strong strong{font-weight:400}
126.keyseq{color:rgba(51,51,51,.8)}
127kbd{font-family:"Droid Sans Mono","DejaVu Sans Mono",monospace;display:inline-block;color:rgba(0,0,0,.8);font-size:.65em;line-height:1.45;background-color:#f7f7f7;border:1px solid #ccc;-webkit-border-radius:3px;border-radius:3px;-webkit-box-shadow:0 1px 0 rgba(0,0,0,.2),0 0 0 .1em white inset;box-shadow:0 1px 0 rgba(0,0,0,.2),0 0 0 .1em #fff inset;margin:0 .15em;padding:.2em .5em;vertical-align:middle;position:relative;top:-.1em;white-space:nowrap}
128.keyseq kbd:first-child{margin-left:0}
129.keyseq kbd:last-child{margin-right:0}
130.menuseq,.menuref{color:#000}
131.menuseq b:not(.caret),.menuref{font-weight:inherit}
132.menuseq{word-spacing:-.02em}
133.menuseq b.caret{font-size:1.25em;line-height:.8}
134.menuseq i.caret{font-weight:bold;text-align:center;width:.45em}
135b.button::before,b.button::after{position:relative;top:-1px;font-weight:400}
136b.button::before{content:"[";padding:0 3px 0 2px}
137b.button::after{content:"]";padding:0 2px 0 3px}
138p a>code:hover{color:rgba(0,0,0,.9)}
139#header,#content,#footnotes,#footer{width:100%;margin-left:auto;margin-right:auto;margin-top:0;margin-bottom:0;max-width:62.5em;*zoom:1;position:relative;padding-left:.9375em;padding-right:.9375em}
140#header::before,#header::after,#content::before,#content::after,#footnotes::before,#footnotes::after,#footer::before,#footer::after{content:" ";display:table}
141#header::after,#content::after,#footnotes::after,#footer::after{clear:both}
142#content{margin-top:1.25em}
143#content::before{content:none}
144#header>h1:first-child{color:rgba(0,0,0,.85);margin-top:2.25rem;margin-bottom:0}
145#header>h1:first-child+#toc{margin-top:8px;border-top:1px solid #ddddd8}
146#header>h1:only-child,body.toc2 #header>h1:nth-last-child(2){border-bottom:1px solid #ddddd8;padding-bottom:8px}
147#header .details{border-bottom:1px solid #ddddd8;line-height:1.45;padding-top:.25em;padding-bottom:.25em;padding-left:.25em;color:rgba(0,0,0,.6);display:-ms-flexbox;display:-webkit-flex;display:flex;-ms-flex-flow:row wrap;-webkit-flex-flow:row wrap;flex-flow:row wrap}
148#header .details span:first-child{margin-left:-.125em}
149#header .details span.email a{color:rgba(0,0,0,.85)}
150#header .details br{display:none}
151#header .details br+span::before{content:"\00a0\2013\00a0"}
152#header .details br+span.author::before{content:"\00a0\22c5\00a0";color:rgba(0,0,0,.85)}
153#header .details br+span#revremark::before{content:"\00a0|\00a0"}
154#header #revnumber{text-transform:capitalize}
155#header #revnumber::after{content:"\00a0"}
156#content>h1:first-child:not([class]){color:rgba(0,0,0,.85);border-bottom:1px solid #ddddd8;padding-bottom:8px;margin-top:0;padding-top:1rem;margin-bottom:1.25rem}
157#toc{border-bottom:1px solid #efefed;padding-bottom:.5em}
158#toc>ul{margin-left:.125em}
159#toc ul.sectlevel0>li>a{font-style:italic}
160#toc ul.sectlevel0 ul.sectlevel1{margin:.5em 0}
161#toc ul{font-family:"Open Sans","DejaVu Sans",sans-serif;list-style-type:none}
162#toc li{line-height:1.3334;margin-top:.3334em}
163#toc a{text-decoration:none}
164#toc a:active{text-decoration:underline}
165#toctitle{color:#7a2518;font-size:1.2em}
166@media screen and (min-width:768px){#toctitle{font-size:1.375em}
167body.toc2{padding-left:15em;padding-right:0}
168#toc.toc2{margin-top:0!important;background-color:#f8f8f7;position:fixed;width:15em;left:0;top:0;border-right:1px solid #efefed;border-top-width:0!important;border-bottom-width:0!important;z-index:1000;padding:1.25em 1em;height:100%;overflow:auto}
169#toc.toc2 #toctitle{margin-top:0;margin-bottom:.8rem;font-size:1.2em}
170#toc.toc2>ul{font-size:.9em;margin-bottom:0}
171#toc.toc2 ul ul{margin-left:0;padding-left:1em}
172#toc.toc2 ul.sectlevel0 ul.sectlevel1{padding-left:0;margin-top:.5em;margin-bottom:.5em}
173body.toc2.toc-right{padding-left:0;padding-right:15em}
174body.toc2.toc-right #toc.toc2{border-right-width:0;border-left:1px solid #efefed;left:auto;right:0}}
175@media screen and (min-width:1280px){body.toc2{padding-left:20em;padding-right:0}
176#toc.toc2{width:20em}
177#toc.toc2 #toctitle{font-size:1.375em}
178#toc.toc2>ul{font-size:.95em}
179#toc.toc2 ul ul{padding-left:1.25em}
180body.toc2.toc-right{padding-left:0;padding-right:20em}}
181#content #toc{border-style:solid;border-width:1px;border-color:#e0e0dc;margin-bottom:1.25em;padding:1.25em;background:#f8f8f7;-webkit-border-radius:4px;border-radius:4px}
182#content #toc>:first-child{margin-top:0}
183#content #toc>:last-child{margin-bottom:0}
184#footer{max-width:100%;background-color:rgba(0,0,0,.8);padding:1.25em}
185#footer-text{color:rgba(255,255,255,.8);line-height:1.44}
186#content{margin-bottom:.625em}
187.sect1{padding-bottom:.625em}
188@media screen and (min-width:768px){#content{margin-bottom:1.25em}
189.sect1{padding-bottom:1.25em}}
190.sect1:last-child{padding-bottom:0}
191.sect1+.sect1{border-top:1px solid #efefed}
192#content h1>a.anchor,h2>a.anchor,h3>a.anchor,#toctitle>a.anchor,.sidebarblock>.content>.title>a.anchor,h4>a.anchor,h5>a.anchor,h6>a.anchor{position:absolute;z-index:1001;width:1.5ex;margin-left:-1.5ex;display:block;text-decoration:none!important;visibility:hidden;text-align:center;font-weight:400}
193#content h1>a.anchor::before,h2>a.anchor::before,h3>a.anchor::before,#toctitle>a.anchor::before,.sidebarblock>.content>.title>a.anchor::before,h4>a.anchor::before,h5>a.anchor::before,h6>a.anchor::before{content:"\00A7";font-size:.85em;display:block;padding-top:.1em}
194#content h1:hover>a.anchor,#content h1>a.anchor:hover,h2:hover>a.anchor,h2>a.anchor:hover,h3:hover>a.anchor,#toctitle:hover>a.anchor,.sidebarblock>.content>.title:hover>a.anchor,h3>a.anchor:hover,#toctitle>a.anchor:hover,.sidebarblock>.content>.title>a.anchor:hover,h4:hover>a.anchor,h4>a.anchor:hover,h5:hover>a.anchor,h5>a.anchor:hover,h6:hover>a.anchor,h6>a.anchor:hover{visibility:visible}
195#content h1>a.link,h2>a.link,h3>a.link,#toctitle>a.link,.sidebarblock>.content>.title>a.link,h4>a.link,h5>a.link,h6>a.link{color:#ba3925;text-decoration:none}
196#content h1>a.link:hover,h2>a.link:hover,h3>a.link:hover,#toctitle>a.link:hover,.sidebarblock>.content>.title>a.link:hover,h4>a.link:hover,h5>a.link:hover,h6>a.link:hover{color:#a53221}
197.audioblock,.imageblock,.literalblock,.listingblock,.stemblock,.videoblock{margin-bottom:1.25em}
198.admonitionblock td.content>.title,.audioblock>.title,.exampleblock>.title,.imageblock>.title,.listingblock>.title,.literalblock>.title,.stemblock>.title,.openblock>.title,.paragraph>.title,.quoteblock>.title,table.tableblock>.title,.verseblock>.title,.videoblock>.title,.dlist>.title,.olist>.title,.ulist>.title,.qlist>.title,.hdlist>.title{text-rendering:optimizeLegibility;text-align:left;font-family:"Noto Serif","DejaVu Serif",serif;font-size:1rem;font-style:italic}
199table.tableblock.fit-content>caption.title{white-space:nowrap;width:0}
200.paragraph.lead>p,#preamble>.sectionbody>[class="paragraph"]:first-of-type p{font-size:1.21875em;line-height:1.6;color:rgba(0,0,0,.85)}
201table.tableblock #preamble>.sectionbody>[class="paragraph"]:first-of-type p{font-size:inherit}
202.admonitionblock>table{border-collapse:separate;border:0;background:none;width:100%}
203.admonitionblock>table td.icon{text-align:center;width:80px}
204.admonitionblock>table td.icon img{max-width:none}
205.admonitionblock>table td.icon .title{font-weight:bold;font-family:"Open Sans","DejaVu Sans",sans-serif;text-transform:uppercase}
206.admonitionblock>table td.content{padding-left:1.125em;padding-right:1.25em;border-left:1px solid #ddddd8;color:rgba(0,0,0,.6)}
207.admonitionblock>table td.content>:last-child>:last-child{margin-bottom:0}
208.exampleblock>.content{border-style:solid;border-width:1px;border-color:#e6e6e6;margin-bottom:1.25em;padding:1.25em;background:#fff;-webkit-border-radius:4px;border-radius:4px}
209.exampleblock>.content>:first-child{margin-top:0}
210.exampleblock>.content>:last-child{margin-bottom:0}
211.sidebarblock{border-style:solid;border-width:1px;border-color:#e0e0dc;margin-bottom:1.25em;padding:1.25em;background:#f8f8f7;-webkit-border-radius:4px;border-radius:4px}
212.sidebarblock>:first-child{margin-top:0}
213.sidebarblock>:last-child{margin-bottom:0}
214.sidebarblock>.content>.title{color:#7a2518;margin-top:0;text-align:center}
215.exampleblock>.content>:last-child>:last-child,.exampleblock>.content .olist>ol>li:last-child>:last-child,.exampleblock>.content .ulist>ul>li:last-child>:last-child,.exampleblock>.content .qlist>ol>li:last-child>:last-child,.sidebarblock>.content>:last-child>:last-child,.sidebarblock>.content .olist>ol>li:last-child>:last-child,.sidebarblock>.content .ulist>ul>li:last-child>:last-child,.sidebarblock>.content .qlist>ol>li:last-child>:last-child{margin-bottom:0}
216.literalblock pre,.listingblock pre:not(.highlight),.listingblock pre[class="highlight"],.listingblock pre[class^="highlight "],.listingblock pre.CodeRay,.listingblock pre.prettyprint{background:#f7f7f8}
217.sidebarblock .literalblock pre,.sidebarblock .listingblock pre:not(.highlight),.sidebarblock .listingblock pre[class="highlight"],.sidebarblock .listingblock pre[class^="highlight "],.sidebarblock .listingblock pre.CodeRay,.sidebarblock .listingblock pre.prettyprint{background:#f2f1f1}
218.literalblock pre,.literalblock pre[class],.listingblock pre,.listingblock pre[class]{-webkit-border-radius:4px;border-radius:4px;word-wrap:break-word;padding:1em;font-size:.8125em}
219.literalblock pre.nowrap,.literalblock pre[class].nowrap,.listingblock pre.nowrap,.listingblock pre[class].nowrap{overflow-x:auto;white-space:pre;word-wrap:normal}
220@media screen and (min-width:768px){.literalblock pre,.literalblock pre[class],.listingblock pre,.listingblock pre[class]{font-size:.90625em}}
221@media screen and (min-width:1280px){.literalblock pre,.literalblock pre[class],.listingblock pre,.listingblock pre[class]{font-size:1em}}
222.literalblock.output pre{color:#f7f7f8;background-color:rgba(0,0,0,.9)}
223.listingblock pre.highlightjs{padding:0}
224.listingblock pre.highlightjs>code{padding:1em;-webkit-border-radius:4px;border-radius:4px}
225.listingblock pre.prettyprint{border-width:0}
226.listingblock>.content{position:relative}
227.listingblock code[data-lang]::before{display:none;content:attr(data-lang);position:absolute;font-size:.75em;top:.425rem;right:.5rem;line-height:1;text-transform:uppercase;color:#999}
228.listingblock:hover code[data-lang]::before{display:block}
229.listingblock.terminal pre .command::before{content:attr(data-prompt);padding-right:.5em;color:#999}
230.listingblock.terminal pre .command:not([data-prompt])::before{content:"$"}
231table.pyhltable{border-collapse:separate;border:0;margin-bottom:0;background:none}
232table.pyhltable td{vertical-align:top;padding-top:0;padding-bottom:0;line-height:1.45}
233table.pyhltable td.code{padding-left:.75em;padding-right:0}
234pre.pygments .lineno,table.pyhltable td:not(.code){color:#999;padding-left:0;padding-right:.5em;border-right:1px solid #ddddd8}
235pre.pygments .lineno{display:inline-block;margin-right:.25em}
236table.pyhltable .linenodiv{background:none!important;padding-right:0!important}
237.quoteblock{margin:0 1em 1.25em 1.5em;display:table}
238.quoteblock>.title{margin-left:-1.5em;margin-bottom:.75em}
239.quoteblock blockquote,.quoteblock blockquote p{color:rgba(0,0,0,.85);font-size:1.15rem;line-height:1.75;word-spacing:.1em;letter-spacing:0;font-style:italic;text-align:justify}
240.quoteblock blockquote{margin:0;padding:0;border:0}
241.quoteblock blockquote::before{content:"\201c";float:left;font-size:2.75em;font-weight:bold;line-height:.6em;margin-left:-.6em;color:#7a2518;text-shadow:0 1px 2px rgba(0,0,0,.1)}
242.quoteblock blockquote>.paragraph:last-child p{margin-bottom:0}
243.quoteblock .attribution{margin-top:.5em;margin-right:.5ex;text-align:right}
244.quoteblock .quoteblock{margin-left:0;margin-right:0;padding:.5em 0;border-left:3px solid rgba(0,0,0,.6)}
245.quoteblock .quoteblock blockquote{padding:0 0 0 .75em}
246.quoteblock .quoteblock blockquote::before{display:none}
247.verseblock{margin:0 1em 1.25em}
248.verseblock pre{font-family:"Open Sans","DejaVu Sans",sans;font-size:1.15rem;color:rgba(0,0,0,.85);font-weight:300;text-rendering:optimizeLegibility}
249.verseblock pre strong{font-weight:400}
250.verseblock .attribution{margin-top:1.25rem;margin-left:.5ex}
251.quoteblock .attribution,.verseblock .attribution{font-size:.9375em;line-height:1.45;font-style:italic}
252.quoteblock .attribution br,.verseblock .attribution br{display:none}
253.quoteblock .attribution cite,.verseblock .attribution cite{display:block;letter-spacing:-.025em;color:rgba(0,0,0,.6)}
254.quoteblock.abstract{margin:0 1em 1.25em;display:block}
255.quoteblock.abstract>.title{margin:0 0 .375em;font-size:1.15em;text-align:center}
256.quoteblock.abstract blockquote,.quoteblock.abstract blockquote p{word-spacing:0;line-height:1.6}
257.quoteblock.abstract blockquote::before,.quoteblock.abstract p::before{display:none}
258table.tableblock{max-width:100%;border-collapse:separate}
259p.tableblock:last-child{margin-bottom:0}
260td.tableblock>.content{margin-bottom:-1.25em}
261table.tableblock,th.tableblock,td.tableblock{border:0 solid #dedede}
262table.grid-all>thead>tr>.tableblock,table.grid-all>tbody>tr>.tableblock{border-width:0 1px 1px 0}
263table.grid-all>tfoot>tr>.tableblock{border-width:1px 1px 0 0}
264table.grid-cols>*>tr>.tableblock{border-width:0 1px 0 0}
265table.grid-rows>thead>tr>.tableblock,table.grid-rows>tbody>tr>.tableblock{border-width:0 0 1px}
266table.grid-rows>tfoot>tr>.tableblock{border-width:1px 0 0}
267table.grid-all>*>tr>.tableblock:last-child,table.grid-cols>*>tr>.tableblock:last-child{border-right-width:0}
268table.grid-all>tbody>tr:last-child>.tableblock,table.grid-all>thead:last-child>tr>.tableblock,table.grid-rows>tbody>tr:last-child>.tableblock,table.grid-rows>thead:last-child>tr>.tableblock{border-bottom-width:0}
269table.frame-all{border-width:1px}
270table.frame-sides{border-width:0 1px}
271table.frame-topbot,table.frame-ends{border-width:1px 0}
272table.stripes-all tr,table.stripes-odd tr:nth-of-type(odd){background:#f8f8f7}
273table.stripes-none tr,table.stripes-odd tr:nth-of-type(even){background:none}
274th.halign-left,td.halign-left{text-align:left}
275th.halign-right,td.halign-right{text-align:right}
276th.halign-center,td.halign-center{text-align:center}
277th.valign-top,td.valign-top{vertical-align:top}
278th.valign-bottom,td.valign-bottom{vertical-align:bottom}
279th.valign-middle,td.valign-middle{vertical-align:middle}
280table thead th,table tfoot th{font-weight:bold}
281tbody tr th{display:table-cell;line-height:1.6;background:#f7f8f7}
282tbody tr th,tbody tr th p,tfoot tr th,tfoot tr th p{color:rgba(0,0,0,.8);font-weight:bold}
283p.tableblock>code:only-child{background:none;padding:0}
284p.tableblock{font-size:1em}
285td>div.verse{white-space:pre}
286ol{margin-left:1.75em}
287ul li ol{margin-left:1.5em}
288dl dd{margin-left:1.125em}
289dl dd:last-child,dl dd:last-child>:last-child{margin-bottom:0}
290ol>li p,ul>li p,ul dd,ol dd,.olist .olist,.ulist .ulist,.ulist .olist,.olist .ulist{margin-bottom:.625em}
291ul.checklist,ul.none,ol.none,ul.no-bullet,ol.no-bullet,ol.unnumbered,ul.unstyled,ol.unstyled{list-style-type:none}
292ul.no-bullet,ol.no-bullet,ol.unnumbered{margin-left:.625em}
293ul.unstyled,ol.unstyled{margin-left:0}
294ul.checklist{margin-left:.625em}
295ul.checklist li>p:first-child>.fa-square-o:first-child,ul.checklist li>p:first-child>.fa-check-square-o:first-child{width:1.25em;font-size:.8em;position:relative;bottom:.125em}
296ul.checklist li>p:first-child>input[type="checkbox"]:first-child{margin-right:.25em}
297ul.inline{display:-ms-flexbox;display:-webkit-box;display:flex;-ms-flex-flow:row wrap;-webkit-flex-flow:row wrap;flex-flow:row wrap;list-style:none;margin:0 0 .625em -1.25em}
298ul.inline>li{margin-left:1.25em}
299.unstyled dl dt{font-weight:400;font-style:normal}
300ol.arabic{list-style-type:decimal}
301ol.decimal{list-style-type:decimal-leading-zero}
302ol.loweralpha{list-style-type:lower-alpha}
303ol.upperalpha{list-style-type:upper-alpha}
304ol.lowerroman{list-style-type:lower-roman}
305ol.upperroman{list-style-type:upper-roman}
306ol.lowergreek{list-style-type:lower-greek}
307.hdlist>table,.colist>table{border:0;background:none}
308.hdlist>table>tbody>tr,.colist>table>tbody>tr{background:none}
309td.hdlist1,td.hdlist2{vertical-align:top;padding:0 .625em}
310td.hdlist1{font-weight:bold;padding-bottom:1.25em}
311.literalblock+.colist,.listingblock+.colist{margin-top:-.5em}
312.colist td:not([class]):first-child{padding:.4em .75em 0;line-height:1;vertical-align:top}
313.colist td:not([class]):first-child img{max-width:none}
314.colist td:not([class]):last-child{padding:.25em 0}
315.thumb,.th{line-height:0;display:inline-block;border:solid 4px #fff;-webkit-box-shadow:0 0 0 1px #ddd;box-shadow:0 0 0 1px #ddd}
316.imageblock.left,.imageblock[style*="float: left"]{margin:.25em .625em 1.25em 0}
317.imageblock.right,.imageblock[style*="float: right"]{margin:.25em 0 1.25em .625em}
318.imageblock>.title{margin-bottom:0}
319.imageblock.thumb,.imageblock.th{border-width:6px}
320.imageblock.thumb>.title,.imageblock.th>.title{padding:0 .125em}
321.image.left,.image.right{margin-top:.25em;margin-bottom:.25em;display:inline-block;line-height:0}
322.image.left{margin-right:.625em}
323.image.right{margin-left:.625em}
324a.image{text-decoration:none;display:inline-block}
325a.image object{pointer-events:none}
326sup.footnote,sup.footnoteref{font-size:.875em;position:static;vertical-align:super}
327sup.footnote a,sup.footnoteref a{text-decoration:none}
328sup.footnote a:active,sup.footnoteref a:active{text-decoration:underline}
329#footnotes{padding-top:.75em;padding-bottom:.75em;margin-bottom:.625em}
330#footnotes hr{width:20%;min-width:6.25em;margin:-.25em 0 .75em;border-width:1px 0 0}
331#footnotes .footnote{padding:0 .375em 0 .225em;line-height:1.3334;font-size:.875em;margin-left:1.2em;margin-bottom:.2em}
332#footnotes .footnote a:first-of-type{font-weight:bold;text-decoration:none;margin-left:-1.05em}
333#footnotes .footnote:last-of-type{margin-bottom:0}
334#content #footnotes{margin-top:-.625em;margin-bottom:0;padding:.75em 0}
335.gist .file-data>table{border:0;background:#fff;width:100%;margin-bottom:0}
336.gist .file-data>table td.line-data{width:99%}
337div.unbreakable{page-break-inside:avoid}
338.big{font-size:larger}
339.small{font-size:smaller}
340.underline{text-decoration:underline}
341.overline{text-decoration:overline}
342.line-through{text-decoration:line-through}
343.aqua{color:#00bfbf}
344.aqua-background{background-color:#00fafa}
345.black{color:#000}
346.black-background{background-color:#000}
347.blue{color:#0000bf}
348.blue-background{background-color:#0000fa}
349.fuchsia{color:#bf00bf}
350.fuchsia-background{background-color:#fa00fa}
351.gray{color:#606060}
352.gray-background{background-color:#7d7d7d}
353.green{color:#006000}
354.green-background{background-color:#007d00}
355.lime{color:#00bf00}
356.lime-background{background-color:#00fa00}
357.maroon{color:#600000}
358.maroon-background{background-color:#7d0000}
359.navy{color:#000060}
360.navy-background{background-color:#00007d}
361.olive{color:#606000}
362.olive-background{background-color:#7d7d00}
363.purple{color:#600060}
364.purple-background{background-color:#7d007d}
365.red{color:#bf0000}
366.red-background{background-color:#fa0000}
367.silver{color:#909090}
368.silver-background{background-color:#bcbcbc}
369.teal{color:#006060}
370.teal-background{background-color:#007d7d}
371.white{color:#bfbfbf}
372.white-background{background-color:#fafafa}
373.yellow{color:#bfbf00}
374.yellow-background{background-color:#fafa00}
375span.icon>.fa{cursor:default}
376a span.icon>.fa{cursor:inherit}
377.admonitionblock td.icon [class^="fa icon-"]{font-size:2.5em;text-shadow:1px 1px 2px rgba(0,0,0,.5);cursor:default}
378.admonitionblock td.icon .icon-note::before{content:"\f05a";color:#19407c}
379.admonitionblock td.icon .icon-tip::before{content:"\f0eb";text-shadow:1px 1px 2px rgba(155,155,0,.8);color:#111}
380.admonitionblock td.icon .icon-warning::before{content:"\f071";color:#bf6900}
381.admonitionblock td.icon .icon-caution::before{content:"\f06d";color:#bf3400}
382.admonitionblock td.icon .icon-important::before{content:"\f06a";color:#bf0000}
383.conum[data-value]{display:inline-block;color:#fff!important;background-color:rgba(0,0,0,.8);-webkit-border-radius:100px;border-radius:100px;text-align:center;font-size:.75em;width:1.67em;height:1.67em;line-height:1.67em;font-family:"Open Sans","DejaVu Sans",sans-serif;font-style:normal;font-weight:bold}
384.conum[data-value] *{color:#fff!important}
385.conum[data-value]+b{display:none}
386.conum[data-value]::after{content:attr(data-value)}
387pre .conum[data-value]{position:relative;top:-.125em}
388b.conum *{color:inherit!important}
389.conum:not([data-value]):empty{display:none}
390dt,th.tableblock,td.content,div.footnote{text-rendering:optimizeLegibility}
391h1,h2,p,td.content,span.alt{letter-spacing:-.01em}
392p strong,td.content strong,div.footnote strong{letter-spacing:-.005em}
393p,blockquote,dt,td.content,span.alt{font-size:1.0625rem}
394p{margin-bottom:1.25rem}
395.sidebarblock p,.sidebarblock dt,.sidebarblock td.content,p.tableblock{font-size:1em}
396.exampleblock>.content{background-color:#fffef7;border-color:#e0e0dc;-webkit-box-shadow:0 1px 4px #e0e0dc;box-shadow:0 1px 4px #e0e0dc}
397.print-only{display:none!important}
398@page{margin:1.25cm .75cm}
399@media print{*{-webkit-box-shadow:none!important;box-shadow:none!important;text-shadow:none!important}
400html{font-size:80%}
401a{color:inherit!important;text-decoration:underline!important}
402a.bare,a[href^="#"],a[href^="mailto:"]{text-decoration:none!important}
403a[href^="http:"]:not(.bare)::after,a[href^="https:"]:not(.bare)::after{content:"(" attr(href) ")";display:inline-block;font-size:.875em;padding-left:.25em}
404abbr[title]::after{content:" (" attr(title) ")"}
405pre,blockquote,tr,img,object,svg{page-break-inside:avoid}
406thead{display:table-header-group}
407svg{max-width:100%}
408p,blockquote,dt,td.content{font-size:1em;orphans:3;widows:3}
409h2,h3,#toctitle,.sidebarblock>.content>.title{page-break-after:avoid}
410#toc,.sidebarblock,.exampleblock>.content{background:none!important}
411#toc{border-bottom:1px solid #ddddd8!important;padding-bottom:0!important}
412body.book #header{text-align:center}
413body.book #header>h1:first-child{border:0!important;margin:2.5em 0 1em}
414body.book #header .details{border:0!important;display:block;padding:0!important}
415body.book #header .details span:first-child{margin-left:0!important}
416body.book #header .details br{display:block}
417body.book #header .details br+span::before{content:none!important}
418body.book #toc{border:0!important;text-align:left!important;padding:0!important;margin:0!important}
419body.book #toc,body.book #preamble,body.book h1.sect0,body.book .sect1>h2{page-break-before:always}
420.listingblock code[data-lang]::before{display:block}
421#footer{padding:0 .9375em}
422.hide-on-print{display:none!important}
423.print-only{display:block!important}
424.hide-for-print{display:none!important}
425.show-for-print{display:inherit!important}}
426@media print,amzn-kf8{#header>h1:first-child{margin-top:1.25rem}
427.sect1{padding:0!important}
428.sect1+.sect1{border:0}
429#footer{background:none}
430#footer-text{color:rgba(0,0,0,.6);font-size:.9em}}
431@media amzn-kf8{#header,#content,#footnotes,#footer{padding:0}}
432</style>
433<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/font-awesome/4.7.0/css/font-awesome.min.css">
434<style>
435/* Stylesheet for CodeRay to match GitHub theme | MIT License | http://foundation.zurb.com */
436/*pre.CodeRay {background-color:#f7f7f8;}*/
437.CodeRay .line-numbers{border-right:1px solid #d8d8d8;padding:0 0.5em 0 .25em}
438.CodeRay span.line-numbers{display:inline-block;margin-right:.5em;color:rgba(0,0,0,.3)}
439.CodeRay .line-numbers strong{color:rgba(0,0,0,.4)}
440table.CodeRay{border-collapse:separate;border-spacing:0;margin-bottom:0;border:0;background:none}
441table.CodeRay td{vertical-align: top;line-height:1.45}
442table.CodeRay td.line-numbers{text-align:right}
443table.CodeRay td.line-numbers>pre{padding:0;color:rgba(0,0,0,.3)}
444table.CodeRay td.code{padding:0 0 0 .5em}
445table.CodeRay td.code>pre{padding:0}
446.CodeRay .debug{color:#fff !important;background:#000080 !important}
447.CodeRay .annotation{color:#007}
448.CodeRay .attribute-name{color:#000080}
449.CodeRay .attribute-value{color:#700}
450.CodeRay .binary{color:#509}
451.CodeRay .comment{color:#998;font-style:italic}
452.CodeRay .char{color:#04d}
453.CodeRay .char .content{color:#04d}
454.CodeRay .char .delimiter{color:#039}
455.CodeRay .class{color:#458;font-weight:bold}
456.CodeRay .complex{color:#a08}
457.CodeRay .constant,.CodeRay .predefined-constant{color:#008080}
458.CodeRay .color{color:#099}
459.CodeRay .class-variable{color:#369}
460.CodeRay .decorator{color:#b0b}
461.CodeRay .definition{color:#099}
462.CodeRay .delimiter{color:#000}
463.CodeRay .doc{color:#970}
464.CodeRay .doctype{color:#34b}
465.CodeRay .doc-string{color:#d42}
466.CodeRay .escape{color:#666}
467.CodeRay .entity{color:#800}
468.CodeRay .error{color:#808}
469.CodeRay .exception{color:inherit}
470.CodeRay .filename{color:#099}
471.CodeRay .function{color:#900;font-weight:bold}
472.CodeRay .global-variable{color:#008080}
473.CodeRay .hex{color:#058}
474.CodeRay .integer,.CodeRay .float{color:#099}
475.CodeRay .include{color:#555}
476.CodeRay .inline{color:#000}
477.CodeRay .inline .inline{background:#ccc}
478.CodeRay .inline .inline .inline{background:#bbb}
479.CodeRay .inline .inline-delimiter{color:#d14}
480.CodeRay .inline-delimiter{color:#d14}
481.CodeRay .important{color:#555;font-weight:bold}
482.CodeRay .interpreted{color:#b2b}
483.CodeRay .instance-variable{color:#008080}
484.CodeRay .label{color:#970}
485.CodeRay .local-variable{color:#963}
486.CodeRay .octal{color:#40e}
487.CodeRay .predefined{color:#369}
488.CodeRay .preprocessor{color:#579}
489.CodeRay .pseudo-class{color:#555}
490.CodeRay .directive{font-weight:bold}
491.CodeRay .type{font-weight:bold}
492.CodeRay .predefined-type{color:inherit}
493.CodeRay .reserved,.CodeRay .keyword {color:#000;font-weight:bold}
494.CodeRay .key{color:#808}
495.CodeRay .key .delimiter{color:#606}
496.CodeRay .key .char{color:#80f}
497.CodeRay .value{color:#088}
498.CodeRay .regexp .delimiter{color:#808}
499.CodeRay .regexp .content{color:#808}
500.CodeRay .regexp .modifier{color:#808}
501.CodeRay .regexp .char{color:#d14}
502.CodeRay .regexp .function{color:#404;font-weight:bold}
503.CodeRay .string{color:#d20}
504.CodeRay .string .string .string{background:#ffd0d0}
505.CodeRay .string .content{color:#d14}
506.CodeRay .string .char{color:#d14}
507.CodeRay .string .delimiter{color:#d14}
508.CodeRay .shell{color:#d14}
509.CodeRay .shell .delimiter{color:#d14}
510.CodeRay .symbol{color:#990073}
511.CodeRay .symbol .content{color:#a60}
512.CodeRay .symbol .delimiter{color:#630}
513.CodeRay .tag{color:#008080}
514.CodeRay .tag-special{color:#d70}
515.CodeRay .variable{color:#036}
516.CodeRay .insert{background:#afa}
517.CodeRay .delete{background:#faa}
518.CodeRay .change{color:#aaf;background:#007}
519.CodeRay .head{color:#f8f;background:#505}
520.CodeRay .insert .insert{color:#080}
521.CodeRay .delete .delete{color:#800}
522.CodeRay .change .change{color:#66f}
523.CodeRay .head .head{color:#f4f}
524</style>
525<link rel="stylesheet" href="../katex/katex.min.css">
526<script src="../katex/katex.min.js"></script>
527<script src="../katex/contrib/auto-render.min.js"></script>
528    <!-- Use KaTeX to render math once document is loaded, see
529         https://github.com/Khan/KaTeX/tree/master/contrib/auto-render -->
530<script>
531    document.addEventListener("DOMContentLoaded", function () {
532        renderMathInElement(
533            document.body,
534            {
535                delimiters: [
536                    { left: "$$", right: "$$", display: true},
537                    { left: "\\[", right: "\\]", display: true},
538                    { left: "$", right: "$", display: false},
539                    { left: "\\(", right: "\\)", display: false}
540                ]
541            }
542        );
543    });
544</script></head>
545<body class="book toc2 toc-left" style="max-width: 100;">
546<div id="header">
547<h1>The OpenGL ES<sup>&#174;</sup> Shading Language, Version 3.20.6</h1>
548<div class="details">
549<span id="author" class="author">Robert J. Simpson, Qualcomm (Editor), John Kessenich, Google (Editor), Dave Baldwin and Randi Rost (Version 1.1 Authors)</span><br>
550<span id="revnumber">version 3.20.6,</span>
551<span id="revdate">Wed, 10 Jul 2019 20:42:56 +0000</span>
552<br><span id="revremark">Git branch information not available</span>
553</div>
554<div id="toc" class="toc2">
555<div id="toctitle">Table of Contents</div>
556<ul class="sectlevel1">
557<li><a href="#introduction">1. Introduction</a>
558<ul class="sectlevel2">
559<li><a href="#changes">1.1. Changes</a></li>
560<li><a href="#overview">1.2. Overview</a></li>
561<li><a href="#error-handling">1.3. Error Handling</a></li>
562<li><a href="#typographical-conventions">1.4. Typographical Conventions</a></li>
563<li><a href="#compatibility">1.5. Compatibility</a></li>
564</ul>
565</li>
566<li><a href="#overview-of-opengl-shading">2. Overview of Shading</a>
567<ul class="sectlevel2">
568<li><a href="#vertex-processor">2.1. Vertex Processor</a></li>
569<li><a href="#tessellation-control-processor">2.2. Tessellation Control Processor</a></li>
570<li><a href="#tessellation-evaluation-processor">2.3. Tessellation Evaluation Processor</a></li>
571<li><a href="#geometry-processor">2.4. Geometry Processor</a></li>
572<li><a href="#fragment-processor">2.5. Fragment Processor</a></li>
573<li><a href="#compute-processor">2.6. Compute Processor</a></li>
574</ul>
575</li>
576<li><a href="#basics">3. Basics</a>
577<ul class="sectlevel2">
578<li><a href="#character-set">3.1. Character Set</a></li>
579<li><a href="#source-strings">3.2. Source Strings</a></li>
580<li><a href="#version-declaration">3.3. Version Declaration</a></li>
581<li><a href="#preprocessor">3.4. Preprocessor</a></li>
582<li><a href="#comments">3.5. Comments</a></li>
583<li><a href="#tokens">3.6. Tokens</a></li>
584<li><a href="#keywords">3.7. Keywords</a></li>
585<li><a href="#identifiers">3.8. Identifiers</a></li>
586<li><a href="#definitions">3.9. Definitions</a></li>
587<li><a href="#logical-phases-of-compilation">3.10. Logical Phases of Compilation</a></li>
588</ul>
589</li>
590<li><a href="#variables-and-types">4. Variables and Types</a>
591<ul class="sectlevel2">
592<li><a href="#basic-types">4.1. Basic Types</a></li>
593<li><a href="#scoping">4.2. Scoping</a></li>
594<li><a href="#storage-qualifiers">4.3. Storage Qualifiers</a></li>
595<li><a href="#layout-qualifiers">4.4. Layout Qualifiers</a></li>
596<li><a href="#interpolation-qualifiers">4.5. Interpolation Qualifiers</a></li>
597<li><a href="#parameter-qualifiers">4.6. Parameter Qualifiers</a></li>
598<li><a href="#precision-and-precision-qualifiers">4.7. Precision and Precision Qualifiers</a></li>
599<li><a href="#variance-and-the-invariant-qualifier">4.8. Variance and the Invariant Qualifier</a></li>
600<li><a href="#the-precise-qualifier">4.9. The Precise Qualifier</a></li>
601<li><a href="#memory-qualifiers">4.10. Memory Qualifiers</a></li>
602<li><a href="#specialization-constant-qualifier">4.11. Specialization-Constant Qualifier</a></li>
603<li><a href="#order-of-qualification">4.12. Order and Repetition of Qualification</a></li>
604<li><a href="#empty-declarations">4.13. Empty Declarations</a></li>
605</ul>
606</li>
607<li><a href="#operators-and-expressions">5. Operators and Expressions</a>
608<ul class="sectlevel2">
609<li><a href="#operators">5.1. Operators</a></li>
610<li><a href="#array-operations">5.2. Array Operations</a></li>
611<li><a href="#function-calls">5.3. Function Calls</a></li>
612<li><a href="#constructors">5.4. Constructors</a></li>
613<li><a href="#vector-components">5.5. Vector Components</a></li>
614<li><a href="#matrix-components">5.6. Matrix Components</a></li>
615<li><a href="#structure-and-array-operations">5.7. Structure and Array Operations</a></li>
616<li><a href="#assignments">5.8. Assignments</a></li>
617<li><a href="#expressions">5.9. Expressions</a></li>
618<li><a href="#vector-and-matrix-operations">5.10. Vector and Matrix Operations</a></li>
619<li><a href="#specialization-constant-operations">5.11. Specialization-Constant Operations</a></li>
620<li><a href="#evaluation-of-expressions">5.12. Evaluation of Expressions</a></li>
621</ul>
622</li>
623<li><a href="#statements-and-structure">6. Statements and Structure</a>
624<ul class="sectlevel2">
625<li><a href="#function-definitions">6.1. Function Definitions</a></li>
626<li><a href="#selection">6.2. Selection</a></li>
627<li><a href="#iteration">6.3. Iteration</a></li>
628<li><a href="#jumps">6.4. Jumps</a></li>
629</ul>
630</li>
631<li><a href="#built-in-variables">7. Built-In Variables</a>
632<ul class="sectlevel2">
633<li><a href="#built-in-language-variables">7.1. Built-In Language Variables</a></li>
634<li><a href="#built-in-constants">7.2. Built-In Constants</a></li>
635<li><a href="#built-in-uniform-state">7.3. Built-In Uniform State</a></li>
636<li><a href="#redeclaring-built-in-blocks">7.4. Redeclaring Built-In Blocks</a></li>
637</ul>
638</li>
639<li><a href="#built-in-functions">8. Built-In Functions</a>
640<ul class="sectlevel2">
641<li><a href="#angle-and-trigonometry-functions">8.1. Angle and Trigonometry Functions</a></li>
642<li><a href="#exponential-functions">8.2. Exponential Functions</a></li>
643<li><a href="#common-functions">8.3. Common Functions</a></li>
644<li><a href="#floating-point-pack-and-unpack-functions">8.4. Floating-Point Pack and Unpack Functions</a></li>
645<li><a href="#geometric-functions">8.5. Geometric Functions</a></li>
646<li><a href="#matrix-functions">8.6. Matrix Functions</a></li>
647<li><a href="#vector-relational-functions">8.7. Vector Relational Functions</a></li>
648<li><a href="#integer-functions">8.8. Integer Functions</a></li>
649<li><a href="#texture-functions">8.9. Texture Functions</a></li>
650<li><a href="#atomic-counter-functions">8.10. Atomic Counter Functions</a></li>
651<li><a href="#atomic-memory-functions">8.11. Atomic Memory Functions</a></li>
652<li><a href="#image-functions">8.12. Image Functions</a></li>
653<li><a href="#geometry-shader-functions">8.13. Geometry Shader Functions</a></li>
654<li><a href="#fragment-processing-functions">8.14. Fragment Processing Functions</a></li>
655<li><a href="#shader-invocation-control-functions">8.15. Shader Invocation Control Functions</a></li>
656<li><a href="#shader-memory-control-functions">8.16. Shader Memory Control Functions</a></li>
657<li><a href="#_subpass_input_functions">8.17. Subpass-Input Functions</a></li>
658</ul>
659</li>
660<li><a href="#shader-interface-matching">9. Shader Interface Matching</a>
661<ul class="sectlevel2">
662<li><a href="#input-output-matching-by-name-in-linked-programs">9.1. Input Output Matching by Name in Linked Programs</a></li>
663<li><a href="#matching-of-qualifiers">9.2. Matching of Qualifiers</a></li>
664</ul>
665</li>
666<li><a href="#shading-language-grammar">10. Shading Language Grammar</a></li>
667<li><a href="#counting-of-inputs-and-outputs">11. Counting of Inputs and Outputs</a></li>
668<li><a href="#acknowledgments">12. Acknowledgments</a></li>
669<li><a href="#references">13. Normative References</a></li>
670<li><a href="#_non_normative_spir_v_mappings">14. Non-Normative SPIR-V Mappings</a>
671<ul class="sectlevel2">
672<li><a href="#_feature_comparisons">14.1. Feature Comparisons</a></li>
673<li><a href="#_mapping_from_glsl_to_spir_v">14.2. Mapping from GLSL to SPIR-V</a></li>
674</ul>
675</li>
676</ul>
677</div>
678</div>
679<div id="content">
680<div id="preamble">
681<div class="sectionbody">
682<div style="page-break-after: always;"></div>
683<div class="paragraph">
684<p>Copyright &#169; 2008-2018 The Khronos Group Inc. All Rights Reserved.</p>
685</div>
686<div class="paragraph">
687<p>This specification is protected by copyright laws and contains material
688proprietary to the Khronos Group, Inc. It or any components may not be
689reproduced, republished, distributed, transmitted, displayed, broadcast,
690or otherwise exploited in any manner without the express prior written
691permission of Khronos Group. You may use this specification for
692implementing the functionality therein, without altering or removing any
693trademark, copyright or other notice from the specification, but the
694receipt or possession of this specification does not convey any rights
695to reproduce, disclose, or distribute its contents, or to manufacture,
696use, or sell anything that it may describe, in whole or in part.</p>
697</div>
698<div class="paragraph">
699<p>Khronos Group grants express permission to any current Promoter,
700Contributor or Adopter member of Khronos to copy and redistribute
701UNMODIFIED versions of this specification in any fashion, provided that
702NO CHARGE is made for the specification and the latest available update
703of the specification for any version of the API is used whenever
704possible. Such distributed specification may be reformatted AS LONG AS
705the contents of the specification are not changed in any way. The
706specification may be incorporated into a product that is sold as long as
707such product includes significant independent work developed by the
708seller. A link to the current version of this specification on the
709Khronos Group website should be included whenever possible with
710specification distributions.</p>
711</div>
712<div class="paragraph">
713<p>Khronos Group makes no, and expressly disclaims any, representations or
714warranties, express or implied, regarding this specification, including,
715without limitation, any implied warranties of merchantability or fitness
716for a particular purpose or noninfringement of any intellectual
717property. Khronos Group makes no, and expressly disclaims any,
718warranties, express or implied, regarding the correctness, accuracy,
719completeness, timeliness, and reliability of the specification. Under no
720circumstances will the Khronos Group, or any of its Promoters,
721Contributors or Members or their respective partners, officers,
722directors, employees, agents, or representatives be liable for any
723damages, whether direct, indirect, special or consequential damages for
724lost revenues, lost profits, or otherwise, arising from or in connection
725with these materials.</p>
726</div>
727<div class="paragraph">
728<p>Khronos, Vulkan, SYCL, SPIR, WebGL, EGL, COLLADA, StreamInput, OpenVX,
729OpenKCam, glTF, OpenKODE, OpenVG, OpenWF, OpenSL ES, OpenMAX, OpenMAX
730AL, OpenMAX IL and OpenMAX DL are trademarks and WebCL is a
731certification mark of the Khronos Group Inc. OpenCL is a trademark of
732Apple Inc. and OpenGL and OpenML are registered trademarks and the
733OpenGL ES and OpenGL SC logos are trademarks of Silicon Graphics
734International used under license by Khronos. All other product names,
735trademarks, and/or company names are used solely for identification and
736belong to their respective owners.</p>
737</div>
738<div style="page-break-after: always;"></div>
739<!-- toc disabled -->
740</div>
741</div>
742<div class="sect1">
743<h2 id="introduction">1. Introduction</h2>
744<div class="sectionbody">
745<div class="paragraph">
746<p>This document specifies only version 3.20 of the OpenGL ES Shading Language (GLSL ES).
747It requires __VERSION__ to substitute 320, and requires
748<strong>#version</strong> to accept only
749<code>320 es</code>.
750If <strong>#version</strong> is declared with a smaller number, the language accepted is a
751previous version of the shading language, which will be supported depending
752on the version and type of context in the API.
753See the <a href="#references">normative references</a> for details on what language
754versions are supported.</p>
755</div>
756<div class="paragraph">
757<p>Throughout, when generating SPIR-V for consumption by the Vulkan API
758(see <a href="#references">normative references</a>), this will be said to be
759<em>targeting Vulkan</em>.</p>
760</div>
761<div class="paragraph">
762<p>While this specification and the OpenGL ES Specification are normative for OpenGL ES Shading Language, for
763SPIR-V generation it is still the SPIR-V specification and the SPIR-V client
764API specification that are normative for the generated SPIR-V.
765See the <a href="#references">normative references</a> for further detail.</p>
766</div>
767<div class="paragraph">
768<p>For SPIR-V generation, the SPIR-V client API specifies the commands used to
769manipulate SPIR-V shaders.</p>
770</div>
771<div class="paragraph">
772<p>Independent offline tool chains will compile GLSL ES down to the SPIR-V
773intermediate language.
774SPIR-V generation is not enabled with a <strong>#extension</strong>, <strong>#version</strong>, or a
775profile.
776Instead, use of GLSL ES for SPIR-V is determined by offline tool-chain use.
777See the documentation of such tools to see how to request generation of
778SPIR-V for its client API.</p>
779</div>
780<div class="paragraph">
781<p>GLSL ES &#8594; SPIR-V compilers must be directed as to what SPIR-V <strong>Capabilities</strong>
782are legal at run-time and give errors for GLSL ES feature use outside those
783capabilities.
784This is also true for implementation-dependent limits that can be error
785checked by the front-end against built-in constants present in the GLSL ES
786source: the front-end can be informed of such limits, and report errors when
787they are exceeded.</p>
788</div>
789<div class="paragraph">
790<p>All references in this specification to the <a href="#references">OpenGL ES Specification</a> are to
791version 3.2.</p>
792</div>
793<div class="sect2">
794<h3 id="changes">1.1. Changes</h3>
795<div class="sect3">
796<h4 id="_changes_from_glsl_es_3_2_revision_5">1.1.1. Changes from GLSL ES 3.2 revision 5</h4>
797<div class="ulist">
798<ul>
799<li>
800<p>Incorporated the GL_KHR_vulkan_glsl specification.</p>
801</li>
802<li>
803<p>Clarify it is same location that triggers default-uniform block matching
804rules.
805See <a href="#uniform-variable-layout-qualifiers">Uniform Variable Layout Qualifiers</a>.</p>
806</li>
807</ul>
808</div>
809</div>
810<div class="sect3">
811<h4 id="changes-from-glsl-es-3.2-revision-4">1.1.2. Changes from GLSL ES 3.2 revision 4</h4>
812<div class="ulist">
813<ul>
814<li>
815<p>Clarified that this specification completely defines the OpenGL ES Shading Language.
816Normatively reference C++ only for the preprocessor.</p>
817</li>
818<li>
819<p>Private GLSL issues 7, 38: Corrected the values of some builtin constants.
820The values were given correctly in the OpenGL ES Specification.</p>
821</li>
822<li>
823<p>Private GLSL issue 30: Clarify that output packing rules apply to the last
824vertex pipeline stage, not necessarily the vertex stage.</p>
825</li>
826<li>
827<p>Private GLSL issue 15: Clarify the ordering of bindings for arrays of arrays.</p>
828</li>
829<li>
830<p>Private GLSL issue 14: Uniform variables need only match at link time if they
831are statically used.</p>
832</li>
833<li>
834<p>For <strong>precise</strong> computations, the controlling expressions for
835control flow and ternary operators (<strong>?:</strong>) are not included.</p>
836</li>
837</ul>
838</div>
839</div>
840<div class="sect3">
841<h4 id="changes-from-glsl-es-3.2-revision-3">1.1.3. Changes from GLSL ES 3.2 revision 3</h4>
842<div class="ulist">
843<ul>
844<li>
845<p>Matching of default uniforms when shaders are linked.</p>
846</li>
847<li>
848<p><em>gl_DepthRange</em> is only guaranteed to be available in the fragment
849stage.</p>
850</li>
851<li>
852<p>Clarification of definition of static use.</p>
853</li>
854<li>
855<p>Sampling behavior in the absence of <strong>sample</strong> and <strong>centroid</strong>.</p>
856</li>
857<li>
858<p>Clarified the requirements when the specification uses the terms
859<em>should</em>/<em>should not</em> and <em>undefined behavior</em>.</p>
860</li>
861<li>
862<p>Arrayed blocks cannot have layout location qualifiers on members</p>
863</li>
864<li>
865<p><strong>barrier</strong>() defines a partial order which includes tessellation control
866shader outputs.</p>
867</li>
868<li>
869<p>Vertex shader integer output qualification.</p>
870</li>
871<li>
872<p>Incorrect use of predefined pragmas.</p>
873</li>
874<li>
875<p>Clarified use of <strong>readonly</strong> and <strong>writeonly</strong> qualifiers.</p>
876</li>
877<li>
878<p>USAMPLERBUFFER added to grammar.</p>
879</li>
880<li>
881<p>Clarified precision qualifiers can be used in interface blocks.</p>
882</li>
883<li>
884<p>Clarified <strong>memoryBarrierShared</strong> only applies to the current workgroup.</p>
885</li>
886<li>
887<p>The layout qualifier <em>invocations</em> must not be zero.</p>
888</li>
889<li>
890<p>The layout qualifier <em>local_size</em> must not be zero.</p>
891</li>
892<li>
893<p>Clarified the definition of static assignment.</p>
894</li>
895<li>
896<p>Removed list of types with no default precision.</p>
897</li>
898<li>
899<p>Removed scoping rules from the grammar. Refer instead to the <a href="#scoping">scoping</a> section.</p>
900</li>
901<li>
902<p>Require a statement after the final label of a switch.</p>
903</li>
904<li>
905<p>Define <strong>gl_BoundingBox</strong>.</p>
906</li>
907<li>
908<p><strong>length</strong>() expressions returning a constant-value may not include side effects.</p>
909</li>
910<li>
911<p>Clarified that variables may be declared <strong>readonly writeonly</strong>.</p>
912</li>
913<li>
914<p>Use of constant expressions within <strong>#line</strong> directives is undefined.</p>
915</li>
916<li>
917<p><strong>gl_in</strong> can be redeclared using unsized-array syntax.</p>
918</li>
919<li>
920<p>Clarified which sampler types may be used for depth and stencil textures.</p>
921</li>
922<li>
923<p>Added order-of-operation and other explanations to the
924<a href="#the-precise-qualifier">Precise Qualifier</a> section.</p>
925</li>
926<li>
927<p>The <strong>precise</strong> qualifier applied to a block/struct applies recursively to the members.</p>
928</li>
929</ul>
930</div>
931</div>
932<div class="sect3">
933<h4 id="changes-from-glsl-es-3.2-revision-2">1.1.4. Changes from GLSL ES 3.2 revision 2</h4>
934<div class="ulist">
935<ul>
936<li>
937<p>Updated value for <em>gl_MaxTessControlTotalOutputComponents</em></p>
938</li>
939<li>
940<p>Clarified the allowed character set for pre-processing</p>
941</li>
942<li>
943<p>Integer division wrapping behavior</p>
944</li>
945<li>
946<p>Clarified pre-processor expressions (<em>pp-constant-expression</em>)</p>
947</li>
948<li>
949<p>UBO and SSBO precisions do not need to match for linked shaders
950(consistent with GLSL ES 3.1)</p>
951</li>
952<li>
953<p><strong>modf</strong> function</p>
954</li>
955<li>
956<p>Sequence and ternary operators with <strong>void</strong> type</p>
957</li>
958<li>
959<p>Sequence and ternary operators with array types</p>
960</li>
961</ul>
962</div>
963</div>
964<div class="sect3">
965<h4 id="changes-from-glsl-es-3.2-revision-1">1.1.5. Changes from GLSL ES 3.2 revision 1</h4>
966<div class="ulist">
967<ul>
968<li>
969<p>Signed zeros must be supported</p>
970</li>
971<li>
972<p>Layout qualifier table</p>
973</li>
974<li>
975<p>Allowed optimizations when evaluating expressions</p>
976</li>
977<li>
978<p>Updated value for <em>gl_MaxTessControlInputComponents</em></p>
979</li>
980<li>
981<p>Updated value for <em>gl_MaxTessControlOutputComponents</em></p>
982</li>
983<li>
984<p>Updated value for <em>gl_MaxTessEvaluationInputComponents</em></p>
985</li>
986<li>
987<p>Updated value for <em>gl_MaxTessEvaluationOutputComponents</em></p>
988</li>
989<li>
990<p>Updated value for <em>gl_MaxGeometryOutputComponents</em></p>
991</li>
992<li>
993<p>Require precisions in blocks to match when linking</p>
994</li>
995<li>
996<p>Updated conclusions in issues section</p>
997</li>
998</ul>
999</div>
1000</div>
1001<div class="sect3">
1002<h4 id="changes-from-glsl-es-3.1-revision-4">1.1.6. Changes from GLSL ES 3.1 revision 4</h4>
1003<div class="ulist">
1004<ul>
1005<li>
1006<p>Added the following extensions:</p>
1007<div class="ulist">
1008<ul>
1009<li>
1010<p><a href="https://www.opengl.org/registry/specs/KHR/blend_equation_advanced.txt">KHR_blend_equation_advanced</a></p>
1011</li>
1012<li>
1013<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_sample_variables.txt">OES_sample_variables</a></p>
1014</li>
1015<li>
1016<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_shader_image_atomic.txt">OES_shader_image_atomic</a></p>
1017</li>
1018<li>
1019<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_shader_multisample_interpolation.txt">OES_shader_multisample_interpolation</a></p>
1020</li>
1021<li>
1022<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_texture_storage_multisample_2d_array.txt">OES_texture_storage_multisample_2d_array</a></p>
1023</li>
1024<li>
1025<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_geometry_shader.txt">OES_geometry_shader</a></p>
1026</li>
1027<li>
1028<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_gpu_shader5.txt">OES_gpu_shader5</a></p>
1029</li>
1030<li>
1031<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_primitive_bounding_box.txt">OES_primitive_bounding_box</a></p>
1032</li>
1033<li>
1034<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_shader_io_blocks.txt">OES_shader_io_blocks</a></p>
1035</li>
1036<li>
1037<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_tessellation_shader.txt">OES_tessellation_shader</a></p>
1038</li>
1039<li>
1040<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_texture_buffer.txt">OES_texture_buffer</a></p>
1041</li>
1042<li>
1043<p><a href="https://www.khronos.org/registry/gles/extensions/OES/OES_texture_cube_map_array.txt">OES_texture_cube_map_array</a></p>
1044</li>
1045<li>
1046<p><a href="https://www.opengl.org/registry/specs/KHR/robustness.txt">KHR_robustness</a></p>
1047</li>
1048</ul>
1049</div>
1050</li>
1051</ul>
1052</div>
1053</div>
1054</div>
1055<div class="sect2">
1056<h3 id="overview">1.2. Overview</h3>
1057<div class="paragraph">
1058<p>This document describes <em>The OpenGL ES Shading Language, version 3.20</em>.</p>
1059</div>
1060<div class="paragraph">
1061<p>Independent compilation units written in this language are called <em>shaders</em>.
1062A <em>program</em> is a set of shaders that are compiled and linked
1063together.
1064The aim of this document is to thoroughly specify the programming language.
1065The <a href="#references">normative references</a> will specify the API entry points
1066used to manipulate and communicate with programs and shaders.</p>
1067</div>
1068</div>
1069<div class="sect2">
1070<h3 id="error-handling">1.3. Error Handling</h3>
1071<div class="paragraph">
1072<p>Compilers, in general, accept programs that are ill-formed, due to the
1073impossibility of detecting all ill-formed programs.
1074Portability is only ensured for well-formed programs, which this
1075specification describes.
1076Compilers are encouraged to detect ill-formed programs and issue diagnostic
1077messages, but are not required to do so for all cases.</p>
1078</div>
1079<div class="paragraph">
1080<p>The compilation process is implementation-dependent but is generally split
1081into a number of stages, each of which occurs at one of the following times:</p>
1082</div>
1083<div class="ulist">
1084<ul>
1085<li>
1086<p>A call to <em>glCompileShader</em></p>
1087</li>
1088<li>
1089<p>A call to <em>glLinkProgram</em></p>
1090</li>
1091<li>
1092<p>A draw call or a call to <em>glValidateProgram</em></p>
1093</li>
1094</ul>
1095</div>
1096<div class="paragraph">
1097<p>The implementation should report errors as early a possible but in any case
1098must satisfy the following:</p>
1099</div>
1100<div class="ulist">
1101<ul>
1102<li>
1103<p>All lexical, grammatical and semantic errors must have been detected
1104following a call to <em>glLinkProgram</em></p>
1105</li>
1106<li>
1107<p>Errors due to mismatch between the shaders (link-time errors) must have
1108been detected following a call to <em>glLinkProgram</em></p>
1109</li>
1110<li>
1111<p>Errors due to exceeding resource limits must have been detected
1112following any draw call or a call to <em>glValidateProgram</em></p>
1113</li>
1114<li>
1115<p>A call to <em>glValidateProgram</em> must report all errors associated with a
1116program object given the current GL state.</p>
1117</li>
1118</ul>
1119</div>
1120<div class="paragraph">
1121<p>Where the specification uses the terms <em>required</em>, <em>must</em>/<em>must</em> <em>not</em>,
1122<em>does</em>/<em>does</em> <em>not</em>, <em>disallowed</em>, or <em>not</em> <em>supported</em>, the compiler or
1123linker is required to detect and report any violations.
1124Similarly when a condition or situation is an <strong>error</strong>, it must be reported.
1125Use of any feature marked as <em>reserved</em> is an error.
1126Where the specification uses the terms <em>should</em>/<em>should</em> <em>not</em>, <em>undefined</em>
1127<em>behavior</em>, <em>undefined</em> <em>value</em> or <em>undefined</em> <em>*results*</em>, implementations
1128will not produce a compile-time error but are encouraged to issue a warning
1129for violations.
1130The run-time behavior of the program in these cases is not constrained (and
1131so may include termination or system instability).
1132It is expected that systems will be designed to handle these cases
1133gracefully but specification of this is outside the scope of this
1134specification.</p>
1135</div>
1136<div class="paragraph">
1137<p>Implementations may not in general support functionality beyond the mandated
1138parts of the specification without use of the relevant extension.
1139The only exceptions are:</p>
1140</div>
1141<div class="olist arabic">
1142<ol class="arabic">
1143<li>
1144<p>If a feature is marked as optional.</p>
1145</li>
1146<li>
1147<p>Where a maximum value is stated (e.g. the maximum number of vertex
1148outputs), the implementation may support a higher value than that
1149specified.</p>
1150</li>
1151</ol>
1152</div>
1153<div class="paragraph">
1154<p>Where the implementation supports more than the mandated specification,
1155off-target compilers are encouraged to issue warnings if these features are
1156used.</p>
1157</div>
1158<div class="paragraph">
1159<p>The compilation process is split between the compiler and linker.
1160The allocation of tasks between the compiler and linker is implementation
1161dependent.
1162Consequently there are many errors which may be detected either at compile
1163or link time, depending on the implementation.</p>
1164</div>
1165</div>
1166<div class="sect2">
1167<h3 id="typographical-conventions">1.4. Typographical Conventions</h3>
1168<div class="paragraph">
1169<p>Italic, bold, and font choices have been used in this specification
1170primarily to improve readability.
1171Code fragments use a fixed width font.
1172Identifiers embedded in text are italicized.
1173Keywords embedded in text are bold.
1174Operators are called by their name, followed by their symbol in bold in
1175parentheses.
1176The clarifying grammar fragments in the text use bold for literals and
1177italics for non-terminals.
1178The official grammar in &#8220;<a href="#shading-language-grammar">Shading Language
1179Grammar</a>&#8221; uses all capitals for terminals and lower case for
1180non-terminals.</p>
1181</div>
1182</div>
1183<div class="sect2">
1184<h3 id="compatibility">1.5. Compatibility</h3>
1185<div class="paragraph">
1186<p>The OpenGL ES 3.2 API is designed to work with GLSL ES v1.00,
1187GLSL ES 3.00, GLSL ES 3.10 and GLSL ES 3.20.
1188In general a shader written for versions prior to OpenGL ES 3.2
1189should work without modification in OpenGL ES 3.2.</p>
1190</div>
1191<div class="paragraph">
1192<p>When porting applications from an earlier to later version of the GLSL ES,
1193the following points should be noted:</p>
1194</div>
1195<div class="ulist">
1196<ul>
1197<li>
1198<p>Not all language constructs present in earlier versions of the language
1199are available in later versions e.g. attribute and varying qualifiers
1200are present in v1.00 but not v3.00.
1201However, the functionality of GLSL ES 3.20 is a super-set of
1202GLSL ES 3.10.</p>
1203</li>
1204<li>
1205<p>Some features of later versions of the API require language features
1206that are not present in earlier version of the language.</p>
1207</li>
1208<li>
1209<p>It is an error to link shaders if they are written in different versions
1210of the language.</p>
1211</li>
1212<li>
1213<p>The OpenGL ES 2.0 and 3.0 APIs do not support shaders written in GLSL ES
12143.20.</p>
1215</li>
1216<li>
1217<p>Using GLSL ES 1.00 shaders within OpenGL ES 3.x may extend the resources
1218available beyond the minima specified in GLSL ES 1.0.
1219Shaders which make use of this will not necessarily run on an OpenGL ES
12202.0 implementation: Similarly for GLSL ES 3.00 shaders running within
1221OpenGL ES 3.2.</p>
1222</li>
1223<li>
1224<p>Support of line continuation and support of UTF-8 characters within
1225comments is optional in GLSL ES 1.00 when used with the OpenGL ES 2.0
1226API.
1227However, support is mandated for both of these when a GLSL ES 1.00
1228shader is used with the OpenGL ES 3.x APIs.</p>
1229</li>
1230</ul>
1231</div>
1232</div>
1233</div>
1234</div>
1235<div class="sect1">
1236<h2 id="overview-of-opengl-shading">2. Overview of Shading</h2>
1237<div class="sectionbody">
1238<div class="paragraph">
1239<p>The OpenGL ES Shading Language is actually several closely related languages.
1240These languages are used to create shaders for each of the programmable
1241processors contained in the API&#8217;s processing pipeline.
1242Currently, these processors are the vertex, tessellation control,
1243tessellation evaluation, geometry, fragment, and compute processors.</p>
1244</div>
1245<div class="paragraph">
1246<p>Compilation units for these processors are referred to as <em>shaders</em> and the
1247processors themselves are also referred to as <em>shader stages</em>.
1248Only one shader can be run on a processor at any one time; there is no
1249support for linking multiple compilation units together for a single shader
1250stage.
1251One or more shaders are linked together to form a single program and each
1252program contains shader <em>executables</em> for one or more consecutive shader
1253stages.</p>
1254</div>
1255<div class="paragraph">
1256<p>Unless otherwise noted in this paper, a language feature applies to all
1257languages, and common usage will refer to these languages as a single
1258language.
1259The specific languages will be referred to by the name of the processor they
1260target: vertex, tessellation control, tessellation evaluation, geometry,
1261fragment, or compute.</p>
1262</div>
1263<div class="paragraph">
1264<p>Most API state is not tracked or made available to shaders.
1265Typically, user-defined variables will be used for communicating between
1266different stages of the API pipeline.
1267However, a small amount of state is still tracked and automatically made
1268available to shaders, and there are a few built-in variables for interfaces
1269between different stages of the API pipeline.</p>
1270</div>
1271<div class="sect2">
1272<h3 id="vertex-processor">2.1. Vertex Processor</h3>
1273<div class="paragraph">
1274<p>The <em>vertex processor</em> is a programmable unit that operates on incoming
1275vertices and their associated data.
1276Compilation units written in the OpenGL ES Shading Language to run on this processor are called
1277<em>vertex shaders</em>.</p>
1278</div>
1279<div class="paragraph">
1280<p>The vertex processor operates on one vertex at a time.
1281It does not replace graphics operations that require knowledge of several
1282vertices at a time.</p>
1283</div>
1284</div>
1285<div class="sect2">
1286<h3 id="tessellation-control-processor">2.2. Tessellation Control Processor</h3>
1287<div class="paragraph">
1288<p>The <em>tessellation control processor</em> is a programmable unit that operates on
1289a patch of incoming vertices and their associated data, emitting a new
1290output patch.
1291Compilation units written in the OpenGL ES Shading Language to run on this processor are called
1292tessellation control shaders.</p>
1293</div>
1294<div class="paragraph">
1295<p>The tessellation control shader is invoked for each vertex of the output
1296patch.
1297Each invocation can read the attributes of any vertex in the input or output
1298patches, but can only write per-vertex attributes for the corresponding
1299output patch vertex.
1300The shader invocations collectively produce a set of per-patch attributes
1301for the output patch.</p>
1302</div>
1303<div class="paragraph">
1304<p>After all tessellation control shader invocations have completed, the output
1305vertices and per-patch attributes are assembled to form a patch to be used
1306by subsequent pipeline stages.</p>
1307</div>
1308<div class="paragraph">
1309<p>Tessellation control shader invocations run mostly independently, with
1310undefined relative execution order.
1311However, the built-in function <strong>barrier</strong>() can be used to control execution
1312order by synchronizing invocations, effectively dividing tessellation
1313control shader execution into a set of phases.
1314Tessellation control shaders will get undefined results if one invocation
1315reads from a per-vertex or per-patch attribute written by another invocation
1316at any point during the same phase, or if two invocations attempt to write
1317different values to the same per-patch output
1318in a single phase.</p>
1319</div>
1320</div>
1321<div class="sect2">
1322<h3 id="tessellation-evaluation-processor">2.3. Tessellation Evaluation Processor</h3>
1323<div class="paragraph">
1324<p>The <em>tessellation evaluation processor</em> is a programmable unit that
1325evaluates the position and other attributes of a vertex generated by the
1326tessellation primitive generator, using a patch of incoming vertices and
1327their associated data.
1328Compilation units written in the OpenGL ES Shading Language to run on this processor are called
1329tessellation evaluation shaders.</p>
1330</div>
1331<div class="paragraph">
1332<p>Each invocation of the tessellation evaluation executable computes the
1333position and attributes of a single vertex generated by the tessellation
1334primitive generator.
1335The executable can read the attributes of any vertex in the input patch,
1336plus the tessellation coordinate, which is the relative location of the
1337vertex in the primitive being tessellated.
1338The executable writes the position and other attributes of the vertex.</p>
1339</div>
1340</div>
1341<div class="sect2">
1342<h3 id="geometry-processor">2.4. Geometry Processor</h3>
1343<div class="paragraph">
1344<p>The <em>geometry processor</em> is a programmable unit that operates on data for
1345incoming vertices for a primitive assembled after vertex processing and
1346outputs a sequence of vertices forming output primitives.
1347Compilation units written in the OpenGL ES Shading Language to run on this processor are called
1348<em>geometry shaders</em>.</p>
1349</div>
1350<div class="paragraph">
1351<p>A single invocation of the geometry shader executable on the geometry
1352processor will operate on a declared input primitive with a fixed number of
1353vertices.
1354This single invocation can emit a variable number of vertices that are
1355assembled into primitives of a declared output primitive type and passed to
1356subsequent pipeline stages.</p>
1357</div>
1358</div>
1359<div class="sect2">
1360<h3 id="fragment-processor">2.5. Fragment Processor</h3>
1361<div class="paragraph">
1362<p>The <em>fragment processor</em> is a programmable unit that operates on fragment
1363values and their associated data.
1364Compilation units written in the OpenGL ES Shading Language to run on this processor are called
1365<em>fragment shaders</em>.</p>
1366</div>
1367<div class="paragraph">
1368<p>A fragment shader cannot change a fragment&#8217;s (<em>x</em>, <em>y</em>) position.
1369Access to neighboring fragments is not allowed.
1370The values computed by the fragment shader are ultimately used to update
1371framebuffer memory or texture memory, depending on the current API
1372state and the API command that caused the fragments to be generated.</p>
1373</div>
1374</div>
1375<div class="sect2">
1376<h3 id="compute-processor">2.6. Compute Processor</h3>
1377<div class="paragraph">
1378<p>The <em>compute processor</em> is a programmable unit that operates independently
1379from the other shader processors.
1380Compilation units written in the OpenGL ES Shading Language to run on this processor are called
1381<em>compute shaders</em>.</p>
1382</div>
1383<div class="paragraph">
1384<p>A compute shader has access to many of the same resources as fragment and
1385other shader processors, such as textures, buffers, image variables, and
1386atomic counters.
1387It does not have fixed-function outputs.
1388It is not part of the graphics pipeline and its visible side effects are
1389through changes to images, storage buffers, and atomic counters.</p>
1390</div>
1391<div class="paragraph">
1392<p>A compute shader operates on a group of work items called a <em>workgroup</em>.
1393A workgroup is a collection of shader invocations that execute the same
1394code, potentially in parallel.
1395An invocation within a workgroup may share data with other members of the
1396same workgroup through shared variables and issue memory and control flow
1397barriers to synchronize with other members of the same workgroup.</p>
1398</div>
1399</div>
1400</div>
1401</div>
1402<div class="sect1">
1403<h2 id="basics">3. Basics</h2>
1404<div class="sectionbody">
1405<div class="sect2">
1406<h3 id="character-set">3.1. Character Set</h3>
1407<div class="paragraph">
1408<p>The source character set used for the OpenGL ES Shading Language is Unicode in the UTF-8
1409encoding scheme.
1410Invalid UTF-8 characters are ignored.
1411During pre-processing, the following applies:</p>
1412</div>
1413<div class="ulist">
1414<ul>
1415<li>
1416<p>A byte with the value zero is always interpreted as the end of the
1417string</p>
1418</li>
1419<li>
1420<p>Backslash ('\'), is used to indicate line continuation when immediately
1421preceding a new-line.</p>
1422</li>
1423<li>
1424<p>White space consists of one or more of the following characters: the
1425space character, horizontal tab, vertical tab, form feed,
1426carriage-return, line-feed.</p>
1427</li>
1428<li>
1429<p>The number sign (<strong>#</strong>) is used for preprocessor directives</p>
1430</li>
1431<li>
1432<p>Macro names are restricted to:</p>
1433<div class="ulist">
1434<ul>
1435<li>
1436<p>The letters <strong>a-z</strong>, <strong>A-Z</strong>, and the underscore (<strong>_</strong>).</p>
1437</li>
1438<li>
1439<p>The numbers <strong>0-9</strong>, except for the first character of a macro name.</p>
1440</li>
1441</ul>
1442</div>
1443</li>
1444</ul>
1445</div>
1446<div class="paragraph">
1447<p>After preprocessing, only the following characters are allowed in the
1448resulting stream of GLSL ES tokens:</p>
1449</div>
1450<div class="ulist">
1451<ul>
1452<li>
1453<p>The letters <strong>a-z</strong>, <strong>A-Z</strong>, and the underscore (<strong>_</strong>).</p>
1454</li>
1455<li>
1456<p>The numbers <strong>0-9</strong>.</p>
1457</li>
1458<li>
1459<p>The symbols period (<strong>.</strong>), plus (<strong>+</strong>), dash (<strong>-</strong>), slash (<strong>/</strong>), asterisk
1460(<strong>*</strong>), percent (<strong>%</strong>), angled brackets (<strong>&lt;</strong> and <strong>&gt;</strong>), square brackets
1461(<strong>[</strong> and <strong>]</strong>), parentheses (<strong>(</strong> and <strong>)</strong>), braces (<strong>{</strong> and <strong>}</strong>), caret
1462(<strong>^</strong>), vertical bar (<strong>|</strong>), ampersand (<strong>&amp;</strong>), tilde (<strong>~</strong>), equals (<strong>=</strong>),
1463exclamation point (<strong>!</strong>), colon (<strong>:</strong>), semicolon (<strong>;</strong>), comma (<strong>,</strong>), and
1464question mark (<strong>?</strong>).</p>
1465</li>
1466</ul>
1467</div>
1468<div class="paragraph">
1469<p>There are no digraphs or trigraphs.
1470There are no escape sequences or other uses of the backslash beyond use as
1471the line-continuation character.</p>
1472</div>
1473<div class="paragraph">
1474<p>Lines are relevant for compiler diagnostic messages and the preprocessor.
1475They are terminated by carriage-return or line-feed.
1476If both are used together, it will count as only a single line termination.
1477For the remainder of this document, any of these combinations is simply
1478referred to as a new-line.
1479Lines may be of arbitrary length.</p>
1480</div>
1481<div class="paragraph">
1482<p>In general, the language&#8217;s use of this character set is case sensitive.</p>
1483</div>
1484<div class="paragraph">
1485<p>There are no character or string data types, so no quoting characters are
1486included.</p>
1487</div>
1488<div class="paragraph">
1489<p>There is no end-of-file character.</p>
1490</div>
1491</div>
1492<div class="sect2">
1493<h3 id="source-strings">3.2. Source Strings</h3>
1494<div class="paragraph">
1495<p>The source for a single shader is an array of strings of characters from the
1496character set.
1497A single shader is made from the concatenation of these strings.
1498Each string can contain multiple lines, separated by new-lines.
1499No new-lines need be present in a string; a single line can be formed from
1500multiple strings.
1501No new-lines or other characters are inserted by the implementation when it
1502concatenates the strings to form a single shader.</p>
1503</div>
1504<div class="paragraph">
1505<p>Diagnostic messages returned from compiling a shader must identify both the
1506line number within a string and which source string the message applies to.
1507Source strings are counted sequentially with the first string being string
15080.
1509Line numbers are one more than the number of new-lines that have been
1510processed, including counting the new-lines that will be removed by the
1511line-continuation character (<strong>\</strong>).</p>
1512</div>
1513<div class="paragraph">
1514<p>Lines separated by the line-continuation character preceding a new-line are
1515concatenated together before either comment processing or preprocessing.
1516This means that no white space is substituted for the line-continuation
1517character.
1518That is, a single token could be formed by the concatenation by taking the
1519characters at the end of one line concatenating them with the characters at
1520the beginning of the next line.</p>
1521</div>
1522<div class="listingblock">
1523<div class="content">
1524<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span> f\
1525oo;
1526<span class="comment">// forms a single line equivalent to &quot;float foo;&quot;</span>
1527<span class="comment">// (assuming '\' is the last character before the new-line and &quot;oo&quot; are</span>
1528<span class="comment">// the first two characters of the next line)</span></code></pre>
1529</div>
1530</div>
1531</div>
1532<div class="sect2">
1533<h3 id="version-declaration">3.3. Version Declaration</h3>
1534<div class="paragraph">
1535<p>Shaders must declare the version of the language they are written to.
1536The version is specified in the first line of a shader by a character
1537string:</p>
1538</div>
1539<div class="listingblock">
1540<div class="content">
1541<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#version</span> number es</code></pre>
1542</div>
1543</div>
1544<div class="paragraph">
1545<p>where <em>number</em> must be a version of the language, following the same
1546convention as __VERSION__ above.
1547The directive &#8220;<strong>#version 320 es</strong>&#8221; is required in any shader that
1548uses version 3.20 of the language.
1549Any <em>number</em> representing a version of the language a compiler does not
1550support will cause an error to be generated.
1551Version 1.00 of the language does not require shaders to include this
1552directive, and shaders that do not include a <strong>#version</strong> directive will be
1553treated as targeting version 1.00.</p>
1554</div>
1555<div class="paragraph">
1556<p>Shaders declaring version 3.20 of the shading language cannot be
1557linked with shaders declaring a previous version.</p>
1558</div>
1559<div class="paragraph">
1560<p>The <strong>#version</strong> directive must be present in the first line of a shader and
1561must be followed by a newline.
1562It may contain optional white-space as specified below but no other
1563characters are allowed.
1564The directive is only permitted in the first line of a shader.</p>
1565</div>
1566<div class="paragraph">
1567<p>Processing of the #version directive occurs before all other preprocessing,
1568including line concatenation and comment processing.</p>
1569</div>
1570<div class="openblock bnf">
1571<div class="content">
1572<div class="dlist">
1573<dl>
1574<dt class="hdlist1"><em>version-declaration</em> : </dt>
1575<dd>
1576<p><em>whitespace<sub>opt</sub></em> POUND <em>whitespace<sub>opt</sub></em> VERSION <em>whitespace</em> <em>number</em>
1577<em>whitespace</em> ES <em>whitespace<sub>opt</sub></em></p>
1578</dd>
1579</dl>
1580</div>
1581</div>
1582</div>
1583<div class="paragraph">
1584<p>Tokens:</p>
1585</div>
1586<div class="dlist">
1587<dl>
1588<dt class="hdlist1"></dt>
1589<dd>
1590<p>POUND <strong>#</strong><br>
1591VERSION <strong>version</strong><br>
1592ES <strong>es</strong></p>
1593</dd>
1594</dl>
1595</div>
1596</div>
1597<div class="sect2">
1598<h3 id="preprocessor">3.4. Preprocessor</h3>
1599<div class="paragraph">
1600<p>There is a preprocessor that processes the source strings as part of the
1601compilation process.
1602Except as noted below, it behaves as the C++ standard preprocessor (see
1603&#8220;<a href="#references">Normative References</a>&#8221;).</p>
1604</div>
1605<div class="paragraph">
1606<p>The complete list of preprocessor directives is as follows.</p>
1607</div>
1608<div class="dlist">
1609<dl>
1610<dt class="hdlist1"></dt>
1611<dd>
1612<p>#<br>
1613#define<br>
1614#undef<br></p>
1615</dd>
1616<dt class="hdlist1"></dt>
1617<dd>
1618<p>#if<br>
1619#ifdef<br>
1620#ifndef<br>
1621#else<br>
1622#elif<br>
1623#endif<br></p>
1624</dd>
1625<dt class="hdlist1"></dt>
1626<dd>
1627<p>#error<br>
1628#pragma<br></p>
1629</dd>
1630<dt class="hdlist1"></dt>
1631<dd>
1632<p>#extension<br></p>
1633</dd>
1634<dt class="hdlist1"></dt>
1635<dd>
1636<p>#line</p>
1637</dd>
1638</dl>
1639</div>
1640<div class="paragraph">
1641<p>The following
1642operator is
1643also available:</p>
1644</div>
1645<div class="dlist">
1646<dl>
1647<dt class="hdlist1"></dt>
1648<dd>
1649<p>defined<br></p>
1650</dd>
1651</dl>
1652</div>
1653<div class="paragraph">
1654<p>Note that the version directive is not considered to be a preprocessor
1655directive and so is not listed here.</p>
1656</div>
1657<div class="paragraph">
1658<p>Each number sign (<strong>#</strong>) can be preceded in its line only by spaces or
1659horizontal tabs.
1660It may also be followed by spaces and horizontal tabs, preceding the
1661directive.
1662Each directive is terminated by a new-line.
1663Preprocessing does not change the number or relative location of new-lines
1664in a source string.</p>
1665</div>
1666<div class="paragraph">
1667<p>The number sign (<strong>#</strong>) on a line by itself is ignored.
1668Any directive not listed above will cause an error.</p>
1669</div>
1670<div class="paragraph">
1671<p><strong>#define</strong> and <strong>#undef</strong> functionality are defined as is standard for C++
1672preprocessors for macro definitions both with and without macro parameters.</p>
1673</div>
1674<div class="paragraph">
1675<p>The following predefined macros are available:</p>
1676</div>
1677<div class="dlist">
1678<dl>
1679<dt class="hdlist1"></dt>
1680<dd>
1681<p>__LINE__<br>
1682__FILE__<br>
1683__VERSION__<br>
1684GL_ES</p>
1685</dd>
1686</dl>
1687</div>
1688<div class="paragraph">
1689<p>__LINE__ will substitute a decimal integer constant that is one more than
1690the number of preceding new-lines in the current source string.</p>
1691</div>
1692<div class="paragraph">
1693<p>__FILE__ will substitute a decimal integer constant that says which source
1694string number is currently being processed.</p>
1695</div>
1696<div class="paragraph">
1697<p>__VERSION__ will substitute a decimal integer reflecting the version
1698number of the OpenGL ES Shading Language.
1699The version of the shading language described in this document will have
1700__VERSION__ substitute the decimal integer 320.</p>
1701</div>
1702<div class="paragraph">
1703<p>GL_ES will be defined and set to 1.
1704This is not true for the non-ES OpenGL Shading Language, so it can be used
1705to do a compile time test to determine if a shader is compiling as an
1706GLSL ES shader.</p>
1707</div>
1708<div class="paragraph">
1709<p>By convention, all macro names containing two consecutive underscores (__)
1710are reserved for use by underlying software layers.
1711Defining
1712such a name in a shader does not itself result in an error, but may
1713result in unintended behaviors that stem from having multiple definitions of
1714the same name.
1715All macro names prefixed with &#8220;GL_&#8221; (&#8220;GL&#8221; followed by a single
1716underscore) are also reserved, and defining
1717such a name results in a compile-time error.</p>
1718</div>
1719<div class="paragraph">
1720<p>It is an error to undefine or to redefine a built-in (pre-defined) macro
1721name.</p>
1722</div>
1723<div class="paragraph">
1724<p>Implementations must support macro-name lengths of up to 1024 characters.
1725It is an error to declare a name with a length greater than this.</p>
1726</div>
1727<div class="paragraph">
1728<p><strong>#if</strong>, <strong>#ifdef</strong>, <strong>#ifndef</strong>, <strong>#else</strong>, <strong>#elif</strong>, and <strong>#endif</strong> are defined to
1729operate as is standard for C++ preprocessors except for the following:</p>
1730</div>
1731<div class="ulist">
1732<ul>
1733<li>
1734<p>Expressions following <strong>#if</strong> and <strong>#elif</strong> are
1735restricted to <em>pp-constant-expressions</em> as defined below.</p>
1736</li>
1737<li>
1738<p>Undefined identifiers not consumed by the <strong>defined</strong> operator do not
1739default to '0'.
1740Use of such identifiers causes an error.</p>
1741</li>
1742<li>
1743<p>Character constants are not supported.</p>
1744</li>
1745</ul>
1746</div>
1747<div class="paragraph">
1748<p>As in C++, a macro name defined with an empty replacement list does not
1749default to '0' when used in a preprocessor expression.</p>
1750</div>
1751<div class="paragraph">
1752<p>A <em>pp-constant-expression</em> is an integral expression, evaluated at
1753compile-time during preprocessing and formed from literal integer constants
1754and the following operators:</p>
1755</div>
1756<table class="tableblock frame-all grid-all stretch">
1757<colgroup>
1758<col style="width: 25%;">
1759<col style="width: 25%;">
1760<col style="width: 25%;">
1761<col style="width: 25%;">
1762</colgroup>
1763<thead>
1764<tr>
1765<th class="tableblock halign-left valign-top">Precedence</th>
1766<th class="tableblock halign-left valign-top">Operator class</th>
1767<th class="tableblock halign-left valign-top">Operators</th>
1768<th class="tableblock halign-left valign-top">Associativity</th>
1769</tr>
1770</thead>
1771<tbody>
1772<tr>
1773<td class="tableblock halign-left valign-top"><p class="tableblock">1 (highest)</p></td>
1774<td class="tableblock halign-left valign-top"><p class="tableblock">parenthetical grouping</p></td>
1775<td class="tableblock halign-left valign-top"><p class="tableblock">( )</p></td>
1776<td class="tableblock halign-left valign-top"><p class="tableblock">NA</p></td>
1777</tr>
1778<tr>
1779<td class="tableblock halign-left valign-top"><p class="tableblock">2</p></td>
1780<td class="tableblock halign-left valign-top"><p class="tableblock">unary</p></td>
1781<td class="tableblock halign-left valign-top"><p class="tableblock">defined<br>
1782                                         + - ~ !</p></td>
1783<td class="tableblock halign-left valign-top"><p class="tableblock">Right to Left</p></td>
1784</tr>
1785<tr>
1786<td class="tableblock halign-left valign-top"><p class="tableblock">3</p></td>
1787<td class="tableblock halign-left valign-top"><p class="tableblock">multiplicative</p></td>
1788<td class="tableblock halign-left valign-top"><p class="tableblock">* / %</p></td>
1789<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1790</tr>
1791<tr>
1792<td class="tableblock halign-left valign-top"><p class="tableblock">4</p></td>
1793<td class="tableblock halign-left valign-top"><p class="tableblock">additive</p></td>
1794<td class="tableblock halign-left valign-top"><p class="tableblock">+ -</p></td>
1795<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1796</tr>
1797<tr>
1798<td class="tableblock halign-left valign-top"><p class="tableblock">5</p></td>
1799<td class="tableblock halign-left valign-top"><p class="tableblock">bit-wise shift</p></td>
1800<td class="tableblock halign-left valign-top"><p class="tableblock">&lt;&lt; &gt;&gt;</p></td>
1801<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1802</tr>
1803<tr>
1804<td class="tableblock halign-left valign-top"><p class="tableblock">6</p></td>
1805<td class="tableblock halign-left valign-top"><p class="tableblock">relational</p></td>
1806<td class="tableblock halign-left valign-top"><p class="tableblock">&lt; &gt; &lt;= &gt;=</p></td>
1807<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1808</tr>
1809<tr>
1810<td class="tableblock halign-left valign-top"><p class="tableblock">7</p></td>
1811<td class="tableblock halign-left valign-top"><p class="tableblock">equality</p></td>
1812<td class="tableblock halign-left valign-top"><p class="tableblock">== !=</p></td>
1813<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1814</tr>
1815<tr>
1816<td class="tableblock halign-left valign-top"><p class="tableblock">8</p></td>
1817<td class="tableblock halign-left valign-top"><p class="tableblock">bit-wise and</p></td>
1818<td class="tableblock halign-left valign-top"><p class="tableblock">&amp;</p></td>
1819<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1820</tr>
1821<tr>
1822<td class="tableblock halign-left valign-top"><p class="tableblock">9</p></td>
1823<td class="tableblock halign-left valign-top"><p class="tableblock">bit-wise exclusive or</p></td>
1824<td class="tableblock halign-left valign-top"><p class="tableblock">^</p></td>
1825<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1826</tr>
1827<tr>
1828<td class="tableblock halign-left valign-top"><p class="tableblock">10</p></td>
1829<td class="tableblock halign-left valign-top"><p class="tableblock">bit-wise inclusive or</p></td>
1830<td class="tableblock halign-left valign-top"><p class="tableblock">|</p></td>
1831<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1832</tr>
1833<tr>
1834<td class="tableblock halign-left valign-top"><p class="tableblock">11</p></td>
1835<td class="tableblock halign-left valign-top"><p class="tableblock">logical and</p></td>
1836<td class="tableblock halign-left valign-top"><p class="tableblock">&amp;&amp;</p></td>
1837<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1838</tr>
1839<tr>
1840<td class="tableblock halign-left valign-top"><p class="tableblock">12 (lowest)</p></td>
1841<td class="tableblock halign-left valign-top"><p class="tableblock">logical inclusive or</p></td>
1842<td class="tableblock halign-left valign-top"><p class="tableblock">||</p></td>
1843<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
1844</tr>
1845</tbody>
1846</table>
1847<div class="paragraph">
1848<p>The <strong>defined</strong> operator can be used in either of the following ways:</p>
1849</div>
1850<div class="listingblock">
1851<div class="content">
1852<pre class="CodeRay highlight"><code data-lang="c++">defined identifier
1853defined ( identifier )</code></pre>
1854</div>
1855</div>
1856<div class="paragraph">
1857<p>There are no number sign based operators (e.g. no <strong><mark></strong> or <strong></mark>@</strong>), no <strong>##</strong>
1858operator, nor is there a <strong>sizeof</strong> operator.</p>
1859</div>
1860<div class="paragraph">
1861<p>The semantics of applying operators in the preprocessor match those standard
1862in the C++ preprocessor with the following exceptions:</p>
1863</div>
1864<div class="ulist">
1865<ul>
1866<li>
1867<p>The 2<sup>nd</sup> operand in a logical and ('&amp;&amp;') operation is evaluated if and
1868only if the 1<sup>st</sup> operand evaluates to non-zero.</p>
1869</li>
1870<li>
1871<p>The 2<sup>nd</sup> operand in a logical or ('||') operation is evaluated if and
1872only if the 1<sup>st</sup> operand evaluates to zero.</p>
1873</li>
1874<li>
1875<p>There is no boolean type and no boolean literals.
1876A <em>true</em> or <em>false</em> result is returned as integer <em>one</em> or <em>zero</em>
1877respectively.
1878Wherever a boolean operand is expected, any non-zero integer is
1879interpreted as <em>true</em> and a zero integer as <em>false</em>.</p>
1880</li>
1881</ul>
1882</div>
1883<div class="paragraph">
1884<p>If an operand is not evaluated, the presence of undefined identifiers in the
1885operand will not cause an error.</p>
1886</div>
1887<div class="paragraph">
1888<p><strong>#error</strong> will cause the implementation to put a compile-time diagnostic message
1889into the shader object&#8217;s information log (see section 7.12 &#8220;Shader, Program
1890and Program Pipeline Queries&#8221; of the <a href="#references">OpenGL ES Specification</a> for how to
1891access a shader object&#8217;s information log).
1892The message will be the tokens following the <strong>#error</strong> directive, up to the
1893first new-line.
1894The implementation must treat the presence of a <strong>#error</strong> directive as a
1895compile-time error.</p>
1896</div>
1897<div class="paragraph">
1898<p><strong>#pragma</strong> allows implementation-dependent compiler control.
1899Tokens following <strong>#pragma</strong> are not subject to preprocessor macro expansion.
1900If an implementation does not recognize the tokens following <strong>#pragma</strong>, then
1901it will ignore that pragma.
1902The following pragmas are defined as part of the language.</p>
1903</div>
1904<div class="listingblock">
1905<div class="content">
1906<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#pragma</span> STDGL</code></pre>
1907</div>
1908</div>
1909<div class="paragraph">
1910<p>The <strong>STDGL</strong> pragma is used to reserve pragmas for use by future revisions of
1911this language.
1912No implementation may use a pragma whose first token is <strong>STDGL</strong>.</p>
1913</div>
1914<div class="listingblock">
1915<div class="content">
1916<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#pragma</span> optimize(on)
1917<span class="preprocessor">#pragma</span> optimize(off)</code></pre>
1918</div>
1919</div>
1920<div class="paragraph">
1921<p>can be used to turn off optimizations as an aid in developing and debugging
1922shaders.
1923It can only be used outside function definitions.
1924By default, optimization is turned on for all shaders.
1925The debug pragma</p>
1926</div>
1927<div class="listingblock">
1928<div class="content">
1929<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#pragma</span> debug(on)
1930<span class="preprocessor">#pragma</span> debug(off)</code></pre>
1931</div>
1932</div>
1933<div class="paragraph">
1934<p>can be used to enable compiling and annotating a shader with debug
1935information, so that it can be used with a debugger.
1936It can only be used outside function definitions.
1937By default, debug is turned off.</p>
1938</div>
1939<div class="paragraph">
1940<p>The scope as well as the effect of the optimize and debug pragmas is
1941implementation-dependent except that their use must not generate an error.
1942Incorrect use of predefined pragmas does not cause an error.</p>
1943</div>
1944<div class="paragraph">
1945<p>By default, compilers of this language must issue compile-time
1946syntactic, semantic,
1947and
1948grammatical errors for shaders that do not conform to this specification.
1949Any extended behavior must first be enabled.
1950Directives to control the behavior of the compiler with respect to
1951extensions are declared with the <strong>#extension</strong> directive</p>
1952</div>
1953<div class="listingblock">
1954<div class="content">
1955<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#extension</span> extension_name : behavior
1956<span class="preprocessor">#extension</span> all : behavior</code></pre>
1957</div>
1958</div>
1959<div class="paragraph">
1960<p>where <em>extension_name</em> is the name of an extension.
1961Extension names are not documented in this specification.
1962The token <strong>all</strong> means the behavior applies to all extensions supported by
1963the compiler.
1964The <em>behavior</em> can be one of the following:</p>
1965</div>
1966<table class="tableblock frame-all grid-all stretch">
1967<colgroup>
1968<col style="width: 50%;">
1969<col style="width: 50%;">
1970</colgroup>
1971<thead>
1972<tr>
1973<th class="tableblock halign-left valign-top">Behavior</th>
1974<th class="tableblock halign-left valign-top">Effect</th>
1975</tr>
1976</thead>
1977<tbody>
1978<tr>
1979<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>require</strong></p></td>
1980<td class="tableblock halign-left valign-top"><p class="tableblock">Behave as specified by the extension <em>extension_name</em>.<br>
1981              Give a compile-time error on the <strong>#extension</strong> if the extension
1982              <em>extension_name</em> is not supported, or if <strong>all</strong> is specified.</p></td>
1983</tr>
1984<tr>
1985<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>enable</strong></p></td>
1986<td class="tableblock halign-left valign-top"><p class="tableblock">Behave as specified by the extension <em>extension_name</em>.<br>
1987              Warn on the <strong>#extension</strong> if the extension <em>extension_name</em> is
1988              not supported.<br>
1989              Give an error on the <strong>#extension</strong> if <strong>all</strong> is
1990              specified.</p></td>
1991</tr>
1992<tr>
1993<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>warn</strong></p></td>
1994<td class="tableblock halign-left valign-top"><p class="tableblock">Behave as specified by the extension <em>extension_name</em>,
1995              except issue warnings on any detectable use of that extension,
1996              unless such use is supported by other enabled or required
1997              extensions.<br>
1998              If <strong>all</strong> is specified, then warn on all detectable uses of any
1999              extension used.<br>
2000              Warn on the <strong>#extension</strong> if the extension <em>extension_name</em> is
2001              not supported.</p></td>
2002</tr>
2003<tr>
2004<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>disable</strong></p></td>
2005<td class="tableblock halign-left valign-top"><p class="tableblock">Behave (including issuing errors and warnings) as if the
2006              extension <em>extension_name</em> is not part of the language
2007              definition.<br>
2008              If <strong>all</strong> is specified, then behavior must revert back to that
2009              of the non-extended core version of the language being
2010              compiled to.<br>
2011              Warn on the <strong>#extension</strong> if the extension <em>extension_name</em> is
2012              not supported.</p></td>
2013</tr>
2014</tbody>
2015</table>
2016<div class="paragraph">
2017<p>The <strong>extension</strong> directive is a simple, low-level mechanism to set the
2018behavior for each extension.
2019It does not define policies such as which combinations are appropriate,
2020those must be defined elsewhere.
2021Order of directives matters in setting the behavior for each extension:
2022Directives that occur later override those seen earlier.
2023The <strong>all</strong> variant sets the behavior for all extensions, overriding all
2024previously issued <strong>extension</strong> directives, but only for the <em>behaviors</em>
2025<strong>warn</strong> and <strong>disable</strong>.</p>
2026</div>
2027<div class="paragraph">
2028<p>The initial state of the compiler is as if the directive</p>
2029</div>
2030<div class="listingblock">
2031<div class="content">
2032<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#extension</span> all : disable</code></pre>
2033</div>
2034</div>
2035<div class="paragraph">
2036<p>was issued, telling the compiler that all error and warning reporting must
2037be done according to this specification, ignoring any extensions.</p>
2038</div>
2039<div class="paragraph">
2040<p>Each extension can define its allowed granularity of scope.
2041If nothing is said, the granularity is a shader (that is, a single
2042compilation unit), and the extension directives must occur before any
2043non-preprocessor tokens.
2044If necessary, the linker can enforce granularities larger than a single
2045compilation unit, in which case each involved shader will have to contain
2046the necessary extension directive.</p>
2047</div>
2048<div class="paragraph">
2049<p>Macro expansion is not done on lines containing <strong>#extension</strong> and <strong>#version</strong>
2050directives.</p>
2051</div>
2052<div class="paragraph">
2053<p>For each extension there is an associated macro.
2054The macro is always defined in an implementation that supports the
2055extension.
2056This allows the following construct to be used:</p>
2057</div>
2058<div class="listingblock">
2059<div class="content">
2060<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#ifdef</span> OES_extension_name
2061    <span class="preprocessor">#extension</span> OES_extension_name : enable
2062    <span class="comment">// code that requires the extension</span>
2063<span class="preprocessor">#else</span>
2064    <span class="comment">// alternative code</span>
2065<span class="preprocessor">#endif</span></code></pre>
2066</div>
2067</div>
2068<div class="paragraph">
2069<p><strong>#line</strong> must have, after macro substitution, one of the following forms:</p>
2070</div>
2071<div class="listingblock">
2072<div class="content">
2073<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#line</span> line
2074<span class="preprocessor">#line</span> line source-<span class="predefined-type">string</span>-number</code></pre>
2075</div>
2076</div>
2077<div class="paragraph">
2078<p>where <em>line</em> and <em>source-string-number</em> are
2079<em>pp-constant-expressions</em>.
2080If these constant expressions are not integer literals then behavior is undefined.
2081After processing this directive (including its new-line), the implementation
2082will behave as if it is compiling at line number <em>line</em> and source string
2083number <em>source-string-number</em>.
2084Subsequent source strings will be numbered sequentially, until another
2085<strong>#line</strong> directive overrides that numbering.</p>
2086</div>
2087<div class="admonitionblock note">
2088<table>
2089<tr>
2090<td class="icon">
2091<i class="fa icon-note" title="Note"></i>
2092</td>
2093<td class="content">
2094<div class="title">Note</div>
2095<div class="paragraph">
2096<p>Some implementations have allowed constant expressions in #line directives and
2097some have not. Even where expressions are supported the grammar is ambiguous and so
2098results are implementation dependent. For example,
2099+ #line +2 +2               // Line number set to 4, or file to 2 and line to 2</p>
2100</div>
2101</td>
2102</tr>
2103</table>
2104</div>
2105<div class="paragraph">
2106<p>When targeting Vulkan, the following predefined macro is available:</p>
2107</div>
2108<div class="listingblock">
2109<div class="content">
2110<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#define</span> VULKAN <span class="integer">100</span></code></pre>
2111</div>
2112</div>
2113<div class="paragraph">
2114<p>If during macro expansion a preprocessor directive is encountered, the
2115results are undefined; the compiler may or may not report an error in such
2116cases.</p>
2117</div>
2118</div>
2119<div class="sect2">
2120<h3 id="comments">3.5. Comments</h3>
2121<div class="paragraph">
2122<p>Comments are delimited by <strong>/*</strong> and <strong>*/</strong>, or by <strong>//</strong> and a new-line.
2123// style comments include the initial // marker and continue up to, but
2124not including, the terminating newline.
2125/*...*/ comments include both the start and end marker.
2126The begin comment delimiters (/* or //) are not recognized as comment
2127delimiters inside of a comment, hence comments cannot be nested.
2128Comments are treated syntactically as a single space.</p>
2129</div>
2130</div>
2131<div class="sect2">
2132<h3 id="tokens">3.6. Tokens</h3>
2133<div class="paragraph">
2134<p>The language, after preprocessing, is a sequence of tokens.
2135A token can be</p>
2136</div>
2137<div class="openblock bnf">
2138<div class="content">
2139<div class="dlist">
2140<dl>
2141<dt class="hdlist1"><em>token</em> : </dt>
2142<dd>
2143<p><em>keyword</em><br>
2144<em>identifier</em><br>
2145<em>integer-constant</em><br>
2146<em>floating-constant</em><br>
2147<em>operator</em><br>
2148<strong>;</strong> <strong>{</strong> <strong>}</strong></p>
2149</dd>
2150</dl>
2151</div>
2152</div>
2153</div>
2154</div>
2155<div class="sect2">
2156<h3 id="keywords">3.7. Keywords</h3>
2157<div class="paragraph">
2158<p>The following are the keywords in the language and (after
2159preprocessing) can only be used as described in this specification,
2160or an error results:</p>
2161</div>
2162<div class="dlist">
2163<dl>
2164<dt class="hdlist1"></dt>
2165<dd>
2166<p><strong>const</strong> <strong>uniform</strong> <strong>buffer</strong> <strong>shared</strong></p>
2167</dd>
2168<dt class="hdlist1"></dt>
2169<dd>
2170<p><strong>coherent</strong> <strong>volatile</strong> <strong>restrict</strong> <strong>readonly</strong> <strong>writeonly</strong></p>
2171</dd>
2172<dt class="hdlist1"></dt>
2173<dd>
2174<p><strong>atomic_uint</strong></p>
2175</dd>
2176<dt class="hdlist1"></dt>
2177<dd>
2178<p><strong>layout</strong></p>
2179</dd>
2180<dt class="hdlist1"></dt>
2181<dd>
2182<p><strong>centroid</strong> <strong>flat</strong> <strong>smooth</strong></p>
2183</dd>
2184<dt class="hdlist1"></dt>
2185<dd>
2186<p><strong>patch</strong> <strong>sample</strong></p>
2187</dd>
2188<dt class="hdlist1"></dt>
2189<dd>
2190<p><strong>invariant</strong> <strong>precise</strong></p>
2191</dd>
2192<dt class="hdlist1"></dt>
2193<dd>
2194<p><strong>break</strong> <strong>continue</strong> <strong>do</strong> <strong>for</strong> <strong>while</strong> <strong>switch</strong> <strong>case</strong> <strong>default</strong></p>
2195</dd>
2196<dt class="hdlist1"></dt>
2197<dd>
2198<p><strong>if</strong> <strong>else</strong></p>
2199</dd>
2200<dt class="hdlist1"></dt>
2201<dd>
2202<p><strong>in</strong> <strong>out</strong> <strong>inout</strong></p>
2203</dd>
2204<dt class="hdlist1"></dt>
2205<dd>
2206<p><strong>int</strong> <strong>void</strong> <strong>bool</strong> <strong>true</strong> <strong>false</strong> <strong>float</strong></p>
2207</dd>
2208<dt class="hdlist1"></dt>
2209<dd>
2210<p><strong>discard</strong> <strong>return</strong></p>
2211</dd>
2212<dt class="hdlist1"></dt>
2213<dd>
2214<p><strong>vec2</strong> <strong>vec3</strong> <strong>vec4</strong> <strong>ivec2</strong> <strong>ivec3</strong> <strong>ivec4</strong> <strong>bvec2</strong> <strong>bvec3</strong> <strong>bvec4</strong></p>
2215</dd>
2216<dt class="hdlist1"></dt>
2217<dd>
2218<p><strong>uint</strong> <strong>uvec2</strong> <strong>uvec3</strong> <strong>uvec4</strong></p>
2219</dd>
2220<dt class="hdlist1"></dt>
2221<dd>
2222<p><strong>mat2</strong> <strong>mat3</strong> <strong>mat4</strong></p>
2223</dd>
2224<dt class="hdlist1"></dt>
2225<dd>
2226<p><strong>mat2x2</strong> <strong>mat2x3</strong> <strong>mat2x4</strong></p>
2227</dd>
2228<dt class="hdlist1"></dt>
2229<dd>
2230<p><strong>mat3x2</strong> <strong>mat3x3</strong> <strong>mat3x4</strong></p>
2231</dd>
2232<dt class="hdlist1"></dt>
2233<dd>
2234<p><strong>mat4x2</strong> <strong>mat4x3</strong> <strong>mat4x4</strong></p>
2235</dd>
2236<dt class="hdlist1"></dt>
2237<dd>
2238<p><strong>lowp</strong> <strong>mediump</strong> <strong>highp</strong> <strong>precision</strong></p>
2239</dd>
2240<dt class="hdlist1"></dt>
2241<dd>
2242<p><strong>sampler2D</strong> <strong>sampler2DShadow</strong> <strong>sampler2DArray</strong> <strong>sampler2DArrayShadow</strong></p>
2243</dd>
2244<dt class="hdlist1"></dt>
2245<dd>
2246<p><strong>isampler2D</strong> <strong>isampler2DArray</strong> <strong>usampler2D</strong> <strong>usampler2DArray</strong></p>
2247</dd>
2248<dt class="hdlist1"></dt>
2249<dd>
2250<p><strong>sampler2DMS</strong> <strong>isampler2DMS</strong> <strong>usampler2DMS</strong></p>
2251</dd>
2252<dt class="hdlist1"></dt>
2253<dd>
2254<p><strong>sampler2DMSArray</strong> <strong>isampler2DMSArray</strong> <strong>usampler2DMSArray</strong></p>
2255</dd>
2256<dt class="hdlist1"></dt>
2257<dd>
2258<p><strong>sampler3D</strong> <strong>isampler3D</strong> <strong>usampler3D</strong></p>
2259</dd>
2260<dt class="hdlist1"></dt>
2261<dd>
2262<p><strong>samplerCube</strong> <strong>samplerCubeShadow</strong> <strong>isamplerCube</strong> <strong>usamplerCube</strong></p>
2263</dd>
2264<dt class="hdlist1"></dt>
2265<dd>
2266<p><strong>samplerCubeArray</strong> <strong>samplerCubeArrayShadow</strong></p>
2267</dd>
2268<dt class="hdlist1"></dt>
2269<dd>
2270<p><strong>isamplerCubeArray</strong> <strong>usamplerCubeArray</strong></p>
2271</dd>
2272<dt class="hdlist1"></dt>
2273<dd>
2274<p><strong>samplerBuffer</strong> <strong>isamplerBuffer</strong> <strong>usamplerBuffer</strong></p>
2275</dd>
2276<dt class="hdlist1"></dt>
2277<dd>
2278<p><strong>image2D</strong> <strong>iimage2D</strong> <strong>uimage2D</strong></p>
2279</dd>
2280<dt class="hdlist1"></dt>
2281<dd>
2282<p><strong>image2DArray</strong> <strong>iimage2DArray</strong> <strong>uimage2DArray</strong></p>
2283</dd>
2284<dt class="hdlist1"></dt>
2285<dd>
2286<p><strong>image3D</strong> <strong>iimage3D</strong> <strong>uimage3D</strong></p>
2287</dd>
2288<dt class="hdlist1"></dt>
2289<dd>
2290<p><strong>imageCube</strong> <strong>iimageCube</strong> <strong>uimageCube</strong></p>
2291</dd>
2292<dt class="hdlist1"></dt>
2293<dd>
2294<p><strong>imageCubeArray</strong> <strong>iimageCubeArray</strong> <strong>uimageCubeArray</strong></p>
2295</dd>
2296<dt class="hdlist1"></dt>
2297<dd>
2298<p><strong>imageBuffer</strong> <strong>iimageBuffer</strong> <strong>uimageBuffer</strong></p>
2299</dd>
2300<dt class="hdlist1"></dt>
2301<dd>
2302<p><strong>struct</strong></p>
2303</dd>
2304</dl>
2305</div>
2306<div class="paragraph">
2307<p>In addition, when targeting Vulkan, the following keywords also exist:</p>
2308</div>
2309<div class="dlist">
2310<dl>
2311<dt class="hdlist1"></dt>
2312<dd>
2313<p><strong>texture2D</strong> <strong>texture2DArray</strong></p>
2314</dd>
2315<dt class="hdlist1"></dt>
2316<dd>
2317<p><strong>itexture2D</strong> <strong>itexture2DArray</strong> <strong>utexture2D</strong> <strong>utexture2DArray</strong></p>
2318</dd>
2319<dt class="hdlist1"></dt>
2320<dd>
2321<p><strong>texture2DMS</strong> <strong>itexture2DMS</strong> <strong>utexture2DMS</strong></p>
2322</dd>
2323<dt class="hdlist1"></dt>
2324<dd>
2325<p><strong>texture2DMSArray</strong> <strong>itexture2DMSArray</strong> <strong>utexture2DMSArray</strong></p>
2326</dd>
2327<dt class="hdlist1"></dt>
2328<dd>
2329<p><strong>texture3D</strong> <strong>itexture3D</strong> <strong>utexture3D</strong></p>
2330</dd>
2331<dt class="hdlist1"></dt>
2332<dd>
2333<p><strong>textureCube</strong> <strong>itextureCube</strong> <strong>utextureCube</strong></p>
2334</dd>
2335<dt class="hdlist1"></dt>
2336<dd>
2337<p><strong>textureCubeArray</strong> <strong>itextureCubeArray</strong> <strong>utextureCubeArray</strong></p>
2338</dd>
2339<dt class="hdlist1"></dt>
2340<dd>
2341<p><strong>textureBuffer</strong> <strong>itextureBuffer</strong> <strong>utextureBuffer</strong></p>
2342</dd>
2343<dt class="hdlist1"></dt>
2344<dd>
2345<p><strong>sampler</strong> <strong>samplerShadow</strong></p>
2346</dd>
2347<dt class="hdlist1"></dt>
2348<dd>
2349<p><strong>subpassInput</strong> <strong>isubpassInput</strong> <strong>usubpassInput</strong></p>
2350</dd>
2351<dt class="hdlist1"></dt>
2352<dd>
2353<p><strong>subpassInputMS</strong> <strong>isubpassInputMS</strong> <strong>usubpassInputMS</strong></p>
2354</dd>
2355</dl>
2356</div>
2357<div class="paragraph">
2358<p>The following are the keywords reserved for future use.
2359Using them will result in an error:</p>
2360</div>
2361<div class="dlist">
2362<dl>
2363<dt class="hdlist1"></dt>
2364<dd>
2365<p><strong>attribute</strong> <strong>varying</strong></p>
2366</dd>
2367<dt class="hdlist1"></dt>
2368<dd>
2369<p><strong>noperspective</strong></p>
2370</dd>
2371<dt class="hdlist1"></dt>
2372<dd>
2373<p><strong>subroutine</strong></p>
2374</dd>
2375<dt class="hdlist1"></dt>
2376<dd>
2377<p><strong>common</strong> <strong>partition</strong> <strong>active</strong></p>
2378</dd>
2379<dt class="hdlist1"></dt>
2380<dd>
2381<p><strong>asm</strong></p>
2382</dd>
2383<dt class="hdlist1"></dt>
2384<dd>
2385<p><strong>class</strong> <strong>union</strong> <strong>enum</strong> <strong>typedef</strong> <strong>template</strong> <strong>this</strong></p>
2386</dd>
2387<dt class="hdlist1"></dt>
2388<dd>
2389<p><strong>resource</strong></p>
2390</dd>
2391<dt class="hdlist1"></dt>
2392<dd>
2393<p><strong>goto</strong></p>
2394</dd>
2395<dt class="hdlist1"></dt>
2396<dd>
2397<p><strong>inline</strong> <strong>noinline</strong> <strong>public</strong> <strong>static</strong> <strong>extern</strong> <strong>external</strong> <strong>interface</strong></p>
2398</dd>
2399<dt class="hdlist1"></dt>
2400<dd>
2401<p><strong>long</strong> <strong>short</strong> <strong>half</strong> <strong>fixed</strong> <strong>unsigned</strong> <strong>superp</strong>
2402<strong>double</strong></p>
2403</dd>
2404<dt class="hdlist1"></dt>
2405<dd>
2406<p><strong>input</strong> <strong>output</strong></p>
2407</dd>
2408<dt class="hdlist1"></dt>
2409<dd>
2410<p><strong>hvec2</strong> <strong>hvec3</strong> <strong>hvec4</strong> <strong>fvec2</strong> <strong>fvec3</strong> <strong>fvec4</strong></p>
2411</dd>
2412<dt class="hdlist1"></dt>
2413<dd>
2414<p><strong>dvec2</strong> <strong>dvec3</strong> <strong>dvec4</strong></p>
2415</dd>
2416<dt class="hdlist1"></dt>
2417<dd>
2418<p><strong>dmat2</strong> <strong>dmat3</strong> <strong>dmat4</strong></p>
2419</dd>
2420<dt class="hdlist1"></dt>
2421<dd>
2422<p><strong>dmat2x2</strong> <strong>dmat2x3</strong> <strong>dmat2x4</strong></p>
2423</dd>
2424<dt class="hdlist1"></dt>
2425<dd>
2426<p><strong>dmat3x2</strong> <strong>dmat3x3</strong> <strong>dmat3x4</strong></p>
2427</dd>
2428<dt class="hdlist1"></dt>
2429<dd>
2430<p><strong>dmat4x2</strong> <strong>dmat4x3</strong> <strong>dmat4x4</strong></p>
2431</dd>
2432<dt class="hdlist1"></dt>
2433<dd>
2434<p><strong>filter</strong></p>
2435</dd>
2436<dt class="hdlist1"></dt>
2437<dd>
2438<p><strong>sizeof</strong> <strong>cast</strong></p>
2439</dd>
2440<dt class="hdlist1"></dt>
2441<dd>
2442<p><strong>namespace</strong> <strong>using</strong></p>
2443</dd>
2444<dt class="hdlist1"></dt>
2445<dd>
2446<p><strong>sampler1D</strong> <strong>sampler1DShadow</strong> <strong>sampler1DArray</strong> <strong>sampler1DArrayShadow</strong></p>
2447</dd>
2448<dt class="hdlist1"></dt>
2449<dd>
2450<p><strong>isampler1D</strong> <strong>isampler1DArray</strong> <strong>usampler1D</strong> <strong>usampler1DArray</strong></p>
2451</dd>
2452<dt class="hdlist1"></dt>
2453<dd>
2454<p><strong>sampler2DRect</strong> <strong>sampler2DRectShadow</strong> <strong>isampler2DRect</strong> <strong>usampler2DRect</strong></p>
2455</dd>
2456<dt class="hdlist1"></dt>
2457<dd>
2458<p><strong>sampler3DRect</strong></p>
2459</dd>
2460<dt class="hdlist1"></dt>
2461<dd>
2462<p><strong>image1D</strong> <strong>iimage1D</strong> <strong>uimage1D</strong></p>
2463</dd>
2464<dt class="hdlist1"></dt>
2465<dd>
2466<p><strong>image1DArray</strong> <strong>iimage1DArray</strong> <strong>uimage1DArray</strong></p>
2467</dd>
2468<dt class="hdlist1"></dt>
2469<dd>
2470<p><strong>image2DRect</strong> <strong>iimage2DRect</strong> <strong>uimage2DRect</strong></p>
2471</dd>
2472<dt class="hdlist1"></dt>
2473<dd>
2474<p><strong>image2DMS</strong> <strong>iimage2DMS</strong> <strong>uimage2DMS</strong></p>
2475</dd>
2476<dt class="hdlist1"></dt>
2477<dd>
2478<p><strong>image2DMSArray</strong> <strong>iimage2DMSArray</strong> <strong>uimage2DMSArray</strong></p>
2479</dd>
2480</dl>
2481</div>
2482<div class="paragraph">
2483<p>In addition, all identifiers containing two consecutive underscores (__)
2484are reserved for use by underlying software layers.
2485Defining such a name in a shader does not itself result in an error, but may
2486result in unintended behaviors that stem from having multiple definitions of
2487the same name.</p>
2488</div>
2489</div>
2490<div class="sect2">
2491<h3 id="identifiers">3.8. Identifiers</h3>
2492<div class="paragraph">
2493<p>Identifiers are used for variable names, function names, structure names,
2494and field selectors (field selectors select components of
2495<code><a href="#vector-components">vectors</a></code> and <code><a href="#matrix-components">matrices</a></code>,
2496similarly to structure members).
2497Identifiers have the form:</p>
2498</div>
2499<div class="openblock bnf">
2500<div class="content">
2501<div class="dlist">
2502<dl>
2503<dt class="hdlist1"><em>identifier</em> : </dt>
2504<dd>
2505<p><em>nondigit</em><br>
2506<em>identifier</em> <em>nondigit</em><br>
2507<em>identifier</em> <em>digit</em></p>
2508</dd>
2509<dt class="hdlist1"><em>nondigit</em> : one of </dt>
2510<dd>
2511<p><strong>_</strong> <strong>a b c d e f g h i j k l m n o p q r s t u v w x y z</strong><br>
2512<strong>A B C D E F G H I J K L M N O P Q R S T U V W X Y Z</strong></p>
2513</dd>
2514<dt class="hdlist1"><em>digit</em> : one of </dt>
2515<dd>
2516<p><strong>0 1 2 3 4 5 6 7 8 9</strong></p>
2517</dd>
2518</dl>
2519</div>
2520</div>
2521</div>
2522<div class="paragraph">
2523<p>Identifiers starting with &#8220;gl_&#8221; are reserved, and
2524in general, may not be declared in a shader;
2525this results in an error.
2526However, as noted in the specification, there are some cases where
2527previously declared variables can be redeclared, and predeclared &#8220;gl_&#8221;
2528names are allowed to be redeclared in a shader only for these specific
2529purposes.</p>
2530</div>
2531<div class="paragraph">
2532<p>Implementations must support identifier lengths of up to 1024 characters.
2533It is an error if the length exceeds this value.</p>
2534</div>
2535</div>
2536<div class="sect2">
2537<h3 id="definitions">3.9. Definitions</h3>
2538<div class="paragraph">
2539<p>Some language rules described below depend on the following definitions.</p>
2540</div>
2541<div class="sect3">
2542<h4 id="static-use">3.9.1. Static Use</h4>
2543<div class="paragraph">
2544<p>A shader contains a <em>static use</em> of a variable <em>x</em> if, after preprocessing,
2545the shader contains a statement that would access any part of <em>x</em>,
2546whether or not flow of control will cause that statement to be executed.
2547Such a variable is referred to as being <em>statically used</em>. If the access is a
2548write then <em>x</em> is further said to be <em>statically assigned</em>.</p>
2549</div>
2550</div>
2551<div class="sect3">
2552<h4 id="uniform-and-non-uniform-control-flow">3.9.2. Uniform and Non-Uniform Control Flow</h4>
2553<div class="paragraph">
2554<p>When executing statements in a fragment shader, control flow starts as
2555<em>uniform control flow</em>; all fragments enter the same control path into
2556<em>main()</em>.
2557Control flow becomes <em>non-uniform</em> when different fragments take different
2558paths through control-flow statements (selection, iteration, and jumps).
2559Control flow subsequently returns to being uniform after such divergent
2560sub-statements or skipped code completes, until the next time different
2561control paths are taken.</p>
2562</div>
2563<div class="paragraph">
2564<p>For example:</p>
2565</div>
2566<div class="listingblock">
2567<div class="content">
2568<pre class="CodeRay highlight"><code data-lang="c++">main()
2569{
2570    <span class="predefined-type">float</span> a = ...; <span class="comment">// this is uniform control flow</span>
2571    <span class="keyword">if</span> (a &lt; b) {   <span class="comment">// this expression is true for some fragments, not all</span>
2572        ...;       <span class="comment">// non-uniform control flow</span>
2573    } <span class="keyword">else</span> {
2574        ...;       <span class="comment">// non-uniform control flow</span>
2575    }
2576    ...;           <span class="comment">// uniform control flow again</span>
2577}</code></pre>
2578</div>
2579</div>
2580<div class="paragraph">
2581<p>Other examples of non-uniform control flow can occur within switch
2582statements and after conditional breaks, continues, early returns, and after
2583fragment discards, when the condition is true for some fragments but not
2584others.
2585Loop iterations that only some fragments execute are also non-uniform
2586control flow.</p>
2587</div>
2588<div class="paragraph">
2589<p>This is similarly defined for other shader stages, based on the per-instance
2590data items they process.</p>
2591</div>
2592</div>
2593<div class="sect3">
2594<h4 id="dynamically-uniform-expressions">3.9.3. Dynamically Uniform Expressions</h4>
2595<div class="paragraph">
2596<p>A fragment-shader expression is <em>dynamically uniform</em> if all fragments
2597evaluating it get the same resulting value.
2598When loops are involved, this refers to the expression&#8217;s value for the same
2599loop iteration.
2600When functions are involved, this refers to calls from the same call point.</p>
2601</div>
2602<div class="paragraph">
2603<p>This is similarly defined for other shader stages, based on the per-instance
2604data they process.</p>
2605</div>
2606<div class="paragraph">
2607<p>Note that constant expressions are trivially dynamically uniform.
2608It follows that typical loop counters based on these are also dynamically
2609uniform.</p>
2610</div>
2611</div>
2612</div>
2613<div class="sect2">
2614<h3 id="logical-phases-of-compilation">3.10. Logical Phases of Compilation</h3>
2615<div class="paragraph">
2616<p>The compilation units for the shader processors are processed separately
2617before optionally being linked together in the final stage of compilation.
2618The logical phases of compilation are:</p>
2619</div>
2620<div class="olist arabic">
2621<ol class="arabic">
2622<li>
2623<p>Source strings are input as byte sequences.
2624The value 'zero' is interpreted as a terminator.</p>
2625</li>
2626<li>
2627<p>Source strings are concatenated to form a single input.
2628Zero bytes are discarded but all other values are retained.</p>
2629</li>
2630<li>
2631<p>Each string is interpreted according to the UTF-8 standard, with the
2632exception that all invalid byte sequences are retained in their original
2633form for subsequent processing.</p>
2634</li>
2635<li>
2636<p>Each {carriage-return, line-feed} and {line-feed, carriage return}
2637sequence is replaced by a single newline.
2638All remaining carriage-return and line-feed characters are then each
2639replaced by a newline.</p>
2640</li>
2641<li>
2642<p>Line numbering for each character, which is equal to the number of
2643preceding newlines plus one, is noted.
2644Note this can only be subsequently changed by the #line directive and is
2645not affected by the removal of newlines in phase 6 of compilation.</p>
2646</li>
2647<li>
2648<p>Wherever a backslash ('\') occurs immediately before a newline, both are
2649deleted.
2650Note that no whitespace is substituted, thereby allowing a single
2651preprocessing token to span a newline.
2652This operation is not recursive; any new {backslash newline} sequences
2653generated are not removed.</p>
2654</li>
2655<li>
2656<p>All comments are replaced with a single space.
2657All (non-zero) characters and invalid UTF-8 byte sequences are allowed
2658within comments.
2659'//' style comments include the initial '//' marker and continue up to,
2660but not including, the terminating newline.
2661'/<strong>&#8230;&#8203;</strong>/' comments include both the start and end marker.</p>
2662</li>
2663<li>
2664<p>The source string is converted into a sequence of preprocessing tokens.
2665These tokens include preprocessing numbers, identifiers and
2666preprocessing operations.
2667The line number associated with each token is copied from the line
2668number of the first character of the token.</p>
2669</li>
2670<li>
2671<p>The preprocessor is run.
2672Directives are executed and macro expansion is performed.</p>
2673</li>
2674<li>
2675<p>White space and newlines are discarded.</p>
2676</li>
2677<li>
2678<p>Preprocessing tokens are converted into tokens.</p>
2679</li>
2680<li>
2681<p>The syntax is analyzed according to the GLSL ES grammar.</p>
2682</li>
2683<li>
2684<p>The result is checked according to the semantic rules of the language.</p>
2685</li>
2686<li>
2687<p>Optionally, the shaders are linked together to form one or more programs
2688or separable programs.
2689When a pair of shaders from consecutive stages are linked into the same
2690program, any outputs and corresponding inputs not used in both shaders
2691may be discarded.</p>
2692</li>
2693<li>
2694<p>The binary is generated.</p>
2695</li>
2696</ol>
2697</div>
2698</div>
2699</div>
2700</div>
2701<div class="sect1">
2702<h2 id="variables-and-types">4. Variables and Types</h2>
2703<div class="sectionbody">
2704<div class="paragraph">
2705<p>All variables and functions must be declared before being used.
2706Variable and function names are identifiers.</p>
2707</div>
2708<div class="paragraph">
2709<p>There are no default types.
2710All variable and function declarations must have a declared type, and
2711optionally qualifiers.
2712A variable is declared by specifying its type followed by one or more names
2713separated by commas.
2714In many cases, a variable can be initialized as part of its declaration by
2715using the assignment operator (<strong>=</strong>).</p>
2716</div>
2717<div class="paragraph">
2718<p>User-defined types may be defined using <strong>struct</strong> to aggregate a list of
2719existing types into a single name.</p>
2720</div>
2721<div class="paragraph">
2722<p>The OpenGL ES Shading Language is type safe.
2723There are no implicit conversions between types.</p>
2724</div>
2725<div class="sect2">
2726<h3 id="basic-types">4.1. Basic Types</h3>
2727<div class="dlist">
2728<dl>
2729<dt class="hdlist1">Definition</dt>
2730<dd>
2731<p>A <em>basic type</em> is a type defined by a keyword in the language.</p>
2732</dd>
2733</dl>
2734</div>
2735<div class="paragraph">
2736<p>The OpenGL ES Shading Language supports the following basic data types, grouped as follows.</p>
2737</div>
2738<div class="paragraph">
2739<p><strong>Transparent Types</strong></p>
2740</div>
2741<table class="tableblock frame-all grid-all stretch">
2742<colgroup>
2743<col style="width: 50%;">
2744<col style="width: 50%;">
2745</colgroup>
2746<thead>
2747<tr>
2748<th class="tableblock halign-left valign-top">Type</th>
2749<th class="tableblock halign-left valign-top">Meaning</th>
2750</tr>
2751</thead>
2752<tbody>
2753<tr>
2754<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>void</strong></p></td>
2755<td class="tableblock halign-left valign-top"><p class="tableblock">for functions that do not return a value</p></td>
2756</tr>
2757<tr>
2758<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>bool</strong></p></td>
2759<td class="tableblock halign-left valign-top"><p class="tableblock">a conditional type, taking on values of true or false</p></td>
2760</tr>
2761<tr>
2762<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>int</strong></p></td>
2763<td class="tableblock halign-left valign-top"><p class="tableblock">a signed integer</p></td>
2764</tr>
2765<tr>
2766<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>uint</strong></p></td>
2767<td class="tableblock halign-left valign-top"><p class="tableblock">an unsigned integer</p></td>
2768</tr>
2769<tr>
2770<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>float</strong></p></td>
2771<td class="tableblock halign-left valign-top"><p class="tableblock">a single-precision floating-point scalar</p></td>
2772</tr>
2773<tr>
2774<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>vec2</strong></p></td>
2775<td class="tableblock halign-left valign-top"><p class="tableblock">a two-component single-precision floating-point vector</p></td>
2776</tr>
2777<tr>
2778<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>vec3</strong></p></td>
2779<td class="tableblock halign-left valign-top"><p class="tableblock">a three-component single-precision floating-point vector</p></td>
2780</tr>
2781<tr>
2782<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>vec4</strong></p></td>
2783<td class="tableblock halign-left valign-top"><p class="tableblock">a four-component single-precision floating-point vector</p></td>
2784</tr>
2785<tr>
2786<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>bvec2</strong></p></td>
2787<td class="tableblock halign-left valign-top"><p class="tableblock">a two-component Boolean vector</p></td>
2788</tr>
2789<tr>
2790<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>bvec3</strong></p></td>
2791<td class="tableblock halign-left valign-top"><p class="tableblock">a three-component Boolean vector</p></td>
2792</tr>
2793<tr>
2794<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>bvec4</strong></p></td>
2795<td class="tableblock halign-left valign-top"><p class="tableblock">a four-component Boolean vector</p></td>
2796</tr>
2797<tr>
2798<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>ivec2</strong></p></td>
2799<td class="tableblock halign-left valign-top"><p class="tableblock">a two-component signed integer vector</p></td>
2800</tr>
2801<tr>
2802<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>ivec3</strong></p></td>
2803<td class="tableblock halign-left valign-top"><p class="tableblock">a three-component signed integer vector</p></td>
2804</tr>
2805<tr>
2806<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>ivec4</strong></p></td>
2807<td class="tableblock halign-left valign-top"><p class="tableblock">a four-component signed integer vector</p></td>
2808</tr>
2809<tr>
2810<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>uvec2</strong></p></td>
2811<td class="tableblock halign-left valign-top"><p class="tableblock">a two-component unsigned integer vector</p></td>
2812</tr>
2813<tr>
2814<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>uvec3</strong></p></td>
2815<td class="tableblock halign-left valign-top"><p class="tableblock">a three-component unsigned integer vector</p></td>
2816</tr>
2817<tr>
2818<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>uvec4</strong></p></td>
2819<td class="tableblock halign-left valign-top"><p class="tableblock">a four-component unsigned integer vector</p></td>
2820</tr>
2821<tr>
2822<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat2</strong></p></td>
2823<td class="tableblock halign-left valign-top"><p class="tableblock">a 2 × 2 single-precision floating-point matrix</p></td>
2824</tr>
2825<tr>
2826<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat3</strong></p></td>
2827<td class="tableblock halign-left valign-top"><p class="tableblock">a 3 × 3 single-precision floating-point matrix</p></td>
2828</tr>
2829<tr>
2830<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat4</strong></p></td>
2831<td class="tableblock halign-left valign-top"><p class="tableblock">a 4 × 4 single-precision floating-point matrix</p></td>
2832</tr>
2833<tr>
2834<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat2x2</strong></p></td>
2835<td class="tableblock halign-left valign-top"><p class="tableblock">same as a <strong>mat2</strong></p></td>
2836</tr>
2837<tr>
2838<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat2x3</strong></p></td>
2839<td class="tableblock halign-left valign-top"><p class="tableblock">a single-precision floating-point matrix with 2 columns and 3 rows</p></td>
2840</tr>
2841<tr>
2842<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat2x4</strong></p></td>
2843<td class="tableblock halign-left valign-top"><p class="tableblock">a single-precision floating-point matrix with 2 columns and 4 rows</p></td>
2844</tr>
2845<tr>
2846<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat3x2</strong></p></td>
2847<td class="tableblock halign-left valign-top"><p class="tableblock">a single-precision floating-point matrix with 3 columns and 2 rows</p></td>
2848</tr>
2849<tr>
2850<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat3x3</strong></p></td>
2851<td class="tableblock halign-left valign-top"><p class="tableblock">same as a <strong>mat3</strong></p></td>
2852</tr>
2853<tr>
2854<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat3x4</strong></p></td>
2855<td class="tableblock halign-left valign-top"><p class="tableblock">a single-precision floating-point matrix with 3 columns and 4 rows</p></td>
2856</tr>
2857<tr>
2858<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat4x2</strong></p></td>
2859<td class="tableblock halign-left valign-top"><p class="tableblock">a single-precision floating-point matrix with 4 columns and 2 rows</p></td>
2860</tr>
2861<tr>
2862<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat4x3</strong></p></td>
2863<td class="tableblock halign-left valign-top"><p class="tableblock">a single-precision floating-point matrix with 4 columns and 3 rows</p></td>
2864</tr>
2865<tr>
2866<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mat4x4</strong></p></td>
2867<td class="tableblock halign-left valign-top"><p class="tableblock">same as a <strong>mat4</strong></p></td>
2868</tr>
2869</tbody>
2870</table>
2871<div class="paragraph">
2872<p>Note that where the following tables say &#8220;accessing a texture&#8221;, the
2873<strong>sampler*</strong> opaque types access textures, and the <strong>image*</strong> opaque types
2874access images, of a specified type.</p>
2875</div>
2876<div class="paragraph">
2877<p><strong>Floating-Point Opaque Types</strong></p>
2878</div>
2879<table class="tableblock frame-all grid-all stretch">
2880<colgroup>
2881<col style="width: 50%;">
2882<col style="width: 50%;">
2883</colgroup>
2884<thead>
2885<tr>
2886<th class="tableblock halign-left valign-top">Type</th>
2887<th class="tableblock halign-left valign-top">Meaning</th>
2888</tr>
2889</thead>
2890<tbody>
2891<tr>
2892<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sampler2D</strong><br>
2893  <strong>texture2D</strong><br>
2894  <strong>image2D</strong></p></td>
2895<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a 2D texture</p></td>
2896</tr>
2897<tr>
2898<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sampler2DShadow</strong></p></td>
2899<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a 2D depth texture with comparison</p></td>
2900</tr>
2901<tr>
2902<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sampler2DArray</strong><br>
2903  <strong>texture2DArray</strong><br>
2904  <strong>image2DArray</strong></p></td>
2905<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a 2D array texture</p></td>
2906</tr>
2907<tr>
2908<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sampler2DArrayShadow</strong></p></td>
2909<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a 2D array depth texture with comparison</p></td>
2910</tr>
2911<tr>
2912<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sampler2DMS</strong><br>
2913  <strong>texture2DMS</strong></p></td>
2914<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a 2D multisample texture</p></td>
2915</tr>
2916<tr>
2917<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sampler2DMSArray</strong></p></td>
2918<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a 2D multisample array texture</p></td>
2919</tr>
2920<tr>
2921<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sampler3D</strong><br>
2922  <strong>texture3D</strong><br>
2923  <strong>image3D</strong></p></td>
2924<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a 3D texture</p></td>
2925</tr>
2926<tr>
2927<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>samplerCube</strong><br>
2928  <strong>textureCube</strong><br>
2929  <strong>imageCube</strong></p></td>
2930<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a cube mapped texture</p></td>
2931</tr>
2932<tr>
2933<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>samplerCubeShadow</strong></p></td>
2934<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a cube map depth texture with comparison</p></td>
2935</tr>
2936<tr>
2937<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>samplerCubeArray</strong><br>
2938  <strong>textureCubeArray</strong><br>
2939  <strong>imageCubeArray</strong></p></td>
2940<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a cube map array texture</p></td>
2941</tr>
2942<tr>
2943<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>samplerCubeArrayShadow</strong></p></td>
2944<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a cube map array depth texture with comparison</p></td>
2945</tr>
2946<tr>
2947<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>samplerBuffer</strong><br>
2948  <strong>textureBuffer</strong><br>
2949  <strong>imageBuffer</strong></p></td>
2950<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a buffer texture</p></td>
2951</tr>
2952<tr>
2953<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>subpassInput</strong></p></td>
2954<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a floating-point subpass input</p></td>
2955</tr>
2956<tr>
2957<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>subpassInputMS</strong></p></td>
2958<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a multi-sampled floating-point subpass input</p></td>
2959</tr>
2960</tbody>
2961</table>
2962<div class="paragraph">
2963<p><strong>Signed Integer Opaque Types</strong></p>
2964</div>
2965<table class="tableblock frame-all grid-all stretch">
2966<colgroup>
2967<col style="width: 50%;">
2968<col style="width: 50%;">
2969</colgroup>
2970<thead>
2971<tr>
2972<th class="tableblock halign-left valign-top">Type</th>
2973<th class="tableblock halign-left valign-top">Meaning</th>
2974</tr>
2975</thead>
2976<tbody>
2977<tr>
2978<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isampler2D</strong><br>
2979  <strong>itexture2D</strong><br>
2980  <strong>iimage2D</strong></p></td>
2981<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an integer 2D texture</p></td>
2982</tr>
2983<tr>
2984<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isampler2DArray</strong><br>
2985  <strong>itexture2DArray</strong><br>
2986  <strong>iimage2DArray</strong></p></td>
2987<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an integer 2D array texture</p></td>
2988</tr>
2989<tr>
2990<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isampler2DMS</strong><br>
2991  <strong>itexture2DMS</strong></p></td>
2992<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an integer 2D multisample texture</p></td>
2993</tr>
2994<tr>
2995<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isampler2DMSArray</strong><br>
2996  <strong>itexture2DMSArray</strong></p></td>
2997<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an integer 2D multisample array texture</p></td>
2998</tr>
2999<tr>
3000<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isampler3D</strong><br>
3001  <strong>itexture3D</strong><br>
3002  <strong>iimage3D</strong></p></td>
3003<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an integer 3D texture</p></td>
3004</tr>
3005<tr>
3006<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isamplerCube</strong><br>
3007  <strong>itextureCube</strong><br>
3008  <strong>iimageCube</strong></p></td>
3009<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an integer cube mapped texture</p></td>
3010</tr>
3011<tr>
3012<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isamplerCubeArray</strong><br>
3013  <strong>itextureCubeArray</strong><br>
3014  <strong>iimageCubeArray</strong></p></td>
3015<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an integer cube map array texture</p></td>
3016</tr>
3017<tr>
3018<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isamplerBuffer</strong><br>
3019  <strong>itextureBuffer</strong><br>
3020  <strong>iimageBuffer</strong></p></td>
3021<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an integer buffer texture</p></td>
3022</tr>
3023<tr>
3024<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isubpassInput</strong></p></td>
3025<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an integer subpass input</p></td>
3026</tr>
3027<tr>
3028<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>isubpassInputMS</strong></p></td>
3029<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a multi-sampled integer subpass input</p></td>
3030</tr>
3031</tbody>
3032</table>
3033<div class="paragraph">
3034<p><strong>Unsigned Integer Opaque Types</strong></p>
3035</div>
3036<table class="tableblock frame-all grid-all stretch">
3037<colgroup>
3038<col style="width: 50%;">
3039<col style="width: 50%;">
3040</colgroup>
3041<thead>
3042<tr>
3043<th class="tableblock halign-left valign-top">Type</th>
3044<th class="tableblock halign-left valign-top">Meaning</th>
3045</tr>
3046</thead>
3047<tbody>
3048<tr>
3049<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usampler2D</strong><br>
3050  <strong>utexture2D</strong><br>
3051  <strong>uimage2D</strong></p></td>
3052<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned integer 2D texture</p></td>
3053</tr>
3054<tr>
3055<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usampler2DArray</strong><br>
3056  <strong>utexture1DArray</strong><br>
3057  <strong>uimage2DArray</strong></p></td>
3058<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned integer 2D array texture</p></td>
3059</tr>
3060<tr>
3061<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usampler2DMS</strong><br>
3062  <strong>utexture2DMS</strong></p></td>
3063<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned integer 2D multisample texture</p></td>
3064</tr>
3065<tr>
3066<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usampler2DMSArray</strong><br>
3067  <strong>utexture2DMSArray</strong></p></td>
3068<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned integer 2D multisample array texture</p></td>
3069</tr>
3070<tr>
3071<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usampler3D</strong><br>
3072  <strong>utexture3D</strong><br>
3073  <strong>uimage3D</strong></p></td>
3074<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned integer 3D texture</p></td>
3075</tr>
3076<tr>
3077<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usamplerCube</strong><br>
3078  <strong>utextureCube</strong><br>
3079  <strong>uimageCube</strong></p></td>
3080<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned integer cube mapped texture</p></td>
3081</tr>
3082<tr>
3083<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usamplerCubeArray</strong><br>
3084  <strong>utextureCubeArray</strong><br>
3085  <strong>uimageCubeArray</strong></p></td>
3086<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned integer cube map array texture</p></td>
3087</tr>
3088<tr>
3089<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usamplerBuffer</strong><br>
3090  <strong>utextureBuffer</strong><br>
3091  <strong>uimageBuffer</strong></p></td>
3092<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned integer buffer texture</p></td>
3093</tr>
3094<tr>
3095<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>atomic_uint</strong></p></td>
3096<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned integer atomic counter</p></td>
3097</tr>
3098<tr>
3099<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usubpassInput</strong></p></td>
3100<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing an unsigned-integer subpass input</p></td>
3101</tr>
3102<tr>
3103<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>usubpassInputMS</strong></p></td>
3104<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing a multi-sampled unsigned-integer subpass input</p></td>
3105</tr>
3106</tbody>
3107</table>
3108<div class="paragraph">
3109<p><strong>Sampler Opaque Types</strong></p>
3110</div>
3111<table class="tableblock frame-all grid-all stretch">
3112<colgroup>
3113<col style="width: 50%;">
3114<col style="width: 50%;">
3115</colgroup>
3116<thead>
3117<tr>
3118<th class="tableblock halign-left valign-top">Type</th>
3119<th class="tableblock halign-left valign-top">Meaning</th>
3120</tr>
3121</thead>
3122<tbody>
3123<tr>
3124<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sampler</strong></p></td>
3125<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing state describing how to sample a texture</p></td>
3126</tr>
3127<tr>
3128<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>samplerShadow</strong></p></td>
3129<td class="tableblock halign-left valign-top"><p class="tableblock">a handle for accessing state describing how to sample a depth
3130                            texture with comparison</p></td>
3131</tr>
3132</tbody>
3133</table>
3134<div class="paragraph">
3135<p>In addition, a shader can aggregate these basic types using arrays and
3136structures to build more complex types.</p>
3137</div>
3138<div class="paragraph">
3139<p>There are no pointer types.</p>
3140</div>
3141<div class="sect3">
3142<h4 id="void">4.1.1. Void</h4>
3143<div class="paragraph">
3144<p>Functions that do not return a value must be declared as <strong>void</strong>.
3145There is no default function return type.
3146The keyword <strong>void</strong> cannot be used in any other declarations (except for
3147empty formal or actual parameter lists), or an error results.</p>
3148</div>
3149</div>
3150<div class="sect3">
3151<h4 id="booleans">4.1.2. Booleans</h4>
3152<div class="dlist">
3153<dl>
3154<dt class="hdlist1">Definition</dt>
3155<dd>
3156<p>A <em>boolean type</em> is any boolean scalar or vector type (<strong>bool</strong>, <strong>bvec2</strong>,
3157<strong>bvec3</strong>, <strong>bvec4</strong>)</p>
3158</dd>
3159</dl>
3160</div>
3161<div class="paragraph">
3162<p>To make conditional execution of code easier to express, the type <strong>bool</strong> is
3163supported.
3164There is no expectation that hardware directly supports variables of this
3165type.
3166It is a genuine Boolean type, holding only one of two values meaning either
3167true or false.
3168Two keywords <strong>true</strong> and <strong>false</strong> can be used as literal Boolean constants.
3169Booleans are declared and optionally initialized as in the follow example:</p>
3170</div>
3171<div class="listingblock">
3172<div class="content">
3173<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">bool</span> success;      <span class="comment">// declare &quot;success&quot; to be a Boolean</span>
3174<span class="predefined-type">bool</span> done = <span class="predefined-constant">false</span>; <span class="comment">// declare and initialize &quot;done&quot;</span></code></pre>
3175</div>
3176</div>
3177<div class="paragraph">
3178<p>Expressions used for conditional jumps (<strong>if</strong>, <strong>for</strong>, <strong>?:</strong>, <strong>while</strong>,
3179<strong>do</strong>-<strong>while</strong>) must evaluate to the type <strong>bool</strong>.</p>
3180</div>
3181</div>
3182<div class="sect3">
3183<h4 id="integers">4.1.3. Integers</h4>
3184<div class="dlist">
3185<dl>
3186<dt class="hdlist1">Definitions</dt>
3187<dd>
3188<p>An <em>integral type</em> is any signed or unsigned, scalar or vector integer type.
3189It excludes arrays and structures.</p>
3190</dd>
3191<dt class="hdlist1"></dt>
3192<dd>
3193<p>A <em>scalar integral type</em> is a scalar signed or unsigned integer type:</p>
3194</dd>
3195<dt class="hdlist1"></dt>
3196<dd>
3197<p>A <em>vector integral type</em> is a vector of signed or unsigned integers:</p>
3198</dd>
3199</dl>
3200</div>
3201<div class="paragraph">
3202<p>Signed and unsigned integer variables are fully supported.
3203In this document, the term <em>integer</em> is meant to generally include both
3204signed and unsigned integers.</p>
3205</div>
3206<div class="paragraph">
3207<p><strong>highp</strong>
3208unsigned integers have exactly 32 bits of precision.</p>
3209</div>
3210<div class="paragraph">
3211<p><strong>highp</strong>
3212signed integers use 32 bits, including a sign bit, in two&#8217;s complement form.</p>
3213</div>
3214<div class="paragraph">
3215<p>When targeting Vulkan, <strong>mediump</strong> and <strong>lowp</strong> integers are as defined by the
3216SPIR-V <strong>RelaxedPrecision</strong> decoration.
3217Otherwise, <strong>mediump</strong> and <strong>lowp</strong> integers have implementation-defined numbers of bits.
3218See &#8220;<a href="#range-and-precision">Range and Precision</a>&#8221; for details.</p>
3219</div>
3220<div class="paragraph">
3221<p>For all precisions, addition,
3222subtraction and multiplication resulting in overflow or
3223underflow will result in the low-order
3224n
3225bits of the correct result R, where
3226n is the size in bits of the integer and
3227R is computed with enough precision to avoid overflow or underflow.
3228Division resulting in overflow will result in an undefined value.</p>
3229</div>
3230<div class="paragraph">
3231<p>Integers are declared and optionally initialized with integer expressions,
3232as in the following example:</p>
3233</div>
3234<div class="listingblock">
3235<div class="content">
3236<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span> i, j = <span class="integer">42</span>; <span class="comment">// default integer literal type is int</span>
3237uint k = <span class="integer">3</span>u;   <span class="comment">// &quot;u&quot; establishes the type as uint</span></code></pre>
3238</div>
3239</div>
3240<div class="paragraph">
3241<p>Literal integer constants can be expressed in decimal (base 10), octal (base
32428), or hexadecimal (base 16) as follows.</p>
3243</div>
3244<div class="openblock bnf">
3245<div class="content">
3246<div class="dlist">
3247<dl>
3248<dt class="hdlist1"><em>integer-constant</em> : </dt>
3249<dd>
3250<p><em>decimal-constant</em> <em>integer-suffix<sub>opt</sub></em><br>
3251<em>octal-constant</em> <em>integer-suffix<sub>opt</sub></em><br>
3252<em>hexadecimal-constant</em> <em>integer-suffix<sub>opt</sub></em></p>
3253</dd>
3254<dt class="hdlist1"><em>integer-suffix</em> : one of </dt>
3255<dd>
3256<p><strong>u</strong> <strong>U</strong></p>
3257</dd>
3258<dt class="hdlist1"><em>decimal-constant</em> : </dt>
3259<dd>
3260<p><em>nonzero-digit</em><br>
3261<em>decimal-constant</em> <em>digit</em></p>
3262</dd>
3263<dt class="hdlist1"><em>octal-constant</em> : </dt>
3264<dd>
3265<p><strong>0</strong><br>
3266<em>octal-constant</em> <em>octal-digit</em></p>
3267</dd>
3268<dt class="hdlist1"><em>hexadecimal-constant</em> : </dt>
3269<dd>
3270<p><strong>0x</strong> <em>hexadecimal-digit</em><br>
3271<strong>0X</strong> <em>hexadecimal-digit</em><br>
3272<em>hexadecimal-constant</em> <em>hexadecimal-digit</em></p>
3273</dd>
3274<dt class="hdlist1"><em>digit</em> : </dt>
3275<dd>
3276<p><strong>0</strong><br>
3277<em>nonzero-digit</em></p>
3278</dd>
3279<dt class="hdlist1"><em>nonzero-digit</em> : one of </dt>
3280<dd>
3281<p><strong>1 2 3 4 5 6 7 8 9</strong></p>
3282</dd>
3283<dt class="hdlist1"><em>octal-digit</em> : one of </dt>
3284<dd>
3285<p><strong>0 1 2 3 4 5 6 7</strong></p>
3286</dd>
3287<dt class="hdlist1"><em>hexadecimal-digit</em> : one of </dt>
3288<dd>
3289<p><strong>0 1 2 3 4 5 6 7 8 9<br>
3290a b c d e f<br>
3291A B C D E F</strong></p>
3292</dd>
3293</dl>
3294</div>
3295</div>
3296</div>
3297<div class="paragraph">
3298<p>No white space is allowed between the digits of an integer constant,
3299including after the leading <strong>0</strong> or after the leading <strong>0x</strong> or <strong>0X</strong> of a
3300constant, or before the suffix <strong>u</strong> or <strong>U</strong>.
3301When the suffix <strong>u</strong> or <strong>U</strong> is present, the literal has type <strong>uint</strong>,
3302otherwise the type is <strong>int</strong>.
3303A leading unary minus sign (-) is interpreted as an arithmetic unary
3304negation, not as part of the constant.
3305Hence, literals themselves are always expressed with non-negative syntax,
3306though they could result in a negative value.</p>
3307</div>
3308<div class="paragraph">
3309<p>It is an error to provide a literal integer whose bit pattern
3310cannot fit in 32 bits.
3311The bit pattern of the literal is always used unmodified.
3312So a signed literal whose bit pattern includes a set sign bit creates a
3313negative value.</p>
3314</div>
3315<div class="paragraph">
3316<p>For example,</p>
3317</div>
3318<div class="listingblock">
3319<div class="content">
3320<pre class="CodeRay highlight"><code data-lang="c++"><span class="integer">1</span>             <span class="comment">// OK. Signed integer, value 1</span>
3321<span class="integer">1</span>u            <span class="comment">// OK. Unsigned integer, value 1</span>
3322-<span class="integer">1</span>            <span class="comment">// OK. Unary minus applied to signed integer.</span>
3323              <span class="comment">// result is a signed integer, value -1</span>
3324-<span class="integer">1</span>u           <span class="comment">// OK. Unary minus applies to unsigned integer.</span>
3325              <span class="comment">// Result is an unsigned integer, value 0xffffffff</span>
3326<span class="hex">0xA0000000</span>    <span class="comment">// OK. 32-bit signed hexadecimal</span>
3327<span class="hex">0xABcdEF00</span>u   <span class="comment">// OK. 32-bit unsigned hexadecimal</span>
3328<span class="hex">0xffffffff</span>    <span class="comment">// OK. Signed integer, value -1</span>
3329<span class="hex">0x80000000</span>    <span class="comment">// OK. Evaluates to -2147483648</span>
3330<span class="hex">0xffffffff</span>u   <span class="comment">// OK. Unsigned integer, value 0xffffffff</span>
3331<span class="hex">0xfffffffff</span>   <span class="comment">// Error: needs more than 32 bits</span>
3332<span class="integer">3000000000</span>    <span class="comment">// OK. A signed decimal literal taking 32 bits.</span>
3333              <span class="comment">// It evaluates to -1294967296</span>
3334<span class="integer">2147483648</span>    <span class="comment">// OK. Evaluates to -2147483648 (the literal set the sign bit)</span>
3335<span class="integer">5000000000</span>    <span class="comment">// Error: needs more than 32 bits</span></code></pre>
3336</div>
3337</div>
3338</div>
3339<div class="sect3">
3340<h4 id="floats">4.1.4. Floats</h4>
3341<div class="dlist">
3342<dl>
3343<dt class="hdlist1">Definition</dt>
3344<dd>
3345<p>A <em>floating-point type</em> is any floating-point scalar, vector or matrix type.
3346It excludes arrays and structures.</p>
3347</dd>
3348</dl>
3349</div>
3350<div class="paragraph">
3351<p>Floats are available for use in a variety of scalar calculations.
3352Floating-point variables are defined as in the following example:</p>
3353</div>
3354<div class="listingblock">
3355<div class="content">
3356<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span> a, b = <span class="float">1</span><span class="float">.5</span>;</code></pre>
3357</div>
3358</div>
3359<div class="paragraph">
3360<p>As an input value to one of the processing units, a floating-point variable
3361is expected to match the IEEE 754 single precision floating-point definition
3362for precision and dynamic range.
3363<strong>highp</strong> floating-point variables within a shader are encoded according to
3364the IEEE 754 specification for single-precision floating-point values
3365(logically, not necessarily physically).
3366While encodings are logically IEEE 754, operations (addition,
3367multiplication, etc.) are not necessarily performed as required by IEEE 754.
3368See &#8220;<a href="#range-and-precision">Range and Precision</a>&#8221; for more details on
3369precision and usage of NaNs (Not a Number) and Infs (positive or negative
3370infinities).</p>
3371</div>
3372<div class="paragraph">
3373<p>Floating-point constants are defined as follows.</p>
3374</div>
3375<div class="openblock bnf">
3376<div class="content">
3377<div class="dlist">
3378<dl>
3379<dt class="hdlist1"><em>floating-constant</em> : </dt>
3380<dd>
3381<p><em>fractional-constant</em> <em>exponent-part<sub>opt</sub></em> <em>floating-suffix<sub>opt</sub></em><br>
3382<em>digit-sequence</em> <em>exponent-part</em> <em>floating-suffix<sub>opt</sub></em></p>
3383</dd>
3384<dt class="hdlist1"><em>fractional-constant</em> : </dt>
3385<dd>
3386<p><em>digit-sequence</em> <strong>.</strong> <em>digit-sequence</em><br>
3387<em>digit-sequence</em> <strong>.</strong><br>
3388<strong>.</strong> <em>digit-sequence</em></p>
3389</dd>
3390<dt class="hdlist1"><em>exponent-part</em> : </dt>
3391<dd>
3392<p><strong>e</strong> <em>sign<sub>opt</sub></em> <em>digit-sequence</em><br>
3393<strong>E</strong> <em>sign<sub>opt</sub></em> <em>digit-sequence</em></p>
3394</dd>
3395<dt class="hdlist1"><em>sign</em> : one of </dt>
3396<dd>
3397<p><strong>+</strong> <strong>-</strong></p>
3398</dd>
3399<dt class="hdlist1"><em>digit-sequence</em> : </dt>
3400<dd>
3401<p><em>digit</em><br>
3402<em>digit-sequence</em> <em>digit</em></p>
3403</dd>
3404<dt class="hdlist1"><em>floating-suffix</em> : one of </dt>
3405<dd>
3406<p><strong>f</strong> <strong>F</strong></p>
3407</dd>
3408</dl>
3409</div>
3410</div>
3411</div>
3412<div class="paragraph">
3413<p>A decimal point (<strong>.</strong>) is not needed if the exponent part is present.
3414No white space may appear anywhere within a floating-point constant,
3415including before a suffix.
3416A leading unary minus sign (<strong>-</strong>) is interpreted as a unary operator and is
3417not part of the floating-point constant.</p>
3418</div>
3419<div class="paragraph">
3420<p>There is no limit on the number of digits in any <em>digit-sequence</em>.
3421If the value of the floating-point number is too large (small) to be stored
3422as a single precision value, it is converted to positive (negative)
3423infinity.
3424A value with a magnitude too small to be represented as a mantissa and
3425exponent is converted to zero.
3426Implementations may also convert subnormal (denormalized) numbers to zero.</p>
3427</div>
3428</div>
3429<div class="sect3">
3430<h4 id="vectors">4.1.5. Vectors</h4>
3431<div class="paragraph">
3432<p>The OpenGL ES Shading Language includes data types for generic 2-, 3-, and 4-component vectors
3433of floating-point values, integers, and Booleans.
3434Floating-point vector variables can be used to store colors, normals,
3435positions, texture coordinates, texture lookup results and the like.
3436Boolean vectors can be used for component-wise comparisons of numeric
3437vectors.
3438Some examples of vector declarations are:</p>
3439</div>
3440<div class="listingblock">
3441<div class="content">
3442<pre class="CodeRay highlight"><code data-lang="c++">vec2 texcoord1, texcoord2;
3443vec3 position;
3444vec4 myRGBA;
3445ivec2 textureLookup;
3446bvec3 less;</code></pre>
3447</div>
3448</div>
3449<div class="paragraph">
3450<p>Initialization of vectors can be done with constructors.
3451See &#8220;<a href="#vector-and-matrix-constructors">Vector and Matrix Constructors</a>&#8221;.</p>
3452</div>
3453</div>
3454<div class="sect3">
3455<h4 id="matrices">4.1.6. Matrices</h4>
3456<div class="paragraph">
3457<p>The OpenGL ES Shading Language has built-in types for 2 × 2, 2 × 3, 2 × 4, 3
3458× 2, 3 × 3, 3 × 4, 4 × 2, 4 × 3, and 4 ×
34594 matrices of floating-point numbers.
3460The first number in the type is the number of columns, the second is the
3461number of rows.
3462If there is only one number, the matrix is square.
3463Example matrix declarations:</p>
3464</div>
3465<div class="listingblock">
3466<div class="content">
3467<pre class="CodeRay highlight"><code data-lang="c++">mat2 mat2D;
3468mat3 optMatrix;
3469mat4 view, projection;
3470mat4x4 view; <span class="comment">// an alternate way of declaring a mat4</span>
3471mat3x2 m;    <span class="comment">// a matrix with 3 columns and 2 rows</span></code></pre>
3472</div>
3473</div>
3474<div class="paragraph">
3475<p>Initialization of matrix values is done with constructors (described in
3476&#8220;<a href="#vector-and-matrix-constructors">Vector and Matrix Constructors</a>&#8221;) in
3477column-major order.</p>
3478</div>
3479<div class="paragraph">
3480<p><strong>mat2</strong> is an alias for <strong>mat2x2</strong>, not a distinct type.
3481Similarly for <strong>mat3</strong> and <strong>mat4.</strong> The following is legal:</p>
3482</div>
3483<div class="listingblock">
3484<div class="content">
3485<pre class="CodeRay highlight"><code data-lang="c++">mat2 a;
3486mat2x2 b = a;</code></pre>
3487</div>
3488</div>
3489</div>
3490<div class="sect3">
3491<h4 id="opaque-types">4.1.7. Opaque Types</h4>
3492<div class="dlist">
3493<dl>
3494<dt class="hdlist1">Definition</dt>
3495<dd>
3496<p>An <em>opaque type</em> is a type where the internal structure of the type is
3497hidden from the language.</p>
3498</dd>
3499</dl>
3500</div>
3501<div class="paragraph">
3502<p>The opaque types, as listed in the following sections, declare variables
3503that are effectively opaque handles to other objects.
3504These objects are accessed through built-in functions, not through direct
3505reading or writing of the declared variable.
3506They can only be declared as function parameters or in <strong>uniform</strong>-qualified
3507variables (see &#8220;<a href="#uniform-variables">Uniform Variables</a>&#8221;).
3508The only opaque types that take memory qualifiers are the image types.
3509Except for array indexing, structure member selection, and parentheses,
3510opaque variables are not allowed to be operands in expressions; such use
3511results in a compile-time error.</p>
3512</div>
3513<div class="paragraph">
3514<p>When aggregated into arrays within a shader, opaque types can only be
3515indexed with a dynamically uniform integral expression (see &#8220;Dynamically
3516Uniform Expressions&#8221;) unless otherwise noted; otherwise, results are
3517undefined.</p>
3518</div>
3519<div class="paragraph">
3520<p>Opaque variables cannot be treated as l-values; hence cannot be used as
3521<strong>out</strong> or <strong>inout</strong> function parameters, nor can they be assigned into.
3522Any such use results in a compile-time error.
3523However, they can be passed as <strong>in</strong> parameters with matching types and
3524memory qualifiers.
3525They cannot be declared with an initializer.</p>
3526</div>
3527<div class="paragraph">
3528<p>Because a single opaque type declaration effectively declares two objects,
3529the opaque handle itself and the object it is a handle to, there is room for
3530both a storage qualifier and a memory qualifier.
3531The storage qualifier will qualify the opaque handle, while the memory
3532qualifier will qualify the object it is a handle to.</p>
3533</div>
3534<div class="sect4">
3535<h5 id="samplers">Texture-Combined Samplers</h5>
3536<div class="paragraph">
3537<p>Texture-combined sampler types (e.g. <strong>sampler2D</strong>) are the sampler types
3538described in the Basic Types tables as handles for accessing textures.
3539(They do not include <strong>sampler</strong> and <strong>samplerShadow</strong>.)
3540There are distinct texture-combined sampler types for each texture target,
3541and for each of float, integer, and unsigned integer data types.
3542Texture accesses are done through built-in texture functions (described in
3543&#8220;<a href="#texture-functions">Texture Functions</a>&#8221;) and texture-combined samplers
3544are used to specify which texture to access and how it is to be filtered.</p>
3545</div>
3546<div class="paragraph">
3547<p>Texture-combined sampler types are opaque types,
3548declared and behaving as described above for opaque types.</p>
3549</div>
3550</div>
3551<div class="sect4">
3552<h5 id="images">Images</h5>
3553<div class="paragraph">
3554<p>Image types are opaque types, declared and behaving as described above for
3555opaque types.
3556They can be further qualified with memory qualifiers.
3557When aggregated into arrays within a shader, images can only be indexed with
3558a constant integral expression.</p>
3559</div>
3560<div class="paragraph">
3561<p>Image variables are handles to
3562two-, or three-dimensional images
3563corresponding to all or a portion of a single level of a texture image bound
3564to an image unit.
3565There are distinct image variable types for each texture target, and for
3566each of float, integer, and unsigned integer data types.
3567Image accesses should use an image type that matches the target of the
3568texture whose level is bound to the image unit, or for non-layered bindings
3569of 3D or array images should use the image type that matches the
3570dimensionality of the layer of the image (i.e., a layer of 3D, 2DArray,
3571Cube, or CubeArray should use
3572<strong>image2D</strong>).
3573If the image target type does not match the bound image in this manner, if
3574the data type does not match the bound image, or if the format layout
3575qualifier does not match the image unit format as described in section
35768.22
3577&#8220;Texture Image Loads and Stores&#8221; of the <a href="#references">OpenGL ES Specification</a>, the
3578results of image accesses are undefined but cannot include program
3579termination.</p>
3580</div>
3581<div class="paragraph">
3582<p>Image variables are used in the image load, store, and atomic functions
3583described in &#8220;<a href="#image-functions">Image Functions</a>&#8221; to specify an image to
3584access.</p>
3585</div>
3586</div>
3587<div class="sect4">
3588<h5 id="atomic-counters">Atomic Counters</h5>
3589<div class="paragraph">
3590<p>Atomic counter types (e.g. <strong>atomic_uint</strong>) are opaque handles to counters,
3591declared and behaving as described above for opaque types.
3592The variables they declare specify which counter to access when using the
3593built-in atomic counter functions as described in
3594&#8220;<a href="#atomic-counter-functions">Atomic Counter Functions</a>&#8221;.
3595They are bound to buffers as described in
3596&#8220;<a href="#atomic-counter-layout-qualifiers">Atomic Counter Layout Qualifiers</a>&#8221;.</p>
3597</div>
3598<div class="paragraph">
3599<p>Atomic counters aggregated into arrays within a shader can only be indexed
3600with dynamically uniform integral expressions, otherwise results are
3601undefined.</p>
3602</div>
3603<div class="paragraph">
3604<p>Members of structures cannot be declared as atomic counter types.</p>
3605</div>
3606<div class="paragraph">
3607<p>The default precision of all atomic types is <strong>highp</strong>.
3608It is an error to declare an atomic type with a different precision or to
3609specify the default precision for an atomic type to be <strong>lowp</strong> or <strong>mediump</strong>.</p>
3610</div>
3611<div class="paragraph">
3612<p>Atomic counter types are not available when targeting Vulkan.</p>
3613</div>
3614</div>
3615<div class="sect4">
3616<h5 id="_texture_sampler_and_samplershadow_types">Texture, <strong>sampler</strong>, and <strong>samplerShadow</strong> Types</h5>
3617<div class="paragraph">
3618<p>Texture (e.g., <strong>texture2D</strong>), <strong>sampler</strong>, and <strong>samplerShadow</strong> types are opaque
3619types, declared and behaving as described above for opaque types.
3620These types are only available when targeting Vulkan.
3621When aggregated into arrays within a shader, these types can only be indexed
3622with a dynamically uniform expression, or texture lookup will result in
3623undefined values.
3624Texture variables are handles to one-, two-, and three-dimensional textures,
3625cube maps, etc., as enumerated in the basic types tables.
3626There are distinct texture types for each texture target, and for each of
3627float, integer, and unsigned integer data types.
3628Textures can be combined with a variable of type <strong>sampler</strong> or <strong>samplerShadow</strong>
3629to create a texture-combined sampler type (e.g., sampler2D, or sampler2DShadow).
3630This is done with a constructor, e.g., <code>sampler2D(texture2D, sampler)</code>,
3631<code>sampler2DShadow(texture2D, sampler)</code>, <code>sampler2DShadow(texture2D, samplerShadow)</code>,
3632or <code>sampler2D(texture2D, samplerShadow)</code> and is described in more detail
3633in section 5.4 "Constructors".</p>
3634</div>
3635</div>
3636<div class="sect4">
3637<h5 id="_subpass_inputs">Subpass Inputs</h5>
3638<div class="paragraph">
3639<p>Subpass-input types are only available when targeting Vulkan.</p>
3640</div>
3641<div class="paragraph">
3642<p>Subpass-input types (e.g., <strong>subpassInput</strong>) are opaque types, declared
3643and behaving as described above for opaque types.
3644When aggregated into arrays within a shader, they can only be indexed with a
3645dynamically uniform integral expression, otherwise results are undefined.</p>
3646</div>
3647<div class="paragraph">
3648<p>Subpass-input types are handles to two-dimensional single sampled or
3649multi-sampled images, with distinct types for each of float, integer,
3650and unsigned integer data types.</p>
3651</div>
3652<div class="paragraph">
3653<p>Subpass-input types are only available in fragment shaders.  It is
3654an error to use them in any other stage.</p>
3655</div>
3656</div>
3657</div>
3658<div class="sect3">
3659<h4 id="structures">4.1.8. Structures</h4>
3660<div class="paragraph">
3661<p>User-defined types can be created by aggregating other already defined types
3662into a structure using the <strong>struct</strong> keyword.
3663For example,</p>
3664</div>
3665<div class="listingblock">
3666<div class="content">
3667<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">struct</span> light {
3668    <span class="predefined-type">float</span> intensity;
3669    vec3 position;
3670} lightVar;</code></pre>
3671</div>
3672</div>
3673<div class="paragraph">
3674<p>In this example, <em>light</em> becomes the name of the new type, and <em>lightVar</em>
3675becomes a variable of type <em>light</em>.
3676To declare variables of the new type, use its name (without the keyword
3677<strong>struct</strong>).</p>
3678</div>
3679<div class="listingblock">
3680<div class="content">
3681<pre class="CodeRay highlight"><code data-lang="c++">light lightVar2;</code></pre>
3682</div>
3683</div>
3684<div class="paragraph">
3685<p>More formally, structures are declared as follows.
3686However, the definitive grammar is as given in
3687&#8220;<a href="#shading-language-grammar">Shading Language Grammar</a>&#8221;.</p>
3688</div>
3689<div class="openblock bnf">
3690<div class="content">
3691<div class="dlist">
3692<dl>
3693<dt class="hdlist1"><em>struct-definition</em> : </dt>
3694<dd>
3695<p><em>qualifier<sub>opt</sub></em> <strong>struct</strong> name<sub>opt</sub>_ <strong>{</strong> <em>member-list</em> <strong>}</strong>
3696<em>declarators<sub>opt</sub></em> <strong>;</strong></p>
3697</dd>
3698<dt class="hdlist1"><em>member-list</em> : </dt>
3699<dd>
3700<p><em>member-declaration</em> <strong>;</strong><br>
3701<em>member-declaration</em> <em>member-list</em> <strong>;</strong></p>
3702</dd>
3703<dt class="hdlist1"><em>member-declaration</em> : </dt>
3704<dd>
3705<p><em>basic-type</em> <em>declarators</em> <strong>;</strong></p>
3706</dd>
3707</dl>
3708</div>
3709</div>
3710</div>
3711<div class="paragraph">
3712<p>where <em>name</em> becomes the user-defined type, and can be used to declare
3713variables to be of this new type.
3714The <em>name</em> shares the same name space as other variables, types, and
3715functions.
3716All previously visible variables, types, constructors, or functions with
3717that name are hidden.
3718The optional <em>qualifier</em> only applies to any <em>declarators</em>, and is not part
3719of the type being defined for <em>name</em>.</p>
3720</div>
3721<div class="paragraph">
3722<p>Structures must have at least one member declaration.
3723Member declarators may contain precision qualifiers, but use of any other
3724qualifier results in an error.
3725Bit fields are not supported.
3726Member types must be already defined (there are no forward references).
3727Member declarations cannot contain initializers.
3728Member declarators can contain arrays.
3729Such arrays must have a size specified, and the size must be a constant
3730integral expression that&#8217;s greater than zero (see
3731&#8220;<a href="#constant-expressions">Constant Expressions</a>&#8221;).
3732Each level of structure has its own name space for names given in member
3733declarators; such names need only be unique within that name space.</p>
3734</div>
3735<div class="paragraph">
3736<p>Anonymous structures are not supported.
3737Embedded structure definitions are not supported.</p>
3738</div>
3739<div class="listingblock">
3740<div class="content">
3741<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">struct</span> S { <span class="predefined-type">float</span> f; }; <span class="comment">// Allowed: S is defined as a structure.</span>
3742
3743<span class="keyword">struct</span> T {
3744    S;              <span class="comment">// Error: anonymous structures disallowed</span>
3745    <span class="keyword">struct</span> { ... }; <span class="comment">// Error: embedded structures disallowed</span>
3746    S s;            <span class="comment">// Allowed: nested structure with a name.</span>
3747};</code></pre>
3748</div>
3749</div>
3750<div class="paragraph">
3751<p>Structures can be initialized at declaration time using constructors, as
3752discussed in &#8220;<a href="#structure-constructors">Structure Constructors</a>&#8221;.</p>
3753</div>
3754<div class="paragraph">
3755<p>Any restrictions on the usage of a type or qualifier also apply to any
3756structure that contains a member of that type or qualifier.
3757This also applies to structure members that are structures, recursively.</p>
3758</div>
3759<div class="paragraph">
3760<p>Structures can contain variables of any type except:</p>
3761</div>
3762<div class="ulist">
3763<ul>
3764<li>
3765<p><strong>atomic_uint</strong> (since there is no mechanism to specify the binding)</p>
3766</li>
3767<li>
3768<p>image types (since there is no mechanism to specify the format
3769qualifier)</p>
3770</li>
3771</ul>
3772</div>
3773</div>
3774<div class="sect3">
3775<h4 id="arrays">4.1.9. Arrays</h4>
3776<div class="paragraph">
3777<p>Variables of the same type can be aggregated into arrays by declaring a name
3778followed by brackets (<strong>[ ]</strong>) enclosing an optional size.</p>
3779</div>
3780<div class="paragraph">
3781<p>When present, the array size must be a constant integral expression (see
3782&#8220;<a href="#constant-expressions">Constant Expressions</a>&#8221;) greater than zero.
3783The type of the size parameter can be a signed or unsigned integer and the
3784choice of type does not affect the type of the resulting array.
3785Arrays only have a single dimension (a single number within &#8220;[ ]&#8221;),
3786however, arrays of arrays can be declared.
3787Any type can be formed into an array.</p>
3788</div>
3789<div class="paragraph">
3790<p>Arrays are sized either at compile-time or at run-time.
3791To size an array at compile-time, either the size must be specified within
3792the brackets as above or must be inferred from the type of the initializer.</p>
3793</div>
3794<div class="paragraph">
3795<p>If an array is declared as the last member of a shader storage block and the
3796size is not specified at compile-time, it is sized at run-time.
3797In all other cases, arrays are sized only at compile-time.
3798An array declaration sized at compile-time which leaves the size of the
3799array unspecified is an error.</p>
3800</div>
3801<div class="paragraph">
3802<p>For compile-time sized arrays, it is illegal to index an array with a
3803constant integral expression greater than or equal to the declared size or
3804with a negative constant expression.
3805Arrays declared as formal parameters in a function declaration must also
3806specify a size.
3807Undefined behavior results from indexing an array with a non-constant
3808expression that&#8217;s greater than or equal to the array&#8217;s size or less than 0.
3809If robust buffer access is enabled (see section 10.3.5 &#8220;Robust Buffer
3810Access&#8221; of the <a href="#references">OpenGL ES Specification</a>), such indexing must not result in
3811abnormal program termination.
3812The results are still undefined, but implementations are encouraged to
3813produce zero values for such accesses.</p>
3814</div>
3815<div class="paragraph">
3816<p>Some examples are:</p>
3817</div>
3818<div class="listingblock">
3819<div class="content">
3820<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span> frequencies[<span class="integer">3</span>];
3821uniform vec4 lightPosition[<span class="integer">4</span>u];
3822<span class="directive">const</span> <span class="predefined-type">int</span> numLights = <span class="integer">2</span>;
3823light lights[numLights];
3824vec4 a[<span class="integer">3</span>][<span class="integer">2</span>];
3825a.length() <span class="comment">// this is 3</span>
3826a[x].length() <span class="comment">// this is 2</span>
3827<span class="comment">// a shader storage block, introduced in section 4.3.7 &quot;Buffer Variables&quot;</span>
3828buffer b {
3829    <span class="predefined-type">float</span> u[]; <span class="comment">// an error</span>
3830    vec4 v[];  <span class="comment">// okay, v will be sized at run-time</span>
3831} name[<span class="integer">3</span>];     <span class="comment">// when the block is arrayed, all u will be the same size,</span>
3832               <span class="comment">// but not necessarily all v, if sized dynamically</span></code></pre>
3833</div>
3834</div>
3835<div class="paragraph">
3836<p>When the <strong>length</strong>() method will return a compile-time constant, the
3837expression in brackets (x above) will be evaluated and subject to the rules
3838required for array indices, but the array will not be dereferenced.
3839Thus, behavior is well defined even if the run-time value of the expression
3840is out of bounds.</p>
3841</div>
3842<div class="paragraph">
3843<p>An array type can be formed by specifying a non-array type
3844(<a href="#type_specifier_nonarray">[type_specifier_nonarray]</a>) followed by an <a href="#array_specifier">[array_specifier]</a>.</p>
3845</div>
3846<div class="listingblock">
3847<div class="content">
3848<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span>[<span class="integer">5</span>]</code></pre>
3849</div>
3850</div>
3851<div class="paragraph">
3852<p>Note that the construct <em>type [size]</em> does not always result in an array of
3853length <em>size</em> of type <em>type</em>:</p>
3854</div>
3855<div class="listingblock">
3856<div class="content">
3857<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span>[<span class="integer">2</span>][<span class="integer">3</span>] <span class="comment">// an array of size [2] of array of size [3] of float,</span>
3858            <span class="comment">// not size [3] of float[2]</span></code></pre>
3859</div>
3860</div>
3861<div class="paragraph">
3862<p>This type can be used anywhere any other type can be used, including as the
3863return value from a function</p>
3864</div>
3865<div class="listingblock">
3866<div class="content">
3867<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span>[<span class="integer">5</span>] foo() { }</code></pre>
3868</div>
3869</div>
3870<div class="paragraph">
3871<p>as a constructor of an array:</p>
3872</div>
3873<div class="listingblock">
3874<div class="content">
3875<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span>[<span class="integer">5</span>](<span class="float">3</span><span class="float">.4</span>, <span class="float">4</span><span class="float">.2</span>, <span class="float">5</span><span class="float">.0</span>, <span class="float">5</span><span class="float">.2</span>, <span class="float">1</span><span class="float">.1</span>)</code></pre>
3876</div>
3877</div>
3878<div class="paragraph">
3879<p>as an unnamed parameter:</p>
3880</div>
3881<div class="listingblock">
3882<div class="content">
3883<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">void</span> foo(<span class="predefined-type">float</span>[<span class="integer">5</span>])</code></pre>
3884</div>
3885</div>
3886<div class="paragraph">
3887<p>and as an alternate way of declaring a variable or function parameter:</p>
3888</div>
3889<div class="listingblock">
3890<div class="content">
3891<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span>[<span class="integer">5</span>] a;</code></pre>
3892</div>
3893</div>
3894<div class="paragraph">
3895<p>An array type can also be formed without specifying a size if the definition
3896includes an initializer:</p>
3897</div>
3898<div class="listingblock">
3899<div class="content">
3900<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span> x[] = <span class="predefined-type">float</span>[<span class="integer">2</span>] (<span class="float">1</span><span class="float">.0</span>, <span class="float">2</span><span class="float">.0</span>); <span class="comment">// declares an array of size 2</span>
3901<span class="predefined-type">float</span> y[] = <span class="predefined-type">float</span>[] (<span class="float">1</span><span class="float">.0</span>, <span class="float">2</span><span class="float">.0</span>, <span class="float">3</span><span class="float">.0</span>); <span class="comment">// declares an array of size 3</span>
3902<span class="predefined-type">float</span> a[<span class="integer">5</span>];
3903<span class="predefined-type">float</span> b[] = a;</code></pre>
3904</div>
3905</div>
3906<div class="paragraph">
3907<p>Note that the initializer itself does not need to be a constant expression
3908but the length of the initializer will be a constant expression.</p>
3909</div>
3910<div class="paragraph">
3911<p>Arrays can have initializers formed from array constructors:</p>
3912</div>
3913<div class="listingblock">
3914<div class="content">
3915<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span> a[<span class="integer">5</span>] = <span class="predefined-type">float</span>[<span class="integer">5</span>](<span class="float">3</span><span class="float">.4</span>, <span class="float">4</span><span class="float">.2</span>, <span class="float">5</span><span class="float">.0</span>, <span class="float">5</span><span class="float">.2</span>, <span class="float">1</span><span class="float">.1</span>);
3916<span class="predefined-type">float</span> a[<span class="integer">5</span>] = <span class="predefined-type">float</span>[](<span class="float">3</span><span class="float">.4</span>, <span class="float">4</span><span class="float">.2</span>, <span class="float">5</span><span class="float">.0</span>, <span class="float">5</span><span class="float">.2</span>, <span class="float">1</span><span class="float">.1</span>);  <span class="comment">// same thing</span></code></pre>
3917</div>
3918</div>
3919<div class="paragraph">
3920<p>An array of arrays can be declared as:</p>
3921</div>
3922<div class="listingblock">
3923<div class="content">
3924<pre class="CodeRay highlight"><code data-lang="c++">vec4 a[<span class="integer">3</span>][<span class="integer">2</span>]; <span class="comment">// size-3 array of size-2 array of vec4</span></code></pre>
3925</div>
3926</div>
3927<div class="paragraph">
3928<p>which declares a one-dimensional array of size 3 of one-dimensional arrays
3929of size 2 of <strong>vec4</strong>.
3930The following declarations do the same thing:</p>
3931</div>
3932<div class="listingblock">
3933<div class="content">
3934<pre class="CodeRay highlight"><code data-lang="c++">vec4[<span class="integer">2</span>] a[<span class="integer">3</span>]; <span class="comment">// size-3 array of size-2 array of vec4</span>
3935vec4[<span class="integer">3</span>][<span class="integer">2</span>] a; <span class="comment">// size-3 array of size-2 array of vec4</span></code></pre>
3936</div>
3937</div>
3938<div class="paragraph">
3939<p>When in transparent memory (like in a uniform block), the layout is that the
3940inner-most (right-most in declaration) dimensions iterate faster than the
3941outer dimensions.
3942That is, for the above, the order in memory would be:</p>
3943</div>
3944<div class="dlist">
3945<dl>
3946<dt class="hdlist1"></dt>
3947<dd>
3948<p>Low address : a[0][0] : a[0][1] : a[1][0] : a[1][1] : a[2][0] : a[2][1]
3949: High address</p>
3950</dd>
3951</dl>
3952</div>
3953<div class="paragraph">
3954<p>The last member of a shader storage block (see &#8220;<a href="#buffer-variables">Buffer
3955Variables</a>&#8221;), may be declared without specifying a size.
3956For such arrays, the effective array size is inferred at run-time from the
3957size of the data store backing the shader storage block.
3958Such runtime-sized arrays may be indexed with general integer expressions,
3959but may not be passed as an argument to a function or indexed with a
3960negative constant expression.</p>
3961</div>
3962<div class="listingblock">
3963<div class="content">
3964<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">struct</span> S { <span class="predefined-type">float</span> f; };
3965buffer ShaderStorageBlock1
3966{
3967    vec4 a[]; <span class="comment">// illegal</span>
3968    vec4 b[]; <span class="comment">// legal, runtime-sized arrays are last member</span>
3969};
3970buffer ShaderStorageBlock2
3971{
3972    vec4 a[<span class="integer">4</span>]; <span class="comment">// legal, size declared</span>
3973    S b[]; <span class="comment">// legal, runtime-sized arrays are allowed,</span>
3974    <span class="comment">// including arrays of structures</span>
3975};</code></pre>
3976</div>
3977</div>
3978<div class="paragraph">
3979<p>However, it is a compile-time error to assign to a runtime-sized array.
3980Assignments to individual elements of the array is allowed.</p>
3981</div>
3982<div class="paragraph">
3983<p>Arrays have a fixed number of elements.
3984This can be obtained by using the <strong>length</strong>() method:</p>
3985</div>
3986<div class="listingblock">
3987<div class="content">
3988<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span> a[<span class="integer">5</span>];
3989a.length(); <span class="comment">// returns 5</span></code></pre>
3990</div>
3991</div>
3992<div class="paragraph">
3993<p>The return value is a signed integral expression.
3994For compile-time sized arrays, the value returned by the length method is a
3995constant expression.
3996For run-time sized arrays , the value returned will not be constant
3997expression and will be determined at run time based on the size of the
3998buffer object providing storage for the block.</p>
3999</div>
4000<div class="paragraph">
4001<p>The precision is determined using the same rules as for other cases where
4002there is no intrinsic precision.
4003See &#8220;<a href="#precision-qualifiers">Precision Qualifiers</a>&#8221;.</p>
4004</div>
4005<div class="paragraph">
4006<p>Any restrictions on the usage of a type also apply to arrays of that type.
4007This applies recursively.</p>
4008</div>
4009</div>
4010</div>
4011<div class="sect2">
4012<h3 id="scoping">4.2. Scoping</h3>
4013<div class="paragraph">
4014<p>The scope of a declaration determines where the declaration is visible.
4015GLSL ES uses a system of statically nested scopes.
4016This allows names to be redefined within a shader.</p>
4017</div>
4018<div class="sect3">
4019<h4 id="definition-of-terms">4.2.1. Definition of Terms</h4>
4020<div class="paragraph">
4021<p>The term <em>scope</em> refers to a specified region of the program where names are
4022guaranteed to be visible.
4023For example, a <em>compound_statement_with_scope</em> ('{' <em>statement</em> <em>statement</em>
4024&#8230;&#8203;
4025'}') defines a scope.</p>
4026</div>
4027<div class="paragraph">
4028<p>A <em>nested scope</em> is a scope defined within an outer scope.</p>
4029</div>
4030<div class="paragraph">
4031<p>The terms '<em>same scope'</em> and '<em>current scope</em>' are equivalent to the term
4032'<em>scope</em>' but used to emphasize that nested scopes are excluded.</p>
4033</div>
4034<div class="paragraph">
4035<p>The <em>scope of a declaration</em> is the region or regions of the program where
4036that declaration is visible.</p>
4037</div>
4038<div class="paragraph">
4039<p>A <em>name space</em> defines where names may be defined.
4040Within a single name space, a name has at most one entry, specifying it to
4041be one of: structure, variable, or function.</p>
4042</div>
4043<div class="paragraph">
4044<p>In general, each scope has an associated name space.
4045However, in certain cases e.g. for uniforms, multiple scopes share the same
4046name space.
4047In these cases, conflicting declarations are an error, even though the name
4048is only visible in the scopes where it is declared.</p>
4049</div>
4050</div>
4051<div class="sect3">
4052<h4 id="types-of-scope">4.2.2. Types of Scope</h4>
4053<div class="paragraph">
4054<p>The scope of a variable is determined by where it is declared.
4055If it is declared outside all function definitions, it has global scope,
4056which starts from where it is declared and persists to the end of the shader
4057it is declared in.
4058If it is declared in a <strong>while</strong> test or a <strong>for</strong> statement, then it is scoped
4059to the end of the following sub-statement (specified as
4060<em>statement-no-new-scope</em> in the grammar).
4061Otherwise, if it is declared as a statement within a compound statement, it
4062is scoped to the end of that compound statement.
4063If it is declared as a parameter in a function definition, it is scoped
4064until the end of that function definition.
4065A function&#8217;s parameter declarations and body together form a single scope.</p>
4066</div>
4067<div class="listingblock">
4068<div class="content">
4069<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span> f( <span class="comment">/* nested scope begins here */</span> <span class="predefined-type">int</span> k)
4070{
4071    <span class="predefined-type">int</span> k = k + <span class="integer">3</span>; <span class="comment">// redeclaration error of the name k</span>
4072    ...
4073}
4074<span class="predefined-type">int</span> f(<span class="predefined-type">int</span> k)
4075{
4076    {
4077       <span class="predefined-type">int</span> k = k + <span class="integer">3</span>; <span class="comment">// 2nd k is parameter, initializing nested first k</span>
4078       <span class="predefined-type">int</span> m = k <span class="comment">// use of new k, which is hiding the parameter</span>
4079    }
4080}</code></pre>
4081</div>
4082</div>
4083<div class="paragraph">
4084<p>For both for and while loops, the sub-statement itself does not introduce a
4085new scope for variable names, so the following has a redeclaration
4086compile-time error:</p>
4087</div>
4088<div class="listingblock">
4089<div class="content">
4090<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">for</span> ( <span class="comment">/* nested scope begins here */</span> <span class="predefined-type">int</span> i = <span class="integer">0</span>; i &lt; <span class="integer">10</span>; i++)
4091{
4092    <span class="predefined-type">int</span> i; <span class="comment">// redeclaration error</span>
4093}</code></pre>
4094</div>
4095</div>
4096<div class="paragraph">
4097<p>The body of a <strong>do</strong>-<strong>while</strong> loop introduces a new scope lasting only between
4098the do and while (not including the while test expression), whether or not
4099the body is simple or compound:</p>
4100</div>
4101<div class="listingblock">
4102<div class="content">
4103<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span> _i_ = <span class="integer">17</span>;
4104<span class="keyword">do</span>
4105    <span class="predefined-type">int</span> i = <span class="integer">4</span>; <span class="comment">// okay, in nested scope</span>
4106<span class="keyword">while</span> (i == <span class="integer">0</span>); <span class="comment">// i is 17, scoped outside the do-while body</span></code></pre>
4107</div>
4108</div>
4109<div class="paragraph">
4110<p>The statement following a <strong>switch</strong> (&#8230;&#8203;) forms a nested scope.</p>
4111</div>
4112<div class="paragraph">
4113<p>Representing the if construct as:</p>
4114</div>
4115<div class="paragraph">
4116<p><strong>if</strong> if-expression <strong>then</strong> if-statement <strong>else</strong> else-statement,</p>
4117</div>
4118<div class="paragraph">
4119<p>a variable declared in the if-statement is scoped to the end of the
4120if-statement.
4121A variable declared in the else-statement is scoped to the end of the
4122else-statement.
4123This applies both when these statements are simple statements and when they
4124are compound statements.
4125The if-expression does not allow new variables to be declared, hence does
4126not form a new scope.</p>
4127</div>
4128<div class="paragraph">
4129<p>Within a declaration, the scope of a name starts immediately after the
4130initializer if present or immediately after the name being declared if not.
4131Several examples:</p>
4132</div>
4133<div class="listingblock">
4134<div class="content">
4135<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span> x = <span class="integer">1</span>;
4136{
4137    <span class="predefined-type">int</span> x = <span class="integer">2</span>,<span class="comment">/* 2nd x visible here */</span> y = x; <span class="comment">// y is initialized to 2</span>
4138    <span class="predefined-type">int</span> z = z; <span class="comment">// error if z not previously defined.</span>
4139}
4140{
4141<span class="predefined-type">int</span> x = x; <span class="comment">// x is initialized to '1'</span>
4142}</code></pre>
4143</div>
4144</div>
4145<div class="paragraph">
4146<p>A structure name declaration is visible at the end of the <em>struct_specifier</em>
4147in which it was declared:</p>
4148</div>
4149<div class="listingblock">
4150<div class="content">
4151<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">struct</span> S
4152{
4153    <span class="predefined-type">int</span> x;
4154};
4155{
4156    S S = S(<span class="integer">0</span>); <span class="comment">// 'S' is only visible as a struct and constructor</span>
4157    S; <span class="comment">// 'S' is now visible as a variable</span>
4158}
4159<span class="predefined-type">int</span> x = x; <span class="comment">// Error if x has not been previously defined.</span></code></pre>
4160</div>
4161</div>
4162</div>
4163<div class="sect3">
4164<h4 id="redeclaring-names">4.2.3. Redeclaring Names</h4>
4165<div class="paragraph">
4166<p>All variable names, structure type names, and function names in a given
4167scope share the same name space.
4168Function names can be redeclared in the same scope, with the same or
4169different parameters, without error.
4170Otherwise, within a shader, a declared name cannot be redeclared in the same
4171scope; doing so results in a redeclaration error.
4172If a nested scope redeclares a name used in an outer scope, it hides all
4173existing uses of that name.
4174There is no way to access the hidden name or make it unhidden, without
4175exiting the scope that hid it.</p>
4176</div>
4177<div class="paragraph">
4178<p>Names of built-in functions cannot be redeclared as functions.
4179Therefore overloading or redefining built-in functions is an error.</p>
4180</div>
4181<div class="paragraph">
4182<p>A <em>declaration</em> is considered to be a statement that adds a name or
4183signature to the symbol table.
4184A <em>definition</em> is a statement that fully defines that name or signature.
4185E.g.</p>
4186</div>
4187<div class="listingblock">
4188<div class="content">
4189<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span> f();<span class="comment">// declaration;</span>
4190<span class="predefined-type">int</span> f() {<span class="keyword">return</span> <span class="integer">0</span>;}<span class="comment">// declaration and definition</span>
4191<span class="predefined-type">int</span> x; <span class="comment">// declaration and definition</span>
4192<span class="predefined-type">int</span> a[<span class="integer">4</span>];<span class="comment">// array declaration and definition</span>
4193<span class="keyword">struct</span> S {<span class="predefined-type">int</span> x;};<span class="comment">// structure declaration and definition</span></code></pre>
4194</div>
4195</div>
4196<div class="paragraph">
4197<p>The determination of equivalence of two declarations depends on the type of
4198declaration.
4199For functions, the whole function signature must be considered (see
4200&#8220;Function Definitions&#8221;).
4201For variables (including arrays) and structures only the names must match.</p>
4202</div>
4203<div class="paragraph">
4204<p>Within each scope, a name may be declared either as a variable declaration
4205<em>or</em> as function declarations <em>or</em> as a structure.</p>
4206</div>
4207<div class="paragraph">
4208<p>Examples of combinations that are allowed:</p>
4209</div>
4210<div class="paragraph">
4211<p>1.</p>
4212</div>
4213<div class="openblock">
4214<div class="content">
4215<div class="listingblock">
4216<div class="content">
4217<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">void</span> f(<span class="predefined-type">int</span>) {...}
4218<span class="directive">void</span> f(<span class="predefined-type">float</span>) {...}<span class="comment">// function overloading allowed</span></code></pre>
4219</div>
4220</div>
4221</div>
4222</div>
4223<div class="paragraph">
4224<p>2.</p>
4225</div>
4226<div class="openblock">
4227<div class="content">
4228<div class="listingblock">
4229<div class="content">
4230<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">void</span> f(<span class="predefined-type">int</span>);<span class="comment">// 1^st^ declaration (allowed)</span>
4231<span class="directive">void</span> f(<span class="predefined-type">int</span>);<span class="comment">// repeated declaration (allowed)</span>
4232<span class="directive">void</span> f(<span class="predefined-type">int</span>) {...}<span class="comment">// single definition (allowed)</span></code></pre>
4233</div>
4234</div>
4235</div>
4236</div>
4237<div class="paragraph">
4238<p>Examples of combinations that are disallowed:</p>
4239</div>
4240<div class="paragraph">
4241<p>1.</p>
4242</div>
4243<div class="openblock">
4244<div class="content">
4245<div class="listingblock">
4246<div class="content">
4247<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">void</span> f(<span class="predefined-type">int</span>) {...}
4248<span class="directive">void</span> f(<span class="predefined-type">int</span>) {...}<span class="comment">// Error: repeated definition</span></code></pre>
4249</div>
4250</div>
4251</div>
4252</div>
4253<div class="paragraph">
4254<p>2.</p>
4255</div>
4256<div class="openblock">
4257<div class="content">
4258<div class="listingblock">
4259<div class="content">
4260<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">void</span> f(<span class="predefined-type">int</span>);
4261<span class="keyword">struct</span> f {<span class="predefined-type">int</span> x;};<span class="comment">// Error: type 'f' conflicts with function 'f'</span></code></pre>
4262</div>
4263</div>
4264</div>
4265</div>
4266<div class="paragraph">
4267<p>3.</p>
4268</div>
4269<div class="openblock">
4270<div class="content">
4271<div class="listingblock">
4272<div class="content">
4273<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">struct</span> f {<span class="predefined-type">int</span> x;};
4274<span class="predefined-type">int</span> f;<span class="comment">// Error: conflicts with the type 'f'</span></code></pre>
4275</div>
4276</div>
4277</div>
4278</div>
4279<div class="paragraph">
4280<p>4.</p>
4281</div>
4282<div class="openblock">
4283<div class="content">
4284<div class="listingblock">
4285<div class="content">
4286<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span> a[<span class="integer">3</span>];
4287<span class="predefined-type">int</span> a[<span class="integer">3</span>];<span class="comment">// Error: repeated array definition</span></code></pre>
4288</div>
4289</div>
4290</div>
4291</div>
4292<div class="paragraph">
4293<p>5.</p>
4294</div>
4295<div class="openblock">
4296<div class="content">
4297<div class="listingblock">
4298<div class="content">
4299<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span> x;
4300<span class="predefined-type">int</span> x;<span class="comment">// Error: repeated variable definition</span></code></pre>
4301</div>
4302</div>
4303</div>
4304</div>
4305</div>
4306<div class="sect3">
4307<h4 id="global-scope">4.2.4. Global Scope</h4>
4308<div class="paragraph">
4309<p>The built-in functions are scoped in the global scope users declare global
4310variables in.
4311That is, a shader&#8217;s global scope, available for user-defined functions and
4312global variables, is the same as the scope containing the built-in
4313functions.
4314Function declarations (prototypes) cannot occur inside of functions; they
4315must be at global scope.
4316Hence it is not possible to hide a name with a function.</p>
4317</div>
4318</div>
4319<div class="sect3">
4320<h4 id="shared-globals">4.2.5. Shared Globals</h4>
4321<div class="paragraph">
4322<p>Shared globals are variables that can be accessed by multiple compilation
4323units.
4324In GLSL ES the only shared globals are uniforms.
4325Vertex shader outputs are not considered to be shared globals since they
4326must pass through the rasterization stage before they are used as input by
4327the fragment shader.</p>
4328</div>
4329<div class="paragraph">
4330<p>Shared globals share the same name space, and must be declared with the same
4331type and precision.
4332They will share the same storage.
4333Shared global arrays must have the same base type and the same explicit
4334size.
4335Scalars must have exactly the same precision, type name and type definition.
4336Structures must have the same name, sequence of type names, and type
4337definitions, and member names to be considered the same type.
4338This rule applies recursively for nested or embedded types.</p>
4339</div>
4340</div>
4341</div>
4342<div class="sect2">
4343<h3 id="storage-qualifiers">4.3. Storage Qualifiers</h3>
4344<div class="paragraph">
4345<p>Variable declarations may have at most one storage qualifier specified in
4346front of the type.
4347These are summarized as</p>
4348</div>
4349<table class="tableblock frame-all grid-all stretch">
4350<colgroup>
4351<col style="width: 50%;">
4352<col style="width: 50%;">
4353</colgroup>
4354<thead>
4355<tr>
4356<th class="tableblock halign-left valign-top">Storage Qualifier</th>
4357<th class="tableblock halign-left valign-top">Meaning</th>
4358</tr>
4359</thead>
4360<tbody>
4361<tr>
4362<td class="tableblock halign-left valign-top"><p class="tableblock">&lt;none: default&gt;</p></td>
4363<td class="tableblock halign-left valign-top"><p class="tableblock">local read/write memory, or an input parameter to a
4364                      function</p></td>
4365</tr>
4366<tr>
4367<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>const</strong></p></td>
4368<td class="tableblock halign-left valign-top"><p class="tableblock">a compile-time constant</p></td>
4369</tr>
4370<tr>
4371<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>in</strong></p></td>
4372<td class="tableblock halign-left valign-top"><p class="tableblock">linkage into a shader from a previous stage, variable
4373                      is copied in</p></td>
4374</tr>
4375<tr>
4376<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>out</strong></p></td>
4377<td class="tableblock halign-left valign-top"><p class="tableblock">linkage out of a shader to a subsequent stage,
4378                      variable is copied out</p></td>
4379</tr>
4380<tr>
4381<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>uniform</strong></p></td>
4382<td class="tableblock halign-left valign-top"><p class="tableblock">value does not change across the primitive being
4383                      processed, uniforms form the linkage between a shader,
4384                      API, and the application</p></td>
4385</tr>
4386<tr>
4387<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>buffer</strong></p></td>
4388<td class="tableblock halign-left valign-top"><p class="tableblock">value is stored in a buffer object, and can be read or
4389                      written both by shader invocations and the API</p></td>
4390</tr>
4391<tr>
4392<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>shared</strong></p></td>
4393<td class="tableblock halign-left valign-top"><p class="tableblock">compute shader only; variable storage is shared across
4394                      all work items in a workgroup</p></td>
4395</tr>
4396</tbody>
4397</table>
4398<div class="paragraph">
4399<p>Some input and output qualified variables can be qualified with at most one
4400additional auxiliary storage qualifier:</p>
4401</div>
4402<table class="tableblock frame-all grid-all stretch">
4403<colgroup>
4404<col style="width: 50%;">
4405<col style="width: 50%;">
4406</colgroup>
4407<thead>
4408<tr>
4409<th class="tableblock halign-left valign-top">Auxiliary Storage Qualifier</th>
4410<th class="tableblock halign-left valign-top">Meaning</th>
4411</tr>
4412</thead>
4413<tbody>
4414<tr>
4415<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>centroid</strong></p></td>
4416<td class="tableblock halign-left valign-top"><p class="tableblock">centroid-based interpolation</p></td>
4417</tr>
4418<tr>
4419<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sample</strong></p></td>
4420<td class="tableblock halign-left valign-top"><p class="tableblock">per-sample interpolation</p></td>
4421</tr>
4422<tr>
4423<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>patch</strong></p></td>
4424<td class="tableblock halign-left valign-top"><p class="tableblock">per-tessellation-patch attributes</p></td>
4425</tr>
4426</tbody>
4427</table>
4428<div class="paragraph">
4429<p>Local variables can only use the <strong>const</strong> storage qualifier (or use no
4430storage qualifier).</p>
4431</div>
4432<div class="paragraph">
4433<p>Note that function parameters can use <strong>const</strong>, <strong>in</strong>, and <strong>out</strong> qualifiers,
4434but as <em>parameter qualifiers</em>.
4435Parameter qualifiers are discussed in
4436&#8220;<a href="#function-calling-conventions">Function Calling Conventions</a>&#8221;.</p>
4437</div>
4438<div class="paragraph">
4439<p>Function return types and structure members do not use storage qualifiers.</p>
4440</div>
4441<div class="paragraph">
4442<p>Data types for communication from one run of a shader executable to its next
4443run (to communicate between fragments or between vertices) do not exist.
4444This would prevent parallel execution of the same shader executable on
4445multiple vertices or fragments.</p>
4446</div>
4447<div class="paragraph">
4448<p>In declarations of global variables with no storage qualifier or with a
4449const qualifier, any initializer must be a constant expression.
4450Declarations of global variables with other storage qualifiers may not
4451contain initializers.
4452Global variables without storage qualifiers that are not initialized in
4453their declaration or by the application will not be initialized,
4454but rather will enter <em>main()</em> with undefined values.</p>
4455</div>
4456<div class="sect3">
4457<h4 id="default-storage-qualifier">4.3.1. Default Storage Qualifier</h4>
4458<div class="paragraph">
4459<p>If no qualifier is present on a global variable, then the variable has no
4460linkage to the application or shaders running on other pipeline stages.
4461For either global or local unqualified variables, the declaration will
4462appear to allocate memory associated with the processor it targets.
4463This variable will provide read/write access to this allocated memory.</p>
4464</div>
4465</div>
4466<div class="sect3">
4467<h4 id="constant-qualifier">4.3.2. Constant Qualifier</h4>
4468<div class="paragraph">
4469<p>Named compile-time constants
4470can be declared using
4471the <strong>const</strong> qualifier.
4472Any variables qualified as constant are read-only variables for that shader.
4473Declaring variables as constant allows more descriptive shaders than using
4474hard-wired numerical constants.
4475The <strong>const</strong> qualifier can be used with any of the non-void transparent basic
4476data types, as well as with structures and arrays of these.
4477It is an error to write to a <strong>const</strong> variable outside of its
4478declaration, so they must be initialized when declared.
4479For example,</p>
4480</div>
4481<div class="listingblock">
4482<div class="content">
4483<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">const</span> vec3 zAxis = vec3 (<span class="float">0</span><span class="float">.0</span>, <span class="float">0</span><span class="float">.0</span>, <span class="float">1</span><span class="float">.0</span>);</code></pre>
4484</div>
4485</div>
4486<div class="paragraph">
4487<p>Structure members may not be qualified with <strong>const</strong>.
4488Structure variables can be declared as <strong>const</strong>, and initialized with a
4489structure
4490constructor.</p>
4491</div>
4492<div class="paragraph">
4493<p>Initializers for <strong>const</strong> declarations
4494must be constant expressions, as defined in
4495&#8220;<a href="#constant-expressions">Constant Expressions</a>&#8221;.</p>
4496</div>
4497</div>
4498<div class="sect3">
4499<h4 id="constant-expressions">4.3.3. Constant Expressions</h4>
4500<div class="paragraph">
4501<p>SPIR-V specialization constants are expressed in GLSL ES as <strong>const</strong> with the
4502layout qualifier <strong>constant_id</strong>, as described in
4503&#8220;<a href="#specialization-constant-qualifier">Specialization-Constant
4504Qualifier.</a>&#8221;</p>
4505</div>
4506<div class="paragraph">
4507<p>A <em>constant expression</em> is one of</p>
4508</div>
4509<div class="ulist">
4510<ul>
4511<li>
4512<p>A literal value (e.g. <strong>5</strong> or <strong>true</strong>).</p>
4513</li>
4514<li>
4515<p>A variable declared with the <strong>const</strong> qualifier and an initializer, where
4516the initializer is a constant expression.
4517This includes both <strong>const</strong> declared with a specialization-constant
4518layout qualifier, e.g. <strong>layout</strong>(<strong>constant_id</strong> = &#8230;&#8203;), and those declared
4519without a specialization-constant layout qualifier.</p>
4520</li>
4521<li>
4522<p>Built-in variables qualified as <strong>const</strong>.</p>
4523</li>
4524<li>
4525<p>An expression formed by an operator on operands that are all constant
4526expressions, including getting an element of a constant array, or a
4527member of a constant structure, or components of a constant vector.
4528However, the lowest precedence operators of the sequence operator (<strong>,</strong>)
4529and the assignment operators (<strong>=</strong>, <strong>+=</strong>, <strong>&#8230;&#8203;</strong>) are not included in the
4530operators that can create a constant expression.
4531Also, an array access with a specialization constant as an index does
4532not result in a constant expression.</p>
4533</li>
4534<li>
4535<p>The <strong>length</strong>() method on a compile-time sized array, whether or not the
4536object itself is constant.</p>
4537</li>
4538<li>
4539<p>A constructor whose arguments are all constant expressions.</p>
4540</li>
4541<li>
4542<p>For non-specialization constants only:
4543A built-in function call whose arguments are all constant expressions,
4544with the exception of the texture lookup functions.
4545This rule excludes functions with a <strong>void</strong> return or functions that have
4546an <strong>out</strong> parameter.
4547The built-in functions <strong>dFdx</strong>, <strong>dFdy</strong>, and <strong>fwidth</strong> must return 0 when
4548evaluated inside an initializer with an argument that is a constant
4549expression.</p>
4550</li>
4551</ul>
4552</div>
4553<div class="paragraph">
4554<p>Function calls to user-defined functions (non-built-in functions) cannot be
4555used to form constant expressions.</p>
4556</div>
4557<div class="paragraph">
4558<p>Scalar, vector, matrix, array and structure variables are constant
4559expressions if qualified as <strong>const</strong>.
4560Opaque types cannot be constant expressions.</p>
4561</div>
4562<div class="paragraph">
4563<p>A <em>constant integral expression</em> is a constant expression that evaluates to
4564a scalar signed or unsigned integer.</p>
4565</div>
4566<div class="paragraph">
4567<p>Constant expressions will be evaluated in an invariant way so as to create
4568the same value in multiple shaders when the same constant expressions appear
4569in those shaders.
4570See &#8220;<a href="#the-invariant-qualifier">The Invariant Qualifier</a>&#8221; for more details
4571on how to create invariant expressions and
4572&#8220;<a href="#precision-qualifiers">Precision Qualifiers</a>&#8221; for detail on how
4573expressions are evaluated.</p>
4574</div>
4575<div class="paragraph">
4576<p>Constant expressions respect the <strong>precise</strong> and <strong>invariant</strong> qualifiers but
4577will be always be evaluated in an invariant way, independent of the use of
4578such qualification, so as to create the same value in multiple shaders when
4579the same constant expressions appear in those shaders.
4580See &#8220;<a href="#the-invariant-qualifier">The Invariant Qualifier</a>&#8221; and
4581&#8220;<a href="#the-precise-qualifier">The Precise Qualifier</a>&#8221; for more details on how
4582to create invariant expressions.</p>
4583</div>
4584<div class="paragraph">
4585<p>Constant-expressions may be evaluated by a
4586host platform, and are therefore not required to compute the same value that
4587the same expression would evaluate to on the shader execution target.
4588However, the host must use the same or greater precision than the target
4589would use.
4590When the precision qualification cannot be determined, the expression is
4591evaluated at <strong>highp</strong>.
4592See &#8220;<a href="#default-precision-qualifiers">Default Precision Qualifiers</a>&#8221;.</p>
4593</div>
4594<div class="paragraph">
4595<p>Specialization-constant expressions are never evaluated by the compiler
4596front end, but instead retain the expression&#8217;s operations needed to evaluate
4597them later on the host.</p>
4598</div>
4599</div>
4600<div class="sect3">
4601<h4 id="input-variables">4.3.4. Input Variables</h4>
4602<div class="paragraph">
4603<p>Shader input variables are declared with the <strong>in</strong> storage qualifier.
4604They form the input interface between previous stages of the API
4605pipeline and the declaring shader.
4606Input variables must be declared at global scope.
4607Values from the previous pipeline stage are copied into input variables at
4608the beginning of shader execution.
4609It is an error to write to a variable declared as an input.</p>
4610</div>
4611<div class="paragraph">
4612<p>Only the input variables that are
4613actually
4614read need to be written by the
4615previous stage; it is allowed to have superfluous declarations of input
4616variables.</p>
4617</div>
4618<div class="paragraph">
4619<p>See &#8220;<a href="#built-in-variables">Built-In Variables</a>&#8221; for a list of the built-in
4620input names.</p>
4621</div>
4622<div class="paragraph">
4623<p>Vertex shader input variables (or attributes) receive per-vertex data.
4624It is an error to use auxiliary storage or interpolation qualifiers
4625on a vertex shader input.
4626The values copied in are established by the API or through the use
4627of the layout identifier <strong>location</strong>.</p>
4628</div>
4629<div class="paragraph">
4630<p>It is a compile-time error to declare a vertex shader input with, or that
4631contains, any of the following types:</p>
4632</div>
4633<div class="ulist">
4634<ul>
4635<li>
4636<p>A <a href="#booleans">boolean type</a></p>
4637</li>
4638<li>
4639<p>An <a href="#opaque-types">opaque type</a></p>
4640</li>
4641<li>
4642<p>An array</p>
4643</li>
4644<li>
4645<p>A structure</p>
4646</li>
4647</ul>
4648</div>
4649<div class="paragraph">
4650<p>Example declarations in a vertex shader:</p>
4651</div>
4652<div class="listingblock">
4653<div class="content">
4654<pre class="CodeRay highlight"><code data-lang="c++">in vec4 position;
4655in vec3 normal;
4656in vec2 texCoord[<span class="integer">4</span>];</code></pre>
4657</div>
4658</div>
4659<div class="paragraph">
4660<p>It is expected that graphics hardware will have a small number of fixed
4661vector locations for passing vertex inputs.
4662Therefore, the OpenGL ES Shading Language defines each non-matrix input variable as taking up
4663one such vector location.
4664There is an implementation-dependent limit on the number of locations that
4665can be used, and if this is exceeded it will cause a link-time error.
4666(Declared input variables that are not statically used do not count against
4667this limit.) A scalar input counts the same amount against this limit as a
4668<strong>vec4</strong>, so applications may want to consider packing groups of four
4669unrelated float inputs together into a vector to better utilize the
4670capabilities of the underlying hardware.
4671A matrix input will use up multiple locations.
4672The number of locations used will equal the number of columns in the matrix.</p>
4673</div>
4674<div class="paragraph">
4675<p>Tessellation control, evaluation, and geometry shader input variables get
4676the per-vertex values written out by output variables of the same names in
4677the previous active shader stage.
4678For these inputs, <strong>centroid</strong> and interpolation qualifiers are allowed, but
4679have no effect.
4680Since tessellation control, tessellation evaluation, and geometry shaders
4681operate on a set of vertices, each input variable (or input block, see
4682interface blocks below) needs to be declared as an array.
4683For example,</p>
4684</div>
4685<div class="listingblock">
4686<div class="content">
4687<pre class="CodeRay highlight"><code data-lang="c++">in <span class="predefined-type">float</span> foo[]; <span class="comment">// geometry shader input for vertex &quot;out float foo&quot;</span></code></pre>
4688</div>
4689</div>
4690<div class="paragraph">
4691<p>Each element of such an array corresponds to one vertex of the primitive
4692being processed.
4693Each array can optionally have a size declared.
4694For geometry shaders, the array size will be set by, (or if provided must be
4695consistent with) the input <strong>layout</strong> declaration(s) establishing the type of
4696input primitive, as described later in &#8220;<a href="#input-layout-qualifiers">Input
4697Layout Qualifiers</a>&#8221;.</p>
4698</div>
4699<div class="paragraph">
4700<p>Some inputs and outputs are <em>arrayed</em>, meaning that for an interface between
4701two shader stages either the input or output declaration requires an extra
4702level of array indexing for the declarations to match.
4703For example, with the interface between a vertex shader and a geometry
4704shader, vertex shader output variables and geometry shader input variables
4705of the same name must have matching types, except that the geometry shader
4706will have one more array dimension than the vertex shader, to allow for
4707vertex indexing.
4708If such an arrayed interface variable is not declared with the necessary
4709additional input or output array dimension, a link-time error will result.
4710Geometry shader inputs, tessellation control shader inputs and outputs, and
4711tessellation evaluation inputs all have an additional level of arrayness
4712relative to other shader inputs and outputs.
4713These inputs and outputs are known as <em>per-vertex-arrayed</em> inputs and
4714outputs.
4715Component limits for arrayed interfaces (e.g.
4716<em>gl_MaxTessControlInputComponents</em>) are limits per vertex, not limits for
4717the entire interface.</p>
4718</div>
4719<div class="paragraph">
4720<p>For non-arrayed interfaces (meaning array dimensionally stays the same
4721between stages), it is a link-time error if the input variable is not
4722declared with the same type, including array dimensionality, as the matching
4723output variable.</p>
4724</div>
4725<div class="paragraph">
4726<p>The link-time type-matching rules apply to all declared input and output
4727variables, whether or not they are used.</p>
4728</div>
4729<div class="paragraph">
4730<p>Additionally, tessellation evaluation shaders support per-patch input
4731variables declared with the <strong>patch</strong> and <strong>in</strong> qualifiers.
4732Per-patch input variables are filled with the values of per-patch output
4733variables written by the tessellation control shader.
4734Per-patch inputs may be declared as one-dimensional arrays, but are not
4735indexed by vertex number.
4736Applying the <strong>patch</strong> qualifier to inputs can only be done in tessellation
4737evaluation shaders.
4738As with other input variables, per-patch inputs must be declared using the
4739same type and qualification as per-patch outputs from the previous
4740(tessellation control) shader stage.
4741It is a compile-time error to use <strong>patch</strong> with inputs in any other stage.</p>
4742</div>
4743<div class="paragraph">
4744<p>It is a compile-time error to declare a tessellation control, tessellation
4745evaluation or geometry shader input with, or that contains, any of the
4746following types:</p>
4747</div>
4748<div class="ulist">
4749<ul>
4750<li>
4751<p>A <a href="#booleans">boolean type</a></p>
4752</li>
4753<li>
4754<p>An <a href="#opaque-types">opaque type</a></p>
4755</li>
4756<li>
4757<p>A structure containing an array</p>
4758</li>
4759<li>
4760<p>A structure containing a structure</p>
4761</li>
4762<li>
4763<p>For per-vertex-arrayed variables:</p>
4764<div class="ulist">
4765<ul>
4766<li>
4767<p>Per-vertex-arrayed arrays of arrays</p>
4768</li>
4769<li>
4770<p>Per-vertex-arrayed arrays of structures</p>
4771</li>
4772</ul>
4773</div>
4774</li>
4775<li>
4776<p>For non-per-vertex-arrayed variables:</p>
4777<div class="ulist">
4778<ul>
4779<li>
4780<p>An array of arrays</p>
4781</li>
4782<li>
4783<p>An array of structures</p>
4784</li>
4785</ul>
4786</div>
4787</li>
4788</ul>
4789</div>
4790<div class="paragraph">
4791<p>Fragment shader inputs get per-fragment values, typically interpolated from
4792a previous stage&#8217;s outputs.</p>
4793</div>
4794<div class="paragraph">
4795<p>It is a compile-time error to declare a fragment shader input with, or that
4796contains, any of the following types:</p>
4797</div>
4798<div class="ulist">
4799<ul>
4800<li>
4801<p>A <a href="#booleans">boolean type</a></p>
4802</li>
4803<li>
4804<p>An <a href="#opaque-types">opaque type</a></p>
4805</li>
4806<li>
4807<p>An array of arrays</p>
4808</li>
4809<li>
4810<p>An array of structures</p>
4811</li>
4812<li>
4813<p>A structure containing an array</p>
4814</li>
4815<li>
4816<p>A structure containing a structure</p>
4817</li>
4818</ul>
4819</div>
4820<div class="paragraph">
4821<p>Fragment shader inputs that are, or contain, integral
4822types must be
4823qualified with the interpolation qualifier <strong>flat</strong>.</p>
4824</div>
4825<div class="paragraph">
4826<p>Fragment inputs are declared as in the following examples:</p>
4827</div>
4828<div class="listingblock">
4829<div class="content">
4830<pre class="CodeRay highlight"><code data-lang="c++">in vec3 normal;
4831centroid in vec2 TexCoord;
4832flat in vec3 myColor;</code></pre>
4833</div>
4834</div>
4835<div class="paragraph">
4836<p>The fragment shader inputs form an interface with the last active shader in
4837the vertex processing pipeline.
4838For this interface, the last active shader stage output variables and
4839fragment shader input variables of the same name must match in type and
4840qualification, with a few exceptions: The storage qualifiers must, of
4841course, differ (one is <strong>in</strong> and one is <strong>out</strong>).
4842Also,
4843auxiliary qualification (e.g. <strong>centroid</strong>) may differ.
4844When
4845auxiliary qualifiers do not match, those provided in
4846the fragment shader supersede those provided in previous stages.
4847If any such qualifiers are completely missing in the fragment shaders, then
4848the default is used, rather than any qualifiers that may have been declared
4849in previous stages.
4850That is, what matters is what is declared in the fragment shaders, not what
4851is declared in shaders in previous stages.</p>
4852</div>
4853<div class="paragraph">
4854<p>When an interface between shader stages is formed using shaders from two
4855separate program objects, it is not possible to detect mismatches between
4856inputs and outputs when the programs are linked.
4857When there are mismatches between inputs and outputs on such interfaces,
4858attempting to use the two programs in the same program pipeline will result
4859in program pipeline validation failures, as described in section 7.4.1
4860&#8220;Shader Interface Matching&#8221; of the <a href="#references">OpenGL ES Specification</a>.</p>
4861</div>
4862<div class="paragraph">
4863<p>Shaders can ensure matches across such interfaces either by using input and
4864output layout qualifiers (sections &#8220;<a href="#input-layout-qualifiers">Input Layout
4865Qualifiers</a>&#8221; and &#8220;<a href="#output-layout-qualifiers">Output Layout
4866Qualifiers</a>&#8221;) or by using identical input and output declarations of
4867blocks or variables.
4868Complete rules for interface matching are found in section 7.4.1 &#8220;Shader
4869Interface Matching&#8221; of the <a href="#references">OpenGL ES Specification</a>.</p>
4870</div>
4871<div class="paragraph">
4872<p>Compute shaders do not permit user-defined input variables and do not form a
4873formal interface with any other shader stage.
4874See &#8220;<a href="#compute-shader-special-variables">Compute Shader Special
4875Variables</a>&#8221; for a description of built-in compute shader input variables.
4876All other input to a compute shader is retrieved explicitly through image
4877loads, texture fetches, loads from uniforms or uniform buffers, or other
4878user supplied code.</p>
4879</div>
4880</div>
4881<div class="sect3">
4882<h4 id="uniform-variables">4.3.5. Uniform Variables</h4>
4883<div class="paragraph">
4884<p>The <strong>uniform</strong> qualifier is used to declare global variables whose values are
4885the same across the entire primitive being processed.
4886All <strong>uniform</strong> variables are read-only.
4887Except for variables declared within a uniform block, all uniform variables
4888are initialized to 0 at link time and may be updated through the API.
4889When targeting Vulkan, it is an error to declare <strong>uniform</strong>
4890variables outside a block.</p>
4891</div>
4892<div class="paragraph">
4893<p>Example declarations are:</p>
4894</div>
4895<div class="listingblock">
4896<div class="content">
4897<pre class="CodeRay highlight"><code data-lang="c++">uniform vec4 lightPosition;</code></pre>
4898</div>
4899</div>
4900<div class="paragraph">
4901<p>The <strong>uniform</strong> qualifier can be used with any of the basic data types, or
4902when declaring a variable whose type is a structure, or an array of any of
4903these.</p>
4904</div>
4905<div class="paragraph">
4906<p>There is an implementation-dependent limit on the amount of storage for
4907uniforms that can be used for each type of shader and if this is exceeded it
4908will cause a compile-time or link-time error.
4909Uniform variables that are declared but not
4910statically
4911used do not count against this limit.
4912The number of user-defined uniform variables and the number of built-in
4913uniform variables that are used within a shader are added together to
4914determine whether available uniform storage has been exceeded.</p>
4915</div>
4916<div class="paragraph">
4917<p>Uniforms in shaders all share a single global name space when linked into a
4918program or separable program.
4919Hence, the types,
4920precisions,
4921and any location specifiers of all statically used uniform variables with the
4922same name must match across all shaders that are linked into a single program.
4923However it is not required to repeat the
4924location specifier in all the linked shaders.
4925While this single uniform name space is cross stage, a uniform variable
4926name&#8217;s scope is per stage: If a uniform variable name is declared in one
4927stage (e.g. a vertex shader) but not in another (e.g. a fragment shader),
4928then that name is still available in the other stage for a different use.</p>
4929</div>
4930<div class="paragraph">
4931<p>A compile or link-time error is generated if any of the explicitly given or
4932compiler generated uniform locations is greater than the
4933implementation-defined maximum number of uniform locations minus one.</p>
4934</div>
4935<div class="paragraph">
4936<p>Unlike locations for inputs and outputs, uniform locations are logical
4937values, not register locations, and there is no concept of overlap.
4938For example:</p>
4939</div>
4940<div class="listingblock">
4941<div class="content">
4942<pre class="CodeRay highlight"><code data-lang="c++">layout(location = <span class="integer">2</span>) uniform mat4 x;
4943layout(location = <span class="integer">3</span>) uniform mat4 y; <span class="comment">// No overlap with x</span>
4944layout(location = <span class="integer">2</span>) in mat4 x;
4945layout(location = <span class="integer">3</span>) in mat4 y; <span class="comment">// Error, locations conflict with x</span></code></pre>
4946</div>
4947</div>
4948</div>
4949<div class="sect3">
4950<h4 id="output-variables">4.3.6. Output Variables</h4>
4951<div class="paragraph">
4952<p>Shader output variables are declared with the <strong>out</strong> storage qualifier.
4953They form the output interface between the declaring shader and the
4954subsequent stages of the API pipeline.
4955Output variables must be declared at global scope.
4956During shader execution they will behave as normal unqualified global
4957variables.
4958Their values are copied out to the subsequent pipeline stage on shader exit.
4959Only output variables that are read by the subsequent pipeline stage need to
4960be written; it is allowed to have superfluous declarations of output
4961variables.</p>
4962</div>
4963<div class="paragraph">
4964<p>There is <em>not</em> an <strong>inout</strong> storage qualifier for declaring a single variable
4965name as both input and output to a shader.
4966Also, a variable cannot be declared with both the <strong>in</strong> and the <strong>out</strong>
4967qualifiers, this will result in a compile-time or link-time error.
4968Output variables must be declared with different names than input variables.
4969However, nesting an input or output inside an interface block with an
4970instance name allows the same names with one referenced through a block
4971instance name.</p>
4972</div>
4973<div class="paragraph">
4974<p>Vertex, tessellation evaluation, and geometry output variables output
4975per-vertex data and are declared using the <strong>out</strong> storage qualifier.
4976Applying <strong>patch</strong> to an output can only be done in a tessellation control
4977shader.
4978It is a compile-time error to use <strong>patch</strong> on outputs in any other stage.</p>
4979</div>
4980<div class="paragraph">
4981<p>It is a compile-time error to declare a vertex, tessellation evaluation,
4982tessellation control, or geometry shader output with, or that contains, any
4983of the following types:</p>
4984</div>
4985<div class="ulist">
4986<ul>
4987<li>
4988<p>A <a href="#booleans">boolean type</a></p>
4989</li>
4990<li>
4991<p>An <a href="#opaque-types">opaque type</a></p>
4992</li>
4993<li>
4994<p>A structure containing an array</p>
4995</li>
4996<li>
4997<p>A structure containing a structure</p>
4998</li>
4999<li>
5000<p>For per-vertex-arrayed variables (applies to tessellation control,
5001tessellation evaluation and geometry shaders):</p>
5002<div class="ulist">
5003<ul>
5004<li>
5005<p>Per-vertex-arrayed arrays of arrays</p>
5006</li>
5007<li>
5008<p>Per-vertex-arrayed arrays of structures</p>
5009</li>
5010</ul>
5011</div>
5012</li>
5013<li>
5014<p>For non-per-vertex-arrayed variables:</p>
5015<div class="ulist">
5016<ul>
5017<li>
5018<p>An array of arrays</p>
5019</li>
5020<li>
5021<p>An array of structures</p>
5022</li>
5023</ul>
5024</div>
5025</li>
5026</ul>
5027</div>
5028<div class="paragraph">
5029<p>Vertex shader outputs may be qualified with the interpolation qualifier
5030<strong>flat</strong><sup>1</sup>.</p>
5031</div>
5032<div class="dlist">
5033<dl>
5034<dt class="hdlist1">1</dt>
5035<dd>
5036<p>Unlike previous versions of the OpenGL ES Shading Language, there is no requirement for
5037outputs containing integers to be qualified as <strong>flat</strong>, since the vertex
5038shader may interface with the tessellation control shader.
5039However, in all cases, the qualifier must match across interfaces.</p>
5040</dd>
5041</dl>
5042</div>
5043<div class="paragraph">
5044<p>Individual outputs are declared as in the following examples:</p>
5045</div>
5046<div class="listingblock">
5047<div class="content">
5048<pre class="CodeRay highlight"><code data-lang="c++">out vec3 normal;
5049centroid out vec2 TexCoord;
5050invariant centroid out vec4 Color;
5051flat out vec3 myColor;
5052sample out vec4 perSampleColor;</code></pre>
5053</div>
5054</div>
5055<div class="paragraph">
5056<p>These can also appear in interface blocks, as described in
5057&#8220;<a href="#interface-blocks">Interface Blocks</a>&#8221;.
5058Interface blocks allow simpler addition of arrays to the interface from
5059vertex to geometry shader.
5060They also allow a fragment shader to have the same input interface as a
5061geometry shader for a given vertex shader.</p>
5062</div>
5063<div class="paragraph">
5064<p>Tessellation control shader output variables are used to output
5065per-vertex and per-patch data.
5066Per-vertex output variables are arrayed (see <em>arrayed</em> under
5067&#8220;<a href="#input-variables">Input Variables</a>&#8221;) and declared using the <strong>out</strong>
5068qualifier without the <strong>patch</strong> qualifier.
5069Per-patch output variables are declared using the <strong>patch</strong> and <strong>out</strong>
5070qualifiers.</p>
5071</div>
5072<div class="paragraph">
5073<p>Since tessellation control shaders produce an arrayed primitive comprising
5074multiple vertices, each per-vertex output variable (or output block, see
5075interface blocks below) needs to be declared as an array.
5076For example,</p>
5077</div>
5078<div class="listingblock">
5079<div class="content">
5080<pre class="CodeRay highlight"><code data-lang="c++">out <span class="predefined-type">float</span> foo[]; <span class="comment">// feeds next stage input &quot;in float foo[]&quot;</span></code></pre>
5081</div>
5082</div>
5083<div class="paragraph">
5084<p>Each element of such an array corresponds to one vertex of the primitive
5085being produced.
5086Each array can optionally have a size declared.
5087The array size will be set by (or if provided must be consistent with) the
5088output layout declaration(s) establishing the number of vertices in the
5089output patch, as described later in
5090&#8220;<a href="#tessellation-control-outputs">Tessellation Control Outputs</a>&#8221;.</p>
5091</div>
5092<div class="paragraph">
5093<p>Each tessellation control shader invocation has a corresponding output patch
5094vertex, and may assign values to per-vertex outputs only if they belong to
5095that corresponding vertex.
5096If a per-vertex output variable is used as an l-value, it is a compile-time
5097or link-time error if the expression indicating the vertex index is not the
5098identifier <em>gl_InvocationID</em>.</p>
5099</div>
5100<div class="paragraph">
5101<p>The order of execution of a tessellation control shader invocation relative
5102to the other invocations for the same input patch is undefined unless the
5103built-in function <strong>barrier</strong>() is used.
5104This provides some control over relative execution order.
5105When a shader invocation calls <strong>barrier</strong>(), its execution pauses until all
5106other invocations have reached the same point of execution.
5107Output variable assignments performed by any invocation executed prior to
5108calling <strong>barrier</strong>() will be visible to any other invocation after the call
5109to <strong>barrier</strong>() returns.</p>
5110</div>
5111<div class="paragraph">
5112<p>Because tessellation control shader invocations execute in undefined order
5113between barriers, the values of per-vertex or per-patch output variables
5114will sometimes be undefined.
5115Consider the beginning and end of shader execution and each call to
5116<strong>barrier</strong>() as synchronization points.
5117The value of an output variable will be undefined in any of the three
5118following cases:</p>
5119</div>
5120<div class="olist arabic">
5121<ol class="arabic">
5122<li>
5123<p>At the beginning of execution.</p>
5124</li>
5125<li>
5126<p>At each synchronization point, unless</p>
5127<div class="openblock">
5128<div class="content">
5129<div class="ulist">
5130<ul>
5131<li>
5132<p>the value was well-defined after the previous synchronization point and
5133was not written by any invocation since, or</p>
5134</li>
5135<li>
5136<p>the value was written by exactly one shader invocation since the previous
5137synchronization point, or</p>
5138</li>
5139<li>
5140<p>the value was written by multiple shader invocations since the previous
5141synchronization point, and the last write performed by all such
5142invocations wrote the same value.</p>
5143</li>
5144</ul>
5145</div>
5146</div>
5147</div>
5148</li>
5149<li>
5150<p>When read by a shader invocation, if</p>
5151<div class="openblock">
5152<div class="content">
5153<div class="ulist">
5154<ul>
5155<li>
5156<p>the value was undefined at the previous synchronization point and has not
5157been written by the same shader invocation since, or</p>
5158</li>
5159<li>
5160<p>the output variable is written to by any other shader invocation between
5161the previous and next synchronization points, even if that assignment
5162occurs in code following the read.</p>
5163</li>
5164</ul>
5165</div>
5166</div>
5167</div>
5168</li>
5169</ol>
5170</div>
5171<div class="paragraph">
5172<p>Fragment outputs output per-fragment data and are declared using the <strong>out</strong>
5173storage qualifier.
5174It is an error to use auxiliary storage qualifiers or
5175interpolation qualifiers in a fragment shader output declaration.
5176It is a compile-time error to declare a fragment shader output with, or that
5177contains, any of the following types:</p>
5178</div>
5179<div class="ulist">
5180<ul>
5181<li>
5182<p>A <a href="#booleans">boolean type</a></p>
5183</li>
5184<li>
5185<p>An <a href="#opaque-types">opaque type</a></p>
5186</li>
5187<li>
5188<p>A matrix type</p>
5189</li>
5190<li>
5191<p>A structure</p>
5192</li>
5193<li>
5194<p>An array of arrays</p>
5195</li>
5196</ul>
5197</div>
5198<div class="paragraph">
5199<p>Fragment shader outputs declared as arrays may only be indexed by a constant
5200integral expression.</p>
5201</div>
5202<div class="paragraph">
5203<p>Fragment outputs are declared as in the following examples:</p>
5204</div>
5205<div class="listingblock">
5206<div class="content">
5207<pre class="CodeRay highlight"><code data-lang="c++">out vec4 FragmentColor;
5208out uint Luminosity;</code></pre>
5209</div>
5210</div>
5211<div class="paragraph">
5212<p>Compute shaders have no built-in output variables, do not support
5213user-defined output variables and do not form a formal interface with any
5214other shader stage.
5215All outputs from a compute shader take the form of the side effects such as
5216image stores and operations on atomic counters.</p>
5217</div>
5218</div>
5219<div class="sect3">
5220<h4 id="buffer-variables">4.3.7. Buffer Variables</h4>
5221<div class="paragraph">
5222<p>The <strong>buffer</strong> qualifier is used to declare global variables whose values are
5223stored in the data store of a buffer object bound through the API.
5224Buffer variables can be read and written, with the underlying storage shared
5225among all active shader invocations.
5226Buffer variable memory reads and writes within a single shader invocation
5227are processed in order.
5228However, the order of reads and writes performed in one invocation relative
5229to those performed by another invocation is largely undefined.
5230Buffer variables may be qualified with memory qualifiers affecting how the
5231underlying memory is accessed, as described in &#8220;<a href="#memory-qualifiers">Memory
5232Qualifiers</a>&#8221;.</p>
5233</div>
5234<div class="paragraph">
5235<p>The <strong>buffer</strong> qualifier can be used to declare interface blocks (see
5236&#8220;<a href="#interface-blocks">Interface Blocks</a>&#8221;), which are then referred to as
5237shader storage blocks.
5238It is a compile-time error to declare buffer variables outside a block.
5239Buffer variables cannot have initializers.</p>
5240</div>
5241<div class="admonitionblock note">
5242<table>
5243<tr>
5244<td class="icon">
5245<i class="fa icon-note" title="Note"></i>
5246</td>
5247<td class="content">
5248<div class="title">Note</div>
5249<div class="paragraph">
5250<p>The terms <em>shader storage buffer (object)</em>, and <em>shader storage block</em> are
5251often used interchangeably.
5252The former generally refers to the underlying storage whereas the latter
5253refers only to the interface definition in the shader.
5254Similarly for <em>uniform buffer objects</em> and _uniform blocks.</p>
5255</div>
5256</td>
5257</tr>
5258</table>
5259</div>
5260<div class="listingblock">
5261<div class="content">
5262<pre class="CodeRay highlight"><code data-lang="c++"><span class="comment">// use buffer to create a buffer block (shader storage block)</span>
5263buffer BufferName { <span class="comment">// externally visible name of buffer</span>
5264    <span class="predefined-type">int</span> count;      <span class="comment">// typed, shared memory...</span>
5265    ...             <span class="comment">// ...</span>
5266    vec4 v[];       <span class="comment">// last member may be an array that is not sized</span>
5267                    <span class="comment">// until after link time (dynamically sized)</span>
5268} Name;             <span class="comment">// name of block within the shader</span></code></pre>
5269</div>
5270</div>
5271<div class="paragraph">
5272<p>There are implementation-dependent limits on the number of shader storage
5273blocks used for each type of shader, the combined number of shader storage
5274blocks used for a program, and the amount of storage required by each
5275individual shader storage block.
5276If any of these limits are exceeded, it will cause a compile-time or
5277link-time error.</p>
5278</div>
5279<div class="paragraph">
5280<p>If multiple shaders are linked together, then they will share a single
5281global buffer variable name space.
5282Hence, the types of all declared buffer variables with the same name must
5283match across all shaders that are linked into a single program.</p>
5284</div>
5285<div class="paragraph">
5286<p>Precision qualifiers for such variables need not match.</p>
5287</div>
5288</div>
5289<div class="sect3">
5290<h4 id="shared-variables">4.3.8. Shared Variables</h4>
5291<div class="paragraph">
5292<p>The <strong>shared</strong> qualifier is used to declare global variables that have storage
5293shared between all work items in a compute shader workgroup.
5294Variables declared as <strong>shared</strong> may only be used in compute shaders (see
5295&#8220;<a href="#compute-processor">Compute Processor</a>&#8221;).
5296Any other declaration of a <strong>shared</strong> variable is an error.
5297Shared variables are implicitly coherent (see
5298&#8220;<a href="#memory-qualifiers">Memory Qualifiers</a>&#8221;).</p>
5299</div>
5300<div class="paragraph">
5301<p>Variables declared as <strong>shared</strong> may not have initializers and their contents
5302are undefined at the beginning of shader execution.
5303Any data written to <strong>shared</strong> variables will be visible to other work items
5304(executing the same shader) within the same workgroup.</p>
5305</div>
5306<div class="paragraph">
5307<p>In the absence of synchronization, the order of reads and writes to the same
5308<strong>shared</strong> variable by different invocations of a shader is not defined.</p>
5309</div>
5310<div class="paragraph">
5311<p>In order to achieve ordering with respect to reads and writes to <strong>shared</strong>
5312variables, control flow barriers must be employed using the <strong>barrier</strong>() function
5313(see &#8220;<a href="#shader-invocation-control-functions">Shader Invocation Control
5314Functions</a>&#8221;).</p>
5315</div>
5316<div class="paragraph">
5317<p>There is a limit to the total size of all variables declared as <strong>shared</strong> in a
5318single program.
5319This limit, expressed in units of basic machine units may be determined by
5320using the OpenGL ES API to query the value of
5321MAX_COMPUTE_SHARED_MEMORY_SIZE.</p>
5322</div>
5323</div>
5324<div class="sect3">
5325<h4 id="interface-blocks">4.3.9. Interface Blocks</h4>
5326<div class="paragraph">
5327<p>Input, output, uniform, and buffer variable declarations can be grouped into
5328named interface blocks to provide coarser granularity backing than is
5329achievable with individual declarations.
5330They can have an optional instance name, used in the shader to reference
5331their members.
5332An output block of one programmable stage is backed by a corresponding input
5333block in the subsequent programmable stage.
5334A <em>uniform block</em> is backed by the application with a buffer object.
5335A block of buffer variables, called a <em>shader storage block</em>, is also backed
5336by the application with a buffer object.
5337It is a compile-time error to have an input block in a vertex shader or an
5338output block in a fragment shader.
5339These uses are reserved for future use.</p>
5340</div>
5341<div class="paragraph">
5342<p>An interface block is started by an <strong>in</strong>, <strong>out</strong>, <strong>uniform</strong>, or <strong>buffer</strong>
5343keyword, followed by a block name, followed by an open curly brace (<strong>{</strong>) as
5344follows:</p>
5345</div>
5346<div class="openblock bnf">
5347<div class="content">
5348<div class="dlist">
5349<dl>
5350<dt class="hdlist1"><em>interface-block</em> : </dt>
5351<dd>
5352<p><em>layout-qualifier<sub>opt</sub></em> <em>interface-qualifier</em> <em>block-name</em> <strong>{</strong>
5353<em>member-list</em> <strong>}</strong> <em>instance-name<sub>opt</sub></em> <strong>;</strong></p>
5354</dd>
5355</dl>
5356</div>
5357<div class="dlist">
5358<dl>
5359<dt class="hdlist1"><em>interface-qualifier</em> : </dt>
5360<dd>
5361<p><strong>in</strong><br>
5362<strong>out</strong><br>
5363<strong>patch</strong> <strong>in</strong> // Note: Qualifiers can be in any order.<br>
5364<strong>patch</strong> <strong>out</strong><br>
5365<strong>uniform</strong><br>
5366<strong>buffer</strong></p>
5367</dd>
5368</dl>
5369</div>
5370<div class="dlist">
5371<dl>
5372<dt class="hdlist1"><em>member-list</em> : </dt>
5373<dd>
5374<p><em>member-declaration</em><br>
5375<em>member-declaration</em> <em>member-list</em></p>
5376</dd>
5377<dt class="hdlist1"><em>member-declaration</em> : </dt>
5378<dd>
5379<p><em>layout-qualifier<sub>opt</sub></em> <em>qualifiers<sub>opt</sub></em> <em>type</em> <em>declarators</em> <strong>;</strong></p>
5380</dd>
5381</dl>
5382</div>
5383<div class="dlist">
5384<dl>
5385<dt class="hdlist1"><em>instance-name</em> : </dt>
5386<dd>
5387<p><em>identifier</em><br>
5388<em>identifier</em> <strong>[</strong> <strong>]</strong><br>
5389<em>identifier</em> <strong>[</strong> <em>constant-integral-expression</em> <strong>]</strong></p>
5390</dd>
5391</dl>
5392</div>
5393</div>
5394</div>
5395<div class="paragraph">
5396<p>Each of the above elements is discussed below, with the exception of layout
5397qualifiers (<em>layout-qualifier</em>), which are defined in the next section.</p>
5398</div>
5399<div class="paragraph">
5400<p>First, an example,</p>
5401</div>
5402<div class="listingblock">
5403<div class="content">
5404<pre class="CodeRay highlight"><code data-lang="c++">uniform Transform {
5405    mat4 ModelViewMatrix;
5406    mat4 ModelViewProjectionMatrix;
5407    uniform mat3 NormalMatrix;      <span class="comment">// allowed restatement of qualifier</span>
5408    <span class="predefined-type">float</span> Deformation;
5409};</code></pre>
5410</div>
5411</div>
5412<div class="paragraph">
5413<p>The above establishes a uniform block named &#8220;Transform&#8221; with four uniforms
5414grouped inside it.</p>
5415</div>
5416<div class="paragraph">
5417<p>Types and declarators are the same as for other input, output, uniform, and
5418buffer variable declarations outside blocks, with these exceptions:</p>
5419</div>
5420<div class="ulist">
5421<ul>
5422<li>
5423<p>Opaque types are not allowed</p>
5424</li>
5425<li>
5426<p>Structure definitions cannot be nested inside a block</p>
5427</li>
5428<li>
5429<p>Arrays of arrays of blocks are not allowed, except for the case in the
5430tessellation pipe where the declaration is a per-vertex-array of arrays
5431of blocks.</p>
5432</li>
5433</ul>
5434</div>
5435<div class="paragraph">
5436<p>If no optional qualifier is used in a member-declaration, the qualification
5437of the member includes all <strong>in</strong>, <strong>out</strong>, <strong>patch</strong>, <strong>uniform</strong>, or <strong>buffer</strong> as
5438determined by <em>interface-qualifier</em>.
5439If optional qualifiers are used, they can include interpolation qualifiers,
5440auxiliary storage qualifiers,
5441precision qualifiers,
5442and storage qualifiers and they must declare
5443an input, output, or uniform member consistent with the interface qualifier
5444of the block: Input variables, output variables, uniform variables, and
5445<strong>buffer</strong> members can only be in <strong>in</strong> blocks, <strong>out</strong> blocks, <strong>uniform</strong> blocks,
5446and shader storage blocks, respectively.</p>
5447</div>
5448<div class="paragraph">
5449<p>Repeating the <strong>in</strong>, <strong>out</strong>, <strong>patch</strong>, <strong>uniform</strong>, or <strong>buffer</strong> interface
5450qualifier for a member&#8217;s storage qualifier is optional.
5451For example,</p>
5452</div>
5453<div class="listingblock">
5454<div class="content">
5455<pre class="CodeRay highlight"><code data-lang="c++">in Material {
5456    smooth in vec4 Color1; <span class="comment">// legal, input inside in block</span>
5457    smooth vec4 Color2;    <span class="comment">// legal, 'in' inherited from 'in Material'</span>
5458    vec2 TexCoord;         <span class="comment">// legal, TexCoord is an input</span>
5459    uniform <span class="predefined-type">float</span> Atten;   <span class="comment">// illegal, mismatched storage qualifier</span>
5460};</code></pre>
5461</div>
5462</div>
5463<div class="paragraph">
5464<p>A <em>shader interface</em> is defined to be one of these:</p>
5465</div>
5466<div class="ulist">
5467<ul>
5468<li>
5469<p>All the uniform variables and uniform blocks declared in a program.
5470This spans all compilation units linked together within one program.</p>
5471</li>
5472<li>
5473<p>All the <strong>buffer</strong> blocks declared in a program.</p>
5474</li>
5475<li>
5476<p>The boundary between adjacent programmable pipeline stages: This spans
5477all the outputs declared in all compilation units of the first stage and
5478all the inputs declared in all compilation units of the second stage.
5479Note that for the purposes of this definition, the fragment shader and
5480the preceding shader are considered to have a shared boundary even
5481though in practice, all values passed to the fragment shader first pass
5482through the rasterizer and interpolator.</p>
5483</li>
5484</ul>
5485</div>
5486<div class="paragraph">
5487<p>The block name (<em>block-name</em>) is used to match within shader interfaces: an
5488output block of one pipeline stage will be matched to an input block with
5489the same name in the subsequent pipeline stage.
5490For uniform or shader storage blocks, the application uses the block name to
5491identify the block.
5492Block names have no other use within a shader beyond interface matching; it
5493is an error
5494to use a block name at global scope for anything other than as a
5495block name (e.g. use of a block name for a global variable name or function
5496name is currently reserved).
5497It is a compile-time error to use the same block name for more than one
5498block declaration in the same shader interface (as defined above) within one
5499shader, even if the block contents are identical.</p>
5500</div>
5501<div class="paragraph">
5502<p>Matched block names within a shader interface (as defined above) must match
5503in terms of having the same number of declarations with the same sequence of
5504types and the same sequence of member names, as well as having matching
5505member-wise layout qualification
5506as defined in &#8220;<a href="#matching-of-qualifiers">Matching of Qualifiers</a>&#8221;.
5507Matched uniform or shader storage block names (but not input or output block
5508names) must also either all be lacking an instance name or all having an
5509instance name, putting their members at the same scoping level.
5510When instance names are present on matched block names, it is allowed for
5511the instance names to differ; they need not match for the blocks to match.
5512Furthermore, if a matching block is declared as an array, then the array
5513sizes must also match (or follow array matching rules for the shader
5514interface between consecutive shader stages).
5515Any mismatch will generate a link-time error.
5516A block name is allowed to have different definitions in different shader
5517interfaces within the same shader, allowing, for example, an input block and
5518output block to have the same name.</p>
5519</div>
5520<div class="paragraph">
5521<p>If an instance name (<em>instance-name</em>) is not used, the names declared inside
5522the block are scoped at the global level and accessed as if they were
5523declared outside the block.
5524If an instance name (<em>instance-name</em>) is used, then it puts all the members
5525inside a scope within its own name space, accessed with the field selector
5526(<strong>.</strong>) operator (analogously to structures).
5527For example,</p>
5528</div>
5529<div class="listingblock">
5530<div class="content">
5531<pre class="CodeRay highlight"><code data-lang="c++">in Light {
5532    vec4 LightPos;
5533    vec3 LightColor;
5534};
5535in ColoredTexture {
5536    vec4 Color;
5537    vec2 TexCoord;
5538} Material;           <span class="comment">// instance name</span>
5539vec3 Color;           <span class="comment">// different Color than Material.Color</span>
5540vec4 LightPos;        <span class="comment">// illegal, already defined</span>
5541...
5542... = LightPos;       <span class="comment">// accessing LightPos</span>
5543... = Material.Color; <span class="comment">// accessing Color in ColoredTexture block</span></code></pre>
5544</div>
5545</div>
5546<div class="paragraph">
5547<p>Outside the shading language (i.e., in the API), members are similarly
5548identified except the block name is always used in place of the instance
5549name (API accesses are to shader interfaces, not to shaders).
5550If there is no instance name, then the API does not use the block name to
5551access a member, just the member name.</p>
5552</div>
5553<div class="paragraph">
5554<p>Within a shader interface, all declarations of the same global name must be
5555for the same object and must match in type and in whether they declare a
5556variable or member of a block with no instance name.
5557The API also needs this name to uniquely identify an object in the shader
5558interface.
5559It is a link-time error if any particular shader interface contains</p>
5560</div>
5561<div class="ulist">
5562<ul>
5563<li>
5564<p>two different blocks, each having no instance name, and each having a
5565member of the same name, or</p>
5566</li>
5567<li>
5568<p>a variable outside a block, and a block with no instance name, where the
5569variable has the same name as a member in the block.</p>
5570</li>
5571</ul>
5572</div>
5573<div class="listingblock">
5574<div class="content">
5575<pre class="CodeRay highlight"><code data-lang="c++">out Vertex {
5576    vec4 Position;  <span class="comment">// API transform/feedback will use &quot;Vertex.Position&quot;</span>
5577    vec2 Texture;
5578} Coords;           <span class="comment">// shader will use &quot;Coords.Position&quot;</span>
5579out Vertex2 {
5580    vec4 Color;     <span class="comment">// API will use &quot;Color&quot;</span>
5581    <span class="predefined-type">float</span> Color2;
5582};
5583
5584<span class="comment">// in same program as Vertex2 above:</span>
5585out Vertex3 {
5586    <span class="predefined-type">float</span> Intensity;
5587    vec4 Color;     <span class="comment">// ERROR, name collision with Color in Vertex2</span>
5588};
5589<span class="predefined-type">float</span> Color2;       <span class="comment">// ERROR, collides with Color2 in Vertex2</span></code></pre>
5590</div>
5591</div>
5592<div class="paragraph">
5593<p>For blocks declared as arrays, the array index must also be included when
5594accessing members, as in this example</p>
5595</div>
5596<div class="listingblock">
5597<div class="content">
5598<pre class="CodeRay highlight"><code data-lang="c++">uniform Transform { <span class="comment">// API uses &quot;Transform[2]&quot; to refer to instance 2</span>
5599    mat4 ModelViewMatrix;
5600    mat4 ModelViewProjectionMatrix;
5601    <span class="predefined-type">float</span> Deformation;
5602} transforms[<span class="integer">4</span>];
5603...
5604... = transforms[<span class="integer">2</span>].ModelViewMatrix; <span class="comment">// shader access of instance 2</span>
5605<span class="comment">// API uses &quot;Transform.ModelViewMatrix&quot; to query an offset or other query</span></code></pre>
5606</div>
5607</div>
5608<div class="paragraph">
5609<p>For uniform or shader storage blocks declared as an array, each individual
5610array element corresponds to a separate buffer object bind range, backing
5611one instance of the block.
5612As the array size indicates the number of buffer objects needed, uniform and
5613shader storage block array declarations must specify an array size.
5614All indices used to index a shader storage block array must be constant
5615integral expressions.
5616A uniform block array can only be indexed with a dynamically uniform
5617integral expression, otherwise results are undefined.</p>
5618</div>
5619<div class="paragraph">
5620<p>When using OpenGL ES API entry points to identify the name of an individual
5621block in an array of blocks, the name string may include an array index
5622(e.g. <em>Transform[2]</em>).
5623When using OpenGL ES API entry points to refer to offsets or other
5624characteristics of a block member, an array index must not be specified
5625(e.g. <em>Transform.ModelViewMatrix</em>).
5626See section 7.3.1 &#8220;Program Interfaces&#8221; of the <a href="#references">OpenGL ES Specification</a> for
5627details.</p>
5628</div>
5629<div class="paragraph">
5630<p>Tessellation control, tessellation evaluation and geometry shader input
5631blocks must be declared as arrays and follow the array declaration and
5632linking rules for all shader inputs for the respective stages.
5633All other input and output block arrays must specify an array size.</p>
5634</div>
5635<div class="paragraph">
5636<p>There are implementation-dependent limits on the number of uniform blocks
5637and the number of shader storage blocks that can be used per stage.
5638If either limit is exceeded, it will cause a link-time error.</p>
5639</div>
5640</div>
5641</div>
5642<div class="sect2">
5643<h3 id="layout-qualifiers">4.4. Layout Qualifiers</h3>
5644<div class="paragraph">
5645<p>Layout qualifiers can appear in several forms of declaration.
5646They can appear as part of an interface block definition or block member, as
5647shown in the grammar in the previous section.
5648They can also appear with just an <em>interface-qualifier</em> to establish layouts
5649of other declarations made with that qualifier:</p>
5650</div>
5651<div class="openblock bnf">
5652<div class="content">
5653<div class="paragraph">
5654<p><em>layout-qualifier</em> <em>interface-qualifier</em> <strong>;</strong></p>
5655</div>
5656</div>
5657</div>
5658<div class="paragraph">
5659<p>Or, they can appear with an individual variable declared with an interface
5660qualifier:</p>
5661</div>
5662<div class="openblock bnf">
5663<div class="content">
5664<div class="paragraph">
5665<p><em>layout-qualifier</em> <em>interface-qualifier</em> <em>declaration</em> <strong>;</strong></p>
5666</div>
5667</div>
5668</div>
5669<div class="paragraph">
5670<p>Declarations of layouts can only be made at global scope or block members,
5671and only where indicated in the following subsections; their details are
5672specific to what the interface qualifier is, and are discussed individually.</p>
5673</div>
5674<div class="paragraph">
5675<p>The <em>layout-qualifier</em> expands to:</p>
5676</div>
5677<div class="openblock bnf">
5678<div class="content">
5679<div class="dlist">
5680<dl>
5681<dt class="hdlist1"><em>layout-qualifier</em> : </dt>
5682<dd>
5683<p><strong>layout</strong> <strong>(</strong> <em>layout-qualifier-id-list</em> <strong>)</strong></p>
5684</dd>
5685<dt class="hdlist1"><em>layout-qualifier-id-list</em> : </dt>
5686<dd>
5687<p><em>layout-qualifier-id</em><br>
5688<em>layout-qualifier-id</em> <strong>,</strong> <em>layout-qualifier-id-list</em></p>
5689</dd>
5690<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
5691<dd>
5692<p><em>layout-qualifier-name</em><br>
5693<em>layout-qualifier-name</em> <strong>=</strong> <em>layout-qualifier-value</em><br>
5694<strong>shared</strong></p>
5695</dd>
5696</dl>
5697</div>
5698</div>
5699</div>
5700<div class="openblock bnf">
5701<div class="content">
5702<div class="dlist">
5703<dl>
5704<dt class="hdlist1"><em>layout-qualifier-value</em> : </dt>
5705<dd>
5706<p><em>integer-constant</em></p>
5707</dd>
5708</dl>
5709</div>
5710</div>
5711</div>
5712<div class="paragraph">
5713<p>The tokens used for <em>layout-qualifier-name</em> are identifiers, not keywords,
5714however, the <strong>shared</strong> keyword is allowed as a <em>layout-qualifier-id</em>.
5715Generally, they can be listed in any order.
5716Order-dependent meanings exist only if explicitly called out below.
5717As for other identifiers, they are case sensitive.</p>
5718</div>
5719<div class="paragraph">
5720<p>The set of allowed layout qualifiers depends on the shader, the interface
5721and the variable type as specified in the following sections.
5722For example, a sampler in the default uniform block in a fragment shader can
5723have <strong>location</strong> and <strong>binding</strong> layout qualifiers but no others.
5724Invalid use of layout qualifiers is an error.</p>
5725</div>
5726<div class="paragraph">
5727<p>The following table summarizes the use of layout qualifiers.
5728It shows for each one what kinds of declarations it may be applied to.
5729These are all discussed in detail in the following sections.</p>
5730</div>
5731<table class="tableblock frame-all grid-all stretch">
5732<colgroup>
5733<col style="width: 16.6666%;">
5734<col style="width: 16.6666%;">
5735<col style="width: 16.6666%;">
5736<col style="width: 16.6666%;">
5737<col style="width: 16.6666%;">
5738<col style="width: 16.667%;">
5739</colgroup>
5740<thead>
5741<tr>
5742<th class="tableblock halign-left valign-middle">Layout Qualifier</th>
5743<th class="tableblock halign-center valign-middle">Qualifier Only</th>
5744<th class="tableblock halign-center valign-middle">Individual Variable</th>
5745<th class="tableblock halign-center valign-middle">Block</th>
5746<th class="tableblock halign-center valign-middle">Block Member</th>
5747<th class="tableblock halign-left valign-middle">Allowed Interfaces</th>
5748</tr>
5749</thead>
5750<tbody>
5751<tr>
5752<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>shared</strong><br>
5753  <strong>packed</strong><br>
5754  <strong>std140</strong><br>
5755  <strong>std430</strong></p></td>
5756<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5757<td class="tableblock halign-center valign-middle"></td>
5758<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5759<td class="tableblock halign-center valign-middle"></td>
5760<td class="tableblock halign-left valign-middle" rowspan="5"><p class="tableblock"><strong>uniform</strong> / <strong>buffer</strong></p></td>
5761</tr>
5762<tr>
5763<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>row_major</strong><br>
5764  <strong>column_major</strong></p></td>
5765<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5766<td class="tableblock halign-center valign-middle"></td>
5767<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5768<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5769</tr>
5770<tr>
5771<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>binding</strong> =</p></td>
5772<td class="tableblock halign-center valign-middle"></td>
5773<td class="tableblock halign-center valign-middle"><p class="tableblock">opaque types only</p></td>
5774<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5775<td class="tableblock halign-center valign-middle"></td>
5776</tr>
5777<tr>
5778<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>offset</strong> =</p></td>
5779<td class="tableblock halign-center valign-middle"></td>
5780<td class="tableblock halign-center valign-middle"><p class="tableblock">atomic counters only</p></td>
5781<td class="tableblock halign-center valign-middle"></td>
5782<td class="tableblock halign-center valign-middle"><p class="tableblock">Vulkan only</p></td>
5783</tr>
5784<tr>
5785<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>align</strong> =</p></td>
5786<td class="tableblock halign-center valign-middle"></td>
5787<td class="tableblock halign-center valign-middle"></td>
5788<td class="tableblock halign-center valign-middle"><p class="tableblock">Vulkan only</p></td>
5789<td class="tableblock halign-center valign-middle"><p class="tableblock">Vulkan only</p></td>
5790</tr>
5791<tr>
5792<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>set</strong> =</p></td>
5793<td class="tableblock halign-center valign-middle"></td>
5794<td class="tableblock halign-center valign-middle"><p class="tableblock">opaque types only</p></td>
5795<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5796<td class="tableblock halign-center valign-middle"></td>
5797<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>uniform</strong> / <strong>buffer</strong> (Vulkan only)</p></td>
5798</tr>
5799<tr>
5800<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>push_constant</strong></p></td>
5801<td class="tableblock halign-center valign-middle"></td>
5802<td class="tableblock halign-center valign-middle"></td>
5803<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5804<td class="tableblock halign-center valign-middle"></td>
5805<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>uniform</strong> (Vulkan only)</p></td>
5806</tr>
5807<tr>
5808<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>input_attachment_index</strong> =</p></td>
5809<td class="tableblock halign-center valign-middle"></td>
5810<td class="tableblock halign-center valign-middle"><p class="tableblock">subpass types only</p></td>
5811<td class="tableblock halign-center valign-middle"></td>
5812<td class="tableblock halign-center valign-middle"></td>
5813<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>uniform</strong> (Vulkan only)</p></td>
5814</tr>
5815<tr>
5816<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>location</strong> =</p></td>
5817<td class="tableblock halign-center valign-middle"></td>
5818<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5819<td class="tableblock halign-center valign-middle"></td>
5820<td class="tableblock halign-center valign-middle"></td>
5821<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>uniform</strong> / <strong>buffer</strong></p></td>
5822</tr>
5823<tr>
5824<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>location</strong> =</p></td>
5825<td class="tableblock halign-center valign-middle"></td>
5826<td class="tableblock halign-center valign-middle"><p class="tableblock">X<sup>1</sup></p></td>
5827<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5828<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5829<td class="tableblock halign-left valign-middle"><p class="tableblock">all <strong>in</strong> / <strong>out</strong>, except for compute</p></td>
5830</tr>
5831<tr>
5832<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>triangles</strong><br>
5833  <strong>quads</strong><br>
5834  <strong>isolines</strong></p></td>
5835<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5836<td class="tableblock halign-center valign-middle"></td>
5837<td class="tableblock halign-center valign-middle"></td>
5838<td class="tableblock halign-center valign-middle"></td>
5839<td class="tableblock halign-left valign-middle"><p class="tableblock">tessellation evaluation <strong>in</strong></p></td>
5840</tr>
5841<tr>
5842<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>equal_spacing</strong><br>
5843  <strong>fractional_even_spacing</strong><br>
5844  <strong>fractional_odd_spacing</strong></p></td>
5845<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5846<td class="tableblock halign-center valign-middle"></td>
5847<td class="tableblock halign-center valign-middle"></td>
5848<td class="tableblock halign-center valign-middle"></td>
5849<td class="tableblock halign-left valign-middle"><p class="tableblock">tessellation evaluation <strong>in</strong></p></td>
5850</tr>
5851<tr>
5852<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>cw</strong><br>
5853  <strong>ccw</strong></p></td>
5854<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5855<td class="tableblock halign-center valign-middle"></td>
5856<td class="tableblock halign-center valign-middle"></td>
5857<td class="tableblock halign-center valign-middle"></td>
5858<td class="tableblock halign-left valign-middle"><p class="tableblock">tessellation evaluation <strong>in</strong></p></td>
5859</tr>
5860<tr>
5861<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>point_mode</strong></p></td>
5862<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5863<td class="tableblock halign-center valign-middle"></td>
5864<td class="tableblock halign-center valign-middle"></td>
5865<td class="tableblock halign-center valign-middle"></td>
5866<td class="tableblock halign-left valign-middle"><p class="tableblock">tessellation evaluation <strong>in</strong></p></td>
5867</tr>
5868<tr>
5869<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>points</strong></p></td>
5870<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5871<td class="tableblock halign-center valign-middle"></td>
5872<td class="tableblock halign-center valign-middle"></td>
5873<td class="tableblock halign-center valign-middle"></td>
5874<td class="tableblock halign-left valign-middle"><p class="tableblock">geometry <strong>in</strong>/<strong>out</strong></p></td>
5875</tr>
5876<tr>
5877<td class="tableblock halign-left valign-middle"><p class="tableblock">[ <strong>points</strong> ]<br>
5878  <strong>lines</strong><br>
5879  <strong>lines_adjacency</strong><br>
5880  <strong>triangles</strong><br>
5881  <strong>triangles_adjacency</strong></p></td>
5882<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5883<td class="tableblock halign-center valign-middle"></td>
5884<td class="tableblock halign-center valign-middle"></td>
5885<td class="tableblock halign-center valign-middle"></td>
5886<td class="tableblock halign-left valign-middle"><p class="tableblock">geometry <strong>in</strong></p></td>
5887</tr>
5888<tr>
5889<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>invocations</strong> =</p></td>
5890<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5891<td class="tableblock halign-center valign-middle"></td>
5892<td class="tableblock halign-center valign-middle"></td>
5893<td class="tableblock halign-center valign-middle"></td>
5894<td class="tableblock halign-left valign-middle"><p class="tableblock">geometry <strong>in</strong></p></td>
5895</tr>
5896<tr>
5897<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>early_fragment_tests</strong></p></td>
5898<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5899<td class="tableblock halign-center valign-middle"></td>
5900<td class="tableblock halign-center valign-middle"></td>
5901<td class="tableblock halign-center valign-middle"></td>
5902<td class="tableblock halign-left valign-middle"><p class="tableblock">fragment <strong>in</strong></p></td>
5903</tr>
5904<tr>
5905<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>local_size_x</strong> =<br>
5906  <strong>local_size_y</strong> =<br>
5907  <strong>local_size_z</strong> =</p></td>
5908<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5909<td class="tableblock halign-center valign-middle"></td>
5910<td class="tableblock halign-center valign-middle"></td>
5911<td class="tableblock halign-center valign-middle"></td>
5912<td class="tableblock halign-left valign-middle"><p class="tableblock">compute <strong>in</strong></p></td>
5913</tr>
5914<tr>
5915<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>local_size_x_id</strong> =<br>
5916  <strong>local_size_y_id</strong> =<br>
5917  <strong>local_size_z_id</strong> =</p></td>
5918<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5919<td class="tableblock halign-center valign-middle"></td>
5920<td class="tableblock halign-center valign-middle"></td>
5921<td class="tableblock halign-center valign-middle"></td>
5922<td class="tableblock halign-left valign-middle"><p class="tableblock">compute <strong>in</strong> (SPIR-V only)</p></td>
5923</tr>
5924<tr>
5925<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>vertices</strong> =</p></td>
5926<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5927<td class="tableblock halign-center valign-middle"></td>
5928<td class="tableblock halign-center valign-middle"></td>
5929<td class="tableblock halign-center valign-middle"></td>
5930<td class="tableblock halign-left valign-middle"><p class="tableblock">tessellation control <strong>out</strong></p></td>
5931</tr>
5932<tr>
5933<td class="tableblock halign-left valign-middle"><p class="tableblock">[ <strong>points</strong> ]<br>
5934  <strong>line_strip</strong><br></p>
5935<p class="tableblock">  <strong>triangle_strip</strong></p></td>
5936<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5937<td class="tableblock halign-center valign-middle"></td>
5938<td class="tableblock halign-center valign-middle"></td>
5939<td class="tableblock halign-center valign-middle"></td>
5940<td class="tableblock halign-left valign-middle" rowspan="2"><p class="tableblock">geometry <strong>out</strong></p></td>
5941</tr>
5942<tr>
5943<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>max_vertices</strong> =</p></td>
5944<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5945<td class="tableblock halign-center valign-middle"></td>
5946<td class="tableblock halign-center valign-middle"></td>
5947<td class="tableblock halign-center valign-middle"></td>
5948</tr>
5949<tr>
5950<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>constant_id</strong> =</p></td>
5951<td class="tableblock halign-center valign-middle"></td>
5952<td class="tableblock halign-center valign-middle"><p class="tableblock">scalar only</p></td>
5953<td class="tableblock halign-center valign-middle"></td>
5954<td class="tableblock halign-center valign-middle"></td>
5955<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>const</strong> (SPIR-V only)</p></td>
5956</tr>
5957<tr>
5958<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>rgba32f</strong><br>
5959  <strong>rgba16f</strong><br>
5960  <strong>r32f</strong><br>
5961  <strong>rgba8</strong><br>
5962  <strong>rgba8_snorm</strong><br>
5963  <strong>rgba32i</strong><br>
5964  <strong>rgba16i</strong><br>
5965  <strong>rgba8i</strong><br>
5966  <strong>r32i</strong><br>
5967  <strong>rgba32ui</strong><br>
5968  <strong>rgba16ui</strong><br>
5969  <strong>rgba8ui</strong><br>
5970  <strong>r32ui</strong></p></td>
5971<td class="tableblock halign-center valign-middle"></td>
5972<td class="tableblock halign-center valign-middle"><p class="tableblock">image types only</p></td>
5973<td class="tableblock halign-center valign-middle"></td>
5974<td class="tableblock halign-center valign-middle"></td>
5975<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>uniform</strong></p></td>
5976</tr>
5977<tr>
5978<td class="tableblock halign-left valign-middle"><p class="tableblock"><strong>blend_support_multiply</strong><br>
5979  <strong>blend_support_screen</strong><br>
5980  <strong>blend_support_overlay</strong><br>
5981  <strong>blend_support_darken</strong><br>
5982  <strong>blend_support_lighten</strong><br>
5983  <strong>blend_support_colordodge blend_support_colorburn</strong><br>
5984  <strong>blend_support_hardlight</strong><br>
5985  <strong>blend_support_softlight</strong><br>
5986  <strong>blend_support_difference</strong><br>
5987  <strong>blend_support_exclusion</strong><br>
5988  <strong>blend_support_hsl_hue</strong><br>
5989  <strong>blend_support_hsl_saturation</strong><br>
5990  <strong>blend_support_hsl_color blend_support_hsl_luminosity</strong><br>
5991  <strong>blend_support_all_equations</strong></p></td>
5992<td class="tableblock halign-center valign-middle"><p class="tableblock">X</p></td>
5993<td class="tableblock halign-center valign-middle"></td>
5994<td class="tableblock halign-center valign-middle"></td>
5995<td class="tableblock halign-center valign-middle"></td>
5996<td class="tableblock halign-left valign-middle"><p class="tableblock">fragment <strong>out</strong></p></td>
5997</tr>
5998</tbody>
5999</table>
6000<div class="dlist">
6001<dl>
6002<dt class="hdlist1">1</dt>
6003<dd>
6004<p>Location qualifiers are not allowed for members of an arrayed block,
6005except for per-vertex-arrays (see &#8220;<a href="#interface-blocks">Interface
6006Blocks</a>&#8221;).</p>
6007</dd>
6008</dl>
6009</div>
6010<div class="sect3">
6011<h4 id="input-layout-qualifiers">4.4.1. Input Layout Qualifiers</h4>
6012<div class="paragraph">
6013<p>Layout qualifiers specific to a particular shader language are discussed in
6014separate sections below.</p>
6015</div>
6016<div class="paragraph">
6017<p>All shaders except compute shaders allow <strong>location</strong> layout qualifiers on
6018input variable declarations, input block declarations, and input block
6019member declarations.</p>
6020</div>
6021<div class="openblock bnf">
6022<div class="content">
6023<div class="dlist">
6024<dl>
6025<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
6026<dd>
6027<p><strong>location</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
6028</dd>
6029</dl>
6030</div>
6031</div>
6032</div>
6033<div class="paragraph">
6034<p>For example,</p>
6035</div>
6036<div class="listingblock">
6037<div class="content">
6038<pre class="CodeRay highlight"><code data-lang="c++">layout(location = <span class="integer">3</span>) in vec4 normal;</code></pre>
6039</div>
6040</div>
6041<div class="paragraph">
6042<p>will establish that the shader input <em>normal</em> is assigned to vector location
6043number
60443.
6045For vertex shader inputs, the location specifies the number of the
6046vertex attribute from which input values are taken.
6047For inputs of all other shader types, the location specifies a vector number
6048that can be used to match against outputs from a previous shader stage, even
6049if that shader is in a different program object.</p>
6050</div>
6051<div class="paragraph">
6052<p>The following language describes how many locations are consumed by a given
6053type.
6054However, geometry shader inputs, tessellation control shader inputs and
6055outputs, and tessellation evaluation inputs all have an additional level of
6056arrayness relative to other shader inputs and outputs.
6057This outer array level is removed from the type before considering how many
6058locations the type consumes.</p>
6059</div>
6060<div class="paragraph">
6061<p>If a shader input is any scalar or vector type, it will consume a single
6062location.</p>
6063</div>
6064<div class="paragraph">
6065<p>If the declared input (after potentially removing an outer array level as
6066just described above) is an array of size <em>n</em> and each of the elements takes
6067<em>m</em> locations, it will be assigned <em>m</em> * <em>n</em> consecutive locations starting
6068with the location specified.
6069For example,</p>
6070</div>
6071<div class="listingblock">
6072<div class="content">
6073<pre class="CodeRay highlight"><code data-lang="c++">layout(location = <span class="integer">6</span>) in vec4 colors[<span class="integer">3</span>];</code></pre>
6074</div>
6075</div>
6076<div class="paragraph">
6077<p>will establish that the shader input <em>colors</em> is assigned to vector location
6078numbers 6, 7, and 8.</p>
6079</div>
6080<div class="paragraph">
6081<p>If the declared input is an <em>n</em> × <em>m</em>
6082matrix, it will be assigned multiple locations starting with the location
6083specified.
6084The number of locations assigned for each matrix will be the same as for an
6085<em>n</em>-element array of <em>m</em>-component vectors.
6086For example,</p>
6087</div>
6088<div class="listingblock">
6089<div class="content">
6090<pre class="CodeRay highlight"><code data-lang="c++">layout(location = <span class="integer">9</span>) in mat4 transforms[<span class="integer">2</span>];</code></pre>
6091</div>
6092</div>
6093<div class="paragraph">
6094<p>will establish that shader input <em>transforms</em> is assigned to vector
6095locations 9-16, with <em>transforms[0]</em> being assigned to locations 9-12, and
6096<em>transforms[1]</em> being assigned to locations 13-16.</p>
6097</div>
6098<div class="paragraph">
6099<p>If the declared input is a structure or block, its members will be assigned
6100consecutive locations in their order of declaration, with the first member
6101assigned the location provided in the layout qualifier.
6102For a structure, this process applies to the entire structure.
6103It is a compile-time error to use a <strong>location</strong> qualifier on a member of a
6104structure.
6105For a block, this process applies to the entire block, or until the first
6106member is reached that has a <strong>location</strong> layout qualifier.</p>
6107</div>
6108<div class="paragraph">
6109<p>When a block member is declared with a <strong>location</strong> qualifier, its location
6110comes from that qualifier; the member&#8217;s <strong>location</strong> qualifier overrides the
6111block-level declaration.
6112Subsequent members are again assigned consecutive locations, based on the
6113newest location, until the next member declared with a <strong>location</strong> qualifier.
6114The values used for locations do not have to be declared in increasing
6115order.</p>
6116</div>
6117<div class="paragraph">
6118<p>If a block has no block-level <strong>location</strong> layout qualifier, it is required
6119that either all or none of its members have a <strong>location</strong> layout qualifier,
6120or a compile-time error results.</p>
6121</div>
6122<div class="paragraph">
6123<p>If an input is declared as an array of blocks, excluding per-vertex-arrays
6124as required for tessellation, it is an error to declare a member of the
6125block with a <strong>location</strong> qualifier.</p>
6126</div>
6127<div class="paragraph">
6128<p>When generating SPIR-V, all <strong>in</strong> and <strong>out</strong> qualified user-declared (non
6129built-in) variables and blocks (or all their members) must have a
6130shader-specified <strong>location</strong>.
6131Otherwise, a compile-time error is generated.</p>
6132</div>
6133<div class="paragraph">
6134<p>The locations consumed by block and structure members are determined by
6135applying the rules above recursively as though the structure member were
6136declared as an input variable of the same type.
6137For example:</p>
6138</div>
6139<div class="listingblock">
6140<div class="content">
6141<pre class="CodeRay highlight"><code data-lang="c++">layout(location = <span class="integer">3</span>) in <span class="keyword">struct</span> S
6142{
6143    vec3 a;                      <span class="comment">// gets location 3</span>
6144    mat2 b;                      <span class="comment">// gets locations 4 and 5</span>
6145    vec4 c[<span class="integer">2</span>];                   <span class="comment">// gets locations 6 and 7</span>
6146    layout(location = <span class="integer">8</span>) vec2 A; <span class="comment">// ERROR, can't use on struct member</span>
6147} s;
6148layout(location = <span class="integer">4</span>) in block
6149{
6150    vec4 d;                      <span class="comment">// gets location 4</span>
6151    vec4 e;                      <span class="comment">// gets location 5</span>
6152    layout(location = <span class="integer">7</span>) vec4 f; <span class="comment">// gets location 7</span>
6153    vec4 g;                      <span class="comment">// gets location 8</span>
6154    layout(location = <span class="integer">1</span>) vec4 h; <span class="comment">// gets location 1</span>
6155    vec4 i;                      <span class="comment">// gets location 2</span>
6156    vec4 j;                      <span class="comment">// gets location 3</span>
6157    vec4 k;                      <span class="comment">// ERROR, location 4 already used</span>
6158};</code></pre>
6159</div>
6160</div>
6161<div class="paragraph">
6162<p>The number of input locations available to a shader is limited.
6163For vertex shaders, the limit is the advertised number of vertex attributes.
6164For all other shaders, the limit is implementation-dependent and must be no
6165less than one fourth of the advertised maximum input component count.</p>
6166</div>
6167<div class="paragraph">
6168<p>A program will fail to link if any attached shader uses a location greater
6169than or equal to the number of supported locations, unless device-dependent
6170optimizations are able to make the program fit within available hardware
6171resources.</p>
6172</div>
6173<div class="paragraph">
6174<p>A program will fail to link if explicit location assignments leave the
6175linker unable to find space for other variables without explicit
6176assignments.</p>
6177</div>
6178<div class="paragraph">
6179<p>For the purposes of determining if a non-vertex input matches an output from
6180a previous shader stage, the <strong>location</strong> layout qualifier (if any) must
6181match.</p>
6182</div>
6183<div class="paragraph">
6184<p>If a vertex shader input variable with no location assigned in the shader
6185text has a location specified through the OpenGL ES API, the API-assigned
6186location will be used.
6187Otherwise, such variables will be assigned a location by the linker.
6188See section 11.1.1 &#8220;Vertex Attributes&#8221; of the <a href="#references">OpenGL ES Specification</a> for
6189more details.</p>
6190</div>
6191<div class="paragraph">
6192<p>It is an error if more than one input or element of a matrix input is bound
6193to the same location.</p>
6194</div>
6195<div class="sect4">
6196<h5 id="tessellation-evaluation-inputs">Tessellation Evaluation Inputs</h5>
6197<div class="paragraph">
6198<p>Additional input layout qualifier identifiers allowed for tessellation
6199evaluation shaders are described below.</p>
6200</div>
6201<div class="openblock bnf">
6202<div class="content">
6203<div class="dlist">
6204<dl>
6205<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
6206<dd>
6207<p><em>primitive_mode</em><br>
6208<em>vertex_spacing</em><br>
6209<em>ordering</em><br>
6210<em>point_mode</em></p>
6211</dd>
6212</dl>
6213</div>
6214</div>
6215</div>
6216<div class="paragraph">
6217<p>The <strong>primitive-mode</strong> is used to specify a tessellation primitive mode to be
6218used by the tessellation primitive generator.</p>
6219</div>
6220<div class="openblock bnf">
6221<div class="content">
6222<div class="dlist">
6223<dl>
6224<dt class="hdlist1"><em>primitive-mode</em>: </dt>
6225<dd>
6226<p><strong>triangles</strong><br>
6227<strong>quads</strong><br>
6228<strong>isolines</strong></p>
6229</dd>
6230</dl>
6231</div>
6232</div>
6233</div>
6234<div class="paragraph">
6235<p>If present, the <em>primitive-mode</em> specifies that the tessellation primitive
6236generator should subdivide a triangle into smaller triangles, a quad into
6237triangles, or a quad into a collection of lines, respectively.</p>
6238</div>
6239<div class="paragraph">
6240<p>A second group of layout identifiers, <em>vertex spacing</em>, is used to specify
6241the spacing used by the tessellation primitive generator when subdividing an
6242edge.</p>
6243</div>
6244<div class="openblock bnf">
6245<div class="content">
6246<div class="dlist">
6247<dl>
6248<dt class="hdlist1"><em>vertex-spacing</em>: </dt>
6249<dd>
6250<p><strong>equal_spacing</strong><br>
6251<strong>fractional_even_spacing</strong><br>
6252<strong>fractional_odd_spacing</strong></p>
6253</dd>
6254</dl>
6255</div>
6256</div>
6257</div>
6258<div class="paragraph">
6259<p><strong>equal_spacing</strong> specifies that edges should be divided into a collection of
6260equal-sized segments;</p>
6261</div>
6262<div class="paragraph">
6263<p><strong>fractional_even_spacing</strong> specifies that edges should be divided into an
6264even number of equal-length segments plus two additional shorter
6265&#8220;fractional&#8221; segments; or</p>
6266</div>
6267<div class="paragraph">
6268<p><strong>fractional_odd_spacing</strong> specifies that edges should be divided into an odd
6269number of equal-length segments plus two additional shorter &#8220;fractional&#8221;
6270segments.</p>
6271</div>
6272<div class="paragraph">
6273<p>A third group of layout identifiers, <em>ordering</em>, specifies whether the
6274tessellation primitive generator produces triangles in clockwise or
6275counter-clockwise order, according to the coordinate system depicted in the
6276<a href="#references">OpenGL ES Specification</a>.</p>
6277</div>
6278<div class="openblock bnf">
6279<div class="content">
6280<div class="dlist">
6281<dl>
6282<dt class="hdlist1"><em>ordering</em>: </dt>
6283<dd>
6284<p><strong>cw</strong><br>
6285<strong>ccw</strong></p>
6286</dd>
6287</dl>
6288</div>
6289</div>
6290</div>
6291<div class="paragraph">
6292<p>The identifiers <strong>cw</strong> and <strong>ccw</strong> indicate clockwise and counter-clockwise
6293triangles, respectively.
6294If the tessellation primitive generator does not produce triangles, the
6295order is ignored.</p>
6296</div>
6297<div class="paragraph">
6298<p>Finally, <em>point mode</em> indicates that the tessellation primitive generator
6299should produce one point for each distinct vertex in the subdivided
6300primitive, rather than generating lines or triangles.</p>
6301</div>
6302<div class="openblock bnf">
6303<div class="content">
6304<div class="dlist">
6305<dl>
6306<dt class="hdlist1"><em>point-mode</em>: </dt>
6307<dd>
6308<p><strong>point_mode</strong></p>
6309</dd>
6310</dl>
6311</div>
6312</div>
6313</div>
6314<div class="paragraph">
6315<p>Any or all of these identifiers may be specified one or more times in a
6316single input layout declaration.</p>
6317</div>
6318<div class="paragraph">
6319<p>The tessellation evaluation shader object in a program must declare a
6320primitive mode in its input layout.
6321Declaring vertex spacing, ordering, or point mode identifiers is optional.
6322If spacing or vertex ordering declarations are omitted, the tessellation
6323primitive generator will use equal spacing or counter-clockwise vertex
6324ordering, respectively.
6325If a point mode declaration is omitted, the tessellation primitive generator
6326will produce lines or triangles according to the primitive mode.</p>
6327</div>
6328</div>
6329<div class="sect4">
6330<h5 id="geometry-shader-inputs">Geometry Shader Inputs</h5>
6331<div class="paragraph">
6332<p>Additional layout qualifier identifiers for geometry shader inputs include
6333<em>primitive</em> identifiers and an <em>invocation count</em> identifier:</p>
6334</div>
6335<div class="openblock bnf">
6336<div class="content">
6337<div class="dlist">
6338<dl>
6339<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
6340<dd>
6341<p><strong>points</strong><br>
6342<strong>lines</strong><br>
6343<strong>lines_adjacency</strong><br>
6344<strong>triangles</strong><br>
6345<strong>triangles_adjacency</strong><br>
6346<strong>invocations</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
6347</dd>
6348</dl>
6349</div>
6350</div>
6351</div>
6352<div class="paragraph">
6353<p>The identifiers <strong>points</strong>, <strong>lines</strong>, <strong>lines_adjacency</strong>, <strong>triangles</strong>, and
6354<strong>triangles_adjacency</strong> are used to specify the type of input primitive
6355accepted by the geometry shader, and only one of these is accepted.
6356The geometry shader must declare this input primitive layout.</p>
6357</div>
6358<div class="paragraph">
6359<p>The identifier <strong>invocations</strong> is used to specify the number of times the
6360geometry shader executable is invoked for each input primitive received.
6361Invocation count declarations are optional.
6362If no invocation count is declared in the geometry shader, it will be run
6363once for each input primitive.
6364If an invocation count is declared, all such declarations must specify the
6365same count.
6366If a shader specifies an invocation count greater than the
6367implementation-dependent maximum, or less than or equal to zero,
6368a compile-time error results.</p>
6369</div>
6370<div class="paragraph">
6371<p>For example,</p>
6372</div>
6373<div class="listingblock">
6374<div class="content">
6375<pre class="CodeRay highlight"><code data-lang="c++">layout(triangles, invocations = <span class="integer">6</span>) in;</code></pre>
6376</div>
6377</div>
6378<div class="paragraph">
6379<p>will establish that all inputs to the geometry shader are triangles and that
6380the geometry shader executable is run six times for each triangle processed.</p>
6381</div>
6382<div class="paragraph">
6383<p>All geometry shader input unsized array declarations will be sized by an
6384earlier input primitive layout qualifier, when present, as per the following
6385table.</p>
6386</div>
6387<table class="tableblock frame-all grid-all stretch">
6388<colgroup>
6389<col style="width: 50%;">
6390<col style="width: 50%;">
6391</colgroup>
6392<thead>
6393<tr>
6394<th class="tableblock halign-left valign-top">Layout</th>
6395<th class="tableblock halign-left valign-top">Size of Input Arrays</th>
6396</tr>
6397</thead>
6398<tbody>
6399<tr>
6400<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>points</strong></p></td>
6401<td class="tableblock halign-left valign-top"><p class="tableblock">1</p></td>
6402</tr>
6403<tr>
6404<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>lines</strong></p></td>
6405<td class="tableblock halign-left valign-top"><p class="tableblock">2</p></td>
6406</tr>
6407<tr>
6408<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>lines_adjacency</strong></p></td>
6409<td class="tableblock halign-left valign-top"><p class="tableblock">4</p></td>
6410</tr>
6411<tr>
6412<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>triangles</strong></p></td>
6413<td class="tableblock halign-left valign-top"><p class="tableblock">3</p></td>
6414</tr>
6415<tr>
6416<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>triangles_adjacency</strong></p></td>
6417<td class="tableblock halign-left valign-top"><p class="tableblock">6</p></td>
6418</tr>
6419</tbody>
6420</table>
6421<div class="paragraph">
6422<p>The intrinsically declared input array <em>gl_in[]</em> will also be sized by any
6423input primitive-layout declaration.
6424Hence, the expression</p>
6425</div>
6426<div class="listingblock">
6427<div class="content">
6428<pre class="CodeRay highlight"><code data-lang="c++">gl_in.length()</code></pre>
6429</div>
6430</div>
6431<div class="paragraph">
6432<p>will return the value from the table above.</p>
6433</div>
6434<div class="paragraph">
6435<p>An input can be declared without an array size if there is a previous layout
6436which specifies the size.
6437For built-in inputs (e.g. <em>gl_in[]</em>), a layout must be declared before any
6438use.</p>
6439</div>
6440<div class="paragraph">
6441<p>It is a compile-time error if a layout declaration&#8217;s array size (from the
6442table above) does not match all the explicit array sizes specified in
6443declarations of an input variables in the same shader.
6444The following includes examples of compile-time errors:</p>
6445</div>
6446<div class="listingblock">
6447<div class="content">
6448<pre class="CodeRay highlight"><code data-lang="c++"><span class="comment">// code sequence within one shader...</span>
6449in vec4 Color2[<span class="integer">2</span>];    <span class="comment">// legal, size is 2</span>
6450in vec4 Color3[<span class="integer">3</span>];    <span class="comment">// illegal, input sizes are inconsistent</span>
6451layout(lines) in;     <span class="comment">// legal for Color2, input size is 2, matching Color2</span>
6452in vec4 Color4[<span class="integer">3</span>];    <span class="comment">// illegal, contradicts layout of lines</span>
6453layout(lines) in;     <span class="comment">// legal, matches other layout() declaration</span>
6454layout(triangles) in; <span class="comment">// illegal, does not match earlier layout() declaration</span></code></pre>
6455</div>
6456</div>
6457<div class="paragraph">
6458<p>It is a link-time error if not all provided sizes (sized input arrays and
6459layout size) match in the geometry shader of a program.</p>
6460</div>
6461</div>
6462<div class="sect4">
6463<h5 id="fragment-shader-inputs">Fragment Shader Inputs</h5>
6464<div class="paragraph">
6465<p>Fragment shaders allow the following layout qualifier on <strong>in</strong> only (not with
6466variable declarations):</p>
6467</div>
6468<div class="openblock bnf">
6469<div class="content">
6470<div class="dlist">
6471<dl>
6472<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
6473<dd>
6474<p><strong>early_fragment_tests</strong></p>
6475</dd>
6476</dl>
6477</div>
6478</div>
6479</div>
6480<div class="paragraph">
6481<p>to request that fragment tests be performed before fragment shader
6482execution, as described in section 13.8.4 &#8220;Early Fragment Tests&#8221; of the
6483<a href="#references">OpenGL ES Specification</a>.</p>
6484</div>
6485<div class="paragraph">
6486<p>For example,</p>
6487</div>
6488<div class="listingblock">
6489<div class="content">
6490<pre class="CodeRay highlight"><code data-lang="c++">layout(early_fragment_tests) in;</code></pre>
6491</div>
6492</div>
6493<div class="paragraph">
6494<p>Specifying this will make per-fragment tests be performed before fragment
6495shader execution.
6496In addition it is an error to statically write to <em>gl_FragDepth</em> in the
6497fragment shader.
6498If this is not declared, per-fragment tests will be performed after fragment
6499shader execution.</p>
6500</div>
6501</div>
6502<div class="sect4">
6503<h5 id="compute-shader-inputs">Compute Shader Inputs</h5>
6504<div class="paragraph">
6505<p>There are no layout location qualifiers for compute shader inputs.</p>
6506</div>
6507<div class="paragraph">
6508<p>Layout qualifier identifiers for compute shader inputs are the workgroup
6509size qualifiers:</p>
6510</div>
6511<div class="openblock bnf">
6512<div class="content">
6513<div class="dlist">
6514<dl>
6515<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
6516<dd>
6517<p><strong>local_size_x</strong> <strong>=</strong> <em>layout-qualifier-value</em><br>
6518<strong>local_size_y</strong> <strong>=</strong> <em>layout-qualifier-value</em><br>
6519<strong>local_size_z</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
6520</dd>
6521</dl>
6522</div>
6523</div>
6524</div>
6525<div class="paragraph">
6526<p>The <strong>local_size_x</strong>, <strong>local_size_y</strong>, and <strong>local_size_z</strong> qualifiers are used
6527to declare a fixed workgroup size by the compute shader in the first,
6528second, and third dimension, respectively.
6529If a shader does not specify a size for one of the dimensions, that
6530dimension will have a size of 1.</p>
6531</div>
6532<div class="paragraph">
6533<p>For example, the following declaration in a compute shader</p>
6534</div>
6535<div class="listingblock">
6536<div class="content">
6537<pre class="CodeRay highlight"><code data-lang="c++">layout(local_size_x = <span class="integer">32</span>, local_size_y = <span class="integer">32</span>) in;</code></pre>
6538</div>
6539</div>
6540<div class="paragraph">
6541<p>is used to declare a two-dimensional compute shader with a workgroup size of 32
6542X 32 elements, which is equivalent to a three-dimensional compute shader
6543where the third dimension has size one.</p>
6544</div>
6545<div class="paragraph">
6546<p>As another example, the declaration</p>
6547</div>
6548<div class="listingblock">
6549<div class="content">
6550<pre class="CodeRay highlight"><code data-lang="c++">layout(local_size_x = <span class="integer">8</span>) in;</code></pre>
6551</div>
6552</div>
6553<div class="paragraph">
6554<p>effectively specifies that a one-dimensional compute shader is being
6555compiled, and its size is 8 elements.</p>
6556</div>
6557<div class="paragraph">
6558<p>If the fixed workgroup size of the shader in any dimension is less than
6559or equal to zero or greater than the maximum size supported by the
6560implementation for that dimension, a compile-time error results.
6561Also, if such a layout qualifier is declared more than once in the same
6562shader, all those declarations must set the same set of workgroup
6563sizes and set them to the same values; otherwise a compile-time error
6564results.</p>
6565</div>
6566<div class="paragraph">
6567<p>Furthermore, if a program object contains a compute shader, that shader must
6568contain an input layout qualifier specifying a fixed workgroup size for
6569the program, or a link-time error will occur.</p>
6570</div>
6571</div>
6572</div>
6573<div class="sect3">
6574<h4 id="output-layout-qualifiers">4.4.2. Output Layout Qualifiers</h4>
6575<div class="paragraph">
6576<p>Some output layout qualifiers apply to all shader stages and some apply only
6577to specific stages.
6578The latter are discussed in separate sections below.</p>
6579</div>
6580<div class="paragraph">
6581<p>As with input layout qualifiers, all shaders except compute shaders allow
6582<strong>location</strong> layout qualifiers on output variable declarations, output block
6583declarations, and output block member declarations.</p>
6584</div>
6585<div class="openblock bnf">
6586<div class="content">
6587<div class="dlist">
6588<dl>
6589<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
6590<dd>
6591<p><strong>location</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
6592</dd>
6593</dl>
6594</div>
6595</div>
6596</div>
6597<div class="paragraph">
6598<p>The usage and rules for applying the <strong>location</strong> qualifier
6599to blocks and structures are exactly as described in
6600&#8220;<a href="#input-layout-qualifiers">Input Layout Qualifiers</a>&#8221;.</p>
6601</div>
6602<div class="paragraph">
6603<p>The qualifier may appear at most once within a declaration.
6604For example, in a fragment shader,</p>
6605</div>
6606<div class="listingblock">
6607<div class="content">
6608<pre class="CodeRay highlight"><code data-lang="c++">layout(location = <span class="integer">3</span>) out vec4 color;</code></pre>
6609</div>
6610</div>
6611<div class="paragraph">
6612<p>will establish that the fragment shader output <em>color</em> is assigned to</p>
6613</div>
6614<div class="paragraph">
6615<p>fragment color 3.</p>
6616</div>
6617<div class="paragraph">
6618<p>For fragment shader outputs, the location
6619specifies
6620the color output number
6621receiving the values of the output.
6622For outputs of all other shader stages, the location specifies a vector
6623number that can be used to match against inputs in a subsequent shader
6624stage, even if that shader is in a different program object.</p>
6625</div>
6626<div class="paragraph">
6627<p>Declared outputs of scalar or vector type consume a single location.</p>
6628</div>
6629<div class="paragraph">
6630<p>If the declared output is an array, it will be assigned consecutive
6631locations starting with the location specified.
6632For example,</p>
6633</div>
6634<div class="listingblock">
6635<div class="content">
6636<pre class="CodeRay highlight"><code data-lang="c++">layout(location = <span class="integer">2</span>) out vec4 colors[<span class="integer">3</span>];</code></pre>
6637</div>
6638</div>
6639<div class="paragraph">
6640<p>will establish that <em>colors</em> is assigned to vector location numbers 2, 3,
6641and 4.</p>
6642</div>
6643<div class="paragraph">
6644<p>If the declared output is an <em>n</em> × <em>m</em>
6645matrix, it will be assigned multiple locations starting with the location
6646specified.
6647The number of locations assigned will be the same as for an <em>n</em>-element
6648array of <em>m</em>-component vectors.</p>
6649</div>
6650<div class="paragraph">
6651<p>If the declared output is a structure, its members will be assigned
6652consecutive locations in the order of declaration, with the first member
6653assigned the location specified for the structure.
6654The number of locations consumed by a structure member is determined by
6655applying the rules above recursively as though the structure member were
6656declared as an output variable of the same type.</p>
6657</div>
6658<div class="paragraph">
6659<p><strong>location</strong> layout qualifiers may be used on output variables declared as
6660structures.
6661However, it is a compile-time error to use a <strong>location</strong> qualifier on a
6662structure member.
6663Location layout qualifiers may be used on output blocks and output block
6664members.</p>
6665</div>
6666<div class="paragraph">
6667<p>If an output is declared as an array of blocks, excluding per-vertex-arrays
6668as required for tessellation, it is an error to declare a member of the
6669block with a <strong>location</strong> qualifier.</p>
6670</div>
6671<div class="paragraph">
6672<p>The number of output locations available to a shader is limited.
6673For fragment shaders, the limit is the advertised number of draw buffers.</p>
6674</div>
6675<div class="paragraph">
6676<p>For all other shaders, the limit is implementation-dependent and must be no
6677less than one fourth of the advertised maximum output component count
6678(compute shaders have no outputs).
6679A program will fail to link if any attached shader uses a location greater
6680than or equal to the number of supported locations, unless device-dependent
6681optimizations are able to make the program fit within available hardware
6682resources.</p>
6683</div>
6684<div class="paragraph">
6685<p>Compile-time errors may also be given if at compile time it is known the
6686link will fail.
6687A negative output location will result in an error.</p>
6688</div>
6689<div class="paragraph">
6690<p>It is a compile-time or link-time error if any of the following occur:</p>
6691</div>
6692<div class="ulist">
6693<ul>
6694<li>
6695<p>any two fragment shader output variables are assigned to the same
6696location.</p>
6697</li>
6698<li>
6699<p>if any two output variables from the same vertex, tessellation or
6700geometry shader stage are assigned to the same location.</p>
6701</li>
6702</ul>
6703</div>
6704<div class="paragraph">
6705<p>For all shader types, a program will fail to link if explicit location
6706assignments leave the linker unable to find space for other variables
6707without explicit assignments.</p>
6708</div>
6709<div class="paragraph">
6710<p>If an output variable has no location assigned in the shader text, it will
6711be assigned a location by the linker.
6712See section 11.1.3 &#8220;Shader Execution&#8221; of the <a href="#references">OpenGL ES Specification</a> for
6713more details.</p>
6714</div>
6715<div class="paragraph">
6716<p>If there is only a single output, the location does not need to be
6717specified, in which case it defaults to zero.
6718This applies for all output types, including arrays.
6719If there is more than one fragment output, the location must be specified
6720for all outputs.</p>
6721</div>
6722<div class="paragraph">
6723<p>For the purposes of determining if a non-fragment output matches an input
6724from a subsequent shader stage, the <strong>location</strong> layout qualifier (if any)
6725must match.</p>
6726</div>
6727<div class="sect4">
6728<h5 id="tessellation-control-outputs">Tessellation Control Outputs</h5>
6729<div class="paragraph">
6730<p>Tessellation control shaders allow output layout qualifiers only on the
6731interface
6732qualifier <strong>out</strong>, not on an output block, block member, or variable
6733declaration.
6734The output layout qualifier identifiers allowed for tessellation control
6735shaders are:</p>
6736</div>
6737<div class="openblock bnf">
6738<div class="content">
6739<div class="dlist">
6740<dl>
6741<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
6742<dd>
6743<p><strong>vertices</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
6744</dd>
6745</dl>
6746</div>
6747</div>
6748</div>
6749<div class="paragraph">
6750<p>The identifier <strong>vertices</strong> specifies the number of vertices in the output
6751patch produced by the tessellation control shader, which also specifies the
6752number of times the tessellation control shader is invoked.
6753It is a compile- or link-time error for the output vertex count to be less
6754than or equal to zero, or greater than the implementation-dependent maximum
6755patch size.</p>
6756</div>
6757<div class="paragraph">
6758<p>The intrinsically declared tessellation control output array <em>gl_out[]</em> will
6759also be sized by any output layout declaration.
6760Hence, the expression</p>
6761</div>
6762<div class="listingblock">
6763<div class="content">
6764<pre class="CodeRay highlight"><code data-lang="c++">gl_out.length()</code></pre>
6765</div>
6766</div>
6767<div class="paragraph">
6768<p>will return the output patch vertex count specified in a previous output
6769layout qualifier.
6770For outputs declared without an array size, including intrinsically declared
6771outputs (i.e., <em>gl_out</em>), a layout must be declared before any use of the
6772method <strong>length</strong>() or other array use that requires its size to be known.</p>
6773</div>
6774<div class="paragraph">
6775<p>It is a compile-time error if the output patch vertex count specified in an
6776output layout qualifier does not match the array size specified in any
6777output variable declaration in the same shader.</p>
6778</div>
6779<div class="paragraph">
6780<p>All tessellation control shader layout declarations in a program must
6781specify the same output patch vertex count.
6782There must be at least one layout qualifier specifying an output patch
6783vertex count in any program containing a tessellation control shader.</p>
6784</div>
6785</div>
6786<div class="sect4">
6787<h5 id="geometry-outputs">Geometry Outputs</h5>
6788<div class="paragraph">
6789<p>Geometry shaders can have two additional types of output layout identifiers:
6790an output primitive type and a maximum output vertex count.
6791The primitive type and vertex count identifiers are allowed only on the
6792interface qualifier <strong>out</strong>, not on an output block, block member, or variable
6793declaration.</p>
6794</div>
6795<div class="paragraph">
6796<p>The layout qualifier identifiers for geometry shader outputs are</p>
6797</div>
6798<div class="openblock bnf">
6799<div class="content">
6800<div class="dlist">
6801<dl>
6802<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
6803<dd>
6804<p><strong>points</strong><br>
6805<strong>line_strip</strong><br>
6806<strong>triangle_strip</strong><br>
6807<strong>max_vertices</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
6808</dd>
6809</dl>
6810</div>
6811</div>
6812</div>
6813<div class="paragraph">
6814<p>The primitive type identifiers <strong>points</strong>, <strong>line_strip</strong>, and <strong>triangle_strip</strong>
6815are used to specify the type of output primitive produced by the geometry
6816shader, and only one of these is accepted.
6817The geometry shader object in a program must declare an output primitive
6818type, and all geometry shader output primitive type declarations in a
6819program must declare the same primitive type.</p>
6820</div>
6821<div class="paragraph">
6822<p>The vertex count identifier <strong>max_vertices</strong> is used to specify the maximum
6823number of vertices the shader will ever emit in a single invocation.
6824The geometry shader object in a program must declare a maximum output vertex
6825count, and all geometry shader output vertex count declarations in a program
6826must declare the same count.</p>
6827</div>
6828<div class="paragraph">
6829<p>In this example,</p>
6830</div>
6831<div class="listingblock">
6832<div class="content">
6833<pre class="CodeRay highlight"><code data-lang="c++">layout(triangle_strip, max_vertices = <span class="integer">60</span>) out; <span class="comment">// order does not matter</span>
6834layout(max_vertices = <span class="integer">60</span>) out;  <span class="comment">// redeclaration okay</span>
6835layout(triangle_strip) out;     <span class="comment">// redeclaration okay</span>
6836layout(points) out;             <span class="comment">// error, contradicts triangle_strip</span>
6837layout(max_vertices = <span class="integer">30</span>) out;  <span class="comment">// error, contradicts 60</span></code></pre>
6838</div>
6839</div>
6840<div class="paragraph">
6841<p>all outputs from the geometry shader are triangles and at most 60 vertices
6842will be emitted by the shader.
6843It is an error for the maximum number of vertices to be greater than
6844<em>gl_MaxGeometryOutputVertices</em>.</p>
6845</div>
6846<div class="paragraph">
6847<p>All geometry shader output layout declarations in a program must declare the
6848same layout and same value for <strong>max_vertices</strong>.
6849If geometry shaders are in a program, there must be at least one geometry
6850output layout declaration somewhere in that
6851program.</p>
6852</div>
6853</div>
6854<div class="sect4">
6855<h5 id="fragment-outputs">Fragment Outputs</h5>
6856<div class="paragraph">
6857<p>Fragment shaders can have an output layout for redeclaring the built-in
6858variable <em>gl_FragDepth</em>:</p>
6859</div>
6860<div class="listingblock">
6861<div class="content">
6862<pre class="CodeRay highlight"><code data-lang="c++"><span class="comment">// redeclaration that changes nothing is allowed +</span>
6863out <span class="predefined-type">float</span> gl_FragDepth;</code></pre>
6864</div>
6865</div>
6866<div class="paragraph">
6867<p>The built-in <em>gl_FragDepth</em> is only predeclared in fragment shaders, so
6868redeclaring it in any other shader language results in an error.</p>
6869</div>
6870<div class="paragraph">
6871<p>Fragment shaders additionally support layout qualifiers specifying a set of
6872advanced blend equations supported when the fragment shader is used.
6873These layout qualifiers are only permitted on the interface qualifier <strong>out</strong>,
6874and use the identifiers specified in the &#8220;Layout Qualifier&#8221; column of the
6875table below.</p>
6876</div>
6877<div class="paragraph">
6878<p>If a layout qualifier in the table below is specified in the fragment
6879shader, the fragment shader may be used with the corresponding advanced
6880blend equation in the &#8220;Blend Equation(s) Supported&#8221; column.
6881Additionally, the special qualifier <strong>blend_support_all_equations</strong> indicates
6882that the shader may be used with any advanced blending equation supported by
6883the <a href="#references">OpenGL ES Specification</a>.
6884It is not an error to specify more than one of these identifiers in any
6885fragment shader.
6886Specifying more than one qualifier or <strong>blend_support_all_equations</strong> means
6887that the fragment shader may be used with multiple advanced blend equations.
6888Additionally, it is not an error to specify any single one of these layout
6889qualifiers more than once.</p>
6890</div>
6891<table class="tableblock frame-all grid-all stretch">
6892<colgroup>
6893<col style="width: 50%;">
6894<col style="width: 50%;">
6895</colgroup>
6896<tbody>
6897<tr>
6898<td class="tableblock halign-left valign-top"><p class="tableblock">Layout Qualifier</p></td>
6899<td class="tableblock halign-left valign-top"><p class="tableblock">Blend Equations(s) Supported</p></td>
6900</tr>
6901<tr>
6902<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_multiply</strong></p></td>
6903<td class="tableblock halign-left valign-top"><p class="tableblock">MULTIPLY</p></td>
6904</tr>
6905<tr>
6906<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_screen</strong></p></td>
6907<td class="tableblock halign-left valign-top"><p class="tableblock">SCREEN</p></td>
6908</tr>
6909<tr>
6910<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_overlay</strong></p></td>
6911<td class="tableblock halign-left valign-top"><p class="tableblock">OVERLAY</p></td>
6912</tr>
6913<tr>
6914<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_darken</strong></p></td>
6915<td class="tableblock halign-left valign-top"><p class="tableblock">DARKEN</p></td>
6916</tr>
6917<tr>
6918<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_lighten</strong></p></td>
6919<td class="tableblock halign-left valign-top"><p class="tableblock">LIGHTEN</p></td>
6920</tr>
6921<tr>
6922<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_colordodge</strong></p></td>
6923<td class="tableblock halign-left valign-top"><p class="tableblock">COLORDODGE</p></td>
6924</tr>
6925<tr>
6926<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_colorburn</strong></p></td>
6927<td class="tableblock halign-left valign-top"><p class="tableblock">COLORBURN</p></td>
6928</tr>
6929<tr>
6930<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_hardlight</strong></p></td>
6931<td class="tableblock halign-left valign-top"><p class="tableblock">HARDLIGHT</p></td>
6932</tr>
6933<tr>
6934<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_softlight</strong></p></td>
6935<td class="tableblock halign-left valign-top"><p class="tableblock">SOFTLIGHT</p></td>
6936</tr>
6937<tr>
6938<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_difference</strong></p></td>
6939<td class="tableblock halign-left valign-top"><p class="tableblock">DIFFERENCE</p></td>
6940</tr>
6941<tr>
6942<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_exclusion</strong></p></td>
6943<td class="tableblock halign-left valign-top"><p class="tableblock">EXCLUSION</p></td>
6944</tr>
6945<tr>
6946<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_hsl_hue</strong></p></td>
6947<td class="tableblock halign-left valign-top"><p class="tableblock">HSL_HUE</p></td>
6948</tr>
6949<tr>
6950<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_hsl_saturation</strong></p></td>
6951<td class="tableblock halign-left valign-top"><p class="tableblock">HSL_SATURATION</p></td>
6952</tr>
6953<tr>
6954<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_hsl_color</strong></p></td>
6955<td class="tableblock halign-left valign-top"><p class="tableblock">HSL_COLOR</p></td>
6956</tr>
6957<tr>
6958<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_hsl_luminosity</strong></p></td>
6959<td class="tableblock halign-left valign-top"><p class="tableblock">HSL_LUMINOSITY</p></td>
6960</tr>
6961<tr>
6962<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>blend_support_all_equations</strong></p></td>
6963<td class="tableblock halign-left valign-top"><p class="tableblock">all blend equations</p></td>
6964</tr>
6965</tbody>
6966</table>
6967</div>
6968</div>
6969<div class="sect3">
6970<h4 id="uniform-variable-layout-qualifiers">4.4.3. Uniform Variable Layout Qualifiers</h4>
6971<div class="paragraph">
6972<p>The following layout qualifier can be used for all default-block uniform
6973variables but not for variables in uniform or shader storage blocks.
6974The layout qualifier identifier for uniform variables is:</p>
6975</div>
6976<div class="openblock bnf">
6977<div class="content">
6978<div class="dlist">
6979<dl>
6980<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
6981<dd>
6982<p><strong>location</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
6983</dd>
6984</dl>
6985</div>
6986</div>
6987</div>
6988<div class="paragraph">
6989<p>The location specifies the location by which the API can reference
6990the uniform and update its value.
6991Individual elements of a uniform array are assigned consecutive locations
6992with the first element taking location <strong>location</strong>.
6993Default-block uniform variable declarations sharing the same location
6994linked in the program have to match by name, type, qualifiers and arrayness.
6995For arrays their array dimensionality and array sizes must match.
6996For structs this rule applies recursively to all members.
6997Valid locations for default-block uniform variable locations are in the
6998range of 0 to the implementation-defined maximum number of uniform locations
6999minus one.</p>
7000</div>
7001<div class="paragraph">
7002<p>Locations can be assigned to default-block uniform arrays and structures.
7003The first inner-most scalar, vector or matrix member or element takes the
7004specified <strong>location</strong> and the compiler assigns the next inner-most member or
7005element the next incremental location value.
7006Each subsequent inner-most member or element gets incremental locations for
7007the entire structure or array.
7008This rule applies to nested structures and arrays and gives each inner-most
7009scalar, vector, or matrix member a unique location.
7010When the linker generates locations for uniforms without an explicit
7011location, it assumes for all uniforms with an explicit location all their
7012array elements and structure members are used and the linker will not
7013generate a conflicting location, even if that element or member is deemed
7014unused.</p>
7015</div>
7016<div class="paragraph">
7017<p>When targeting Vulkan, the <strong>push_constant</strong> qualifier is used to
7018declare an entire block, and represents a set of <em>push constants</em>, as defined
7019by the Vulkan API.
7020It is an error to apply this to anything other than a uniform block
7021declaration, or when not targeting Vulkan.
7022The values in the block will be initialized as per the Vulkan API specification.
7023A block declared with <code>layout(push_constant)</code> may optionally include an
7024<em>instance-name</em>.
7025There can be only one <strong>push_constant</strong> block per stage, or a compile-time or
7026link-time error will result.
7027A push-constant array can only be indexed with dynamically uniform indices.
7028Uniform blocks declared with <strong>push_constant</strong> use different resources
7029than those without; and are accounted for separately.</p>
7030</div>
7031</div>
7032<div class="sect3">
7033<h4 id="uniform-and-shader-storage-block-layout-qualifiers">4.4.4. Uniform and Shader Storage Block Layout Qualifiers</h4>
7034<div class="paragraph">
7035<p>Layout qualifiers can be used for uniform and shader storage blocks, but not
7036for non-block uniform declarations.
7037The layout qualifier identifiers (and <strong>shared</strong> keyword) for uniform and
7038shader storage blocks are:</p>
7039</div>
7040<div class="openblock bnf">
7041<div class="content">
7042<div class="dlist">
7043<dl>
7044<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
7045<dd>
7046<p><strong>shared</strong><br>
7047<strong>packed</strong><br>
7048<strong>std140</strong><br>
7049<strong>std430</strong><br>
7050<strong>row_major</strong><br>
7051<strong>column_major</strong><br>
7052<strong>binding</strong> <strong>=</strong> <em>layout-qualifier-value</em><br>
7053<strong>offset</strong> <strong>=</strong> <em>layout-qualifier-value</em>   // Vulkan only<br>
7054<strong>align</strong> <strong>=</strong> <em>layout-qualifier-value</em>    // Vulkan only</p>
7055</dd>
7056</dl>
7057</div>
7058</div>
7059</div>
7060<div class="paragraph">
7061<p>None of these have any semantic effect at all on the usage of the variables
7062being declared; they only describe how data is laid out in memory.
7063For example, matrix semantics are always column-based, as described in the
7064rest of this specification, no matter what layout qualifiers are being used.</p>
7065</div>
7066<div class="paragraph">
7067<p>Uniform and shader storage block layout qualifiers can be declared for
7068global scope, on a single uniform or shader storage block, or on a single
7069block member declaration.</p>
7070</div>
7071<div class="paragraph">
7072<p>Default layouts are established at global scope for uniform blocks as:</p>
7073</div>
7074<div class="listingblock">
7075<div class="content">
7076<pre class="CodeRay highlight"><code data-lang="c++">layout(layout-qualifier-id-list) uniform;</code></pre>
7077</div>
7078</div>
7079<div class="paragraph">
7080<p>and for shader storage blocks as:</p>
7081</div>
7082<div class="listingblock">
7083<div class="content">
7084<pre class="CodeRay highlight"><code data-lang="c++">layout(layout-qualifier-id-list) buffer;</code></pre>
7085</div>
7086</div>
7087<div class="paragraph">
7088<p>When this is done, the previous default qualification is first inherited and
7089then overridden as per the override rules listed below for each qualifier
7090listed in the declaration.
7091The result becomes the new default qualification scoped to subsequent
7092uniform or shader storage block definitions.</p>
7093</div>
7094<div class="paragraph">
7095<p>The initial state of compilation when generating SPIR-V is as if the
7096following were declared:</p>
7097</div>
7098<div class="listingblock">
7099<div class="content">
7100<pre class="CodeRay highlight"><code data-lang="c++">layout(std140, column_major) uniform;
7101layout(std430, column_major) buffer;</code></pre>
7102</div>
7103</div>
7104<div class="paragraph">
7105<p>However, when <strong>push_constant</strong> is declared, the default layout of the
7106buffer will be <strong>std430</strong>. There is no method to globally set this default.</p>
7107</div>
7108<div class="paragraph">
7109<p>The initial state of compilation when not generating SPIR-V is as if the
7110following were declared:</p>
7111</div>
7112<div class="listingblock">
7113<div class="content">
7114<pre class="CodeRay highlight"><code data-lang="c++">layout(shared, column_major) uniform;
7115layout(shared, column_major) buffer;</code></pre>
7116</div>
7117</div>
7118<div class="paragraph">
7119<p>Uniform and shader storage blocks can be declared with optional layout
7120qualifiers, and so can their individual member declarations.
7121Such block layout qualification is scoped only to the content of the block.
7122As with global layout declarations, block layout qualification first
7123inherits from the current default qualification and then overrides it.
7124Similarly, individual member layout qualification is scoped just to the
7125member declaration, and inherits from and overrides the block&#8217;s
7126qualification.</p>
7127</div>
7128<div class="paragraph">
7129<p>The <strong>shared</strong> qualifier overrides only the <strong>std140</strong>, <strong>std430</strong>, and <strong>packed</strong>
7130qualifiers; other qualifiers are inherited.
7131The compiler/linker will ensure that multiple programs and programmable
7132stages containing this definition will share the same memory layout for this
7133block, as long as they also matched in their <strong>row_major</strong> and/or
7134<strong>column_major</strong> qualifications.
7135This allows use of the same buffer to back the same block definition across
7136different programs.
7137It is a compile-time error to use the <strong>shared</strong> qualifier when generating
7138SPIR-V.</p>
7139</div>
7140<div class="paragraph">
7141<p>The <strong>packed</strong> qualifier overrides only <strong>std140</strong>, <strong>std430</strong>, and <strong>shared</strong>;
7142other qualifiers are inherited.
7143When <strong>packed</strong> is used, no shareable layout is guaranteed.
7144The compiler and linker can optimize memory use based on what variables
7145actively get used and on other criteria.
7146Offsets must be queried, as there is no other way of guaranteeing where (and
7147which) variables reside within the block.</p>
7148</div>
7149<div class="paragraph">
7150<p>It is a link-time error to access the same packed uniform or shader storage
7151block in multiple stages within a program.
7152Attempts to access the same packed uniform or shader storage block across
7153programs can result in conflicting member offsets and in undefined values
7154being read.
7155However, implementations may aid application management of packed blocks by
7156using canonical layouts for packed blocks.
7157It is a compile-time error to use the <strong>packed</strong> qualifier when generating
7158SPIR-V.</p>
7159</div>
7160<div class="paragraph">
7161<p>The <strong>std140</strong> and <strong>std430</strong> qualifiers override only the <strong>packed</strong>, <strong>shared</strong>,
7162<strong>std140</strong>, and <strong>std430</strong> qualifiers; other qualifiers are inherited.
7163The <strong>std430</strong> qualifier is supported only for shader storage blocks; a shader
7164using the <strong>std430</strong> qualifier on a uniform block will result in
7165an error, unless it is also declared with <strong>push_constant</strong>.</p>
7166</div>
7167<div class="paragraph">
7168<p>The layout is explicitly determined by this, as described in section 7.6.2.2
7169&#8220;Standard Uniform Block Layout&#8221; of the <a href="#references">OpenGL ES Specification</a>.
7170Hence, as in <strong>shared</strong> above, the resulting layout is shareable across
7171programs.</p>
7172</div>
7173<div class="paragraph">
7174<p>Layout qualifiers on member declarations cannot use the <strong>shared</strong>, <strong>packed</strong>,
7175<strong>std140</strong>, or <strong>std430</strong> qualifiers.
7176These can only be used at global scope (without an object) or on a block
7177declaration, or an error results.</p>
7178</div>
7179<div class="paragraph">
7180<p>The <strong>row_major</strong> and <strong>column_major</strong> qualifiers only affect the layout of
7181matrices, including all matrices contained in structures and arrays they are
7182applied to, to all depths of nesting.
7183These qualifiers can be applied to other types, but will have no effect.</p>
7184</div>
7185<div class="paragraph">
7186<p>The <strong>row_major</strong> qualifier overrides only the <strong>column_major</strong> qualifier; other
7187qualifiers are inherited.
7188Elements within a matrix row will be contiguous in memory.</p>
7189</div>
7190<div class="paragraph">
7191<p>The <strong>column_major</strong> qualifier overrides only the <strong>row_major</strong> qualifier; other
7192qualifiers are inherited.
7193Elements within a matrix column will be contiguous in memory.</p>
7194</div>
7195<div class="paragraph">
7196<p>The <strong>binding</strong> qualifier specifies the uniform buffer binding point
7197corresponding to the uniform or shader storage block, which will be used to
7198obtain the values of the member variables of the block.
7199It is a compile-time error to specify the <strong>binding</strong> qualifier for the global
7200scope or for block member declarations.
7201Any uniform or shader storage block declared without a <strong>binding</strong> qualifier
7202is initially assigned to block binding point zero.
7203After a program is linked, the binding points used for uniform
7204(but not shader storage) blocks
7205declared with or without a <strong>binding</strong> qualifier can be updated
7206by the API.</p>
7207</div>
7208<div class="paragraph">
7209<p>When used with OpenGL ES,
7210if the <strong>binding</strong> qualifier is used with a uniform block or shader storage
7211block instanced as an array, the first element of the array takes the
7212specified block binding and each subsequent element takes the next
7213consecutive binding point.
7214For an array of arrays, each element (e.g. 6 elements for a[2][3]) gets a
7215binding point, and they are ordered per the array of array ordering
7216described in &#8220;<a href="#arrays">Arrays.</a>&#8221;</p>
7217</div>
7218<div class="paragraph">
7219<p>When targeting Vulkan,
7220if the <strong>binding</strong> qualifier is used with a uniform block or buffer block
7221instanced as an array, the entire array takes only the provided binding
7222number.
7223The next consecutive binding number is available for a different
7224object.
7225For an array of arrays, descriptor set array element numbers used
7226in descriptor set accesses are ordered per the array-of-array ordering
7227described in &#8220;<a href="#arrays">Arrays.</a>&#8221;</p>
7228</div>
7229<div class="paragraph">
7230<p>If the binding point for any uniform or shader storage block instance is
7231less than zero, or greater than or equal to the corresponding
7232implementation-dependent maximum number of buffer bindings, a compile-time
7233error will occur.
7234When the <strong>binding</strong> qualifier is used with a uniform or shader storage block
7235instanced as an array of size <em>N</em>, all elements of the array from <strong>binding</strong>
7236through <em>binding + N - 1</em> must be within this range.</p>
7237</div>
7238<div class="paragraph">
7239<p>The <strong>set</strong> qualifier is only available when targeting Vulkan.
7240It specifies the descriptor set this object belongs to.
7241It is an error to apply <strong>set</strong> to a standalone qualifier, to
7242a member of a block, or when not targeting an API that supports descriptor sets.
7243It is an error to apply <strong>set</strong> to a block qualified as <strong>push_constant</strong>.
7244By default, any non-push-constant uniform or shader storage block declared
7245without a <strong>set</strong> identifier is assigned to descriptor set 0.
7246Similarly, any sampler, texture, or subpass-input type declared as a uniform
7247without a <strong>set</strong> identifier is also assigned to descriptor set 0.</p>
7248</div>
7249<div class="paragraph">
7250<p>If applied to an object declared as an array, all elements of the array
7251belong to the specified <strong>set</strong>.</p>
7252</div>
7253<div class="paragraph">
7254<p>When generating SPIR-V, it is an error for either the <strong>set</strong> or
7255<strong>binding</strong> value to exceed a front-end-configuration supplied maximum value.</p>
7256</div>
7257<div class="paragraph">
7258<p>When multiple arguments are listed in a <strong>layout</strong> declaration, the effect
7259will be the same as if they were declared one at a time, in order from left
7260to right, each in turn inheriting from and overriding the result from the
7261previous qualification.</p>
7262</div>
7263<div class="paragraph">
7264<p>For example</p>
7265</div>
7266<div class="listingblock">
7267<div class="content">
7268<pre class="CodeRay highlight"><code data-lang="c++">layout(row_major, column_major)</code></pre>
7269</div>
7270</div>
7271<div class="paragraph">
7272<p>results in the qualification being <strong>column_major</strong>.
7273Other examples:</p>
7274</div>
7275<div class="listingblock">
7276<div class="content">
7277<pre class="CodeRay highlight"><code data-lang="c++">layout(shared, row_major) uniform; <span class="comment">// default is now shared and row_major</span>
7278
7279layout(std140) uniform Transform { <span class="comment">// layout of this block is std140</span>
7280    mat4 M1;                       <span class="comment">// row major</span>
7281    layout(column_major) mat4 M2;  <span class="comment">// column major</span>
7282    mat3 N1;                       <span class="comment">// row major</span>
7283};
7284
7285uniform T2 {                       <span class="comment">// layout of this block is shared</span>
7286    ...
7287};
7288
7289layout(column_major) uniform T3 {  <span class="comment">// shared and column major</span>
7290    mat4 M3;                       <span class="comment">// column major</span>
7291    layout(row_major) mat4 m4;     <span class="comment">// row major</span>
7292    mat3 N2;                       <span class="comment">// column major</span>
7293};</code></pre>
7294</div>
7295</div>
7296<div class="paragraph">
7297<p>When targeting Vulkan, the <strong>offset</strong> and <strong>align</strong> qualifiers for blocks and
7298block members can only be used with <strong>uniform</strong> and <strong>buffer</strong> blocks.
7299When not targeting Vulkan, they cannot be used with blocks or block members.</p>
7300</div>
7301<div class="paragraph">
7302<p>The <strong>offset</strong> qualifier can only be used on block members.
7303The <strong>offset</strong> qualifier forces the qualified member to start at or after the
7304specified <em>layout-qualifier-value</em>, which will be its byte offset from
7305the beginning of the buffer.
7306It is a compile-time error to have any offset, explicit or assigned, that
7307lies within another member of the block.
7308Two blocks linked together in the same program with the same block name must
7309have the exact same set of members qualified with <strong>offset</strong> and their
7310<em>layout-qualifier-value</em> values must be the same, or a link-time error
7311results.
7312The specified offset must be a multiple of the base alignment of the type of
7313the block member it qualifies, or a compile-time error results.</p>
7314</div>
7315<div class="paragraph">
7316<p>The <strong>align</strong> qualifier makes the start of each block member have a minimum
7317byte alignment.
7318It does not affect the internal layout within each member, which will still
7319follow the <strong>std140</strong> or <strong>std430</strong> rules.
7320The specified alignment must be greater than 0 and a power of 2, or a
7321compile-time error results.</p>
7322</div>
7323<div class="paragraph">
7324<p>The <em>actual alignment</em> of a member will be the greater of the specified
7325<strong>align</strong> alignment and the standard (e.g. <strong>std140</strong>) base alignment for the
7326member&#8217;s type.
7327The <em>actual offset</em> of a member is computed as follows: If <strong>offset</strong> was
7328declared, start with that offset, otherwise start with the offset immediately
7329following the preceding member (in declaration order).
7330If the resulting offset is not a multiple of the <em>actual alignment</em>,
7331increase it to the first offset that is a multiple of the <em>actual
7332alignment</em>.
7333This results in the <em>actual offset</em> the member will have.</p>
7334</div>
7335<div class="paragraph">
7336<p>When <strong>align</strong> is applied to an array, it affects only the start of the array,
7337not the array&#8217;s internal stride.
7338Both an <strong>offset</strong> and an <strong>align</strong> qualifier can be specified on a declaration.</p>
7339</div>
7340<div class="paragraph">
7341<p>The <strong>align</strong> qualifier, when used on a block, has the same effect as
7342qualifying each member with the same <strong>align</strong> value as declared on the block,
7343and gets the same compile-time results and errors as if this had been done.
7344As described in general earlier, an individual member can specify its own
7345<strong>align</strong>, which overrides the block-level <strong>align</strong>, but just for that member.</p>
7346</div>
7347<div class="paragraph">
7348<p>Examples:</p>
7349</div>
7350<div class="listingblock">
7351<div class="content">
7352<pre class="CodeRay highlight"><code data-lang="c++">layout(std140) uniform block {
7353 vec4 a;                         <span class="comment">// a takes offsets 0-15</span>
7354 layout(offset = <span class="integer">32</span>) vec3 b;     <span class="comment">// b takes offsets 32-43</span>
7355 layout(offset = <span class="integer">40</span>) vec2 c;     <span class="comment">// ERROR, lies within previous member</span>
7356 layout(offset = <span class="integer">48</span>) vec2 d;     <span class="comment">// d takes offsets 48-55</span>
7357 layout(align = <span class="integer">16</span>) <span class="predefined-type">float</span> e;     <span class="comment">// e takes offsets 64-67</span>
7358 layout(align = <span class="integer">2</span>) <span class="predefined-type">double</span> f;     <span class="comment">// f takes offsets 72-79</span>
7359 layout(align = <span class="integer">6</span>) <span class="predefined-type">double</span> g;     <span class="comment">// ERROR, 6 is not a power of 2</span>
7360 layout(offset = <span class="integer">80</span>) <span class="predefined-type">float</span> h;    <span class="comment">// h takes offsets 80-83</span>
7361 layout(align = <span class="integer">64</span>) dvec3 i;     <span class="comment">// i takes offsets 128-151</span>
7362 layout(offset = <span class="integer">164</span>, align = <span class="integer">8</span>)
7363 <span class="predefined-type">float</span> j;                        <span class="comment">// j takes offsets 168-171</span>
7364};</code></pre>
7365</div>
7366</div>
7367</div>
7368<div class="sect3">
7369<h4 id="opaque-uniform-layout-qualifiers">4.4.5. Opaque Uniform Layout Qualifiers</h4>
7370<div class="paragraph">
7371<p>Opaque uniform variables can take the uniform layout qualifier for binding:</p>
7372</div>
7373<div class="openblock bnf">
7374<div class="content">
7375<div class="dlist">
7376<dl>
7377<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
7378<dd>
7379<p><strong>binding</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
7380</dd>
7381</dl>
7382</div>
7383</div>
7384</div>
7385<div class="paragraph">
7386<p>The <strong>binding</strong> qualifier specifies the point where the variable will be bound.
7387Any opaque variable declared without a binding qualifier has a default binding
7388of zero.</p>
7389</div>
7390<div class="paragraph">
7391<p>When used with OpenGL ES,
7392if the <strong>binding</strong> qualifier is used with an array, the first element of the
7393array takes the specified binding point and each subsequent element takes the
7394next consecutive binding point.
7395For an array of arrays, each element (e.g. 6 elements for a[2][3]) gets a
7396binding point, and they are ordered per the array of array ordering
7397described in &#8220;<a href="#arrays">Arrays.</a>&#8221;</p>
7398</div>
7399<div class="paragraph">
7400<p>When targeting Vulkan,
7401if the <strong>binding</strong> qualifier is used with an array, the entire array
7402takes only the provided binding number. The next consecutive binding
7403number is available for a different object.</p>
7404</div>
7405<div class="paragraph">
7406<p>If the <strong>binding</strong> is less than zero, or greater than or equal to the
7407implementation-dependent maximum supported number of binding points,
7408a compile-time error will occur.
7409When the <strong>binding</strong> qualifier is used with an array of size <em>N</em>, all elements
7410of the array from <strong>binding</strong> through <em>binding + N - 1</em> must be within this
7411range.</p>
7412</div>
7413<div class="paragraph">
7414<p>A link-time error will result if two shaders in a program specify different
7415<em>layout-qualifier-value</em> bindings for the same opaque-uniform name.
7416However, it is not an error to specify a binding on some but not all
7417declarations for the same name, as shown in the examples below.</p>
7418</div>
7419<div class="listingblock">
7420<div class="content">
7421<pre class="CodeRay highlight"><code data-lang="c++"><span class="comment">// in one shader...</span>
7422layout(binding=<span class="integer">3</span>) uniform sampler2D s; <span class="comment">// s bound to point 3</span>
7423
7424<span class="comment">// in another shader...</span>
7425uniform sampler2D s;                   <span class="comment">// okay, s still bound at 3</span>
7426
7427<span class="comment">// in another shader...</span>
7428layout(binding=<span class="integer">4</span>) uniform sampler2D s; <span class="comment">// ERROR: contradictory bindings</span></code></pre>
7429</div>
7430</div>
7431</div>
7432<div class="sect3">
7433<h4 id="atomic-counter-layout-qualifiers">4.4.6. Atomic Counter Layout Qualifiers</h4>
7434<div class="paragraph">
7435<p>Atomic counters are not available when targeting Vulkan.</p>
7436</div>
7437<div class="paragraph">
7438<p>Atomic counter layout qualifiers can be used on atomic counter declarations.
7439The atomic counter qualifiers are:</p>
7440</div>
7441<div class="openblock bnf">
7442<div class="content">
7443<div class="dlist">
7444<dl>
7445<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
7446<dd>
7447<p><strong>binding</strong> <strong>=</strong> <em>layout-qualifier-value</em><br>
7448<strong>offset</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
7449</dd>
7450</dl>
7451</div>
7452</div>
7453</div>
7454<div class="paragraph">
7455<p>For example,</p>
7456</div>
7457<div class="listingblock">
7458<div class="content">
7459<pre class="CodeRay highlight"><code data-lang="c++">layout(binding = <span class="integer">2</span>, offset = <span class="integer">4</span>) uniform atomic_uint a;</code></pre>
7460</div>
7461</div>
7462<div class="paragraph">
7463<p>will establish that the opaque handle to the atomic counter <em>a</em> will be
7464bound to atomic counter buffer binding point 2 at an offset of 4 basic
7465machine units into that buffer.
7466The default <em>offset</em> for binding point 2 will be post incremented by 4 (the
7467size of an atomic counter).</p>
7468</div>
7469<div class="paragraph">
7470<p>A subsequent atomic counter declaration will inherit the previous (post
7471incremented) offset.
7472For example, a subsequent declaration of</p>
7473</div>
7474<div class="listingblock">
7475<div class="content">
7476<pre class="CodeRay highlight"><code data-lang="c++">layout(binding = <span class="integer">2</span>) uniform atomic_uint bar;</code></pre>
7477</div>
7478</div>
7479<div class="paragraph">
7480<p>will establish that the atomic counter <em>bar</em> has a binding to buffer binding
7481point 2 at an offset of 8 basic machine units into that buffer.
7482The offset for binding point 2 will again be post-incremented by 4 (the size
7483of an atomic counter).</p>
7484</div>
7485<div class="paragraph">
7486<p>When multiple variables are listed in a layout declaration, the effect will
7487be the same as if they were declared one at a time, in order from left to
7488right.</p>
7489</div>
7490<div class="paragraph">
7491<p>Binding points are not inherited, only offsets.
7492Each binding point tracks its own current default <em>offset</em> for inheritance
7493of subsequent variables using the same <strong>binding</strong>.
7494The initial state of compilation is that all <strong>binding</strong> points have an
7495<em>offset</em> of 0.
7496The <em>offset</em> can be set per binding point at global scope (without declaring
7497a variable).
7498For example,</p>
7499</div>
7500<div class="listingblock">
7501<div class="content">
7502<pre class="CodeRay highlight"><code data-lang="c++">layout(binding = <span class="integer">2</span>, offset = <span class="integer">4</span>) uniform atomic_uint;</code></pre>
7503</div>
7504</div>
7505<div class="paragraph">
7506<p>Establishes that the next <strong>atomic_uint</strong> declaration for binding point 2 will
7507inherit <em>offset</em> 4 (but does not establish a default <strong>binding</strong>):</p>
7508</div>
7509<div class="listingblock">
7510<div class="content">
7511<pre class="CodeRay highlight"><code data-lang="c++">layout(binding = <span class="integer">2</span>) uniform atomic_uint bar; <span class="comment">// offset is 4</span>
7512layout(offset = <span class="integer">8</span>) uniform atomic_uint bar;  <span class="comment">// error, no default binding</span></code></pre>
7513</div>
7514</div>
7515<div class="paragraph">
7516<p>Atomic counters may share the same binding point, but if a binding is
7517shared, their offsets must be either explicitly or implicitly (from
7518inheritance) unique and non overlapping.</p>
7519</div>
7520<div class="paragraph">
7521<p>Example valid uniform declarations, assuming top of shader:</p>
7522</div>
7523<div class="listingblock">
7524<div class="content">
7525<pre class="CodeRay highlight"><code data-lang="c++">layout(binding=<span class="integer">3</span>, offset=<span class="integer">4</span>) uniform atomic_uint a; <span class="comment">// offset = 4</span>
7526layout(binding=<span class="integer">2</span>) uniform atomic_uint b;           <span class="comment">// offset = 0</span>
7527layout(binding=<span class="integer">3</span>) uniform atomic_uint c;           <span class="comment">// offset = 8</span>
7528layout(binding=<span class="integer">2</span>) uniform atomic_uint d;           <span class="comment">// offset = 4</span></code></pre>
7529</div>
7530</div>
7531<div class="paragraph">
7532<p>Example of an invalid uniform declaration:</p>
7533</div>
7534<div class="listingblock">
7535<div class="content">
7536<pre class="CodeRay highlight"><code data-lang="c++">layout(offset=<span class="integer">4</span>) ...               <span class="comment">// error, must include binding</span>
7537layout(binding=<span class="integer">1</span>, offset=<span class="integer">0</span>) ... a; <span class="comment">// okay</span>
7538layout(binding=<span class="integer">2</span>, offset=<span class="integer">0</span>) ... b; <span class="comment">// okay</span>
7539layout(binding=<span class="integer">1</span>, offset=<span class="integer">0</span>) ... c; <span class="comment">// error, offsets must not be shared</span>
7540                                   <span class="comment">// between a and c</span>
7541layout(binding=<span class="integer">1</span>, offset=<span class="integer">2</span>) ... d; <span class="comment">// error, overlaps offset 0 of a</span></code></pre>
7542</div>
7543</div>
7544<div class="paragraph">
7545<p>It is a compile-time error to bind an atomic counter with a binding value
7546greater than or equal to <em>gl_MaxAtomicCounterBindings</em>.</p>
7547</div>
7548</div>
7549<div class="sect3">
7550<h4 id="format-layout-qualifiers">4.4.7. Format Layout Qualifiers</h4>
7551<div class="paragraph">
7552<p>Format layout qualifiers can be used on image variable declarations (those
7553declared with a basic type having &#8220;<strong>image</strong>&#8221; in its keyword).
7554The format layout qualifier identifiers for image variable declarations are:</p>
7555</div>
7556<div class="openblock bnf">
7557<div class="content">
7558<div class="dlist">
7559<dl>
7560<dt class="hdlist1"><em>layout-qualifier-id</em> : </dt>
7561<dd>
7562<p><em>float-image-format-qualifier</em><br>
7563<em>int-image-format-qualifier</em><br>
7564<em>uint-image-format-qualifier</em><br>
7565<strong>binding</strong> <strong>=</strong> <em>layout-qualifier-value</em></p>
7566</dd>
7567</dl>
7568</div>
7569</div>
7570</div>
7571<div class="openblock bnf">
7572<div class="content">
7573<div class="dlist">
7574<dl>
7575<dt class="hdlist1"><em>float-image-format-qualifier</em> : </dt>
7576<dd>
7577<p><strong>rgba32f</strong><br>
7578<strong>rgba16f</strong><br>
7579<strong>r32f</strong><br>
7580<strong>rgba8</strong><br>
7581<strong>rgba8_snorm</strong></p>
7582</dd>
7583<dt class="hdlist1"><em>int-image-format-qualifier</em> : </dt>
7584<dd>
7585<p><strong>rgba32i</strong><br>
7586<strong>rgba16i</strong><br>
7587<strong>rgba8i</strong><br>
7588<strong>r32i</strong></p>
7589</dd>
7590<dt class="hdlist1"><em>uint-image-format-qualifier</em> : </dt>
7591<dd>
7592<p><strong>rgba32ui</strong><br>
7593<strong>rgba16ui</strong><br>
7594<strong>rgba8ui</strong><br>
7595<strong>r32ui</strong></p>
7596</dd>
7597</dl>
7598</div>
7599</div>
7600</div>
7601<div class="paragraph">
7602<p>A format layout qualifier specifies the image format associated with a
7603declared image variable.
7604Only one format qualifier may be specified for any image variable
7605declaration.
7606For image variables with floating-point component types (keywords starting
7607with &#8220;<strong>image</strong>&#8221;), signed integer component types (keywords starting with
7608&#8220;<strong>iimage</strong>&#8221;), or unsigned integer component types (keywords starting with
7609&#8220;<strong>uimage</strong>&#8221;), the format qualifier used must match the
7610<em>float-image-format-qualifier</em>, <em>int-image-format-qualifier</em>, or
7611<em>uint-image-format-qualifier</em> grammar rules, respectively.
7612It is an error to declare an image variable where the format
7613qualifier does not match the image variable type.</p>
7614</div>
7615<div class="paragraph">
7616<p>Any image variable must specify a format layout qualifier.</p>
7617</div>
7618<div class="paragraph">
7619<p>The <strong>binding</strong> qualifier was described in
7620&#8220;<a href="#opaque-uniform-layout-qualifiers">Opaque Uniform Layout Qualifiers</a>&#8221;.</p>
7621</div>
7622</div>
7623<div class="sect3">
7624<h4 id="_subpass_input_qualifier">4.4.8. Subpass Input Qualifier</h4>
7625<div class="paragraph">
7626<p>Subpass inputs are only available when targeting Vulkan.</p>
7627</div>
7628<div class="paragraph">
7629<p>Subpass inputs are declared with the basic <strong>subpassInput</strong> types.
7630They must be declared with the layout qualifier
7631<strong>input_attachment_index</strong>, or an error results.
7632For example:</p>
7633</div>
7634<div class="listingblock">
7635<div class="content">
7636<pre class="CodeRay highlight"><code data-lang="c++">layout(input_attachment_index = <span class="integer">2</span>) uniform subpassInput t;</code></pre>
7637</div>
7638</div>
7639<div class="paragraph">
7640<p>This selects which subpass input is being read from. The value assigned
7641to <strong>input_attachment_index</strong>, say <em>i</em> (<code>input_attachment_index = i</code>), selects
7642that entry (<em>i</em> th entry) in the input list for the pass.  See the API
7643documentation for more detail about passes and the input list.</p>
7644</div>
7645<div class="paragraph">
7646<p>If an array of size <em>N</em> is declared, it consumes <em>N</em> consecutive
7647<strong>input_attachment_index</strong> values, starting with the one provided.</p>
7648</div>
7649<div class="paragraph">
7650<p>It is a compile-time or link-time error to have different variables
7651declared with the same <strong>input_attachment_index</strong>.
7652This includes any overlap in the implicit <strong>input_attachment_index</strong> consumed by
7653array declarations.</p>
7654</div>
7655<div class="paragraph">
7656<p>It is an error if the value assigned to an <strong>input_attachment_index</strong>
7657is greater than or equal to <em>gl_MaxInputAttachments</em>.</p>
7658</div>
7659</div>
7660</div>
7661<div class="sect2">
7662<h3 id="interpolation-qualifiers">4.5. Interpolation Qualifiers</h3>
7663<div class="paragraph">
7664<p>Inputs and outputs that could be interpolated can be further qualified by at
7665most one of the following interpolation qualifiers:</p>
7666</div>
7667<table class="tableblock frame-all grid-all stretch">
7668<colgroup>
7669<col style="width: 50%;">
7670<col style="width: 50%;">
7671</colgroup>
7672<thead>
7673<tr>
7674<th class="tableblock halign-left valign-top">Qualifier</th>
7675<th class="tableblock halign-left valign-top">Meaning</th>
7676</tr>
7677</thead>
7678<tbody>
7679<tr>
7680<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>smooth</strong></p></td>
7681<td class="tableblock halign-left valign-top"><p class="tableblock">perspective correct interpolation</p></td>
7682</tr>
7683<tr>
7684<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>flat</strong></p></td>
7685<td class="tableblock halign-left valign-top"><p class="tableblock">no interpolation</p></td>
7686</tr>
7687</tbody>
7688</table>
7689<div class="paragraph">
7690<p>The presence of and type of interpolation is controlled by the above
7691interpolation qualifiers as well as the auxiliary storage qualifiers
7692<strong>centroid</strong> and <strong>sample</strong>.
7693When no interpolation qualifier is present, smooth interpolation is used.
7694It is a compile-time error to use more than one interpolation qualifier.
7695The auxiliary storage qualifier <strong>patch</strong> is not used for interpolation; it is
7696a compile-time error to use interpolation qualifiers with <strong>patch</strong>.</p>
7697</div>
7698<div class="paragraph">
7699<p>A variable qualified as <strong>flat</strong> will not be interpolated.
7700Instead, it will have the same value for every fragment within a primitive.
7701This value will come from a single provoking vertex, as described by the
7702<a href="#references">API</a>.
7703A variable qualified as <strong>flat</strong> may also be qualified as <strong>centroid</strong> or
7704<strong>sample</strong>, which will mean the same thing as qualifying it only as <strong>flat</strong>.</p>
7705</div>
7706<div class="paragraph">
7707<p>A variable qualified as <strong>smooth</strong> will be interpolated in a
7708perspective-correct manner over the primitive being rendered.</p>
7709</div>
7710<div class="paragraph">
7711<p>Interpolation in a perspective correct manner is specified in equations 13.4
7712of the <a href="#references">OpenGL ES Specification</a>, section 13.4.1 &#8220;Line Segments&#8221; and
7713equation 13.7, section 13.5.1 &#8220;Polygon Interpolation&#8221;.</p>
7714</div>
7715<div class="paragraph">
7716<p>When multisample rasterization is disabled, or for fragment shader input
7717variables qualified with neither <strong>centroid</strong> nor <strong>sample</strong>, the value of the
7718assigned variable may be interpolated anywhere within the pixel and a single
7719value may be assigned to each sample within the pixel, to the extent
7720permitted by the <a href="#references">OpenGL ES Specification</a>.</p>
7721</div>
7722<div class="paragraph">
7723<p>When multisample rasterization is enabled, <strong>centroid</strong> and <strong>sample</strong> may be
7724used to control the location and frequency of the sampling of the qualified
7725fragment shader input.
7726If a fragment shader input is qualified with <strong>centroid</strong>, a single value may
7727be assigned to that variable for all samples in the pixel, but that value
7728must be interpolated at a location that lies in both the pixel and in the
7729primitive being rendered, including any of the pixel&#8217;s samples covered by
7730the primitive.
7731Because the location at which the variable is interpolated may be different
7732in neighboring pixels, and derivatives may be computed by computing
7733differences between neighboring pixels, derivatives of centroid-sampled
7734inputs may be less accurate than those for non-centroid interpolated
7735variables.
7736If a fragment shader input is qualified with <strong>sample</strong>, a separate value must
7737be assigned to that variable for each covered sample in the pixel, and that
7738value must be sampled at the location of the individual sample.</p>
7739</div>
7740</div>
7741<div class="sect2">
7742<h3 id="parameter-qualifiers">4.6. Parameter Qualifiers</h3>
7743<div class="paragraph">
7744<p>In addition to precision qualifiers and memory qualifiers, parameters can
7745have these parameter qualifiers.</p>
7746</div>
7747<table class="tableblock frame-all grid-all stretch">
7748<colgroup>
7749<col style="width: 50%;">
7750<col style="width: 50%;">
7751</colgroup>
7752<thead>
7753<tr>
7754<th class="tableblock halign-left valign-top">Qualifier</th>
7755<th class="tableblock halign-left valign-top">Meaning</th>
7756</tr>
7757</thead>
7758<tbody>
7759<tr>
7760<td class="tableblock halign-left valign-top"><p class="tableblock">&lt;none: default&gt;</p></td>
7761<td class="tableblock halign-left valign-top"><p class="tableblock">same as <strong>in</strong></p></td>
7762</tr>
7763<tr>
7764<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>in</strong></p></td>
7765<td class="tableblock halign-left valign-top"><p class="tableblock">for function parameters passed into a function</p></td>
7766</tr>
7767<tr>
7768<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>out</strong></p></td>
7769<td class="tableblock halign-left valign-top"><p class="tableblock">for function parameters passed back out of a function,
7770                   but not initialized for use when passed in</p></td>
7771</tr>
7772<tr>
7773<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>inout</strong></p></td>
7774<td class="tableblock halign-left valign-top"><p class="tableblock">for function parameters passed both into and out of a
7775                   function</p></td>
7776</tr>
7777</tbody>
7778</table>
7779<div class="paragraph">
7780<p>Parameter qualifiers are discussed in more detail in
7781&#8220;<a href="#function-calling-conventions">Function Calling Conventions</a>&#8221;.</p>
7782</div>
7783</div>
7784<div class="sect2">
7785<h3 id="precision-and-precision-qualifiers">4.7. Precision and Precision Qualifiers</h3>
7786<div class="paragraph">
7787<p>When targeting Vulkan:
7788For interface matching, uniform variables and uniform and buffer block
7789members must have the same precision qualification.
7790Global variables declared in different compilation units linked into the
7791same shader stage must be declared with the same precision qualification.</p>
7792</div>
7793<div class="paragraph">
7794<p>For the purposes of determining if an output from one shader stage matches
7795an input of the next stage, the precision qualifier need not match.</p>
7796</div>
7797<div class="sect3">
7798<h4 id="range-and-precision">4.7.1. Range and Precision</h4>
7799<div class="paragraph">
7800<p>The precision of <strong>highp</strong>
7801floating-point variables is defined by the IEEE 754 standard for
780232-bit
7803floating-point numbers.</p>
7804</div>
7805<div class="paragraph">
7806<p>This includes support for NaNs (Not a Number) and Infs (positive or negative
7807infinities) and positive and negative zeros.</p>
7808</div>
7809<div class="paragraph">
7810<p>The following rules apply to <strong>highp</strong>
7811operations:
7812Signed infinities and zeros are generated as dictated by IEEE, but subject
7813to the precisions allowed in the following table.
7814Any subnormal (denormalized) value input into a shader or potentially
7815generated by any operation in a shader can be flushed to 0.
7816The rounding mode cannot be set and is undefined but must not affect the
7817result by more than 1 ULP.
7818NaNs are not required to be generated.
7819Support for signaling NaNs is not required and exceptions are never raised.
7820Operations including built-in functions that operate on a NaN are not
7821required to return a NaN as the result.
7822However if NaNs are generated, <strong>isnan</strong>() must return the correct value.</p>
7823</div>
7824<div class="paragraph">
7825<p>Precisions are expressed in terms of maximum relative error in units of ULP
7826(units in the last place), unless otherwise noted.</p>
7827</div>
7828<div class="paragraph">
7829<p>For single precision operations, precisions are required as follows:</p>
7830</div>
7831<table class="tableblock frame-all grid-all stretch">
7832<colgroup>
7833<col style="width: 50%;">
7834<col style="width: 50%;">
7835</colgroup>
7836<thead>
7837<tr>
7838<th class="tableblock halign-left valign-top">Operation</th>
7839<th class="tableblock halign-left valign-top">Precision</th>
7840</tr>
7841</thead>
7842<tbody>
7843<tr>
7844<td class="tableblock halign-left valign-top"><p class="tableblock"><em>a</em> + <em>b</em>, <em>a</em> - <em>b</em>, <em>a</em> * <em>b</em></p></td>
7845<td class="tableblock halign-left valign-top"><p class="tableblock">Correctly rounded.</p></td>
7846</tr>
7847<tr>
7848<td class="tableblock halign-left valign-top"><p class="tableblock">&lt;, &lt;=, ==, &gt;, &gt;=</p></td>
7849<td class="tableblock halign-left valign-top"><p class="tableblock">Correct result.</p></td>
7850</tr>
7851<tr>
7852<td class="tableblock halign-left valign-top"><p class="tableblock"><em>a</em> / <em>b</em>, 1.0 / <em>b</em></p></td>
7853<td class="tableblock halign-left valign-top"><p class="tableblock">2.5 ULP for <span class="eq">|b|</span> in the range <span class="eq">[2<sup>-126</sup>, 2<sup>126</sup>]</span>.</p></td>
7854</tr>
7855<tr>
7856<td class="tableblock halign-left valign-top"><p class="tableblock"><em>a</em> * <em>b</em> + <em>c</em></p></td>
7857<td class="tableblock halign-left valign-top"><p class="tableblock">Correctly rounded single operation or
7858                                    sequence of two correctly rounded operations.</p></td>
7859</tr>
7860<tr>
7861<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>fma</strong>()</p></td>
7862<td class="tableblock halign-left valign-top"><p class="tableblock">Inherited from <em>a</em> * <em>b</em> + <em>c</em>.</p></td>
7863</tr>
7864<tr>
7865<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>pow</strong>(<em>x</em>, <em>y</em>)</p></td>
7866<td class="tableblock halign-left valign-top"><p class="tableblock">Inherited from <strong>exp2</strong>(<em>y</em> * <strong>log2</strong>(<em>x</em>)).</p></td>
7867</tr>
7868<tr>
7869<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>exp</strong>(<em>x</em>), <strong>exp2</strong>(<em>x</em>)</p></td>
7870<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">(3 + 2 · |x|)</span> ULP.</p></td>
7871</tr>
7872<tr>
7873<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>log</strong>(), <strong>log2</strong>()</p></td>
7874<td class="tableblock halign-left valign-top"><p class="tableblock">3 ULP outside the range <span class="eq">[0.5,2.0]</span>.<br>
7875                                    Absolute error &lt; 2<sup>-21</sup> inside the range
7876                                    <span class="eq">[0.5,2.0]</span>.</p></td>
7877</tr>
7878<tr>
7879<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>sqrt</strong>()</p></td>
7880<td class="tableblock halign-left valign-top"><p class="tableblock">Inherited from 1.0 / <strong>inversesqrt</strong>().</p></td>
7881</tr>
7882<tr>
7883<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>inversesqrt</strong>()</p></td>
7884<td class="tableblock halign-left valign-top"><p class="tableblock">2 ULP.</p></td>
7885</tr>
7886<tr>
7887<td class="tableblock halign-left valign-top"><p class="tableblock">implicit and explicit
7888  conversions between types</p></td>
7889<td class="tableblock halign-left valign-top"><p class="tableblock">Correctly rounded.</p></td>
7890</tr>
7891</tbody>
7892</table>
7893<div class="paragraph">
7894<p>Built-in functions defined in the specification with an equation built from
7895the above operations inherit the above errors.
7896These include, for example, the geometric functions, the common functions,
7897and many of the matrix functions.
7898Built-in functions not listed above and not defined as equations of the
7899above have undefined precision.
7900These include, for example, the trigonometric functions and determinant.</p>
7901</div>
7902<div class="paragraph">
7903<p>Storage requirements are declared through use of <em>precision qualifiers</em>.
7904The precision of operations must preserve the storage precisions of the
7905variables involved.</p>
7906</div>
7907<div class="paragraph">
7908<p><strong>highp</strong> floating-point values are stored in IEEE 754 single precision
7909floating-point format.
7910<strong>mediump</strong> and <strong>lowp</strong> floating-point values have minimum range and precision
7911requirements as detailed below and have maximum range and precision as
7912defined by IEEE 754.</p>
7913</div>
7914<div class="paragraph">
7915<p>All integral types are assumed to be implemented as integers and so may not
7916be emulated by floating-point values.
7917<strong>highp</strong> signed integers are represented as twos-complement 32-bit signed
7918integers.
7919<strong>highp</strong> unsigned integers are represented as unsigned 32-bit integers.
7920<strong>mediump</strong> integers (signed and unsigned) must be represented as an integer
7921with between 16 and 32 bits inclusive.
7922<strong>lowp</strong> integers (signed and unsigned) must be represented as an integer with
7923between 9 and 32 bits inclusive.</p>
7924</div>
7925<div class="paragraph">
7926<p>The required ranges and precisions for precision qualifiers are:</p>
7927</div>
7928<table class="tableblock frame-all grid-all stretch">
7929<colgroup>
7930<col style="width: 16.6666%;">
7931<col style="width: 16.6666%;">
7932<col style="width: 16.6666%;">
7933<col style="width: 16.6666%;">
7934<col style="width: 16.6666%;">
7935<col style="width: 16.667%;">
7936</colgroup>
7937<thead>
7938<tr>
7939<th class="tableblock halign-left valign-top">Qualifier</th>
7940<th class="tableblock halign-left valign-top">Floating-Point Range</th>
7941<th class="tableblock halign-left valign-top">Floating-Point Magnitude Range</th>
7942<th class="tableblock halign-left valign-top">Floating-Point Precision</th>
7943<th class="tableblock halign-left valign-top">Signed Integer Range</th>
7944<th class="tableblock halign-left valign-top">Unsigned Integer Range</th>
7945</tr>
7946</thead>
7947<tbody>
7948<tr>
7949<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>highp</strong></p></td>
7950<td class="tableblock halign-left valign-top"><p class="tableblock">Subset of IEEE-754<br>
7951      <span class="eq">(-2<sup>128</sup>,2<sup>128</sup>)</span></p></td>
7952<td class="tableblock halign-left valign-top"><p class="tableblock">Subset of IEEE-754<br>
7953      0.0, <span class="eq">[2<sup>-126</sup>,2<sup>128</sup>)</span></p></td>
7954<td class="tableblock halign-left valign-top"><p class="tableblock">Subset of IEEE-754 relative:<br>
7955      <span class="eq">2<sup>-23</sup></span></p></td>
7956<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">[-2<sup>31</sup>,2<sup>31</sup>-1]</span></p></td>
7957<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">[0,2<sup>32</sup>-1]</span></p></td>
7958</tr>
7959<tr>
7960<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mediump</strong> (minimum requirements)</p></td>
7961<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">(-2<sup>14</sup>,2<sup>14</sup>)</span></p></td>
7962<td class="tableblock halign-left valign-top"><p class="tableblock">0.0, <span class="eq">[2<sup>-14</sup>,2<sup>14</sup>)</span></p></td>
7963<td class="tableblock halign-left valign-top"><p class="tableblock">Relative:<br>
7964      <span class="eq">2<sup>-10</sup></span></p></td>
7965<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">[-2<sup>15</sup>,2<sup>15</sup>-1]</span></p></td>
7966<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">[0,2<sup>16</sup>-1]</span></p></td>
7967</tr>
7968<tr>
7969<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>lowp</strong> (minimum requirements)</p></td>
7970<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">(-2,2)</span></p></td>
7971<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">(2<sup>-8</sup>,2)</span></p></td>
7972<td class="tableblock halign-left valign-top"><p class="tableblock">Absolute:<br>
7973      <span class="eq">2<sup>-8</sup></span></p></td>
7974<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">[-2<sup>8</sup>,2<sup>8</sup>-1]</span></p></td>
7975<td class="tableblock halign-left valign-top"><p class="tableblock"><span class="eq">[0,2<sup>9</sup>-1]</span></p></td>
7976</tr>
7977</tbody>
7978</table>
7979<div class="paragraph">
7980<p>The semi-open interval notation used for the magnitude ranges means that the
7981lower end of the interval is included, but the upper end is excluded. Thus the
7982largest required magnitude is half of the relative precision less than the
7983value given. The largest required value for highp, for example, is <span class="eq">2<sup>128</sup>
7984× (1 - 2<sup>-24</sup>)</span>.</p>
7985</div>
7986<div class="paragraph">
7987<p><em>Relative precision</em> is defined as the worst case (i.e. largest) ratio of
7988the smallest step in relation to the value for all non-zero values in the
7989required ranges, above:</p>
7990</div>
7991<div class="stemblock">
7992<div class="content">
7993\[\begin{aligned}
7994    \mathit{Precision}_{relative} = {
7995        \left| {
7996            { \left| v_1 - v_2 \right| }_{min} } \over { v_1 }
7997        \right|
7998    }_{max}, &amp;
7999    v_1,v_2 \neq 0, v_1 \neq v_2
8000\end{aligned}\]
8001</div>
8002</div>
8003<div class="paragraph">
8004<p>It is therefore twice the maximum rounding error when converting from a real
8005number.
8006Subnormal numbers may be supported and may have lower relative precision.</p>
8007</div>
8008<div class="paragraph">
8009<p>In addition, the range and precision of a <strong>mediump</strong> floating-point value
8010must be the same as or greater than the range and precision of a <strong>lowp</strong>
8011floating-point value.
8012The range and precision of a <strong>highp</strong> floating-point value must be the same
8013as or greater than the range and precision of a <strong>mediump</strong> floating-point
8014value.</p>
8015</div>
8016<div class="paragraph">
8017<p>The range of a <strong>mediump</strong> integer value must be the same as or greater than
8018the range of a <strong>lowp</strong> integer value.
8019The range of a <strong>highp</strong> integer value must be the same as or greater than the
8020range of a <strong>mediump</strong> integer value.</p>
8021</div>
8022<div class="paragraph">
8023<p>Within the above specification, an implementation is allowed to vary the
8024representation of numeric values, both within a shader and between different
8025shaders.
8026If necessary, this variance can be controlled using the invariance
8027qualifier.</p>
8028</div>
8029<div class="paragraph">
8030<p>The actual ranges and precisions provided by an implementation can be
8031queried through the API.
8032See the <a href="#references">normative references</a> for details on how to do
8033this.</p>
8034</div>
8035</div>
8036<div class="sect3">
8037<h4 id="conversion-between-precisions">4.7.2. Conversion between precisions</h4>
8038<div class="paragraph">
8039<p>Within the same type, conversion from a lower to a higher precision must be
8040exact.
8041When converting from a higher precision to a lower precision, if the value
8042is representable by the implementation of the target precision, the
8043conversion must also be exact.
8044If the value is not representable, the behavior is dependent on the type:</p>
8045</div>
8046<div class="ulist">
8047<ul>
8048<li>
8049<p>For signed and unsigned integers, the value is truncated; bits in
8050positions not present in the target precision are set to zero.
8051(Positions start at zero and the least significant bit is considered to
8052be position zero for this purpose.)</p>
8053</li>
8054<li>
8055<p>For floating-point values, the value should either clamp to +Inf or
8056-Inf, or to the maximum or minimum value that the implementation
8057supports.
8058While this behavior is implementation-dependent, it should be consistent
8059for a given implementation.</p>
8060</li>
8061</ul>
8062</div>
8063</div>
8064<div class="sect3">
8065<h4 id="precision-qualifiers">4.7.3. Precision Qualifiers</h4>
8066<div class="paragraph">
8067<p>Any
8068floating-point, integer, or opaque-type declaration can have the type
8069preceded by one of these precision qualifiers:</p>
8070</div>
8071<table class="tableblock frame-all grid-all stretch">
8072<colgroup>
8073<col style="width: 50%;">
8074<col style="width: 50%;">
8075</colgroup>
8076<tbody>
8077<tr>
8078<td class="tableblock halign-left valign-top"><p class="tableblock">Qualifier</p></td>
8079<td class="tableblock halign-left valign-top"><p class="tableblock">Meaning</p></td>
8080</tr>
8081<tr>
8082<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>highp</strong></p></td>
8083<td class="tableblock halign-left valign-top"><p class="tableblock">The variable satisfies the minimum requirements for <strong>highp</strong>
8084              described above. <strong>highp</strong> variables have the maximum range and
8085              precision available but may cause operations to run more
8086              slowly on some implementations.</p></td>
8087</tr>
8088<tr>
8089<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>mediump</strong></p></td>
8090<td class="tableblock halign-left valign-top"><p class="tableblock">SPIR-V <strong>RelaxedPrecision</strong> when targeting Vulkan, otherwise
8091              the variable satisfies the minimum requirements for <strong>mediump</strong>
8092              described above. <strong>mediump</strong> variables may typically be used to
8093              store high dynamic range colors and low precision geometry.</p></td>
8094</tr>
8095<tr>
8096<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>lowp</strong></p></td>
8097<td class="tableblock halign-left valign-top"><p class="tableblock">SPIR-V <strong>RelaxedPrecision</strong> when targeting Vulkan, otherwise
8098              the variable satisfies the minimum requirements for <strong>lowp</strong>
8099              described above. <strong>lowp</strong> variables may typically be used to
8100              store 8-bit color values.</p></td>
8101</tr>
8102</tbody>
8103</table>
8104<div class="paragraph">
8105<p>For example:</p>
8106</div>
8107<div class="listingblock">
8108<div class="content">
8109<pre class="CodeRay highlight"><code data-lang="c++">lowp <span class="predefined-type">float</span> color;
8110out mediump vec2 P;
8111lowp ivec2 foo(lowp mat3);
8112highp mat4 m;</code></pre>
8113</div>
8114</div>
8115<div class="paragraph">
8116<p>Literal constants do not have precision qualifiers.
8117Neither do Boolean variables.
8118Neither do constructors.</p>
8119</div>
8120<div class="paragraph">
8121<p>For this paragraph, &#8220;operation&#8221; includes operators, built-in functions,
8122and constructors, and &#8220;operand&#8221; includes function arguments and
8123constructor arguments.
8124The precision used to internally evaluate an operation, and the precision
8125qualification subsequently associated with any resulting intermediate
8126values, must be at least as high as the highest precision qualification of
8127the operands consumed by the operation.</p>
8128</div>
8129<div class="paragraph">
8130<p>In cases where operands do not have a precision qualifier, the precision
8131qualification will come from the other operands.
8132If no operands have a precision qualifier, then the precision qualifications
8133of the operands of the next consuming operation in the expression will be
8134used.
8135This rule can be applied recursively until a precision qualified operand is
8136found.
8137If necessary, it will also include the precision qualification of l-values
8138for assignments, of the declared variable for initializers, of formal
8139parameters for function call arguments, or of function return types for
8140function return values.
8141If the precision cannot be determined by this method e.g. if an entire
8142expression is composed only of operands with no precision qualifier, and the
8143result is not assigned or passed as an argument, then it is evaluated at the
8144default precision of the type or greater.
8145When this occurs in the fragment shader, the default precision must be
8146defined.</p>
8147</div>
8148<div class="paragraph">
8149<p>For example, consider the statements:</p>
8150</div>
8151<div class="listingblock">
8152<div class="content">
8153<pre class="CodeRay highlight"><code data-lang="c++">uniform highp <span class="predefined-type">float</span> h1;
8154highp <span class="predefined-type">float</span> h2 = <span class="float">2</span><span class="float">.3</span> * <span class="float">4</span><span class="float">.7</span>; <span class="comment">// operation and result are highp</span>
8155precision
8156mediump <span class="predefined-type">float</span> m;
8157m = <span class="float">3</span><span class="float">.7</span> * h1 * h2; <span class="comment">// all operations are highp precision</span>
8158h2 = m * h1; <span class="comment">// operation is highp precision</span>
8159m = h2 - h1; <span class="comment">// operation is highp precision</span>
8160h2 = m + m; <span class="comment">// addition and result at mediump precision</span>
8161<span class="directive">void</span> f(highp <span class="predefined-type">float</span> p);
8162f(<span class="float">3</span><span class="float">.3</span>); <span class="comment">// 3.3 will be passed in at highp precision</span></code></pre>
8163</div>
8164</div>
8165<div class="paragraph">
8166<p>Precision qualifiers, as with other qualifiers, do not affect the basic type
8167of the variable.
8168In particular, there are no constructors for precision conversions;
8169constructors only convert types.
8170Similarly, precision qualifiers, as with other qualifiers, do not contribute
8171to function overloading based on parameter types.
8172As discussed in &#8220;<a href="#function-calling-conventions">Function Calling
8173Conventions</a>&#8221;, function input and output is done through copies, and
8174therefore qualifiers do not have to match.</p>
8175</div>
8176<div class="paragraph">
8177<p>Precision qualifiers for outputs in one shader matched to inputs in another
8178shader need not match when both shaders are linked into the same program.
8179When both shaders are in separate programs, mismatched precision qualifiers
8180will result in a program interface mismatch that will result in program
8181pipeline validation failures, as described in section 7.4.1 &#8220;Shader
8182Interface Matching&#8221; of the <a href="#references">OpenGL ES Specification</a>.</p>
8183</div>
8184<div class="paragraph">
8185<p>The precision of a variable is determined when the variable is declared and
8186cannot be subsequently changed.</p>
8187</div>
8188<div class="paragraph">
8189<p>Where the precision of a constant integral or constant floating-point
8190expression is not specified, evaluation is performed at <strong>highp</strong>.
8191This rule does not affect the precision qualification of the expression.</p>
8192</div>
8193<div class="paragraph">
8194<p>The evaluation of constant expressions must be invariant and will usually be
8195performed at compile time.</p>
8196</div>
8197</div>
8198<div class="sect3">
8199<h4 id="default-precision-qualifiers">4.7.4. Default Precision Qualifiers</h4>
8200<div class="paragraph">
8201<p>The precision statement</p>
8202</div>
8203<div class="listingblock">
8204<div class="content">
8205<pre class="CodeRay highlight"><code data-lang="c++">precision precision-qualifier type;</code></pre>
8206</div>
8207</div>
8208<div class="paragraph">
8209<p>can be used to establish a default precision qualifier.
8210The <em>type</em> field can be either <strong>int</strong>, <strong>float</strong>, or any of the opaque types,
8211and the <em>precision-qualifier</em> can be <strong>lowp</strong>, <strong>mediump</strong>, or <strong>highp</strong>.</p>
8212</div>
8213<div class="paragraph">
8214<p>Any other types or qualifiers will result in an error.
8215If <em>type</em> is <strong>float</strong>, the directive applies to non-precision-qualified
8216floating-point type (scalar, vector, and matrix) declarations.
8217If <em>type</em> is <strong>int</strong>, the directive applies to all non-precision-qualified
8218integer type (scalar, vector, signed, and unsigned) declarations.
8219This includes global variable declarations, function return declarations,
8220function parameter declarations, and local variable declarations.</p>
8221</div>
8222<div class="paragraph">
8223<p>Non-precision qualified declarations will use the precision qualifier
8224specified in the most recent <strong>precision</strong> statement that is still in scope.
8225The <strong>precision</strong> statement has the same scoping rules as variable
8226declarations.
8227If it is declared inside a compound statement, its effect stops at the end
8228of the inner-most statement it was declared in.
8229Precision statements in nested scopes override precision statements in outer
8230scopes.
8231Multiple precision statements for the same basic type can appear inside the
8232same scope, with later statements overriding earlier statements within that
8233scope.</p>
8234</div>
8235<div class="paragraph">
8236<p>All languages except for the fragment language have the following
8237predeclared globally scoped default precision statements:</p>
8238</div>
8239<div class="listingblock">
8240<div class="content">
8241<pre class="CodeRay highlight"><code data-lang="c++">precision highp <span class="predefined-type">float</span>;
8242precision highp <span class="predefined-type">int</span>;
8243precision lowp sampler2D;
8244precision lowp samplerCube;
8245precision highp atomic_uint;</code></pre>
8246</div>
8247</div>
8248<div class="paragraph">
8249<p>The fragment language has the following predeclared globally scoped default
8250precision statements:</p>
8251</div>
8252<div class="listingblock">
8253<div class="content">
8254<pre class="CodeRay highlight"><code data-lang="c++">precision mediump <span class="predefined-type">int</span>;
8255precision lowp sampler2D;
8256precision lowp samplerCube;
8257precision highp atomic_uint;</code></pre>
8258</div>
8259</div>
8260<div class="paragraph">
8261<p>The fragment language has no default precision qualifier for floating-point
8262types.
8263Hence for <strong>float</strong>, floating-point vector and matrix variable declarations,
8264either the declaration must include a precision qualifier or the default
8265float precision must have been previously declared.
8266Similarly, there is no default precision qualifier in any of the languages
8267for any type not listed above.</p>
8268</div>
8269</div>
8270<div class="sect3">
8271<h4 id="available-precision-qualifiers">4.7.5. Available Precision Qualifiers</h4>
8272<div class="paragraph">
8273<p>The built-in macro GL_FRAGMENT_PRECISION_HIGH is defined to one:</p>
8274</div>
8275<div class="listingblock">
8276<div class="content">
8277<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#define</span> GL_FRAGMENT_PRECISION_HIGH <span class="integer">1</span></code></pre>
8278</div>
8279</div>
8280<div class="paragraph">
8281<p>This macro is available in all languages except compute.</p>
8282</div>
8283</div>
8284</div>
8285<div class="sect2">
8286<h3 id="variance-and-the-invariant-qualifier">4.8. Variance and the Invariant Qualifier</h3>
8287<div class="paragraph">
8288<p>In this section, <em>variance</em> refers to the possibility of getting different
8289values from the same expression in different programs.
8290For example, consider the situation where two vertex shaders, in different
8291programs, each set <em>gl_Position</em> with the same expression, and the input
8292values into that expression are the same when both shaders run.
8293It is possible, due to independent compilation of the two shaders, that the
8294values assigned to <em>gl_Position</em> are not exactly the same when the two
8295shaders run.
8296In this example, this can cause problems with alignment of geometry in a
8297multi-pass algorithm.</p>
8298</div>
8299<div class="paragraph">
8300<p>In general, such variance between shaders is allowed.
8301When such variance does not exist for a particular output variable, that
8302variable is said to be <em>invariant</em>.</p>
8303</div>
8304<div class="sect3">
8305<h4 id="the-invariant-qualifier">4.8.1. The Invariant Qualifier</h4>
8306<div class="paragraph">
8307<p>To ensure that a particular output variable is invariant, it is necessary to
8308use the <strong>invariant</strong> qualifier.
8309It can either be used to qualify a previously declared variable as being
8310invariant:</p>
8311</div>
8312<div class="listingblock">
8313<div class="content">
8314<pre class="CodeRay highlight"><code data-lang="c++">invariant gl_Position; <span class="comment">// make existing gl_Position be invariant</span>
8315out vec3 Color;
8316invariant Color;       <span class="comment">// make existing Color be invariant</span></code></pre>
8317</div>
8318</div>
8319<div class="paragraph">
8320<p>or as part of a declaration when a variable is declared:</p>
8321</div>
8322<div class="listingblock">
8323<div class="content">
8324<pre class="CodeRay highlight"><code data-lang="c++">invariant centroid out vec3 Color;</code></pre>
8325</div>
8326</div>
8327<div class="paragraph">
8328<p>Only variables output from a shader can be candidates for invariance.
8329This includes user-defined output variables and the built-in output
8330variables.
8331As only outputs can be declared as invariant, an output from one shader
8332stage will still match an input of a subsequent stage without the input
8333being declared as invariant.</p>
8334</div>
8335<div class="paragraph">
8336<p>The <strong>invariant</strong> keyword can be followed by a comma separated list of
8337previously declared identifiers.
8338All uses of <strong>invariant</strong> must be at global scope or on block
8339members, and before any use of the variables being declared as invariant.</p>
8340</div>
8341<div class="paragraph">
8342<p>To guarantee invariance of a particular output variable across two programs,
8343the following must also be true:</p>
8344</div>
8345<div class="ulist">
8346<ul>
8347<li>
8348<p>The output variable is declared as invariant in both programs.</p>
8349</li>
8350<li>
8351<p>The same values must be input to all shader input variables consumed by
8352expressions and control flow contributing to the value assigned to the
8353output variable.</p>
8354</li>
8355<li>
8356<p>The texture formats, texel values, and texture filtering are set the
8357same way for any texture function calls contributing to the value of the
8358output variable.</p>
8359</li>
8360<li>
8361<p>All input values are all operated on in the same way.
8362    All operations in the consuming expressions and any intermediate
8363    expressions must be the same, with the same order of operands and same
8364    associativity, to give the same order of evaluation.
8365    Intermediate variables and functions must be declared as the same type
8366    with the same explicit or implicit precision
8367qualifiers and the same constant qualifiers.
8368    Any control flow affecting the output value must be the same, and any
8369    expressions consumed to determine this control flow must also follow
8370    these invariance rules.</p>
8371</li>
8372<li>
8373<p>All the data flow and control flow leading to setting the invariant
8374output variable reside in a single compilation unit.</p>
8375</li>
8376</ul>
8377</div>
8378<div class="paragraph">
8379<p>Essentially, all the data flow and control flow leading to an invariant
8380output must match.</p>
8381</div>
8382<div class="paragraph">
8383<p>Initially, by default, all output variables are allowed to be variant.
8384To force all output variables to be invariant, use the pragma</p>
8385</div>
8386<div class="listingblock">
8387<div class="content">
8388<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#pragma</span> STDGL invariant(all)</code></pre>
8389</div>
8390</div>
8391<div class="paragraph">
8392<p>before all declarations in a shader.
8393If this pragma is used after the declaration of any variables or functions,
8394then the set of outputs that behave as invariant is undefined.</p>
8395</div>
8396<div class="paragraph">
8397<p>Generally, invariance is ensured at the cost of flexibility in optimization,
8398so performance can be degraded by use of invariance.
8399Hence, use of this pragma is intended as a debug aid, to avoid individually
8400declaring all output variables as invariant.</p>
8401</div>
8402</div>
8403<div class="sect3">
8404<h4 id="invariance-within-a-shader">4.8.2. Invariance Within a Shader</h4>
8405<div class="paragraph">
8406<p>When a value is stored in a variable, it is usually assumed it will remain
8407constant unless explicitly changed.
8408However, during the process of optimization, it is possible that the
8409compiler may choose to recompute a value rather than store it in a register.
8410Since the precision of operations is not completely specified (e.g. a low
8411precision operation may be done at medium or high precision), it would be
8412possible for the recomputed value to be different from the original value.</p>
8413</div>
8414<div class="paragraph">
8415<p>Values are allowed to be variant within a shader.
8416To prevent this, the invariant qualifier or invariant pragma must be used.</p>
8417</div>
8418<div class="paragraph">
8419<p>Within a shader, there is no invariance for values generated by different
8420non-constant expressions, even if those expressions are identical.</p>
8421</div>
8422<div class="paragraph">
8423<p>Example 1:</p>
8424</div>
8425<div class="listingblock">
8426<div class="content">
8427<pre class="CodeRay highlight"><code data-lang="c++">precision mediump;
8428vec4 col;
8429vec2 a = ...
8430...
8431col = texture(tex, a);<span class="comment">// a has a value _a1 _</span>
8432...
8433col = texture(tex, a); <span class="comment">// a has a value a2 where possibly a1 != a2</span></code></pre>
8434</div>
8435</div>
8436<div class="paragraph">
8437<p>To enforce invariance in this example use:</p>
8438</div>
8439<div class="listingblock">
8440<div class="content">
8441<pre class="CodeRay highlight"><code data-lang="c++"><span class="preprocessor">#pragma</span> STDGL invariant(all)</code></pre>
8442</div>
8443</div>
8444<div class="paragraph">
8445<p>Example 2:</p>
8446</div>
8447<div class="listingblock">
8448<div class="content">
8449<pre class="CodeRay highlight"><code data-lang="c++">vec2 m = ...;
8450vec2 n = ...;
8451vec2 a = m + n;
8452vec2 b = m + n; <span class="comment">// a and b are not guaranteed to be exactly equal</span></code></pre>
8453</div>
8454</div>
8455<div class="paragraph">
8456<p>There is no mechanism to enforce invariance between a and b.</p>
8457</div>
8458</div>
8459<div class="sect3">
8460<h4 id="invariance-of-constant-expressions">4.8.3. Invariance of Constant Expressions</h4>
8461<div class="paragraph">
8462<p>Invariance must be guaranteed for constant expressions.
8463A particular constant expression must evaluate to the same result if it
8464appears again in the same shader or a different shader.
8465This includes the same expression appearing in two shaders of the same
8466language or shaders of two different languages.</p>
8467</div>
8468<div class="paragraph">
8469<p>Constant expressions must evaluate to the same result when operated on as
8470already described above for invariant variables.
8471Constant expressions are not invariant with respect to equivalent
8472non-constant expressions, even when the <strong>invariant</strong> qualifier or pragma is
8473used.</p>
8474</div>
8475</div>
8476<div class="sect3">
8477<h4 id="invariance-of-undefined-values">4.8.4. Invariance of Undefined Values</h4>
8478<div class="paragraph">
8479<p>Undefined values are not invariant nor can they be made invariant by use of
8480the invariant qualifier or pragma.
8481In some implementations, undefined values may cause unexpected behavior if
8482they are used in control-flow expressions e.g. in the following case, one,
8483both or neither functions may be executed and this may not be consistent
8484over multiple invocations of the shader:</p>
8485</div>
8486<div class="listingblock">
8487<div class="content">
8488<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span> x; <span class="comment">// undefined value</span>
8489<span class="keyword">if</span> (x == <span class="integer">1</span>)
8490{
8491    f(); <span class="comment">// Undefined whether f() is executed</span>
8492}
8493<span class="keyword">if</span> (x == <span class="integer">2</span>)
8494{
8495    g(); <span class="comment">// Undefined whether g() is executed.</span>
8496}</code></pre>
8497</div>
8498</div>
8499<div class="paragraph">
8500<p>Note that an undefined value is a value that has not been specified.
8501A value that has been specified but has a potentially large error due to,
8502for example, lack of precision in an expression, is not undefined and so can
8503be made invariant.</p>
8504</div>
8505</div>
8506</div>
8507<div class="sect2">
8508<h3 id="the-precise-qualifier">4.9. The Precise Qualifier</h3>
8509<div class="paragraph">
8510<p>Some algorithms require floating-point computations to exactly follow the
8511order of operations specified in the source code and to treat all operations
8512consistently, even if the implementation supports optimizations that could
8513produce nearly equivalent results with higher performance.
8514For example, many GL implementations support a &#8220;multiply-add&#8221; instruction
8515that can compute a floating-point expression such as</p>
8516</div>
8517<div class="listingblock">
8518<div class="content">
8519<pre class="CodeRay highlight"><code data-lang="c++">result = (a * b) + (c * d);</code></pre>
8520</div>
8521</div>
8522<div class="paragraph">
8523<p>in two operations instead of three operations; one multiply and one
8524multiply-add instead of two multiplies and one add.
8525The result of a floating-point multiply-add might not always be identical to
8526first doing a multiply yielding a floating-point result and then doing a
8527floating-point add.
8528Hence, in this example, the two multiply operations would not be treated
8529consistently; the two multiplies could effectively appear to have differing
8530precisions.</p>
8531</div>
8532<div class="paragraph">
8533<p>The key computation that needs to be made consistent appears when
8534tessellating, where intermediate points for subdivision are synthesized in
8535different directions, yet need to yield the same result, as shown in the
8536diagram below.</p>
8537</div>
8538<div id="img-precise" class="imageblock">
8539<div class="content">
8540<img src="data:image/svg+xml;base64,<?xml version="1.0" encoding="UTF-8"?>

<svg version="1.2" baseProfile="tiny" width="210mm" height="297mm" viewBox="0 0 21000 29700" preserveAspectRatio="xMidYMid" fill-rule="evenodd" clip-path="url(#presentation_clip_path)" stroke-width="28.222" stroke-linejoin="round" xmlns="http://www.w3.org/2000/svg" xmlns:ooo="http://xml.openoffice.org/svg/export" xmlns:xlink="http://www.w3.org/1999/xlink" xml:space="preserve">
 <defs class="ClipPathGroup">
  <clipPath id="presentation_clip_path" clipPathUnits="userSpaceOnUse">
   <rect x="0" y="0" width="21000" height="29700"/>
  </clipPath>
 </defs>
 <defs>
  <font id="EmbeddedFont_1" horiz-adv-x="2048">
   <font-face font-family="Liberation Sans embedded" units-per-em="2048" font-weight="normal" font-style="normal" ascent="1852" descent="423"/>
   <missing-glyph horiz-adv-x="2048" d="M 0,0 L 2047,0 2047,2047 0,2047 0,0 Z"/>
   <glyph unicode="w" horiz-adv-x="1509" d="M 1174,0 L 965,0 792,698 C 787,716 781,738 776,765 770,792 764,818 759,843 752,872 746,903 740,934 734,904 728,874 721,845 716,820 710,793 704,766 697,739 691,715 686,694 L 508,0 300,0 -3,1082 175,1082 358,347 C 363,332 367,313 372,291 377,268 381,246 386,225 391,200 396,175 401,149 406,174 412,199 418,223 423,244 429,265 434,286 439,307 444,325 448,339 L 644,1082 837,1082 1026,339 C 1031,322 1036,302 1041,280 1046,258 1051,237 1056,218 1061,195 1067,172 1072,149 1077,174 1083,199 1088,223 1093,244 1098,265 1103,288 1108,310 1112,330 1117,347 L 1308,1082 1484,1082 1174,0 Z"/>
   <glyph unicode="v" horiz-adv-x="1033" d="M 613,0 L 400,0 7,1082 199,1082 437,378 C 442,363 447,346 454,325 460,304 466,282 473,259 480,236 486,215 492,194 497,173 502,155 506,141 510,155 515,173 522,194 528,215 534,236 541,258 548,280 555,302 562,323 569,344 575,361 580,376 L 826,1082 1017,1082 613,0 Z"/>
   <glyph unicode="u" horiz-adv-x="874" d="M 314,1082 L 314,396 C 314,343 318,299 326,264 333,229 346,200 363,179 380,157 403,142 432,133 460,124 495,119 537,119 580,119 618,127 653,142 687,157 716,178 741,207 765,235 784,270 797,312 810,353 817,401 817,455 L 817,1082 997,1082 997,231 C 997,208 997,185 998,160 998,135 998,111 999,89 1000,66 1000,47 1001,31 1002,15 1002,5 1003,0 L 833,0 C 832,3 832,12 831,27 830,42 830,59 829,78 828,97 827,116 826,136 825,155 825,172 825,185 L 822,185 C 805,154 786,125 765,100 744,75 720,53 693,36 666,18 634,4 599,-6 564,-15 523,-20 476,-20 416,-20 364,-13 321,2 278,17 242,39 214,70 186,101 166,140 153,188 140,236 133,294 133,361 L 133,1082 314,1082 Z"/>
   <glyph unicode="t" horiz-adv-x="531" d="M 554,8 C 527,1 499,-5 471,-10 442,-14 409,-16 372,-16 228,-16 156,66 156,229 L 156,951 31,951 31,1082 163,1082 216,1324 336,1324 336,1082 536,1082 536,951 336,951 336,268 C 336,216 345,180 362,159 379,138 408,127 450,127 467,127 484,128 501,131 517,134 535,137 554,141 L 554,8 Z"/>
   <glyph unicode="s" horiz-adv-x="901" d="M 950,299 C 950,248 940,203 921,164 901,124 872,91 835,64 798,37 752,16 698,2 643,-13 581,-20 511,-20 448,-20 392,-15 342,-6 291,4 247,20 209,41 171,62 139,91 114,126 88,161 69,203 57,254 L 216,285 C 231,227 263,185 311,158 359,131 426,117 511,117 550,117 585,120 618,125 650,130 678,140 701,153 724,166 743,183 756,205 769,226 775,253 775,285 775,318 767,345 752,366 737,387 715,404 688,418 661,432 628,444 589,455 550,465 507,476 460,489 417,500 374,513 331,527 288,541 250,560 216,583 181,606 153,634 132,668 111,702 100,745 100,796 100,895 135,970 206,1022 276,1073 378,1099 513,1099 632,1099 727,1078 798,1036 868,994 912,927 931,834 L 769,814 C 763,842 752,866 736,885 720,904 701,919 678,931 655,942 630,951 602,956 573,961 544,963 513,963 432,963 372,951 333,926 294,901 275,864 275,814 275,785 282,761 297,742 311,723 331,707 357,694 382,681 413,669 449,660 485,650 525,640 568,629 597,622 626,614 656,606 686,597 715,587 744,576 772,564 799,550 824,535 849,519 870,500 889,478 908,456 923,430 934,401 945,372 950,338 950,299 Z"/>
   <glyph unicode="r" horiz-adv-x="530" d="M 142,0 L 142,830 C 142,853 142,876 142,900 141,923 141,946 140,968 139,990 139,1011 138,1030 137,1049 137,1067 136,1082 L 306,1082 C 307,1067 308,1049 309,1030 310,1010 311,990 312,969 313,948 313,929 314,910 314,891 314,874 314,861 L 318,861 C 331,902 344,938 359,969 373,999 390,1024 409,1044 428,1063 451,1078 478,1088 505,1097 537,1102 575,1102 590,1102 604,1101 617,1099 630,1096 641,1094 648,1092 L 648,927 C 636,930 622,933 606,935 590,936 572,937 552,937 511,937 476,928 447,909 418,890 394,865 376,832 357,799 344,759 335,714 326,668 322,618 322,564 L 322,0 142,0 Z"/>
   <glyph unicode="p" horiz-adv-x="953" d="M 1053,546 C 1053,464 1046,388 1033,319 1020,250 998,190 967,140 936,90 895,51 844,23 793,-6 730,-20 655,-20 578,-20 510,-5 452,24 394,53 350,101 319,168 L 314,168 C 315,167 315,161 316,150 316,139 316,126 317,110 317,94 317,76 318,57 318,37 318,17 318,-2 L 318,-425 138,-425 138,861 C 138,887 138,912 138,936 137,960 137,982 136,1002 135,1021 135,1038 134,1052 133,1066 133,1076 132,1082 L 306,1082 C 307,1080 308,1073 309,1061 310,1049 311,1035 312,1018 313,1001 314,982 315,963 316,944 316,925 316,908 L 320,908 C 337,943 356,972 377,997 398,1021 423,1041 450,1057 477,1072 508,1084 542,1091 575,1098 613,1101 655,1101 730,1101 793,1088 844,1061 895,1034 936,997 967,949 998,900 1020,842 1033,774 1046,705 1053,629 1053,546 Z M 864,542 C 864,609 860,668 852,720 844,772 830,816 811,852 791,888 765,915 732,934 699,953 658,962 609,962 569,962 531,956 496,945 461,934 430,912 404,880 377,848 356,804 341,748 326,691 318,618 318,528 318,451 324,387 337,334 350,281 368,238 393,205 417,172 447,149 483,135 519,120 560,113 607,113 657,113 699,123 732,142 765,161 791,189 811,226 830,263 844,308 852,361 860,414 864,474 864,542 Z"/>
   <glyph unicode="o" horiz-adv-x="980" d="M 1053,542 C 1053,353 1011,212 928,119 845,26 724,-20 565,-20 490,-20 422,-9 363,14 304,37 254,71 213,118 172,165 140,223 119,294 97,364 86,447 86,542 86,915 248,1102 571,1102 655,1102 728,1090 789,1067 850,1044 900,1009 939,962 978,915 1006,857 1025,787 1044,717 1053,635 1053,542 Z M 864,542 C 864,626 858,695 845,750 832,805 813,848 788,881 763,914 732,937 696,950 660,963 619,969 574,969 528,969 487,962 450,949 413,935 381,912 355,879 329,846 309,802 296,747 282,692 275,624 275,542 275,458 282,389 297,334 312,279 332,235 358,202 383,169 414,146 449,133 484,120 522,113 563,113 609,113 651,120 688,133 725,146 757,168 783,201 809,234 829,278 843,333 857,388 864,458 864,542 Z"/>
   <glyph unicode="n" horiz-adv-x="874" d="M 825,0 L 825,686 C 825,739 821,783 814,818 806,853 793,882 776,904 759,925 736,941 708,950 679,959 644,963 602,963 559,963 521,956 487,941 452,926 423,904 399,876 374,847 355,812 342,771 329,729 322,681 322,627 L 322,0 142,0 142,851 C 142,874 142,898 142,923 141,948 141,971 140,994 139,1016 139,1035 138,1051 137,1067 137,1077 136,1082 L 306,1082 C 307,1079 307,1070 308,1055 309,1040 310,1024 311,1005 312,986 312,966 313,947 314,927 314,910 314,897 L 317,897 C 334,928 353,957 374,982 395,1007 419,1029 446,1047 473,1064 505,1078 540,1088 575,1097 616,1102 663,1102 723,1102 775,1095 818,1080 861,1065 897,1043 925,1012 953,981 974,942 987,894 1000,845 1006,788 1006,721 L 1006,0 825,0 Z"/>
   <glyph unicode="m" horiz-adv-x="1457" d="M 768,0 L 768,686 C 768,739 765,783 758,818 751,853 740,882 725,904 709,925 688,941 663,950 638,959 607,963 570,963 532,963 498,956 467,941 436,926 410,904 389,876 367,847 350,812 339,771 327,729 321,681 321,627 L 321,0 142,0 142,851 C 142,874 142,898 142,923 141,948 141,971 140,994 139,1016 139,1035 138,1051 137,1067 137,1077 136,1082 L 306,1082 C 307,1079 307,1070 308,1055 309,1040 310,1024 311,1005 312,986 312,966 313,947 314,927 314,910 314,897 L 317,897 C 333,928 350,957 369,982 388,1007 410,1029 435,1047 460,1064 488,1078 521,1088 553,1097 590,1102 633,1102 715,1102 780,1086 828,1053 875,1020 908,968 927,897 L 930,897 C 946,928 964,957 984,982 1004,1007 1027,1029 1054,1047 1081,1064 1111,1078 1144,1088 1177,1097 1215,1102 1258,1102 1313,1102 1360,1095 1400,1080 1439,1065 1472,1043 1497,1012 1522,981 1541,942 1553,894 1565,845 1571,788 1571,721 L 1571,0 1393,0 1393,686 C 1393,739 1390,783 1383,818 1376,853 1365,882 1350,904 1334,925 1313,941 1288,950 1263,959 1232,963 1195,963 1157,963 1123,956 1092,942 1061,927 1035,906 1014,878 992,850 975,815 964,773 952,731 946,682 946,627 L 946,0 768,0 Z"/>
   <glyph unicode="l" horiz-adv-x="187" d="M 138,0 L 138,1484 318,1484 318,0 138,0 Z"/>
   <glyph unicode="k" horiz-adv-x="901" d="M 816,0 L 450,494 318,385 318,0 138,0 138,1484 318,1484 318,557 793,1082 1004,1082 565,617 1027,0 816,0 Z"/>
   <glyph unicode="i" horiz-adv-x="187" d="M 137,1312 L 137,1484 317,1484 317,1312 137,1312 Z M 137,0 L 137,1082 317,1082 317,0 137,0 Z"/>
   <glyph unicode="h" horiz-adv-x="874" d="M 317,897 C 337,934 359,965 382,991 405,1016 431,1037 459,1054 487,1071 518,1083 551,1091 584,1098 622,1102 663,1102 732,1102 789,1093 834,1074 878,1055 913,1029 939,996 964,962 982,922 992,875 1001,828 1006,777 1006,721 L 1006,0 825,0 825,686 C 825,732 822,772 817,807 811,842 800,871 784,894 768,917 745,934 716,946 687,957 649,963 602,963 559,963 521,955 487,940 452,925 423,903 399,875 374,847 355,813 342,773 329,733 322,688 322,638 L 322,0 142,0 142,1484 322,1484 322,1098 C 322,1076 322,1054 321,1032 320,1010 320,990 319,971 318,952 317,937 316,924 315,911 315,902 314,897 L 317,897 Z"/>
   <glyph unicode="g" horiz-adv-x="927" d="M 548,-425 C 486,-425 431,-419 383,-406 335,-393 294,-375 260,-352 226,-328 198,-300 177,-267 156,-234 140,-198 131,-158 L 312,-132 C 324,-182 351,-220 392,-248 433,-274 486,-288 553,-288 594,-288 631,-282 664,-271 697,-260 726,-241 749,-217 772,-191 790,-159 803,-119 816,-79 822,-30 822,27 L 822,201 820,201 C 807,174 790,148 771,123 751,98 727,75 699,56 670,37 637,21 600,10 563,-2 520,-8 472,-8 403,-8 345,4 296,27 247,50 207,84 176,130 145,176 122,233 108,302 93,370 86,449 86,539 86,626 93,704 108,773 122,842 145,901 178,950 210,998 252,1035 304,1061 355,1086 418,1099 492,1099 569,1099 635,1082 692,1047 748,1012 791,962 822,897 L 824,897 C 824,914 825,932 826,953 827,974 828,993 829,1012 830,1030 831,1046 832,1059 833,1072 835,1080 836,1082 L 1007,1082 C 1006,1076 1006,1066 1005,1052 1004,1037 1004,1020 1003,1000 1002,980 1002,958 1002,934 1001,909 1001,884 1001,858 L 1001,31 C 1001,-120 964,-234 890,-311 815,-387 701,-425 548,-425 Z M 822,541 C 822,616 814,681 798,735 781,788 760,832 733,866 706,900 676,925 642,941 607,957 572,965 536,965 490,965 451,957 418,941 385,925 357,900 336,866 314,831 298,787 288,734 277,680 272,616 272,541 272,463 277,398 288,345 298,292 314,249 335,216 356,183 383,160 416,146 449,132 488,125 533,125 569,125 604,133 639,148 673,163 704,188 731,221 758,254 780,297 797,350 814,403 822,466 822,541 Z"/>
   <glyph unicode="f" horiz-adv-x="557" d="M 361,951 L 361,0 181,0 181,951 29,951 29,1082 181,1082 181,1204 C 181,1243 185,1280 192,1314 199,1347 213,1377 233,1402 252,1427 279,1446 313,1461 347,1475 391,1482 445,1482 466,1482 489,1481 512,1479 535,1477 555,1474 572,1470 L 572,1333 C 561,1335 548,1337 533,1339 518,1340 504,1341 492,1341 465,1341 444,1337 427,1330 410,1323 396,1312 387,1299 377,1285 370,1268 367,1248 363,1228 361,1205 361,1179 L 361,1082 572,1082 572,951 361,951 Z"/>
   <glyph unicode="e" horiz-adv-x="980" d="M 276,503 C 276,446 282,394 294,347 305,299 323,258 348,224 372,189 403,163 441,144 479,125 525,115 578,115 656,115 719,131 766,162 813,193 844,233 861,281 L 1019,236 C 1008,206 992,176 972,146 951,115 924,88 890,64 856,39 814,19 763,4 712,-12 650,-20 578,-20 418,-20 296,28 213,123 129,218 87,360 87,548 87,649 100,735 125,806 150,876 185,933 229,977 273,1021 324,1053 383,1073 442,1092 504,1102 571,1102 662,1102 738,1087 799,1058 860,1029 909,988 946,937 983,885 1009,824 1025,754 1040,684 1048,608 1048,527 L 1048,503 276,503 Z M 862,641 C 852,755 823,838 775,891 727,943 658,969 568,969 538,969 507,964 474,955 441,945 410,928 382,903 354,878 330,845 311,803 292,760 281,706 278,641 L 862,641 Z"/>
   <glyph unicode="d" horiz-adv-x="927" d="M 821,174 C 788,105 744,55 689,25 634,-5 565,-20 484,-20 347,-20 247,26 183,118 118,210 86,349 86,536 86,913 219,1102 484,1102 566,1102 634,1087 689,1057 744,1027 788,979 821,914 L 823,914 C 823,921 823,931 823,946 822,960 822,975 822,991 821,1006 821,1021 821,1035 821,1049 821,1059 821,1065 L 821,1484 1001,1484 1001,223 C 1001,197 1001,172 1002,148 1002,124 1002,102 1003,82 1004,62 1004,45 1005,31 1006,16 1006,6 1007,0 L 835,0 C 834,7 833,16 832,29 831,41 830,55 829,71 828,87 827,104 826,122 825,139 825,157 825,174 L 821,174 Z M 275,542 C 275,467 280,403 289,350 298,297 313,253 334,219 355,184 381,159 413,143 445,127 484,119 530,119 577,119 619,127 656,142 692,157 722,182 747,217 771,251 789,296 802,351 815,406 821,474 821,554 821,631 815,696 802,749 789,802 771,844 746,877 721,910 691,933 656,948 620,962 579,969 532,969 488,969 450,961 418,946 386,931 359,906 338,872 317,838 301,794 291,740 280,685 275,619 275,542 Z"/>
   <glyph unicode="c" horiz-adv-x="901" d="M 275,546 C 275,484 280,427 289,375 298,323 313,278 334,241 355,203 384,174 419,153 454,132 497,122 548,122 612,122 666,139 709,173 752,206 778,258 788,328 L 970,328 C 964,283 951,239 931,197 911,155 884,118 850,86 815,54 773,28 724,9 675,-10 618,-20 553,-20 468,-20 396,-6 337,23 278,52 230,91 193,142 156,192 129,251 112,320 95,388 87,462 87,542 87,615 93,679 105,735 117,790 134,839 156,881 177,922 203,957 232,986 261,1014 293,1037 328,1054 362,1071 398,1083 436,1091 474,1098 512,1102 551,1102 612,1102 666,1094 713,1077 760,1060 801,1038 836,1009 870,980 898,945 919,906 940,867 955,824 964,779 L 779,765 C 770,825 746,873 708,908 670,943 616,961 546,961 495,961 452,953 418,936 383,919 355,893 334,859 313,824 298,781 289,729 280,677 275,616 275,546 Z"/>
   <glyph unicode="b" horiz-adv-x="953" d="M 1053,546 C 1053,169 920,-20 655,-20 573,-20 505,-5 451,25 396,54 352,102 318,168 L 316,168 C 316,151 316,133 315,114 314,95 313,78 312,62 311,46 310,32 309,21 308,10 307,3 306,0 L 132,0 C 133,6 133,16 134,31 135,45 135,62 136,82 137,102 137,124 138,148 138,172 138,197 138,223 L 138,1484 318,1484 318,1061 C 318,1041 318,1022 318,1004 317,985 317,969 316,955 315,938 315,923 314,908 L 318,908 C 351,977 396,1027 451,1057 506,1087 574,1102 655,1102 792,1102 892,1056 957,964 1021,872 1053,733 1053,546 Z M 864,540 C 864,615 859,679 850,732 841,785 826,829 805,864 784,898 758,923 726,939 694,955 655,963 609,963 562,963 520,955 484,940 447,925 417,900 393,866 368,832 350,787 337,732 324,677 318,609 318,529 318,452 324,387 337,334 350,281 368,239 393,206 417,173 447,149 483,135 519,120 560,113 607,113 651,113 689,121 721,136 753,151 780,176 801,210 822,244 838,288 849,343 859,397 864,463 864,540 Z"/>
   <glyph unicode="a" horiz-adv-x="1060" d="M 414,-20 C 305,-20 224,9 169,66 114,124 87,203 87,303 87,375 101,434 128,480 155,526 190,562 234,588 277,614 327,632 383,642 439,652 496,657 554,657 L 797,657 797,717 C 797,762 792,800 783,832 774,863 759,889 740,908 721,928 697,942 668,951 639,960 604,965 565,965 530,965 499,963 471,958 443,953 419,944 398,931 377,918 361,900 348,878 335,855 327,827 323,793 L 135,810 C 142,853 154,892 173,928 192,963 218,994 253,1020 287,1046 330,1066 382,1081 433,1095 496,1102 569,1102 705,1102 807,1071 876,1009 945,946 979,856 979,738 L 979,272 C 979,219 986,179 1000,152 1014,125 1041,111 1080,111 1090,111 1100,112 1110,113 1120,114 1130,116 1139,118 L 1139,6 C 1116,1 1094,-3 1072,-6 1049,-9 1025,-10 1000,-10 966,-10 937,-5 913,4 888,13 868,26 853,45 838,63 826,86 818,113 810,140 805,171 803,207 L 797,207 C 778,172 757,141 734,113 711,85 684,61 653,42 622,22 588,7 549,-4 510,-15 465,-20 414,-20 Z M 455,115 C 512,115 563,125 606,146 649,167 684,194 713,226 741,259 762,294 776,332 790,371 797,408 797,443 L 797,531 600,531 C 556,531 514,528 475,522 435,517 400,506 370,489 340,472 316,449 299,418 281,388 272,349 272,300 272,241 288,195 320,163 351,131 396,115 455,115 Z"/>
   <glyph unicode="S" horiz-adv-x="1192" d="M 1272,389 C 1272,330 1261,275 1238,225 1215,175 1179,132 1131,96 1083,59 1023,31 950,11 877,-10 790,-20 690,-20 515,-20 378,11 280,72 182,133 120,222 93,338 L 278,375 C 287,338 302,305 321,275 340,245 367,219 400,198 433,176 473,159 522,147 571,135 629,129 697,129 754,129 806,134 853,144 900,153 941,168 975,188 1009,208 1036,234 1055,266 1074,297 1083,335 1083,379 1083,425 1073,462 1052,491 1031,520 1001,543 963,562 925,581 880,596 827,609 774,622 716,635 652,650 613,659 573,668 534,679 494,689 456,701 420,716 383,730 349,747 317,766 285,785 257,809 234,836 211,863 192,894 179,930 166,965 159,1006 159,1053 159,1120 173,1177 200,1225 227,1272 264,1311 312,1342 360,1373 417,1395 482,1409 547,1423 618,1430 694,1430 781,1430 856,1423 918,1410 980,1396 1032,1375 1075,1348 1118,1321 1152,1287 1178,1247 1203,1206 1224,1159 1239,1106 L 1051,1073 C 1042,1107 1028,1137 1011,1164 993,1191 970,1213 941,1231 912,1249 878,1263 837,1272 796,1281 747,1286 692,1286 627,1286 572,1280 528,1269 483,1257 448,1241 421,1221 394,1201 374,1178 363,1151 351,1124 345,1094 345,1063 345,1021 356,987 377,960 398,933 426,910 462,892 498,874 540,859 587,847 634,835 685,823 738,811 781,801 825,791 868,781 911,770 952,758 991,744 1030,729 1067,712 1102,693 1136,674 1166,650 1191,622 1216,594 1236,561 1251,523 1265,485 1272,440 1272,389 Z"/>
   <glyph unicode="O" horiz-adv-x="1430" d="M 1495,711 C 1495,601 1479,501 1448,411 1416,321 1370,244 1310,180 1250,116 1177,67 1090,32 1003,-3 905,-20 795,-20 679,-20 577,-2 490,35 403,71 330,122 272,187 214,252 170,329 141,418 112,507 97,605 97,711 97,821 112,920 143,1009 174,1098 219,1173 278,1236 337,1298 411,1346 498,1380 585,1413 684,1430 797,1430 909,1430 1009,1413 1096,1379 1183,1345 1256,1297 1315,1234 1374,1171 1418,1096 1449,1007 1480,918 1495,820 1495,711 Z M 1300,711 C 1300,796 1289,873 1268,942 1246,1011 1214,1071 1172,1120 1129,1169 1077,1207 1014,1234 951,1261 879,1274 797,1274 713,1274 639,1261 576,1234 513,1207 460,1169 418,1120 375,1071 344,1011 323,942 302,873 291,796 291,711 291,626 302,549 324,479 345,408 377,348 420,297 462,246 515,206 578,178 641,149 713,135 795,135 883,135 959,149 1023,178 1086,207 1139,247 1180,298 1221,349 1251,409 1271,480 1290,551 1300,628 1300,711 Z"/>
   <glyph unicode="C" horiz-adv-x="1324" d="M 792,1274 C 712,1274 641,1261 580,1234 518,1207 466,1169 425,1120 383,1071 351,1011 330,942 309,873 298,796 298,711 298,626 310,549 333,479 356,408 389,348 432,297 475,246 527,207 590,179 652,151 722,137 800,137 855,137 905,144 950,159 995,173 1035,193 1072,219 1108,245 1140,276 1169,312 1198,347 1223,387 1245,430 L 1401,352 C 1376,299 1344,250 1307,205 1270,160 1226,120 1176,87 1125,54 1068,28 1005,9 941,-10 870,-20 791,-20 677,-20 577,-2 492,35 406,71 334,122 277,187 219,252 176,329 147,418 118,507 104,605 104,711 104,821 119,920 150,1009 180,1098 224,1173 283,1236 341,1298 413,1346 498,1380 583,1413 681,1430 790,1430 940,1430 1065,1401 1166,1342 1267,1283 1341,1196 1388,1081 L 1207,1021 C 1194,1054 1176,1086 1153,1117 1130,1147 1102,1174 1068,1197 1034,1220 994,1239 949,1253 903,1267 851,1274 792,1274 Z"/>
   <glyph unicode=" " horiz-adv-x="556"/>
  </font>
 </defs>
 <defs class="TextShapeIndex">
  <g ooo:slide="id1" ooo:id-list="id3 id4 id5 id6 id7 id8 id9 id10 id11 id12 id13 id14 id15 id16 id17 id18 id19"/>
 </defs>
 <defs class="EmbeddedBulletChars">
  <g id="bullet-char-template(57356)" transform="scale(0.00048828125,-0.00048828125)">
   <path d="M 580,1141 L 1163,571 580,0 -4,571 580,1141 Z"/>
  </g>
  <g id="bullet-char-template(57354)" transform="scale(0.00048828125,-0.00048828125)">
   <path d="M 8,1128 L 1137,1128 1137,0 8,0 8,1128 Z"/>
  </g>
  <g id="bullet-char-template(10146)" transform="scale(0.00048828125,-0.00048828125)">
   <path d="M 174,0 L 602,739 174,1481 1456,739 174,0 Z M 1358,739 L 309,1346 659,739 1358,739 Z"/>
  </g>
  <g id="bullet-char-template(10132)" transform="scale(0.00048828125,-0.00048828125)">
   <path d="M 2015,739 L 1276,0 717,0 1260,543 174,543 174,936 1260,936 717,1481 1274,1481 2015,739 Z"/>
  </g>
  <g id="bullet-char-template(10007)" transform="scale(0.00048828125,-0.00048828125)">
   <path d="M 0,-2 C -7,14 -16,27 -25,37 L 356,567 C 262,823 215,952 215,954 215,979 228,992 255,992 264,992 276,990 289,987 310,991 331,999 354,1012 L 381,999 492,748 772,1049 836,1024 860,1049 C 881,1039 901,1025 922,1006 886,937 835,863 770,784 769,783 710,716 594,584 L 774,223 C 774,196 753,168 711,139 L 727,119 C 717,90 699,76 672,76 641,76 570,178 457,381 L 164,-76 C 142,-110 111,-127 72,-127 30,-127 9,-110 8,-76 1,-67 -2,-52 -2,-32 -2,-23 -1,-13 0,-2 Z"/>
  </g>
  <g id="bullet-char-template(10004)" transform="scale(0.00048828125,-0.00048828125)">
   <path d="M 285,-33 C 182,-33 111,30 74,156 52,228 41,333 41,471 41,549 55,616 82,672 116,743 169,778 240,778 293,778 328,747 346,684 L 369,508 C 377,444 397,411 428,410 L 1163,1116 C 1174,1127 1196,1133 1229,1133 1271,1133 1292,1118 1292,1087 L 1292,965 C 1292,929 1282,901 1262,881 L 442,47 C 390,-6 338,-33 285,-33 Z"/>
  </g>
  <g id="bullet-char-template(9679)" transform="scale(0.00048828125,-0.00048828125)">
   <path d="M 813,0 C 632,0 489,54 383,161 276,268 223,411 223,592 223,773 276,916 383,1023 489,1130 632,1184 813,1184 992,1184 1136,1130 1245,1023 1353,916 1407,772 1407,592 1407,412 1353,268 1245,161 1136,54 992,0 813,0 Z"/>
  </g>
  <g id="bullet-char-template(8226)" transform="scale(0.00048828125,-0.00048828125)">
   <path d="M 346,457 C 273,457 209,483 155,535 101,586 74,649 74,723 74,796 101,859 155,911 209,963 273,989 346,989 419,989 480,963 531,910 582,859 608,796 608,723 608,648 583,586 532,535 482,483 420,457 346,457 Z"/>
  </g>
  <g id="bullet-char-template(8211)" transform="scale(0.00048828125,-0.00048828125)">
   <path d="M -4,459 L 1135,459 1135,606 -4,606 -4,459 Z"/>
  </g>
 </defs>
 <defs class="TextEmbeddedBitmaps"/>
 <g>
  <g id="id2" class="Master_Slide">
   <g id="bg-id2" class="Background"/>
   <g id="bo-id2" class="BackgroundObjects"/>
  </g>
 </g>
 <g class="SlideGroup">
  <g>
   <g id="id1" class="Slide" clip-path="url(#presentation_clip_path)">
    <g class="Page">
     <g class="com.sun.star.drawing.RectangleShape">
      <g id="id3">
       <path fill="none" stroke="rgb(0,0,0)" d="M 8450,7000 L 7000,7000 7000,3000 9900,3000 9900,7000 8450,7000 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.RectangleShape">
      <g id="id4">
       <path fill="none" stroke="rgb(0,0,0)" d="M 12050,7000 L 10100,7000 10100,3000 14000,3000 14000,7000 12050,7000 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.EllipseShape">
      <g id="id5">
       <path fill="none" stroke="rgb(0,0,0)" d="M 8999,3300 C 9105,3312 9195,3383 9288,3531 9387,3690 9459,3889 9516,4165 9574,4441 9600,4712 9600,5031 9600,5350 9574,5621 9516,5897 9459,6173 9402,6328 9319,6478"/>
       <path fill="rgb(0,0,0)" stroke="none" d="M 9141,6701 L 9405,6526 9248,6403 9141,6701 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.EllipseShape">
      <g id="id6">
       <path fill="none" stroke="rgb(0,0,0)" d="M 10794,6710 C 10748,6669 10710,6618 10668,6540 10583,6380 10521,6181 10472,5905 10422,5628 10400,5356 10400,5036 10400,4716 10422,4444 10472,4168 10521,3891 10583,3692 10668,3532 10748,3383 10690,3492 10701,3474"/>
       <path fill="rgb(0,0,0)" stroke="none" d="M 10916,3300 L 10622,3416 10750,3569 10916,3300 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id7">
       <path fill="none" stroke="rgb(0,0,0)" d="M 9899,6900 C 9955,6900 9999,6943 9999,6999 9999,7055 9955,7099 9899,7099 9843,7099 9800,7055 9800,6999 9800,6943 9843,6900 9899,6900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9800,6900 L 9800,6900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10001,7101 L 10001,7101 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id8">
       <path fill="none" stroke="rgb(0,0,0)" d="M 10099,2900 C 10155,2900 10199,2943 10199,2999 10199,3055 10155,3099 10099,3099 10043,3099 10000,3055 10000,2999 10000,2943 10043,2900 10099,2900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10000,2900 L 10000,2900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10201,3101 L 10201,3101 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id9">
       <path fill="none" stroke="rgb(0,0,0)" d="M 9899,2900 C 9955,2900 9999,2943 9999,2999 9999,3055 9955,3099 9899,3099 9843,3099 9800,3055 9800,2999 9800,2943 9843,2900 9899,2900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9800,2900 L 9800,2900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10001,3101 L 10001,3101 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id10">
       <path fill="none" stroke="rgb(0,0,0)" d="M 9899,4900 C 9955,4900 9999,4943 9999,4999 9999,5055 9955,5099 9899,5099 9843,5099 9800,5055 9800,4999 9800,4943 9843,4900 9899,4900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9800,4900 L 9800,4900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10001,5101 L 10001,5101 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id11">
       <path fill="none" stroke="rgb(0,0,0)" d="M 10099,4900 C 10155,4900 10199,4943 10199,4999 10199,5055 10155,5099 10099,5099 10043,5099 10000,5055 10000,4999 10000,4943 10043,4900 10099,4900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10000,4900 L 10000,4900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10201,5101 L 10201,5101 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id12">
       <path fill="none" stroke="rgb(0,0,0)" d="M 10099,6900 C 10155,6900 10199,6943 10199,6999 10199,7055 10155,7099 10099,7099 10043,7099 10000,7055 10000,6999 10000,6943 10043,6900 10099,6900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10000,6900 L 10000,6900 Z"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10201,7101 L 10201,7101 Z"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id13">
       <path fill="rgb(255,255,255)" stroke="none" d="M 5100,5200 L 3500,5200 3500,3000 6700,3000 6700,5200 5100,5200 Z"/>
       <text class="TextShape"><tspan class="TextParagraph" font-family="Liberation Sans, sans-serif" font-size="353px" font-weight="400"><tspan class="TextPosition" x="3750" y="3433"><tspan fill="rgb(0,0,0)" stroke="none">Subdivision </tspan></tspan><tspan class="TextPosition" x="3750" y="3827"><tspan fill="rgb(0,0,0)" stroke="none">points need to </tspan></tspan><tspan class="TextPosition" x="3750" y="4221"><tspan fill="rgb(0,0,0)" stroke="none">land on the same </tspan></tspan><tspan class="TextPosition" x="3750" y="4615"><tspan fill="rgb(0,0,0)" stroke="none">location to </tspan></tspan><tspan class="TextPosition" x="3750" y="5009"><tspan fill="rgb(0,0,0)" stroke="none">prevent cracking</tspan></tspan></tspan></text>
      </g>
     </g>
     <g class="com.sun.star.drawing.LineShape">
      <g id="id14">
       <path fill="none" stroke="rgb(0,0,0)" d="M 6600,4200 L 6651,4213"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 6703,4227 L 6754,4240"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 6806,4253 L 6857,4266"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 6908,4280 L 6960,4293"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7011,4306 L 7063,4319"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7114,4333 L 7165,4346"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7217,4359 L 7268,4372"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7320,4386 L 7371,4399"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7423,4412 L 7474,4426"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7525,4439 L 7577,4452"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7628,4465 L 7680,4479"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7731,4492 L 7782,4505"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7834,4518 L 7885,4532"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 7937,4545 L 7988,4558"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8039,4571 L 8091,4585"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8142,4598 L 8194,4611"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8245,4625 L 8296,4638"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8348,4651 L 8399,4664"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8451,4678 L 8502,4691"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8553,4704 L 8605,4717"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8656,4731 L 8708,4744"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8759,4757 L 8811,4770"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8862,4784 L 8913,4797"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 8965,4810 L 9016,4824"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9068,4837 L 9119,4850"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9170,4863 L 9222,4877"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9273,4890 L 9325,4903"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9376,4916 L 9427,4930"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9479,4943 L 9530,4956"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9582,4969 L 9633,4983"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 9684,4996 L 9700,5000"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id15">
       <path fill="rgb(255,255,255)" stroke="none" d="M 12650,2700 L 10700,2700 10700,1500 14600,1500 14600,2700 12650,2700 Z"/>
       <text class="TextShape"><tspan class="TextParagraph" font-family="Liberation Sans, sans-serif" font-size="353px" font-weight="400"><tspan class="TextPosition" x="10950" y="2024"><tspan fill="rgb(0,0,0)" stroke="none">Corner points start </tspan></tspan><tspan class="TextPosition" x="10950" y="2418"><tspan fill="rgb(0,0,0)" stroke="none">with the same values</tspan></tspan></tspan></text>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id16">
       <path fill="rgb(255,255,255)" stroke="none" d="M 12850,8500 L 10900,8500 10900,7300 14800,7300 14800,8500 12850,8500 Z"/>
       <text class="TextShape"><tspan class="TextParagraph" font-family="Liberation Sans, sans-serif" font-size="353px" font-weight="400"><tspan class="TextPosition" x="11150" y="7824"><tspan fill="rgb(0,0,0)" stroke="none">Corner points start </tspan></tspan><tspan class="TextPosition" x="11150" y="8218"><tspan fill="rgb(0,0,0)" stroke="none">with the same values</tspan></tspan></tspan></text>
      </g>
     </g>
     <g class="com.sun.star.drawing.CustomShape">
      <g id="id17">
       <path fill="rgb(255,255,255)" stroke="none" d="M 12400,5200 L 10900,5200 10900,3300 13900,3300 13900,5200 12400,5200 Z"/>
       <text class="TextShape"><tspan class="TextParagraph" font-family="Liberation Sans, sans-serif" font-size="353px" font-weight="400"><tspan class="TextPosition" x="11150" y="3780"><tspan fill="rgb(0,0,0)" stroke="none">Opposing </tspan></tspan><tspan class="TextPosition" x="11150" y="4174"><tspan fill="rgb(0,0,0)" stroke="none">directions of </tspan></tspan><tspan class="TextPosition" x="11150" y="4568"><tspan fill="rgb(0,0,0)" stroke="none">edge walking </tspan></tspan><tspan class="TextPosition" x="11150" y="4962"><tspan fill="rgb(0,0,0)" stroke="none">for subdivision</tspan></tspan></tspan></text>
      </g>
     </g>
     <g class="com.sun.star.drawing.LineShape">
      <g id="id18">
       <path fill="none" stroke="rgb(0,0,0)" d="M 10300,2900 L 10341,2867"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10383,2834 L 10424,2801"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10466,2767 L 10507,2734"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10549,2701 L 10590,2668"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10632,2635 L 10673,2602"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10715,2568 L 10756,2535"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10797,2502 L 10800,2500"/>
      </g>
     </g>
     <g class="com.sun.star.drawing.LineShape">
      <g id="id19">
       <path fill="none" stroke="rgb(0,0,0)" d="M 10200,7100 L 10243,7131"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10286,7162 L 10330,7193"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10373,7223 L 10416,7254"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10459,7285 L 10502,7316"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10546,7347 L 10589,7378"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10632,7409 L 10675,7439"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10718,7470 L 10762,7501"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10805,7532 L 10848,7563"/>
       <path fill="none" stroke="rgb(0,0,0)" d="M 10891,7594 L 10900,7600"/>
      </g>
     </g>
    </g>
   </g>
  </g>
 </g>
</svg>" alt="precise">
8541</div>
8542</div>
8543<div class="paragraph">
8544<p>As stated in &#8220;<a href="#evaluation-of-expressions">Evaluation of Expressions</a>&#8221;,
8545the compiler may transform expressions even if this changes the resulting
8546value.
8547Without any qualifiers, implementations are permitted to perform
8548optimizations that effectively modify the order or number of operations used
8549to evaluate an expression, even if those optimizations may produce slightly
8550different results relative to unoptimized code.</p>
8551</div>
8552<div class="paragraph">
8553<p>The <strong>precise</strong> qualifier ensures that operations contributing to a
8554variable&#8217;s value are done in their stated order and with operator consistency.
8555The order is determined by operator precedence and parenthesis, as described in
8556&#8220;<a href="#operators">Operators</a>&#8221;.
8557Operator consistency means for each particular operator, for example the
8558multiply operator (<strong>*</strong>), its operation is always computed with the same
8559precision.
8560Specifically, values computed by compiler-generated code must adhere to the
8561following identities:</p>
8562</div>
8563<div class="olist arabic">
8564<ol class="arabic">
8565<li>
8566<p>a + b = b + a</p>
8567</li>
8568<li>
8569<p>a * b = b * a</p>
8570</li>
8571<li>
8572<p>a * b + c * d = b * a + c* d = d * c + b * a = &lt;any other mathematically
8573valid combination&gt;</p>
8574</li>
8575</ol>
8576</div>
8577<div class="paragraph">
8578<p>While the following are prevented:</p>
8579</div>
8580<div class="olist arabic">
8581<ol class="arabic">
8582<li>
8583<p>a + (b + c) is not allowed to become (a + b) + c</p>
8584</li>
8585<li>
8586<p>a * (b * c) is not allowed to become (a * b) * c</p>
8587</li>
8588<li>
8589<p>a * b + c is not allowed to become a single operation <strong>fma</strong>(a, b, c)</p>
8590</li>
8591</ol>
8592</div>
8593<div class="paragraph">
8594<p>Where <em>a</em>, <em>b</em>, <em>c</em>, and <em>d</em>, are scalars or vectors, not matrices.
8595(Matrix multiplication generally does not commute.) It is the shader
8596writer&#8217;s responsibility to express the computation in terms of these rules
8597and the compiler&#8217;s responsibility to follow these rules.
8598See the description of <em>gl_TessCoord</em> for the rules the tessellation stages
8599are responsible for following, which in conjunction with the above allow
8600avoiding cracking when subdividing.</p>
8601</div>
8602<div class="paragraph">
8603<p>For example,</p>
8604</div>
8605<div class="listingblock">
8606<div class="content">
8607<pre class="CodeRay highlight"><code data-lang="c++">precise out vec4 position;</code></pre>
8608</div>
8609</div>
8610<div class="paragraph">
8611<p>declares that operations used to produce the value of <em>position</em> must be
8612performed in exactly the order specified in the source code and with all
8613operators being treated consistently.
8614As with the <strong>invariant</strong> qualifier (see &#8220;<a href="#the-invariant-qualifier">The
8615Invariant Qualifier</a>&#8221;), the <strong>precise</strong> qualifier may be used to qualify a
8616built-in or previously declared user-defined variable as being precise:</p>
8617</div>
8618<div class="listingblock">
8619<div class="content">
8620<pre class="CodeRay highlight"><code data-lang="c++">out vec3 Color;
8621precise Color; <span class="comment">// make existing Color be precise</span></code></pre>
8622</div>
8623</div>
8624<div class="paragraph">
8625<p>When applied to a block, a structure type, or a variable of structure type,
8626<strong>precise</strong> applies to each contained member, recursively.</p>
8627</div>
8628<div class="paragraph">
8629<p>This qualifier will affect the evaluation of an r-value in a particular
8630function if and only if the result is eventually consumed in the same
8631function by an l-value qualified as <strong>precise</strong>.
8632Any other expressions within a function are not affected, including return
8633values and output parameters not declared as <strong>precise</strong> but that are
8634eventually consumed outside the function by a variable qualified as
8635<strong>precise</strong>. Unaffected expressions also include the controlling expressions
8636in selection and iteration statements and the condition in ternary
8637operators (<strong>?:</strong>).</p>
8638</div>
8639<div class="paragraph">
8640<p>Some examples of the use of <strong>precise</strong>:</p>
8641</div>
8642<div class="listingblock">
8643<div class="content">
8644<pre class="CodeRay highlight"><code data-lang="c++">in vec4 a, b, c, d;
8645precise out vec4 v;
8646
8647<span class="predefined-type">float</span> func(<span class="predefined-type">float</span> e, <span class="predefined-type">float</span> f, <span class="predefined-type">float</span> g, <span class="predefined-type">float</span> h)
8648{
8649    <span class="keyword">return</span> (e*f) + (g*h); <span class="comment">// no constraint on order or operator consistency</span>
8650}
8651
8652<span class="predefined-type">float</span> func2(<span class="predefined-type">float</span> e, <span class="predefined-type">float</span> f, <span class="predefined-type">float</span> g, <span class="predefined-type">float</span> h)
8653{
8654    precise <span class="predefined-type">float</span> result = (e*f) + (g*h); <span class="comment">// ensures same precision for the two multiplies</span>
8655    <span class="keyword">return</span> result;
8656}
8657
8658<span class="predefined-type">float</span> func3(<span class="predefined-type">float</span> i, <span class="predefined-type">float</span> j, precise out <span class="predefined-type">float</span> k)
8659{
8660    k = i * i + j;        <span class="comment">// precise, due to &lt;k&gt; declaration</span>
8661}
8662
8663<span class="directive">void</span> main()
8664{
8665    vec3 r = vec3(a * b);             <span class="comment">// precise, used to compute v.xyz</span>
8666    vec3 s = vec3(c * d);             <span class="comment">// precise, used to compute v.xyz</span>
8667    v.xyz = r + s;                    <span class="comment">// precise</span>
8668    v.w = (a.w * b.w) + (c.w * d.w);  <span class="comment">// precise</span>
8669    v.x = func(a.x, b.x, c.x, d.x);   <span class="comment">// values computed in func() are NOT precise</span>
8670    v.x = func2(a.x, b.x, c.x, d.x);  <span class="comment">// precise!</span>
8671    func3(a.x * b.x, c.x * d.x, v.x); <span class="comment">// precise!</span>
8672}</code></pre>
8673</div>
8674</div>
8675<div class="paragraph">
8676<p>For the purposes of determining if an output from one shader stage matches
8677an input of the next stage, the <strong>precise</strong> qualifier need not match between
8678the input and the output.</p>
8679</div>
8680<div class="paragraph">
8681<p>All constant expressions are evaluated as if <strong>precise</strong> was present, whether
8682or not it is present.
8683However, as described in &#8220;<a href="#constant-expressions">Constant Expressions</a>&#8221;,
8684there is no requirement that a compile-time constant expression evaluates to
8685the same value as a corresponding non-constant expression.</p>
8686</div>
8687</div>
8688<div class="sect2">
8689<h3 id="memory-qualifiers">4.10. Memory Qualifiers</h3>
8690<div class="paragraph">
8691<p>Shader storage blocks, variables declared within shader storage blocks and
8692variables declared as image types (the basic opaque types with &#8220;<strong>image</strong>&#8221;
8693in their keyword), can be further qualified with one or more of the
8694following memory qualifiers:</p>
8695</div>
8696<table class="tableblock frame-all grid-all stretch">
8697<colgroup>
8698<col style="width: 50%;">
8699<col style="width: 50%;">
8700</colgroup>
8701<thead>
8702<tr>
8703<th class="tableblock halign-left valign-top">Qualifier</th>
8704<th class="tableblock halign-left valign-top">Meaning</th>
8705</tr>
8706</thead>
8707<tbody>
8708<tr>
8709<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>coherent</strong></p></td>
8710<td class="tableblock halign-left valign-top"><p class="tableblock">memory variable where reads and writes are coherent with
8711                reads and writes from other shader invocations</p></td>
8712</tr>
8713<tr>
8714<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>volatile</strong></p></td>
8715<td class="tableblock halign-left valign-top"><p class="tableblock">memory variable whose underlying value may be changed at any
8716                point during shader execution by some source other than the
8717                current shader invocation</p></td>
8718</tr>
8719<tr>
8720<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>restrict</strong></p></td>
8721<td class="tableblock halign-left valign-top"><p class="tableblock">memory variable where use of that variable is the only way
8722                to read and write the underlying memory in the relevant
8723                shader stage</p></td>
8724</tr>
8725<tr>
8726<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>readonly</strong></p></td>
8727<td class="tableblock halign-left valign-top"><p class="tableblock">memory variable that can be used to read the underlying
8728                memory, but cannot be used to write the underlying memory</p></td>
8729</tr>
8730<tr>
8731<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>writeonly</strong></p></td>
8732<td class="tableblock halign-left valign-top"><p class="tableblock">memory variable that can be used to write the underlying
8733                memory, but cannot be used to read the underlying memory</p></td>
8734</tr>
8735</tbody>
8736</table>
8737<div class="paragraph">
8738<p>Memory accesses to image variables declared using the <strong>coherent</strong> qualifier
8739are performed coherently with accesses to the same location from other
8740shader invocations.</p>
8741</div>
8742<div class="paragraph">
8743<p>As described in section
87447.11
8745&#8220;Shader Memory Access&#8221; of the
8746<a href="#references">OpenGL ES Specification</a>, shader memory reads and writes complete in a
8747largely undefined order.
8748The built-in function <strong>memoryBarrier</strong>() can be used if needed to guarantee
8749the completion and relative ordering of memory accesses performed by a
8750single shader invocation.</p>
8751</div>
8752<div class="paragraph">
8753<p>When accessing memory using variables not declared as <strong>coherent</strong>, the memory
8754accessed by a shader may be cached by the implementation to service future
8755accesses to the same address.
8756Memory stores may be cached in such a way that the values written may not be
8757visible to other shader invocations accessing the same memory.
8758The implementation may cache the values fetched by memory reads and return
8759the same values to any shader invocation accessing the same memory, even if
8760the underlying memory has been modified since the first memory read.
8761While variables not declared as <strong>coherent</strong> may not be useful for
8762communicating between shader invocations, using non-coherent accesses may
8763result in higher performance.</p>
8764</div>
8765<div class="paragraph">
8766<p>Memory accesses to image variables declared using the <strong>volatile</strong> qualifier
8767must treat the underlying memory as though it could be read or written at
8768any point during shader execution by some source other than the executing
8769shader invocation.
8770When a volatile variable is read, its value must be re-fetched from the
8771underlying memory, even if the shader invocation performing the read had
8772previously fetched its value from the same memory.
8773When a volatile variable is written, its value must be written to the
8774underlying memory, even if the compiler can conclusively determine that its
8775value will be overwritten by a subsequent write.
8776Since the external source reading or writing a <strong>volatile</strong> variable may be
8777another shader invocation, variables declared as <strong>volatile</strong> are
8778automatically treated as coherent.</p>
8779</div>
8780<div class="paragraph">
8781<p>Memory accesses to image variables declared using the <strong>restrict</strong> qualifier
8782may be compiled assuming that the variable used to perform the memory access
8783is the only way to access the underlying memory using the shader stage in
8784question.
8785This allows the compiler to coalesce or reorder loads and stores using
8786<strong>restrict</strong>-qualified image variables in ways that wouldn&#8217;t be permitted
8787for image variables not so qualified, because the compiler can assume that
8788the underlying image won&#8217;t be read or written by other code.
8789Applications are responsible for ensuring that image memory referenced by
8790variables qualified with <strong>restrict</strong> will not be referenced using other
8791variables in the same scope; otherwise, accesses to <strong>restrict</strong>-qualified
8792variables will have undefined results.</p>
8793</div>
8794<div class="paragraph">
8795<p>Memory accesses to image variables declared using the <strong>readonly</strong> qualifier
8796may only read the underlying memory, which is treated as read-only memory
8797and cannot be written to.
8798It is an error to pass an image variable qualified with
8799<strong>readonly</strong> to <strong>imageStore</strong>() or other built-in functions that modify image
8800memory.</p>
8801</div>
8802<div class="paragraph">
8803<p>Memory accesses to image variables declared using the <strong>writeonly</strong> qualifier
8804may only write the underlying memory; the underlying memory cannot be read.
8805It is an error to pass an image variable qualified with
8806<strong>writeonly</strong> to <strong>imageLoad</strong>() or other built-in functions that read image
8807memory.</p>
8808</div>
8809<div class="paragraph">
8810<p>A variable could be qualified as both <strong>readonly</strong> and <strong>writeonly</strong>, disallowing
8811both read and write. Such variables can still be used with some queries, for
8812example <strong>imageSize</strong>() and <strong>.length</strong>().</p>
8813</div>
8814<div class="paragraph">
8815<p>Except for image variables qualified with the format qualifiers <strong>r32f</strong>,
8816<strong>r32i</strong>, and <strong>r32ui</strong>, image variables must specify a memory qualifier
8817(<strong>readonly</strong>, <strong>writeonly</strong>, or both).</p>
8818</div>
8819<div class="paragraph">
8820<p>The memory qualifiers <strong>coherent</strong>, <strong>volatile</strong>, <strong>restrict</strong>, <strong>readonly</strong>, and
8821<strong>writeonly</strong> may be used in the declaration of buffer variables (i.e.,
8822members of shader storage blocks).
8823When a buffer variable is declared with a memory qualifier, the behavior
8824specified for memory accesses involving image variables described above
8825applies identically to memory accesses involving that buffer variable.
8826It is a compile-time error to assign to a buffer variable qualified with
8827<strong>readonly</strong> or to read from a buffer variable qualified with <strong>writeonly</strong>.
8828The combination <strong>readonly</strong> <strong>writeonly</strong> is allowed.</p>
8829</div>
8830<div class="paragraph">
8831<p>Additionally, memory qualifiers may be used at the block-level declaration
8832of a shader storage block, including the combination <strong>readonly</strong> <strong>writeonly</strong>.
8833When a block declaration is qualified with a memory qualifier, it is as if
8834all of its members were declared with the same memory qualifier.
8835For example, the block declaration</p>
8836</div>
8837<div class="listingblock">
8838<div class="content">
8839<pre class="CodeRay highlight"><code data-lang="c++">coherent buffer Block {
8840    readonly vec4 member1;
8841    vec4 member2;
8842};</code></pre>
8843</div>
8844</div>
8845<div class="paragraph">
8846<p>is equivalent to</p>
8847</div>
8848<div class="listingblock">
8849<div class="content">
8850<pre class="CodeRay highlight"><code data-lang="c++">buffer Block {
8851    coherent readonly vec4 member1;
8852    coherent vec4 member2;
8853};</code></pre>
8854</div>
8855</div>
8856<div class="paragraph">
8857<p>Memory qualifiers are only supported in the declarations of image variables,
8858buffer variables, and shader storage blocks; it is an error to use such
8859qualifiers in any other declarations.</p>
8860</div>
8861<div class="paragraph">
8862<p>When calling user-defined functions, variables qualified with <strong>coherent</strong>,
8863<strong>volatile</strong>, <strong>readonly</strong>, or <strong>writeonly</strong> may not be passed to functions whose
8864formal parameters lack such qualifiers.
8865(See &#8220;<a href="#function-definitions">Function Definitions</a>&#8221; for more detail on
8866function calling.) It is legal to have any additional memory qualifiers on a
8867formal parameter, but only <strong>restrict</strong> can be taken away from a calling
8868argument, by a formal parameter that lacks the <strong>restrict</strong> qualifier.</p>
8869</div>
8870<div class="paragraph">
8871<p>When a built-in function is called, the code generated is to be based on the
8872actual qualification of the calling argument, not on the list of memory
8873qualifiers specified on the formal parameter in the prototype.</p>
8874</div>
8875<div class="listingblock">
8876<div class="content">
8877<pre class="CodeRay highlight"><code data-lang="c++">vec4 funcA(restrict image2D a) { ... }
8878vec4 funcB(image2D a) { ... }
8879layout(rgba32f) uniform image2D img1;
8880layout(rgba32f) coherent uniform image2D img2;
8881
8882funcA(img1);        <span class="comment">// OK, adding &quot;restrict&quot; is allowed</span>
8883funcB(img2);        <span class="comment">// illegal, stripping &quot;coherent&quot; is not</span></code></pre>
8884</div>
8885</div>
8886<div class="paragraph">
8887<p>Layout qualifiers cannot be used on formal function parameters, and layout
8888qualification is not included in parameter matching.</p>
8889</div>
8890<div class="paragraph">
8891<p>Note that the use of <strong>const</strong> in an image variable declaration is qualifying
8892the const-ness of the variable being declared, not the image it refers to.
8893The qualifier <strong>readonly</strong> qualifies the image memory (as accessed through
8894that variable) while <strong>const</strong> qualifies the variable itself.</p>
8895</div>
8896</div>
8897<div class="sect2">
8898<h3 id="specialization-constant-qualifier">4.11. Specialization-Constant Qualifier</h3>
8899<div class="paragraph">
8900<p>Specialization constants are used only for SPIR-V and declared using the
8901<strong>constant_id</strong> layout qualifier.
8902For example:</p>
8903</div>
8904<div class="listingblock">
8905<div class="content">
8906<pre class="CodeRay highlight"><code data-lang="c++">layout(constant_id = <span class="integer">17</span>) <span class="directive">const</span> <span class="predefined-type">int</span> arraySize = <span class="integer">12</span>;</code></pre>
8907</div>
8908</div>
8909<div class="paragraph">
8910<p>The above makes a specialization constant with a default value of 12.
8911The number 17 is an example author-chosen id by which the API or other tools
8912can later refer to this specific specialization constant.
8913If it is never changed before final lowering, it will retain the value of
891412.
8915It is a compile-time error to use the <strong>constant_id</strong> qualifier on anything
8916but SPIR-V generation of a scalar <strong>bool</strong>, <strong>int</strong>, <strong>uint</strong>, <strong>float</strong>, or
8917<strong>double</strong>.</p>
8918</div>
8919<div class="paragraph">
8920<p>Built-in constants can be declared to be specialization constants.
8921For example:</p>
8922</div>
8923<div class="listingblock">
8924<div class="content">
8925<pre class="CodeRay highlight"><code data-lang="c++">layout(constant_id = <span class="integer">31</span>) gl_MaxClipDistances; <span class="comment">// add specialization_id</span></code></pre>
8926</div>
8927</div>
8928<div class="paragraph">
8929<p>The declaration uses just the name of the previously declared built-in
8930variable, with a <strong>constant_id</strong> layout-qualifier declaration.
8931It is a compile-time error to do this after the constant has been used:
8932Constants are strictly either non-specialization constants or specialization
8933constants, not both.</p>
8934</div>
8935<div class="paragraph">
8936<p>The built-in constant vector <em>gl_WorkGroupSize</em> can be specialized using the
8937<strong>local_size_{xyz}_id</strong> qualifiers, to individually give the components an id.
8938For example:</p>
8939</div>
8940<div class="listingblock">
8941<div class="content">
8942<pre class="CodeRay highlight"><code data-lang="c++">layout(local_size_x_id = <span class="integer">18</span>, local_size_z_id = <span class="integer">19</span>) in;</code></pre>
8943</div>
8944</div>
8945<div class="paragraph">
8946<p>This leaves <em>gl_WorkGroupSize.y</em> as a non-specialization constant, with
8947<em>gl_WorkGroupSize</em> being a partially specialized vector.
8948Its <em>x</em> and <em>z</em> components can be later specialized, after generating
8949SPIR-V, using the ids 18 and 19.
8950These ids are declared independently from declaring the workgroup size:</p>
8951</div>
8952<div class="listingblock">
8953<div class="content">
8954<pre class="CodeRay highlight"><code data-lang="c++">layout(local_size_x = <span class="integer">32</span>, local_size_y = <span class="integer">32</span>) in;   <span class="comment">// size is (32,32,1)</span>
8955layout(local_size_x_id = <span class="integer">18</span>) in;                   <span class="comment">// constant_id for x</span>
8956layout(local_size_z_id = <span class="integer">19</span>) in;                   <span class="comment">// constant_id for z</span></code></pre>
8957</div>
8958</div>
8959<div class="paragraph">
8960<p>Existing rules for declaring <strong>local_size_x</strong>, <strong>local_size_y</strong>, and
8961<strong>local_size_z</strong> are not changed.
8962For the local-size ids, it is a compile-time error to provide different id
8963values for the same local-size id, or to provide them after any use.
8964Otherwise, order, placement, number of statements, and replication do not
8965cause errors.</p>
8966</div>
8967<div class="paragraph">
8968<p>Two arrays sized with specialization constants are the same type only if
8969sized with the same symbol, and involving no operations.
8970For example:</p>
8971</div>
8972<div class="listingblock">
8973<div class="content">
8974<pre class="CodeRay highlight"><code data-lang="c++">layout(constant_id = <span class="integer">51</span>) <span class="directive">const</span> <span class="predefined-type">int</span> aSize = <span class="integer">20</span>;
8975<span class="directive">const</span> <span class="predefined-type">int</span> pad = <span class="integer">2</span>;
8976<span class="directive">const</span> <span class="predefined-type">int</span> total = aSize + pad; <span class="comment">// specialization constant</span>
8977<span class="predefined-type">int</span> a[total], b[total];        <span class="comment">// a and b have the same type</span>
8978<span class="predefined-type">int</span> c[<span class="integer">22</span>];                     <span class="comment">// different type than a or b</span>
8979<span class="predefined-type">int</span> d[aSize + pad];            <span class="comment">// different type than a, b, or c</span>
8980<span class="predefined-type">int</span> e[aSize + <span class="integer">2</span>];              <span class="comment">// different type than a, b, c, or d</span></code></pre>
8981</div>
8982</div>
8983<div class="paragraph">
8984<p>Types containing arrays sized with a specialization constant cannot be
8985compared, assigned as aggregates, declared with an initializer, or used as
8986an initializer.
8987They can, however, be passed as arguments to functions having formal
8988parameters of the same type.
8989Only the outer-most dimension of a variable declared as an array of arrays
8990can be a specialization constant, otherwise a compile-time error results.</p>
8991</div>
8992<div class="paragraph">
8993<p>Arrays inside a block may be sized with a specialization constant, but the
8994block will have a static layout.
8995Changing the specialized size will not re-layout the block.
8996In the absence of explicit offsets, the layout will be based on the default
8997size of the array.</p>
8998</div>
8999</div>
9000<div class="sect2">
9001<h3 id="order-of-qualification">4.12. Order and Repetition of Qualification</h3>
9002<div class="paragraph">
9003<p>When multiple qualifiers are present in a declaration, they may appear in
9004any order, but they must all appear before the type.
9005The <strong>layout</strong> qualifier is the only qualifier that can appear more than once.
9006Further, a declaration can have at most one storage qualifier, at most one
9007auxiliary storage qualifier, and at most one interpolation qualifier.
9008Multiple memory qualifiers can be used.
9009Any violation of these rules will cause a compile-time error.</p>
9010</div>
9011</div>
9012<div class="sect2">
9013<h3 id="empty-declarations">4.13. Empty Declarations</h3>
9014<div class="paragraph">
9015<p><em>Empty declarations</em> are declarations without a variable name, meaning no
9016object is instantiated by the declaration.
9017Generally, empty declarations are allowed.
9018Some are useful when declaring structures, while many others have no effect.
9019For example:</p>
9020</div>
9021<div class="listingblock">
9022<div class="content">
9023<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span>;               <span class="comment">// No effect</span>
9024<span class="keyword">struct</span> S {<span class="predefined-type">int</span> x;}; <span class="comment">// Defines a struct S</span></code></pre>
9025</div>
9026</div>
9027<div class="paragraph">
9028<p>The combinations of qualifiers that cause compile-time or link-time errors
9029are the same whether or not the declaration is empty, for example:</p>
9030</div>
9031<div class="listingblock">
9032<div class="content">
9033<pre class="CodeRay highlight"><code data-lang="c++">invariant in <span class="predefined-type">float</span> x; <span class="comment">// Error. An input cannot be invariant.</span>
9034invariant in <span class="predefined-type">float</span>;   <span class="comment">// Error even though no variable is declared.</span></code></pre>
9035</div>
9036</div>
9037</div>
9038</div>
9039</div>
9040<div class="sect1">
9041<h2 id="operators-and-expressions">5. Operators and Expressions</h2>
9042<div class="sectionbody">
9043<div class="sect2">
9044<h3 id="operators">5.1. Operators</h3>
9045<div class="paragraph">
9046<p>The OpenGL ES Shading Language has the following operators.</p>
9047</div>
9048<table class="tableblock frame-all grid-all stretch">
9049<colgroup>
9050<col style="width: 25%;">
9051<col style="width: 25%;">
9052<col style="width: 25%;">
9053<col style="width: 25%;">
9054</colgroup>
9055<thead>
9056<tr>
9057<th class="tableblock halign-left valign-top">Precedence</th>
9058<th class="tableblock halign-left valign-top">Operator Class</th>
9059<th class="tableblock halign-left valign-top">Operators</th>
9060<th class="tableblock halign-left valign-top">Associativity</th>
9061</tr>
9062</thead>
9063<tbody>
9064<tr>
9065<td class="tableblock halign-left valign-top"><p class="tableblock">1 (highest)</p></td>
9066<td class="tableblock halign-left valign-top"><p class="tableblock">parenthetical grouping</p></td>
9067<td class="tableblock halign-left valign-top"><p class="tableblock">( )</p></td>
9068<td class="tableblock halign-left valign-top"><p class="tableblock">NA</p></td>
9069</tr>
9070<tr>
9071<td class="tableblock halign-left valign-top"><p class="tableblock">2</p></td>
9072<td class="tableblock halign-left valign-top"><p class="tableblock">array subscript<br>
9073                 function call and constructor structure<br>
9074                 field or method selector, swizzle<br>
9075                 post fix increment and decrement</p></td>
9076<td class="tableblock halign-left valign-top"><p class="tableblock">[ ]<br>
9077                                                  ( )<br>
9078                                                  .<br>
9079                                                  ++ --</p></td>
9080<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9081</tr>
9082<tr>
9083<td class="tableblock halign-left valign-top"><p class="tableblock">3</p></td>
9084<td class="tableblock halign-left valign-top"><p class="tableblock">prefix increment and decrement<br>
9085                 unary</p></td>
9086<td class="tableblock halign-left valign-top"><p class="tableblock">++ --<br>
9087                                                  + - ~ !</p></td>
9088<td class="tableblock halign-left valign-top"><p class="tableblock">Right to Left</p></td>
9089</tr>
9090<tr>
9091<td class="tableblock halign-left valign-top"><p class="tableblock">4</p></td>
9092<td class="tableblock halign-left valign-top"><p class="tableblock">multiplicative</p></td>
9093<td class="tableblock halign-left valign-top"><p class="tableblock">* / %</p></td>
9094<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9095</tr>
9096<tr>
9097<td class="tableblock halign-left valign-top"><p class="tableblock">5</p></td>
9098<td class="tableblock halign-left valign-top"><p class="tableblock">additive</p></td>
9099<td class="tableblock halign-left valign-top"><p class="tableblock">+ -</p></td>
9100<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9101</tr>
9102<tr>
9103<td class="tableblock halign-left valign-top"><p class="tableblock">6</p></td>
9104<td class="tableblock halign-left valign-top"><p class="tableblock">bit-wise shift</p></td>
9105<td class="tableblock halign-left valign-top"><p class="tableblock">&lt;&lt; &gt;&gt;</p></td>
9106<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9107</tr>
9108<tr>
9109<td class="tableblock halign-left valign-top"><p class="tableblock">7</p></td>
9110<td class="tableblock halign-left valign-top"><p class="tableblock">relational</p></td>
9111<td class="tableblock halign-left valign-top"><p class="tableblock">&lt; &gt; &lt;= &gt;=</p></td>
9112<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9113</tr>
9114<tr>
9115<td class="tableblock halign-left valign-top"><p class="tableblock">8</p></td>
9116<td class="tableblock halign-left valign-top"><p class="tableblock">equality</p></td>
9117<td class="tableblock halign-left valign-top"><p class="tableblock">== !=</p></td>
9118<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9119</tr>
9120<tr>
9121<td class="tableblock halign-left valign-top"><p class="tableblock">9</p></td>
9122<td class="tableblock halign-left valign-top"><p class="tableblock">bit-wise and</p></td>
9123<td class="tableblock halign-left valign-top"><p class="tableblock">&amp;</p></td>
9124<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9125</tr>
9126<tr>
9127<td class="tableblock halign-left valign-top"><p class="tableblock">10</p></td>
9128<td class="tableblock halign-left valign-top"><p class="tableblock">bit-wise exclusive or</p></td>
9129<td class="tableblock halign-left valign-top"><p class="tableblock">^</p></td>
9130<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9131</tr>
9132<tr>
9133<td class="tableblock halign-left valign-top"><p class="tableblock">11</p></td>
9134<td class="tableblock halign-left valign-top"><p class="tableblock">bit-wise inclusive or</p></td>
9135<td class="tableblock halign-left valign-top"><p class="tableblock">|</p></td>
9136<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9137</tr>
9138<tr>
9139<td class="tableblock halign-left valign-top"><p class="tableblock">12</p></td>
9140<td class="tableblock halign-left valign-top"><p class="tableblock">logical and</p></td>
9141<td class="tableblock halign-left valign-top"><p class="tableblock">&amp;&amp;</p></td>
9142<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9143</tr>
9144<tr>
9145<td class="tableblock halign-left valign-top"><p class="tableblock">13</p></td>
9146<td class="tableblock halign-left valign-top"><p class="tableblock">logical exclusive or</p></td>
9147<td class="tableblock halign-left valign-top"><p class="tableblock">^^</p></td>
9148<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9149</tr>
9150<tr>
9151<td class="tableblock halign-left valign-top"><p class="tableblock">14</p></td>
9152<td class="tableblock halign-left valign-top"><p class="tableblock">logical inclusive or</p></td>
9153<td class="tableblock halign-left valign-top"><p class="tableblock">||</p></td>
9154<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9155</tr>
9156<tr>
9157<td class="tableblock halign-left valign-top"><p class="tableblock">15</p></td>
9158<td class="tableblock halign-left valign-top"><p class="tableblock">selection</p></td>
9159<td class="tableblock halign-left valign-top"><p class="tableblock">? :</p></td>
9160<td class="tableblock halign-left valign-top"><p class="tableblock">Right to Left</p></td>
9161</tr>
9162<tr>
9163<td class="tableblock halign-left valign-top"><p class="tableblock">16</p></td>
9164<td class="tableblock halign-left valign-top"><p class="tableblock">Assignment<br>
9165                 arithmetic assignments</p></td>
9166<td class="tableblock halign-left valign-top"><p class="tableblock">=<br>
9167                                                  += -=<br>
9168                                                  *= /=<br>
9169                                                  %= &lt;&lt;= &gt;&gt;=<br>
9170                                                  &amp;= ^= |=</p></td>
9171<td class="tableblock halign-left valign-top"><p class="tableblock">Right to Left</p></td>
9172</tr>
9173<tr>
9174<td class="tableblock halign-left valign-top"><p class="tableblock">17 (lowest)</p></td>
9175<td class="tableblock halign-left valign-top"><p class="tableblock">sequence</p></td>
9176<td class="tableblock halign-left valign-top"><p class="tableblock">,</p></td>
9177<td class="tableblock halign-left valign-top"><p class="tableblock">Left to Right</p></td>
9178</tr>
9179</tbody>
9180</table>
9181<div class="paragraph">
9182<p>There is no address-of operator nor a dereference operator.
9183There is no typecast operator; constructors are used instead.</p>
9184</div>
9185</div>
9186<div class="sect2">
9187<h3 id="array-operations">5.2. Array Operations</h3>
9188<div class="paragraph">
9189<p>These are now described in &#8220;<a href="#structure-and-array-operations">Structure and
9190Array Operations</a>&#8221;.</p>
9191</div>
9192</div>
9193<div class="sect2">
9194<h3 id="function-calls">5.3. Function Calls</h3>
9195<div class="paragraph">
9196<p>If a function returns a value, then a call to that function may be used as
9197an expression, whose type will be the type that was used to declare or
9198define the function.</p>
9199</div>
9200<div class="paragraph">
9201<p>Function definitions and calling conventions are discussed in
9202&#8220;<a href="#function-definitions">Function Definitions</a>&#8221;.</p>
9203</div>
9204</div>
9205<div class="sect2">
9206<h3 id="constructors">5.4. Constructors</h3>
9207<div class="paragraph">
9208<p>Constructors use the function call syntax, where the function name is a
9209type, and the call makes an object of that type.
9210Constructors are used the same way in both initializers and expressions.
9211(See &#8220;<a href="#shading-language-grammar">Shading Language Grammar</a>&#8221; for details.)
9212The parameters are used to initialize the constructed value.
9213Constructors can be used to request a data type conversion to change from
9214one scalar type to another scalar type, or to build larger types out of
9215smaller types, or to reduce a larger type to a smaller type.</p>
9216</div>
9217<div class="paragraph">
9218<p>In general, constructors are not built-in functions with predetermined
9219prototypes.
9220For arrays and structures, there must be exactly one argument in the
9221constructor for each element or member.
9222For the other types, the arguments must provide a sufficient number of
9223components to perform the initialization, and it is an error to
9224include so many arguments that they cannot all be used.
9225Detailed rules follow.
9226The prototypes actually listed below are merely a subset of examples.</p>
9227</div>
9228<div class="sect3">
9229<h4 id="conversion-and-scalar-constructors">5.4.1. Conversion and Scalar Constructors</h4>
9230<div class="paragraph">
9231<p>Converting between scalar types is done as the following prototypes
9232indicate:</p>
9233</div>
9234<div class="listingblock">
9235<div class="content">
9236<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">int</span>(uint)     <span class="comment">// converts an unsigned integer to a signed integer</span>
9237<span class="predefined-type">int</span>(<span class="predefined-type">bool</span>)     <span class="comment">// converts a Boolean value to an int</span>
9238<span class="predefined-type">int</span>(<span class="predefined-type">float</span>)    <span class="comment">// converts a float value to an int</span>
9239uint(<span class="predefined-type">int</span>)     <span class="comment">// converts a signed integer value to an unsigned integer</span>
9240uint(<span class="predefined-type">bool</span>)    <span class="comment">// converts a Boolean value to an unsigned integer</span>
9241uint(<span class="predefined-type">float</span>)   <span class="comment">// converts a float value to an unsigned integer</span>
9242<span class="predefined-type">bool</span>(<span class="predefined-type">int</span>)     <span class="comment">// converts a signed integer value to a Boolean</span>
9243<span class="predefined-type">bool</span>(uint)    <span class="comment">// converts an unsigned integer value to a Boolean value</span>
9244<span class="predefined-type">bool</span>(<span class="predefined-type">float</span>)   <span class="comment">// converts a float value to a Boolean</span>
9245<span class="predefined-type">float</span>(<span class="predefined-type">int</span>)    <span class="comment">// converts a signed integer value to a float</span>
9246<span class="predefined-type">float</span>(uint)   <span class="comment">// converts an unsigned integer value to a float value</span>
9247<span class="predefined-type">float</span>(<span class="predefined-type">bool</span>)   <span class="comment">// converts a Boolean value to a float</span></code></pre>
9248</div>
9249</div>
9250<div class="paragraph">
9251<p>When constructors are used to convert a floating-point type to an integer
9252type, the fractional part of the floating-point value is dropped.
9253It is undefined to convert a negative floating-point value to an <strong>uint</strong>.</p>
9254</div>
9255<div class="paragraph">
9256<p>Integer values having more bits of precision than a single-precision
9257floating-point mantissa will lose precision when converted to <strong>float</strong>.</p>
9258</div>
9259<div class="paragraph">
9260<p>When a constructor is used to convert any integer or floating-point type to
9261a <strong>bool</strong>, 0 and 0.0 are converted to <strong>false</strong>, and non-zero values are
9262converted to <strong>true</strong>.
9263When a constructor is used to convert a <strong>bool</strong> to any integer or
9264floating-point type, <strong>false</strong> is converted to 0 or 0.0, and <strong>true</strong> is
9265converted to 1 or 1.0.</p>
9266</div>
9267<div class="paragraph">
9268<p>The constructor <strong>int</strong>(<strong>uint</strong>) preserves the bit pattern in the argument,
9269which will change the argument&#8217;s value if its sign bit is set.
9270The constructor <strong>uint</strong>(<strong>int</strong>) preserves the bit pattern in the argument,
9271which will change its value if it is negative.</p>
9272</div>
9273<div class="paragraph">
9274<p>Identity constructors, like <strong>float</strong>(<strong>float</strong>) are also legal, but of little
9275use.</p>
9276</div>
9277<div class="paragraph">
9278<p>Scalar constructors with non-scalar parameters can be used to take the first
9279element from a non-scalar.
9280For example, the constructor <strong>float</strong>(<strong>vec3</strong>) will select the first component
9281of the <strong>vec3</strong> parameter.</p>
9282</div>
9283</div>
9284<div class="sect3">
9285<h4 id="vector-and-matrix-constructors">5.4.2. Vector and Matrix Constructors</h4>
9286<div class="paragraph">
9287<p>Constructors can be used to create vectors or matrices from a set of
9288scalars, vectors, or matrices.
9289This includes the ability to shorten vectors.</p>
9290</div>
9291<div class="paragraph">
9292<p>If there is a single scalar parameter to a vector constructor, it is used to
9293initialize all components of the constructed vector to that scalar&#8217;s value.
9294If there is a single scalar parameter to a matrix constructor, it is used to
9295initialize all the components on the matrix&#8217;s diagonal, with the remaining
9296components initialized to 0.0.</p>
9297</div>
9298<div class="paragraph">
9299<p>If a vector is constructed from multiple scalars, one or more vectors, or
9300one or more matrices, or a mixture of these, the vector&#8217;s components will be
9301constructed in order from the components of the arguments.
9302The arguments will be consumed left to right, and each argument will have
9303all its components consumed, in order, before any components from the next
9304argument are consumed.
9305Similarly for constructing a matrix from multiple scalars or vectors, or a
9306mixture of these.
9307Matrix components will be constructed and consumed in column major order.
9308In these cases, there must be enough components provided in the arguments to
9309provide an initializer for every component in the constructed value.
9310It is an error to provide extra arguments beyond this last used
9311argument.</p>
9312</div>
9313<div class="paragraph">
9314<p>If a matrix is constructed from a matrix, then each component (column <em>i</em>,
9315row <em>j</em>) in the result that has a corresponding component (column <em>i</em>, row
9316<em>j</em>) in the argument will be initialized from there.
9317All other components will be initialized to the identity matrix.
9318If a matrix argument is given to a matrix constructor, it is
9319an error to have any other arguments.</p>
9320</div>
9321<div class="paragraph">
9322<p>If the basic type (<strong>bool</strong>, <strong>int</strong>,
9323or <strong>float</strong>)
9324of a parameter to a
9325constructor does not match the basic type of the object being constructed,
9326the scalar construction rules (above) are used to convert the parameters.</p>
9327</div>
9328<div class="paragraph">
9329<p>Some useful vector constructors are as follows:</p>
9330</div>
9331<div class="listingblock">
9332<div class="content">
9333<pre class="CodeRay highlight"><code data-lang="c++">vec3(<span class="predefined-type">float</span>)          <span class="comment">// initializes each component of the vec3 with the float</span>
9334vec4(ivec4)          <span class="comment">// makes a vec4 with component-wise conversion</span>
9335vec4(mat2)           <span class="comment">// the vec4 is column 0 followed by column 1</span>
9336vec2(<span class="predefined-type">float</span>, <span class="predefined-type">float</span>)   <span class="comment">// initializes a vec2 with 2 floats</span>
9337ivec3(<span class="predefined-type">int</span>, <span class="predefined-type">int</span>, <span class="predefined-type">int</span>) <span class="comment">// initializes an ivec3 with 3 ints</span>
9338bvec4(<span class="predefined-type">int</span>, <span class="predefined-type">int</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>) <span class="comment">// uses 4 Boolean conversions</span>
9339vec2(vec3)           <span class="comment">// drops the third component of a vec3</span>
9340vec3(vec4)           <span class="comment">// drops the fourth component of a vec4</span>
9341vec3(vec2, <span class="predefined-type">float</span>)    <span class="comment">// vec3.x = vec2.x, vec3.y = vec2.y, vec3.z = float</span>
9342vec3(<span class="predefined-type">float</span>, vec2)    <span class="comment">// vec3.x = float, vec3.y = vec2.x, vec3.z = vec2.y</span>
9343vec4(vec3, <span class="predefined-type">float</span>)
9344vec4(<span class="predefined-type">float</span>, vec3)
9345vec4(vec2, vec2)</code></pre>
9346</div>
9347</div>
9348<div class="paragraph">
9349<p>Some examples of these are:</p>
9350</div>
9351<div class="listingblock">
9352<div class="content">
9353<pre class="CodeRay highlight"><code data-lang="c++">vec4 color = vec4(<span class="float">0</span><span class="float">.0</span>, <span class="float">1</span><span class="float">.0</span>, <span class="float">0</span><span class="float">.0</span>, <span class="float">1</span><span class="float">.0</span>);
9354vec4 rgba = vec4(<span class="float">1</span><span class="float">.0</span>);      <span class="comment">// sets each component to 1.0</span>
9355vec3 rgb = vec3(color);     <span class="comment">// drop the 4th component</span></code></pre>
9356</div>
9357</div>
9358<div class="paragraph">
9359<p>To initialize the diagonal of a matrix with all other elements set to zero:</p>
9360</div>
9361<div class="listingblock">
9362<div class="content">
9363<pre class="CodeRay highlight"><code data-lang="c++">mat2(<span class="predefined-type">float</span>)
9364mat3(<span class="predefined-type">float</span>)
9365mat4(<span class="predefined-type">float</span>)</code></pre>
9366</div>
9367</div>
9368<div class="paragraph">
9369<p>That is, <em>result[i][j]</em> is set to the <em>float</em> argument for all \(i
9370= j\) and set to 0 for all \(i \neq j\).</p>
9371</div>
9372<div class="paragraph">
9373<p>To initialize a matrix by specifying vectors or scalars, the components are
9374assigned to the matrix elements in column-major order.</p>
9375</div>
9376<div class="listingblock">
9377<div class="content">
9378<pre class="CodeRay highlight"><code data-lang="c++">mat2(vec2, vec2);                 <span class="comment">// one column per argument</span>
9379mat3(vec3, vec3, vec3);           <span class="comment">// one column per argument</span>
9380mat4(vec4, vec4, vec4, vec4);     <span class="comment">// one column per argument</span>
9381mat3x2(vec2, vec2, vec2);         <span class="comment">// one column per argument</span>
9382mat2(<span class="predefined-type">float</span>, <span class="predefined-type">float</span>,                <span class="comment">// first column</span>
9383     <span class="predefined-type">float</span>, <span class="predefined-type">float</span>);               <span class="comment">// second column</span>
9384mat3(<span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>,         <span class="comment">// first column</span>
9385     <span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>,         <span class="comment">// second column</span>
9386     <span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>);        <span class="comment">// third column</span>
9387mat4(<span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>,  <span class="comment">// first column</span>
9388     <span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>,  <span class="comment">// second column</span>
9389     <span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>,  <span class="comment">// third column</span>
9390     <span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>, <span class="predefined-type">float</span>); <span class="comment">// fourth column</span>
9391mat2x3(vec2, <span class="predefined-type">float</span>,               <span class="comment">// first column</span>
9392       vec2, <span class="predefined-type">float</span>);              <span class="comment">// second column</span></code></pre>
9393</div>
9394</div>
9395<div class="paragraph">
9396<p>A wide range of other possibilities exist, to construct a matrix from
9397vectors and scalars, as long as enough components are present to initialize
9398the matrix.
9399To construct a matrix from a matrix:</p>
9400</div>
9401<div class="listingblock">
9402<div class="content">
9403<pre class="CodeRay highlight"><code data-lang="c++">mat3x3(mat4x4); <span class="comment">// takes the upper-left 3x3 of the mat4x4</span>
9404mat2x3(mat4x2); <span class="comment">// takes the upper-left 2x2 of the mat4x4, last row is 0,0</span>
9405mat4x4(mat3x3); <span class="comment">// puts the mat3x3 in the upper-left, sets the lower right</span>
9406                <span class="comment">// component to 1, and the rest to 0</span></code></pre>
9407</div>
9408</div>
9409</div>
9410<div class="sect3">
9411<h4 id="structure-constructors">5.4.3. Structure Constructors</h4>
9412<div class="paragraph">
9413<p>Once a structure is defined, and its type is given a name, a constructor is
9414available with the same name to construct instances of that structure.
9415For example:</p>
9416</div>
9417<div class="listingblock">
9418<div class="content">
9419<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">struct</span> light {
9420    <span class="predefined-type">float</span> intensity;
9421    vec3 position;
9422};
9423
9424light lightVar = light(<span class="float">3</span><span class="float">.0</span>, vec3(<span class="float">1</span><span class="float">.0</span>, <span class="float">2</span><span class="float">.0</span>, <span class="float">3</span><span class="float">.0</span>));</code></pre>
9425</div>
9426</div>
9427<div class="paragraph">
9428<p>The arguments to the constructor will be used to set the structure&#8217;s
9429members, in order, using one argument per member.
9430Each argument must be the same type as the member it
9431sets.</p>
9432</div>
9433<div class="paragraph">
9434<p>Structure constructors can be used as initializers or in expressions.</p>
9435</div>
9436</div>
9437<div class="sect3">
9438<h4 id="array-constructors">5.4.4. Array Constructors</h4>
9439<div class="paragraph">
9440<p>Array types can also be used as constructor names, which can then be used in
9441expressions or initializers.
9442For example,</p>
9443</div>
9444<div class="listingblock">
9445<div class="content">
9446<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">const</span> <span class="predefined-type">float</span> c[<span class="integer">3</span>] = <span class="predefined-type">float</span>[<span class="integer">3</span>](<span class="float">5</span><span class="float">.0</span>, <span class="float">7</span><span class="float">.2</span>, <span class="float">1</span><span class="float">.1</span>);
9447<span class="directive">const</span> <span class="predefined-type">float</span> d[<span class="integer">3</span>] = <span class="predefined-type">float</span>[](<span class="float">5</span><span class="float">.0</span>, <span class="float">7</span><span class="float">.2</span>, <span class="float">1</span><span class="float">.1</span>);
9448
9449<span class="predefined-type">float</span> g;
9450...
9451<span class="predefined-type">float</span> a[<span class="integer">5</span>] = <span class="predefined-type">float</span>[<span class="integer">5</span>](g, <span class="integer">1</span>, g, <span class="float">2</span><span class="float">.3</span>, g);
9452<span class="predefined-type">float</span> b[<span class="integer">3</span>];
9453
9454b = <span class="predefined-type">float</span>[<span class="integer">3</span>](g, g + <span class="float">1</span><span class="float">.0</span>, g + <span class="float">2</span><span class="float">.0</span>);</code></pre>
9455</div>
9456</div>
9457<div class="paragraph">
9458<p>There must be exactly the same number of arguments as the size of the array
9459being constructed.
9460If no size is present in the constructor, then the array is explicitly sized
9461to the number of arguments provided.
9462The arguments are assigned in order, starting at element 0, to the elements
9463of the constructed array.
9464Each argument must be the same type as the element type of the
9465array.</p>
9466</div>
9467<div class="paragraph">
9468<p>Arrays of arrays are similarly constructed, and the size for any dimension
9469is <strong class="purple">optional</strong></p>
9470</div>
9471<div class="listingblock">
9472<div class="content">
9473<pre class="CodeRay highlight"><code data-lang="c++">vec4 b[<span class="integer">2</span>] = ...;
9474vec4[<span class="integer">3</span>][<span class="integer">2</span>](b, b, b);    <span class="comment">// constructor</span>
9475vec4[][<span class="integer">2</span>](b, b, b);     <span class="comment">// constructor, valid, size deduced</span>
9476vec4[<span class="integer">3</span>][](b, b, b);     <span class="comment">// constructor, valid, size deduced</span>
9477vec4[][](b, b, b);      <span class="comment">// constructor, valid, both sizes deduced</span></code></pre>
9478</div>
9479</div>
9480</div>
9481<div class="sect3">
9482<h4 id="_texture_combined_sampler_constructors">5.4.5. Texture-Combined Sampler Constructors</h4>
9483<div class="paragraph">
9484<p>Texture-combined sampler constructors are only available when targeting Vulkan.</p>
9485</div>
9486<div class="paragraph">
9487<p>Texture-combined sampler types, like <strong>sampler2D</strong>, can be declared with an
9488initializer
9489that is a constructor of the same type, and consuming a texture and a
9490<strong>sampler</strong> or <strong>samplerShadow</strong>.
9491For example:</p>
9492</div>
9493<div class="listingblock">
9494<div class="content">
9495<pre class="CodeRay highlight"><code data-lang="c++">    layout(...) uniform sampler s;   <span class="comment">// handle to filtering information</span>
9496    layout(...) uniform texture2D t; <span class="comment">// handle to a texture</span>
9497    layout(...) in vec2 tCoord;
9498    ...
9499    texture(sampler2D(t, s), tCoord);</code></pre>
9500</div>
9501</div>
9502<div class="paragraph">
9503<p>The result of a texture-combined sampler constructor cannot be assigned to a
9504variable:</p>
9505</div>
9506<div class="listingblock">
9507<div class="content">
9508<pre class="CodeRay highlight"><code data-lang="c++">    ... sampler2D sConstruct = sampler2D(t, s);  <span class="comment">// ERROR</span></code></pre>
9509</div>
9510</div>
9511<div class="paragraph">
9512<p>Texture-combined sampler constructors can only be consumed by a function parameter.</p>
9513</div>
9514<div class="paragraph">
9515<p>Texture-combined sampler constructors of arrays are illegal:</p>
9516</div>
9517<div class="listingblock">
9518<div class="content">
9519<pre class="CodeRay highlight"><code data-lang="c++">    layout(...) uniform texture2D tArray[<span class="integer">6</span>];
9520    ...
9521    ... sampler2D[](tArray, s) ...  <span class="comment">// ERROR</span></code></pre>
9522</div>
9523</div>
9524<div class="paragraph">
9525<p>Formally:</p>
9526</div>
9527<div class="ulist">
9528<ul>
9529<li>
9530<p>every texture-combined sampler type can be used as a constructor</p>
9531</li>
9532<li>
9533<p>the type of the constructor must match the type of the variable being declared</p>
9534</li>
9535<li>
9536<p>the constructor&#8217;s first argument must be a texture type</p>
9537</li>
9538<li>
9539<p>the constructor&#8217;s second argument must be a scalar of type <strong>sampler</strong>
9540or <strong>samplerShadow</strong></p>
9541</li>
9542<li>
9543<p>the dimensionality (1D, 2D, 3D, Cube, Rect, Buffer, MS, and Array)
9544of the texture type must match that of the constructed type
9545(that is, the suffixes of the type of the first argument and the
9546type of the constructor will be spelled the same way)</p>
9547</li>
9548<li>
9549<p>there is no control flow construct (e.g., <code>?:</code>) that consumes any sampler type</p>
9550</li>
9551</ul>
9552</div>
9553<div class="paragraph">
9554<p>Note: Shadow mismatches are allowed between constructors and the second argument.
9555Texture-combined non-shadow samplers can be constructed from <strong>samplerShadow</strong> and
9556texture-combined shadow samplers can be constructed from <strong>sampler</strong>.</p>
9557</div>
9558</div>
9559</div>
9560<div class="sect2">
9561<h3 id="vector-components">5.5. Vector Components</h3>
9562<div class="paragraph">
9563<p>The names of the components of a vector
9564are denoted by a single letter.
9565As a notational convenience, several letters are associated with each
9566component based on common usage of position, color or texture coordinate
9567vectors.
9568The individual components can be selected by following the variable name
9569with period (<strong>.</strong>) and then the component name.</p>
9570</div>
9571<div class="paragraph">
9572<p>The component names supported are:</p>
9573</div>
9574<table class="tableblock frame-all grid-all fit-content">
9575<colgroup>
9576<col>
9577<col>
9578</colgroup>
9579<tbody>
9580<tr>
9581<td class="tableblock halign-left valign-top"><p class="tableblock"><em>{ x, y, z, w }</em></p></td>
9582<td class="tableblock halign-left valign-top"><p class="tableblock">Useful when accessing vectors that represent points or normals</p></td>
9583</tr>
9584<tr>
9585<td class="tableblock halign-left valign-top"><p class="tableblock"><em>{ r, g, b, a }</em></p></td>
9586<td class="tableblock halign-left valign-top"><p class="tableblock">Useful when accessing vectors that represent colors</p></td>
9587</tr>
9588<tr>
9589<td class="tableblock halign-left valign-top"><p class="tableblock"><em>{ s, t, p, q }</em></p></td>
9590<td class="tableblock halign-left valign-top"><p class="tableblock">Useful when accessing vectors that represent texture coordinates</p></td>
9591</tr>
9592</tbody>
9593</table>
9594<div class="paragraph">
9595<p>The component names <em>x</em>, <em>r</em>, and <em>s</em> are, for example, synonyms for the
9596same (first) component in a vector.</p>
9597</div>
9598<div class="paragraph">
9599<p>Note that the third component of the texture coordinate set
9600has been renamed <em>p</em> so as to avoid the confusion with <em>r</em> (for
9601red) in a color.</p>
9602</div>
9603<div class="paragraph">
9604<p>Accessing components beyond those declared for the type is
9605an error so, for example:</p>
9606</div>
9607<div class="listingblock">
9608<div class="content">
9609<pre class="CodeRay highlight"><code data-lang="c++">vec2 pos;
9610pos.x       <span class="comment">// is legal</span>
9611pos.z       <span class="comment">// is illegal</span></code></pre>
9612</div>
9613</div>
9614<div class="paragraph">
9615<p>The component selection syntax allows multiple components to be selected by
9616appending their names (from the same name set) after the period (<strong>.</strong>).</p>
9617</div>
9618<div class="listingblock">
9619<div class="content">
9620<pre class="CodeRay highlight"><code data-lang="c++">vec4 v4;
9621v4.rgba;    <span class="comment">// is a vec4 and the same as just using v4,</span>
9622v4.rgb;     <span class="comment">// is a vec3,</span>
9623v4.b;       <span class="comment">// is a float,</span>
9624v4.xy;      <span class="comment">// is a vec2,</span>
9625v4.xgba;    <span class="comment">// is illegal - the component names do not come from the same set</span></code></pre>
9626</div>
9627</div>
9628<div class="paragraph">
9629<p>No more than 4 components can be selected.</p>
9630</div>
9631<div class="listingblock">
9632<div class="content">
9633<pre class="CodeRay highlight"><code data-lang="c++">vec4 v4;
9634v4.xyzwxy;      <span class="comment">// is illegal since it has 6 components</span>
9635(v4.xyzwxy).xy; <span class="comment">// is illegal since the intermediate value has 6</span>
9636components</code></pre>
9637</div>
9638</div>
9639<div class="paragraph">
9640<p>The order of the components can be different to swizzle them, or replicated:</p>
9641</div>
9642<div class="listingblock">
9643<div class="content">
9644<pre class="CodeRay highlight"><code data-lang="c++">vec4 pos = vec4(<span class="float">1</span><span class="float">.0</span>, <span class="float">2</span><span class="float">.0</span>, <span class="float">3</span><span class="float">.0</span>, <span class="float">4</span><span class="float">.0</span>);
9645vec4 swiz = pos.wzyx;   <span class="comment">// swiz = (4.0, 3.0, 2.0, 1.0)</span>
9646vec4 dup = pos.xxyy;    <span class="comment">// dup = (1.0, 1.0, 2.0, 2.0)</span></code></pre>
9647</div>
9648</div>
9649<div class="paragraph">
9650<p>This notation is more concise than the constructor syntax.
9651To form an r-value, it can be applied to any expression that results in a
9652vector r-value.</p>
9653</div>
9654<div class="paragraph">
9655<p>The component group notation can occur on the left hand side of an
9656expression.</p>
9657</div>
9658<div class="listingblock">
9659<div class="content">
9660<pre class="CodeRay highlight"><code data-lang="c++">vec4 pos = vec4(<span class="float">1</span><span class="float">.0</span>, <span class="float">2</span><span class="float">.0</span>, <span class="float">3</span><span class="float">.0</span>, <span class="float">4</span><span class="float">.0</span>);
9661pos.xw = vec2(<span class="float">5</span><span class="float">.0</span>, <span class="float">6</span><span class="float">.0</span>);        <span class="comment">// pos = (5.0, 2.0, 3.0, 6.0)</span>
9662pos.wx = vec2(<span class="float">7</span><span class="float">.0</span>, <span class="float">8</span><span class="float">.0</span>);        <span class="comment">// pos = (8.0, 2.0, 3.0, 7.0)</span>
9663pos.xx = vec2(<span class="float">3</span><span class="float">.0</span>, <span class="float">4</span><span class="float">.0</span>);        <span class="comment">// illegal - 'x' used twice</span>
9664pos.xy = vec3(<span class="float">1</span><span class="float">.0</span>, <span class="float">2</span><span class="float">.0</span>, <span class="float">3</span><span class="float">.0</span>);   <span class="comment">// illegal - mismatch between vec2 and vec3</span></code></pre>
9665</div>
9666</div>
9667<div class="paragraph">
9668<p>To form an l-value, swizzling must further be applied to an l-value and
9669contain no duplicate components. It results in an l-value of scalar or
9670vector type, depending on number of components specified.</p>
9671</div>
9672<div class="paragraph">
9673<p>Array subscripting syntax can also be applied to vectors (but not to
9674scalars) to provide numeric indexing.
9675So in</p>
9676</div>
9677<div class="listingblock">
9678<div class="content">
9679<pre class="CodeRay highlight"><code data-lang="c++">vec4 pos;</code></pre>
9680</div>
9681</div>
9682<div class="paragraph">
9683<p><em>pos[2]</em> refers to the third element of <em>pos</em> and is equivalent to <em>pos.z</em>.
9684This allows variable indexing into a vector, as well as a generic way of
9685accessing components.
9686Any integer expression can be used as the subscript.
9687The first component is at index zero.
9688Reading from or writing to a vector using a constant integral expression
9689with a value that is negative or greater than or equal to the size of the
9690vector results in an error.
9691When indexing with non-constant expressions, behavior is undefined if the
9692index is negative, or greater than or equal to the size of the vector.</p>
9693</div>
9694<div class="paragraph">
9695<p>Note that scalars are not considered to be single-component vectors and
9696therefore the use of component selection operators on scalars is illegal.</p>
9697</div>
9698</div>
9699<div class="sect2">
9700<h3 id="matrix-components">5.6. Matrix Components</h3>
9701<div class="paragraph">
9702<p>The components of a matrix can be accessed using array subscripting syntax.
9703Applying a single subscript to a matrix treats the matrix as an array of
9704column vectors, and selects a single column, whose type is a vector of the
9705same size as the (column size of the) matrix.
9706The leftmost column is column 0.
9707A second subscript would then operate on the resulting vector, as defined
9708earlier for vectors.
9709Hence, two subscripts select a column and then a row.</p>
9710</div>
9711<div class="listingblock">
9712<div class="content">
9713<pre class="CodeRay highlight"><code data-lang="c++">mat4 m;
9714m[<span class="integer">1</span>] = vec4(<span class="float">2</span><span class="float">.0</span>);   <span class="comment">// sets the second column to all 2.0</span>
9715m[<span class="integer">0</span>][<span class="integer">0</span>] = <span class="float">1</span><span class="float">.0</span>;      <span class="comment">// sets the upper left element to 1.0</span>
9716m[<span class="integer">2</span>][<span class="integer">3</span>] = <span class="float">2</span><span class="float">.0</span>;      <span class="comment">// sets the 4th element of the third column to 2.0</span></code></pre>
9717</div>
9718</div>
9719<div class="paragraph">
9720<p>Behavior is undefined when accessing a component outside the bounds of a
9721matrix with a non-constant expression.
9722It is an error to access a matrix with a constant expression
9723that is outside the bounds of the matrix.</p>
9724</div>
9725</div>
9726<div class="sect2">
9727<h3 id="structure-and-array-operations">5.7. Structure and Array Operations</h3>
9728<div class="paragraph">
9729<p>The members of a structure and the <strong>length</strong>() method of an array are
9730selected using the period (<strong>.</strong>).</p>
9731</div>
9732<div class="paragraph">
9733<p>In total, only the following operators are allowed to operate on arrays and
9734structures as whole entities:</p>
9735</div>
9736<table class="tableblock frame-all grid-all fit-content">
9737<colgroup>
9738<col>
9739<col>
9740</colgroup>
9741<tbody>
9742<tr>
9743<td class="tableblock halign-left valign-top"><p class="tableblock">field selector</p></td>
9744<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>.</strong></p></td>
9745</tr>
9746<tr>
9747<td class="tableblock halign-left valign-top"><p class="tableblock">equality</p></td>
9748<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>==</strong> <strong>!=</strong></p></td>
9749</tr>
9750<tr>
9751<td class="tableblock halign-left valign-top"><p class="tableblock">assignment</p></td>
9752<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>=</strong></p></td>
9753</tr>
9754<tr>
9755<td class="tableblock halign-left valign-top"><p class="tableblock">Ternary operator</p></td>
9756<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>?:</strong></p></td>
9757</tr>
9758<tr>
9759<td class="tableblock halign-left valign-top"><p class="tableblock">Sequence operator</p></td>
9760<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>,</strong></p></td>
9761</tr>
9762<tr>
9763<td class="tableblock halign-left valign-top"><p class="tableblock">indexing (arrays only)</p></td>
9764<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>[</strong> <strong>]</strong></p></td>
9765</tr>
9766</tbody>
9767</table>
9768<div class="paragraph">
9769<p>The equality operators and assignment operator are only allowed if the two
9770operands are same size and type.
9771The operands cannot contain any opaque types.
9772Structure types must be of the same declared structure.
9773Both array operands must be
9774compile-time
9775sized.
9776When using the equality operators, two structures are equal if and only if
9777all the members are component-wise equal, and two arrays are equal if and
9778only if all the elements are element-wise equal.</p>
9779</div>
9780<div class="paragraph">
9781<p>Array elements are accessed using the array subscript operator (<strong>[ ]</strong>).
9782An example of accessing an array element is</p>
9783</div>
9784<div class="listingblock">
9785<div class="content">
9786<pre class="CodeRay highlight"><code data-lang="c++">diffuseColor += lightIntensity[<span class="integer">3</span>] * NdotL;</code></pre>
9787</div>
9788</div>
9789<div class="paragraph">
9790<p>Array indices start at zero.
9791Array elements are accessed using an expression whose type is <strong>int</strong> or
9792<strong>uint</strong>.</p>
9793</div>
9794<div class="paragraph">
9795<p>Behavior is undefined if a shader subscripts an array with an index less
9796than 0 or greater than or equal to the size the array was declared with.</p>
9797</div>
9798<div class="paragraph">
9799<p>Arrays can also be accessed with the method operator (<strong>.</strong>) and the <strong>length</strong>
9800method to query the size of the array:</p>
9801</div>
9802<div class="listingblock">
9803<div class="content">
9804<pre class="CodeRay highlight"><code data-lang="c++">lightIntensity.length() <span class="comment">// return the size of the array</span></code></pre>
9805</div>
9806</div>
9807</div>
9808<div class="sect2">
9809<h3 id="assignments">5.8. Assignments</h3>
9810<div class="paragraph">
9811<p>Assignments of values to variable names are done with the assignment
9812operator (<strong>=</strong>):</p>
9813</div>
9814<div class="dlist">
9815<dl>
9816<dt class="hdlist1"></dt>
9817<dd>
9818<p><em>lvalue-expression</em> = <em>rvalue-expression</em></p>
9819</dd>
9820</dl>
9821</div>
9822<div class="paragraph">
9823<p>The <em>lvalue-expression</em> evaluates to an l-value.
9824The assignment operator stores the value of <em>rvalue-expression</em> into the
9825l-value and returns an r-value with the type and precision of
9826<em>lvalue-expression</em>.
9827The <em>lvalue-expression</em> and <em>rvalue-expression</em> must have the same
9828type.
9829Any type-conversions must be specified explicitly via constructors.
9830It is an error if the l-value is not writable.
9831Variables that are built-in types, entire structures or arrays, structure
9832members, l-values with the field selector (<strong>.</strong>) applied to select components
9833or swizzles without repeated fields, l-values within parentheses, and
9834l-values dereferenced with the array subscript operator (<strong>[ ]</strong>) are all
9835l-values.
9836Other binary or unary expressions, function names, swizzles with repeated
9837fields, and constants cannot be l-values.
9838The ternary operator (<strong>?:</strong>) is also not allowed as an l-value.
9839Using an incorrect expression as an l-value results in an error.</p>
9840</div>
9841<div class="paragraph">
9842<p>Expressions on the left of an assignment are evaluated before expressions on
9843the right of the assignment.</p>
9844</div>
9845<div class="paragraph">
9846<p>The other assignment operators are</p>
9847</div>
9848<div class="ulist">
9849<ul>
9850<li>
9851<p>add into (<strong>+=</strong>)</p>
9852</li>
9853<li>
9854<p>subtract from (<strong>-=</strong>)</p>
9855</li>
9856<li>
9857<p>multiply into (<strong>*=</strong>)</p>
9858</li>
9859<li>
9860<p>divide into (<strong>/=</strong>)</p>
9861</li>
9862<li>
9863<p>modulus into (<strong>%=</strong>)</p>
9864</li>
9865<li>
9866<p>left shift by (<strong>&lt;&lt;=</strong>)</p>
9867</li>
9868<li>
9869<p>right shift by (<strong>&gt;&gt;=</strong>)</p>
9870</li>
9871<li>
9872<p>and into (<strong>&amp;=</strong>)</p>
9873</li>
9874<li>
9875<p>inclusive-or into (<strong>|=</strong>)</p>
9876</li>
9877<li>
9878<p>exclusive-or into (<strong>^=</strong>)</p>
9879</li>
9880</ul>
9881</div>
9882<div class="paragraph">
9883<p>where the general expression</p>
9884</div>
9885<div class="dlist">
9886<dl>
9887<dt class="hdlist1"></dt>
9888<dd>
9889<p><em>lvalue</em> <em>op</em>= <em>expression</em></p>
9890</dd>
9891</dl>
9892</div>
9893<div class="paragraph">
9894<p>is equivalent to</p>
9895</div>
9896<div class="dlist">
9897<dl>
9898<dt class="hdlist1"></dt>
9899<dd>
9900<p><em>lvalue</em> = <em>lvalue</em> <em>op</em> <em>expression</em></p>
9901</dd>
9902</dl>
9903</div>
9904<div class="paragraph">
9905<p>where <em>lvalue</em> is the value returned by <em>lvalue-expression</em>, <em>op</em> is as
9906described below, and the <em>lvalue-expression</em> and <em>expression</em> must satisfy
9907the semantic requirements of both <em>op</em> and equals (<strong>=</strong>).</p>
9908</div>
9909<div class="paragraph">
9910<p>Reading a variable before writing (or initializing) it is legal, however the
9911value is undefined.</p>
9912</div>
9913</div>
9914<div class="sect2">
9915<h3 id="expressions">5.9. Expressions</h3>
9916<div class="paragraph">
9917<p>Expressions in the shading language are built from the following:</p>
9918</div>
9919<div class="ulist">
9920<ul>
9921<li>
9922<p>Constants of type <strong>bool</strong>, all integral types, all floating-point types,
9923all vector types, and all matrix types.</p>
9924</li>
9925<li>
9926<p>Constructors of all types.</p>
9927</li>
9928<li>
9929<p>Variable names of all types.</p>
9930</li>
9931<li>
9932<p>Arrays with the length method applied.</p>
9933</li>
9934<li>
9935<p>Subscripted arrays.</p>
9936</li>
9937<li>
9938<p>Function calls that return values.
9939In some cases, function calls returning <strong>void</strong> are also allowed in
9940expressions as specified below.</p>
9941</li>
9942<li>
9943<p>Component field selectors and array subscript results.</p>
9944</li>
9945<li>
9946<p>Parenthesized expressions.
9947Any expression, including expressions with void type can be
9948parenthesized.
9949Parentheses can be used to group operations.
9950Operations within parentheses are done before operations across
9951parentheses.</p>
9952</li>
9953<li>
9954<p>The arithmetic binary operators add (<strong>+</strong>), subtract (<strong>-</strong>), multiply
9955(<strong>*</strong>), and divide (<strong>/</strong>) operate on integer and floating-point scalars,
9956vectors, and matrices.
9957If the operands are integral types, they must both be signed or both be
9958unsigned.
9959All arithmetic binary operators result in the same fundamental type
9960(signed integer, unsigned integer, or floating-point) as the operands
9961they operate on.
9962The following cases are valid</p>
9963<div class="ulist">
9964<ul>
9965<li>
9966<p>The two operands are scalars.
9967In this case the operation is applied, resulting in a scalar.</p>
9968</li>
9969<li>
9970<p>One operand is a scalar, and the other is a vector or matrix.
9971In this case, the scalar operation is applied independently to each
9972component of the vector or matrix, resulting in the same size vector or
9973matrix.</p>
9974</li>
9975<li>
9976<p>The two operands are vectors of the same size.
9977In this case, the operation is done component-wise resulting in the
9978same size vector.</p>
9979</li>
9980<li>
9981<p>The operator is add (<strong>+</strong>), subtract (<strong>-</strong>), or divide (<strong>/</strong>), and the
9982operands are matrices with the same number of rows and the same number
9983of columns.
9984In this case, the operation is done component-wise resulting in the
9985same size matrix.</p>
9986</li>
9987<li>
9988<p>The operator is multiply (<strong>*</strong>), where both operands are matrices or one
9989operand is a vector and the other a matrix.
9990A right vector operand is treated as a column vector and a left vector
9991operand as a row vector.
9992In all these cases, it is required that the number of columns of the
9993left operand is equal to the number of rows of the right operand.
9994Then, the multiply (<strong>*</strong>) operation does a linear algebraic multiply,
9995yielding an object that has the same number of rows as the left operand
9996and the same number of columns as the right operand.
9997&#8220;<a href="#vector-and-matrix-operations">Vector and Matrix Operations</a>&#8221;
9998explains in more detail how vectors and matrices are operated on.</p>
9999<div class="openblock">
10000<div class="content">
10001<div class="paragraph">
10002<p>All other cases result in an error.</p>
10003</div>
10004<div class="paragraph">
10005<p>Use the built-in functions <strong>dot</strong>, <strong>cross</strong>, <strong>matrixCompMult</strong>, and
10006<strong>outerProduct</strong>, to get, respectively, vector dot product, vector cross
10007product, matrix component-wise multiplication, and the matrix product of a
10008column vector times a row vector.</p>
10009</div>
10010</div>
10011</div>
10012</li>
10013</ul>
10014</div>
10015</li>
10016<li>
10017<p>The operator modulus (<strong>%</strong>) operates on signed or unsigned integers or
10018integer vectors.
10019The operand types must both be signed or both be unsigned.
10020The operands cannot be vectors of differing size; this is
10021an error.
10022If one operand is a scalar and the other vector, then the scalar is
10023applied component-wise to the vector, resulting in the same type as the
10024vector.
10025If both are vectors of the same size, the result is computed
10026component-wise.
10027The resulting value is undefined for any component computed with a
10028second operand that is zero, while results for other components with
10029non-zero second operands remain defined.
10030If both operands are non-negative, then the remainder is non-negative.
10031Results are undefined if one or both operands are negative.
10032The operator modulus (<strong>%</strong>) is not defined for any other data types
10033(non-integer types).</p>
10034</li>
10035<li>
10036<p>The arithmetic unary operators negate (<strong>-</strong>), post- and pre-increment and
10037decrement (<strong>--</strong> and <strong>++</strong>) operate on integer or floating-point values
10038(including vectors and matrices).
10039All unary operators work component-wise on their operands.
10040These result with the same type they operated on.
10041For post- and pre-increment and decrement, the expression must be a writable
10042l-value.
10043Pre-increment and pre-decrement add or subtract 1 or 1.0 to the contents
10044of the expression they operate on, and the value of the pre-increment or
10045pre-decrement expression is the resulting value of that modification.
10046Post-increment and post-decrement expressions add or subtract 1 or 1.0
10047to the contents of the expression they operate on, but the resulting
10048expression has the expression&#8217;s value before the post-increment or
10049post-decrement was executed.</p>
10050</li>
10051<li>
10052<p>The relational operators greater than (<strong>&gt;</strong>), less than (<strong>&lt;</strong>), greater
10053than or equal (<strong>&gt;=</strong>), and less than or equal (<strong>&lt;=</strong>) operate only on
10054scalar integer and scalar floating-point expressions.
10055The result is scalar Boolean.
10056The types of the operands must match.
10057To do component-wise relational comparisons on vectors, use the built-in
10058functions <strong>lessThan</strong>, <strong>lessThanEqual</strong>, <strong>greaterThan</strong>, and
10059<strong>greaterThanEqual.</strong></p>
10060</li>
10061<li>
10062<p>The equality operators <strong>equal</strong> (<strong>==</strong>), and not equal (<strong>!=</strong>) operate on
10063all types except opaque types.
10064They result in a scalar Boolean.
10065The types of the operands must match.
10066For vectors, matrices, structures, and arrays, all components, members,
10067or elements of one operand must equal the corresponding components,
10068members, or elements in the other operand for the operands to be
10069considered equal.
10070To get a vector of component-wise equality results for vectors, use the
10071built-in functions <strong>equal</strong> and <strong>notEqual</strong>.</p>
10072</li>
10073<li>
10074<p>The logical binary operators and (<strong>&amp;&amp;</strong>), or (<strong>||</strong>), and exclusive or
10075(<strong>^^</strong>) operate only on two Boolean expressions and result in a Boolean
10076expression.
10077And (<strong>&amp;&amp;</strong>) will only evaluate the right hand operand if the left hand
10078operand evaluated to <strong>true</strong>.
10079Or (<strong>||</strong>) will only evaluate the right hand operand if the left hand
10080operand evaluated to <strong>false</strong>.
10081Exclusive or (<strong>^^</strong>) will always evaluate both operands.</p>
10082</li>
10083<li>
10084<p>The logical unary operator not (<strong>!</strong>).
10085It operates only on a Boolean expression and results in a Boolean
10086expression.
10087To operate on a vector, use the built-in function <strong>not</strong>.</p>
10088</li>
10089<li>
10090<p>The sequence (<strong>,</strong>) operator that operates on expressions by returning
10091the type and value of the right-most expression in a comma separated
10092list of expressions.
10093All expressions are evaluated, in order, from left to right.
10094The operands to the sequence operator may have <strong>void</strong> type.
10095Opaque types cannot be used with the sequence (,) operator.</p>
10096</li>
10097<li>
10098<p>The ternary selection operator (<strong>?:</strong>).
10099It operates on three expressions (<em>exp1</em> <strong>?</strong> <em>exp2</em> <strong>:</strong> <em>exp3</em>).
10100This operator evaluates the first expression, which must result in a
10101scalar Boolean.
10102If the result is true, it selects to evaluate the second expression,
10103otherwise it selects to evaluate the third expression.
10104Only one of the second and third expressions is evaluated.
10105The second and third expressions cannot be opaque types,
10106or there will be an error.
10107Otherwise,
10108the second and third expressions can be any type, including <strong>void</strong>, as
10109long their types match.
10110This resulting matching type is the type of the entire expression.</p>
10111</li>
10112<li>
10113<p>The one&#8217;s complement operator (<strong>~</strong>).
10114The operand must be of type signed or unsigned integer or integer
10115vector, and the result is the one&#8217;s complement of its operand; each bit
10116of each component is complemented, including any sign bits.</p>
10117</li>
10118<li>
10119<p>The shift operators (<strong>&lt;&lt;</strong>) and (<strong>&gt;&gt;</strong>).
10120For both operators, the operands must be signed or unsigned integers or
10121integer vectors.
10122One operand can be signed while the other is unsigned.
10123In all cases, the resulting type will be the same type as the left
10124operand.
10125If the first operand is a scalar, the second operand has to be a scalar
10126as well.
10127If the first operand is a vector, the second operand must be a scalar or
10128a vector with the same size as the first operand, and the result is
10129computed component-wise.
10130The result is undefined if the right operand is negative, or greater
10131than or equal to the number of bits in the left expression&#8217;s base type.
10132The value of E1 &lt;&lt; E2 is E1 (interpreted as a bit pattern) left-shifted
10133by E2 bits.
10134The value of E1 &gt;&gt; E2 is E1 right-shifted by E2 bit positions.
10135If E1 is a signed integer, the right-shift will extend the sign bit.
10136If E1 is an unsigned integer, the right-shift will zero-extend.</p>
10137</li>
10138<li>
10139<p>The bitwise operators and (<strong>&amp;</strong>), exclusive-or (<strong>^</strong>), and inclusive-or
10140(<strong>|</strong>).
10141The operands must be of type signed or unsigned integers or integer
10142vectors.
10143The operands cannot be vectors of differing size; this is an error.
10144If one operand is a scalar and the other a vector, the scalar is applied
10145component-wise to the vector, resulting in the same type as the vector.
10146The fundamental types of the operands (signed or unsigned) must match,
10147and will be the resulting fundamental type.
10148For and (<strong>&amp;</strong>), the result is the bitwise-and function of the operands.
10149For exclusive-or (<strong>^</strong>), the result is the bitwise exclusive-or function
10150of the operands.
10151For inclusive-or (<strong>|</strong>), the result is the bitwise inclusive-or function
10152of the operands.</p>
10153</li>
10154</ul>
10155</div>
10156<div class="paragraph">
10157<p>For a complete specification of the syntax of expressions, see
10158&#8220;<a href="#shading-language-grammar">Shading Language Grammar</a>&#8221;.</p>
10159</div>
10160</div>
10161<div class="sect2">
10162<h3 id="vector-and-matrix-operations">5.10. Vector and Matrix Operations</h3>
10163<div class="paragraph">
10164<p>With a few exceptions, operations are component-wise.
10165Usually, when an operator operates on a vector or matrix, it is operating
10166independently on each component of the vector or matrix, in a component-wise
10167fashion.
10168For example,</p>
10169</div>
10170<div class="listingblock">
10171<div class="content">
10172<pre class="CodeRay highlight"><code data-lang="c++">vec3 v, u;
10173<span class="predefined-type">float</span> f;
10174v = u + f;</code></pre>
10175</div>
10176</div>
10177<div class="paragraph">
10178<p>will be equivalent to</p>
10179</div>
10180<div class="listingblock">
10181<div class="content">
10182<pre class="CodeRay highlight"><code data-lang="c++">v.x = u.x + f;
10183v.y = u.y + f;
10184v.z = u.z + f;</code></pre>
10185</div>
10186</div>
10187<div class="paragraph">
10188<p>And</p>
10189</div>
10190<div class="listingblock">
10191<div class="content">
10192<pre class="CodeRay highlight"><code data-lang="c++">vec3 v, u, w;
10193w = v + u;</code></pre>
10194</div>
10195</div>
10196<div class="paragraph">
10197<p>will be equivalent to</p>
10198</div>
10199<div class="listingblock">
10200<div class="content">
10201<pre class="CodeRay highlight"><code data-lang="c++">w.x = v.x + u.x;
10202w.y = v.y + u.y;
10203w.z = v.z + u.z;</code></pre>
10204</div>
10205</div>
10206<div class="paragraph">
10207<p>and likewise for most operators and all integer and floating-point vector
10208and matrix types.
10209The exceptions are matrix multiplied by vector, vector multiplied by matrix,
10210and matrix multiplied by matrix.
10211These do not operate component-wise, but rather perform the correct linear
10212algebraic multiply.</p>
10213</div>
10214<div class="listingblock">
10215<div class="content">
10216<pre class="CodeRay highlight"><code data-lang="c++">vec3 v, u;
10217mat3 m;
10218u = v * m;</code></pre>
10219</div>
10220</div>
10221<div class="paragraph">
10222<p>is equivalent to</p>
10223</div>
10224<div class="listingblock">
10225<div class="content">
10226<pre class="CodeRay highlight"><code data-lang="c++">u.x = dot(v, m[<span class="integer">0</span>]); <span class="comment">// m[0] is the left column of m</span>
10227u.y = dot(v, m[<span class="integer">1</span>]); <span class="comment">// dot(a,b) is the inner (dot) product of a and b</span>
10228u.z = dot(v, m[<span class="integer">2</span>]);</code></pre>
10229</div>
10230</div>
10231<div class="paragraph">
10232<p>And</p>
10233</div>
10234<div class="listingblock">
10235<div class="content">
10236<pre class="CodeRay highlight"><code data-lang="c++">u = m * v;</code></pre>
10237</div>
10238</div>
10239<div class="paragraph">
10240<p>is equivalent to</p>
10241</div>
10242<div class="listingblock">
10243<div class="content">
10244<pre class="CodeRay highlight"><code data-lang="c++">u.x = m[<span class="integer">0</span>].x * v.x + m[<span class="integer">1</span>].x * v.y + m[<span class="integer">2</span>].x * v.z;
10245u.y = m[<span class="integer">0</span>].y * v.x + m[<span class="integer">1</span>].y * v.y + m[<span class="integer">2</span>].y * v.z;
10246u.z = m[<span class="integer">0</span>].z * v.x + m[<span class="integer">1</span>].z * v.y + m[<span class="integer">2</span>].z * v.z;</code></pre>
10247</div>
10248</div>
10249<div class="paragraph">
10250<p>And</p>
10251</div>
10252<div class="listingblock">
10253<div class="content">
10254<pre class="CodeRay highlight"><code data-lang="c++">mat3 m, n, r;
10255r = m * n;</code></pre>
10256</div>
10257</div>
10258<div class="paragraph">
10259<p>is equivalent to</p>
10260</div>
10261<div class="listingblock">
10262<div class="content">
10263<pre class="CodeRay highlight"><code data-lang="c++">r[<span class="integer">0</span>].x = m[<span class="integer">0</span>].x * n[<span class="integer">0</span>].x + m[<span class="integer">1</span>].x * n[<span class="integer">0</span>].y + m[<span class="integer">2</span>].x * n[<span class="integer">0</span>].z;
10264r[<span class="integer">1</span>].x = m[<span class="integer">0</span>].x * n[<span class="integer">1</span>].x + m[<span class="integer">1</span>].x * n[<span class="integer">1</span>].y + m[<span class="integer">2</span>].x * n[<span class="integer">1</span>].z;
10265r[<span class="integer">2</span>].x = m[<span class="integer">0</span>].x * n[<span class="integer">2</span>].x + m[<span class="integer">1</span>].x * n[<span class="integer">2</span>].y + m[<span class="integer">2</span>].x * n[<span class="integer">2</span>].z;
10266r[<span class="integer">0</span>].y = m[<span class="integer">0</span>].y * n[<span class="integer">0</span>].x + m[<span class="integer">1</span>].y * n[<span class="integer">0</span>].y + m[<span class="integer">2</span>].y * n[<span class="integer">0</span>].z;
10267r[<span class="integer">1</span>].y = m[<span class="integer">0</span>].y * n[<span class="integer">1</span>].x + m[<span class="integer">1</span>].y * n[<span class="integer">1</span>].y + m[<span class="integer">2</span>].y * n[<span class="integer">1</span>].z;
10268r[<span class="integer">2</span>].y = m[<span class="integer">0</span>].y * n[<span class="integer">2</span>].x + m[<span class="integer">1</span>].y * n[<span class="integer">2</span>].y + m[<span class="integer">2</span>].y * n[<span class="integer">2</span>].z;
10269r[<span class="integer">0</span>].z = m[<span class="integer">0</span>].z * n[<span class="integer">0</span>].x + m[<span class="integer">1</span>].z * n[<span class="integer">0</span>].y + m[<span class="integer">2</span>].z * n[<span class="integer">0</span>].z;
10270r[<span class="integer">1</span>].z = m[<span class="integer">0</span>].z * n[<span class="integer">1</span>].x + m[<span class="integer">1</span>].z * n[<span class="integer">1</span>].y + m[<span class="integer">2</span>].z * n[<span class="integer">1</span>].z;
10271r[<span class="integer">2</span>].z = m[<span class="integer">0</span>].z * n[<span class="integer">2</span>].x + m[<span class="integer">1</span>].z * n[<span class="integer">2</span>].y + m[<span class="integer">2</span>].z * n[<span class="integer">2</span>].z;</code></pre>
10272</div>
10273</div>
10274<div class="paragraph">
10275<p>and similarly for other sizes of vectors and matrices.</p>
10276</div>
10277</div>
10278<div class="sect2">
10279<h3 id="specialization-constant-operations">5.11. Specialization-Constant Operations</h3>
10280<div class="paragraph">
10281<p>Specialization-constant operations are only available when targeting SPIR-V.</p>
10282</div>
10283<div class="paragraph">
10284<p>Only some operations discussed in this section may be applied to a
10285specialization constant and still yield a result that is a specialization
10286constant.
10287The operations that do so are listed below.
10288When a specialization constant is operated on with one of these operators
10289and with another constant or specialization constant, the result is
10290implicitly a specialization constant.</p>
10291</div>
10292<div class="ulist">
10293<ul>
10294<li>
10295<p><strong>int</strong>(), <strong>uint</strong>(), and <strong>bool</strong>() constructors for type conversions from
10296any of the following types to any of the following types:</p>
10297<div class="ulist">
10298<ul>
10299<li>
10300<p><strong>int</strong></p>
10301</li>
10302<li>
10303<p><strong>uint</strong></p>
10304</li>
10305<li>
10306<p><strong>bool</strong></p>
10307</li>
10308</ul>
10309</div>
10310</li>
10311<li>
10312<p>vector versions of the above conversion constructors</p>
10313</li>
10314<li>
10315<p>allowed implicit conversions of the above</p>
10316</li>
10317<li>
10318<p>swizzles (e.g. <code>foo.yx</code>)</p>
10319</li>
10320<li>
10321<p>the following when applied to integer or unsigned integer types:</p>
10322<div class="ulist">
10323<ul>
10324<li>
10325<p>unary negative (<strong>-</strong>)</p>
10326</li>
10327<li>
10328<p>binary operations (<strong>+</strong>, <strong>-</strong>, <strong>*</strong>, <strong>/</strong>, <strong>%</strong>)</p>
10329</li>
10330<li>
10331<p>shift (<strong>&lt;&lt;</strong>, <strong>&gt;&gt;</strong>)</p>
10332</li>
10333<li>
10334<p>bitwise operations (<strong>&amp;</strong>, <strong>|</strong>, <strong>^</strong>)</p>
10335</li>
10336</ul>
10337</div>
10338</li>
10339<li>
10340<p>the following when applied to integer or unsigned integer scalar types:</p>
10341<div class="ulist">
10342<ul>
10343<li>
10344<p>comparison (<strong>==</strong>, <strong>!=</strong>, <strong>&gt;</strong>, <strong>&gt;=</strong>, <strong>&lt;</strong>, <strong>&#8656;</strong>)</p>
10345</li>
10346</ul>
10347</div>
10348</li>
10349<li>
10350<p>The following when applied to the Boolean scalar type:</p>
10351<div class="ulist">
10352<ul>
10353<li>
10354<p>not (<strong>!</strong>)</p>
10355</li>
10356<li>
10357<p>logical operations (<strong>&amp;&amp;</strong>, <strong>||</strong>, <strong>^^</strong>)</p>
10358</li>
10359<li>
10360<p>comparison (<strong>==</strong>, <strong>!=</strong>)</p>
10361</li>
10362</ul>
10363</div>
10364</li>
10365<li>
10366<p>the ternary operator (<strong>?:</strong>)</p>
10367</li>
10368</ul>
10369</div>
10370</div>
10371<div class="sect2">
10372<h3 id="evaluation-of-expressions">5.12. Evaluation of Expressions</h3>
10373<div class="paragraph">
10374<p>In general expressions must be evaluated in the order specified by the
10375precedence of operations and may only be regrouped if the result is the same or
10376where the result is undefined.
10377No other transforms may be applied that affect the result of an operation.
10378GLSL ES relaxes these requirements for scalar operations in the following
10379ways:</p>
10380</div>
10381<div class="ulist">
10382<ul>
10383<li>
10384<p>Addition and multiplication are assumed to be associative.</p>
10385</li>
10386<li>
10387<p>Multiplication is assumed to be distributive over addition.
10388Therefore expressions may be expanded and re-factored.</p>
10389</li>
10390<li>
10391<p>Floating-point division may be replaced by reciprocal and
10392multiplication.</p>
10393</li>
10394<li>
10395<p>Multiplication may be replaced by repeated addition.</p>
10396</li>
10397<li>
10398<p>Within the constraints of invariance (where applicable), the precision
10399used may vary.</p>
10400</li>
10401</ul>
10402</div>
10403<div class="paragraph">
10404<p>These rules also apply to the built-in functions.</p>
10405</div>
10406</div>
10407</div>
10408</div>
10409<div class="sect1">
10410<h2 id="statements-and-structure">6. Statements and Structure</h2>
10411<div class="sectionbody">
10412<div class="paragraph">
10413<p>The fundamental building blocks of the OpenGL ES Shading Language are:</p>
10414</div>
10415<div class="ulist">
10416<ul>
10417<li>
10418<p>statements and declarations</p>
10419</li>
10420<li>
10421<p>function definitions</p>
10422</li>
10423<li>
10424<p>selection (<strong>if</strong>-<strong>else</strong> and <strong>switch</strong>-<strong>case</strong>-<strong>default</strong>)</p>
10425</li>
10426<li>
10427<p>iteration (<strong>for</strong>, <strong>while</strong>, and <strong>do</strong>-<strong>while</strong>)</p>
10428</li>
10429<li>
10430<p>jumps (<strong>discard</strong>, <strong>return</strong>, <strong>break</strong>, and <strong>continue</strong>)</p>
10431</li>
10432</ul>
10433</div>
10434<div class="paragraph">
10435<p>The overall structure of a shader is as follows</p>
10436</div>
10437<div class="openblock bnf">
10438<div class="content">
10439<div class="dlist">
10440<dl>
10441<dt class="hdlist1"><em>translation-unit</em> : </dt>
10442<dd>
10443<p><em>global-declaration</em><br>
10444<em>translation-unit</em> <em>global-declaration</em></p>
10445</dd>
10446<dt class="hdlist1"><em>global-declaration</em> : </dt>
10447<dd>
10448<p><em>function-definition</em><br>
10449<em>declaration</em></p>
10450</dd>
10451</dl>
10452</div>
10453</div>
10454</div>
10455<div class="paragraph">
10456<p>That is, a shader is a sequence of declarations and function bodies.
10457Function bodies are defined as</p>
10458</div>
10459<div class="openblock bnf">
10460<div class="content">
10461<div class="dlist">
10462<dl>
10463<dt class="hdlist1"><em>function-definition</em> : </dt>
10464<dd>
10465<p><em>function-prototype</em> <strong>{</strong> <em>statement-list</em> <strong>}</strong></p>
10466</dd>
10467<dt class="hdlist1"><em>statement-list</em> : </dt>
10468<dd>
10469<p><em>statement</em><br>
10470<em>statement-list</em> <em>statement</em></p>
10471</dd>
10472<dt class="hdlist1"><em>statement</em> : </dt>
10473<dd>
10474<p><em>compound-statement</em><br>
10475<em>simple-statement</em></p>
10476</dd>
10477</dl>
10478</div>
10479</div>
10480</div>
10481<div class="paragraph">
10482<p>Curly braces are used to group sequences of statements into compound
10483statements.</p>
10484</div>
10485<div class="openblock bnf">
10486<div class="content">
10487<div class="dlist">
10488<dl>
10489<dt class="hdlist1"><em>compound-statement</em> : </dt>
10490<dd>
10491<p><strong>{</strong> <em>statement-list</em> <strong>}</strong></p>
10492</dd>
10493<dt class="hdlist1"><em>simple-statement</em> : </dt>
10494<dd>
10495<p><em>declaration-statement</em><br>
10496<em>expression-statement</em><br>
10497<em>selection-statement</em><br>
10498<em>iteration-statement</em><br>
10499<em>jump-statement</em></p>
10500</dd>
10501</dl>
10502</div>
10503</div>
10504</div>
10505<div class="paragraph">
10506<p>Simple declaration, expression, and jump statements end in a semi-colon.</p>
10507</div>
10508<div class="paragraph">
10509<p>This above is slightly simplified, and the complete grammar specified in
10510&#8220;<a href="#shading-language-grammar">Shading Language Grammar</a>&#8221; should be used as
10511the definitive specification.</p>
10512</div>
10513<div class="paragraph">
10514<p>Declarations and expressions have already been discussed.</p>
10515</div>
10516<div class="sect2">
10517<h3 id="function-definitions">6.1. Function Definitions</h3>
10518<div class="paragraph">
10519<p>As indicated by the grammar above, a valid shader is a sequence of global
10520declarations and function definitions.
10521A function is declared as the following example shows:</p>
10522</div>
10523<div class="listingblock">
10524<div class="content">
10525<pre class="CodeRay highlight"><code data-lang="c++"><span class="comment">// prototype</span>
10526returnType functionName (type0 arg0, type1 arg1, ..., typen argn);</code></pre>
10527</div>
10528</div>
10529<div class="paragraph">
10530<p>and a function is defined like</p>
10531</div>
10532<div class="listingblock">
10533<div class="content">
10534<pre class="CodeRay highlight"><code data-lang="c++"><span class="comment">// definition</span>
10535returnType functionName (type0 arg0, type1 arg1, ..., typen argn)
10536{
10537    <span class="comment">// do some computation</span>
10538    <span class="keyword">return</span> returnValue;
10539}</code></pre>
10540</div>
10541</div>
10542<div class="paragraph">
10543<p>where <em>returnType</em> must be present and cannot be void, or:</p>
10544</div>
10545<div class="listingblock">
10546<div class="content">
10547<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">void</span> functionName (type0 arg0, type1 arg1, ..., typen argn)
10548{
10549    <span class="comment">// do some computation</span>
10550    <span class="keyword">return</span>; <span class="comment">// optional</span>
10551}</code></pre>
10552</div>
10553</div>
10554<div class="paragraph">
10555<p>Each of the <em>typeN</em> must include a type and can optionally include a
10556parameter qualifier and/or <strong>const</strong>.</p>
10557</div>
10558<div class="paragraph">
10559<p>A function is called by using its name followed by a list of arguments in
10560parentheses.</p>
10561</div>
10562<div class="paragraph">
10563<p>Arrays are allowed as arguments and as the return type.
10564In both cases, the array must be
10565compile-time
10566sized.
10567An array is passed or returned by using just its name, without brackets, and
10568the size of the array must match the size specified in the function&#8217;s
10569declaration.</p>
10570</div>
10571<div class="paragraph">
10572<p>Structures are also allowed as argument types.
10573The return type can also be a structure.</p>
10574</div>
10575<div class="paragraph">
10576<p>See &#8220;<a href="#shading-language-grammar">Shading Language Grammar</a>&#8221; for the
10577definitive reference on the syntax to declare and define functions.</p>
10578</div>
10579<div class="paragraph">
10580<p>All functions must be either declared with a prototype or defined with a
10581body before they are called.
10582For example:</p>
10583</div>
10584<div class="listingblock">
10585<div class="content">
10586<pre class="CodeRay highlight"><code data-lang="c++"><span class="predefined-type">float</span> myfunc (<span class="predefined-type">float</span> f,      <span class="comment">// f is an input parameter</span>
10587              out <span class="predefined-type">float</span> g); <span class="comment">// g is an output parameter</span></code></pre>
10588</div>
10589</div>
10590<div class="paragraph">
10591<p>Functions that return no value must be declared as <strong>void</strong>.
10592A <strong>void</strong> function can only use <strong>return</strong> without a return argument, even if
10593the return argument has <strong>void</strong> type.
10594Return statements only accept values:</p>
10595</div>
10596<div class="listingblock">
10597<div class="content">
10598<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">void</span> func1() { }
10599<span class="directive">void</span> func2() { <span class="keyword">return</span> func1(); } <span class="comment">// illegal return statement</span></code></pre>
10600</div>
10601</div>
10602<div class="paragraph">
10603<p>Only a precision qualifier is allowed on the return type of a function.
10604Formal parameters can have parameter, precision, and memory qualifiers, but
10605no other qualifiers.</p>
10606</div>
10607<div class="paragraph">
10608<p>Functions that accept no input arguments need not use <strong>void</strong> in the argument
10609list because prototypes (or definitions) are required and therefore there is
10610no ambiguity when an empty argument list &#8220;( )&#8221; is declared.
10611The idiom &#8220;(<strong>void</strong>)&#8221; as a parameter list is provided for convenience.</p>
10612</div>
10613<div class="paragraph">
10614<p>Function names can be overloaded.
10615The same function name can be used for multiple functions, as long as the
10616parameter types differ.
10617If a function name is declared twice with the same parameter types, then the
10618return types and all qualifiers must also match, and it is the same function
10619being declared.
10620When function calls are resolved, an exact type match for all the arguments
10621is required.</p>
10622</div>
10623<div class="paragraph">
10624<p>For example,</p>
10625</div>
10626<div class="listingblock">
10627<div class="content">
10628<pre class="CodeRay highlight"><code data-lang="c++">vec4 f(in vec4 x, out vec4 y);
10629vec4 f(in vec4 x, out uvec4 y);      <span class="comment">// allowed, different argument type</span>
10630<span class="predefined-type">int</span> f(in vec4 x, out vec4 y);        <span class="comment">// error, only return type differs</span>
10631vec4 f(in vec4 x, in vec4 y);        <span class="comment">// error, only qualifier differs</span>
10632vec4 f(<span class="directive">const</span> in vec4 x, out vec4 y); <span class="comment">// error, only qualifier differs</span></code></pre>
10633</div>
10634</div>
10635<div class="paragraph">
10636<p>Calling the first two functions above with the following argument types
10637yields</p>
10638</div>
10639<div class="listingblock">
10640<div class="content">
10641<pre class="CodeRay highlight"><code data-lang="c++">f(vec4, vec4)   <span class="comment">// exact match of vec4 f(in vec4 x, out vec4 y)</span>
10642f(vec4, uvec4)  <span class="comment">// exact match of vec4 f(in vec4 x, out uvec4 y)</span>
10643f(ivec4, vec4)  <span class="comment">// error, no exact match.</span>
10644f(ivec4, uvec4) <span class="comment">// error, no exact match.</span></code></pre>
10645</div>
10646</div>
10647<div class="paragraph">
10648<p>User-defined functions can have multiple declarations, but only one
10649definition.</p>
10650</div>
10651<div class="paragraph">
10652<p>A shader cannot redefine or overload built-in functions.</p>
10653</div>
10654<div class="paragraph">
10655<p>The function <em>main</em> is used as the entry point to a shader executable.
10656All shaders must define a function named <em>main</em>.
10657This function takes no arguments, returns no value, and must be declared as
10658type <strong>void</strong>:</p>
10659</div>
10660<div class="listingblock">
10661<div class="content">
10662<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">void</span> main()
10663{
10664    ...
10665}</code></pre>
10666</div>
10667</div>
10668<div class="paragraph">
10669<p>The function <em>main</em> can contain uses of <strong>return</strong>.
10670See &#8220;<a href="#jumps">Jumps</a>&#8221; for more details.</p>
10671</div>
10672<div class="paragraph">
10673<p>It is a compile-time or link-time error to declare or define a function
10674<strong>main</strong> with any other parameters or return type.</p>
10675</div>
10676<div class="sect3">
10677<h4 id="function-calling-conventions">6.1.1. Function Calling Conventions</h4>
10678<div class="paragraph">
10679<p>Functions are called by value-return.
10680This means input arguments are copied into the function at call time, and
10681output arguments are copied back to the caller before function exit.
10682Because the function works with local copies of parameters, there are no
10683issues regarding aliasing of variables within a function.
10684To control what parameters are copied in and/or out through a function
10685definition or declaration:</p>
10686</div>
10687<div class="ulist">
10688<ul>
10689<li>
10690<p>The keyword <strong>in</strong> is used as a qualifier to denote a parameter is to be
10691copied in, but not copied out.</p>
10692</li>
10693<li>
10694<p>The keyword <strong>out</strong> is used as a qualifier to denote a parameter is to be
10695copied out, but not copied in.
10696This should be used whenever possible to avoid unnecessarily copying
10697parameters in.</p>
10698</li>
10699<li>
10700<p>The keyword <strong>inout</strong> is used as a qualifier to denote the parameter is to
10701be both copied in and copied out.
10702It means the same thing as specifying both <strong>in</strong> and <strong>out</strong>.</p>
10703</li>
10704<li>
10705<p>A function parameter declared with no such qualifier means the same
10706thing as specifying <strong>in</strong>.</p>
10707</li>
10708</ul>
10709</div>
10710<div class="paragraph">
10711<p>All arguments are evaluated at call time, exactly once, in order, from left
10712to right.
10713Evaluation of an <strong>in</strong> parameter results in a value that is copied to the
10714formal parameter.
10715Evaluation of an <strong>out</strong> parameter results in an l-value that is used to copy
10716out a value when the function returns.
10717Evaluation of an <strong>inout</strong> parameter results in both a value and an l-value;
10718the value is copied to the formal parameter at call time and the l-value is
10719used to copy out a value when the function returns.</p>
10720</div>
10721<div class="paragraph">
10722<p>The order in which output parameters are copied back to the caller is
10723undefined.</p>
10724</div>
10725<div class="paragraph">
10726<p>In a function, writing to an input-only parameter is allowed.
10727Only the function&#8217;s copy is modified.
10728This can be prevented by declaring a parameter with the <strong>const</strong> qualifier.</p>
10729</div>
10730<div class="paragraph">
10731<p>When calling a function, expressions that do not evaluate to l-values cannot
10732be passed to parameters declared as <strong>out</strong> or <strong>inout</strong>, or an error
10733results.</p>
10734</div>
10735<div class="paragraph">
10736<p>Only precision qualifiers are allowed on the return type of a function.</p>
10737</div>
10738<div class="paragraph">
10739<p>The syntax for function prototypes can be informally expressed as:</p>
10740</div>
10741<div class="openblock bnf">
10742<div class="content">
10743<div class="dlist">
10744<dl>
10745<dt class="hdlist1"><em>function_prototype</em> : </dt>
10746<dd>
10747<p>[ <em>type_qualifier</em> ] <em>type_specifier</em> <em>IDENTIFIER</em> <em>LEFT_PAREN</em>
10748<em>parameter_declaration</em> , <em>parameter_declaration</em> , &#8230;&#8203;
10749, <em>RIGHT_PAREN</em></p>
10750</dd>
10751<dt class="hdlist1"><em>parameter_declaration</em> : </dt>
10752<dd>
10753<p>[ <em>type_qualifier</em> ] <em>type_specifier</em> [ <em>IDENTIFIER</em> [ <em>array_specifier</em>
10754] ]</p>
10755</dd>
10756<dt class="hdlist1"><em>type_qualifier</em> : </dt>
10757<dd>
10758<p><em>single_type_qualifier</em> , <em>single_type_qualifier</em> , &#8230;&#8203;</p>
10759</dd>
10760</dl>
10761</div>
10762</div>
10763</div>
10764<div class="paragraph">
10765<p>The qualifiers allowed on formal parameters are:</p>
10766</div>
10767<div class="openblock bnf">
10768<div class="content">
10769<div class="dlist">
10770<dl>
10771<dt class="hdlist1"></dt>
10772<dd>
10773<p><em>empty</em><br>
10774<strong>const</strong><br>
10775<strong>in</strong><br>
10776<strong>out</strong><br>
10777<strong>inout</strong><br>
10778<strong>precise</strong><br>
10779<em>memory-qualifier</em><br>
10780<em>precision-qualifier</em></p>
10781</dd>
10782</dl>
10783</div>
10784</div>
10785</div>
10786<div class="paragraph">
10787<p>The <strong>const</strong> qualifier cannot be used with <strong>out</strong> or <strong>inout</strong>, or
10788an error results.
10789The above is used both for function declarations (i.e., prototypes) and for
10790function definitions.
10791Hence, function definitions can have unnamed arguments.</p>
10792</div>
10793<div class="paragraph">
10794<p>Static, and hence dynamic recursion, are not allowed.
10795Static recursion is present if the static function-call graph of a program
10796contains cycles.
10797Dynamic recursion occurs if at any time control flow has entered but not
10798exited a single function more than once.</p>
10799</div>
10800</div>
10801</div>
10802<div class="sect2">
10803<h3 id="selection">6.2. Selection</h3>
10804<div class="paragraph">
10805<p>Conditional control flow in the shading language is done by either <strong>if</strong>,
10806<strong>if</strong>-<strong>else</strong>, or <strong>switch</strong> statements:</p>
10807</div>
10808<div class="openblock bnf">
10809<div class="content">
10810<div class="dlist">
10811<dl>
10812<dt class="hdlist1"><em>selection-statement</em> : </dt>
10813<dd>
10814<p><strong>if</strong> <strong>(</strong> <em>bool-expression</em> <strong>)</strong> <em>statement</em><br>
10815<strong>if</strong> <strong>(</strong> <em>bool-expression</em> <strong>)</strong> <em>statement</em> <strong>else</strong> <em>statement</em><br>
10816<strong>switch</strong> <strong>(</strong> <em>init-expression</em> <strong>)</strong> <strong>{</strong> <em>switch-statement-list<sub>opt</sub></em> <strong>}</strong></p>
10817</dd>
10818</dl>
10819</div>
10820</div>
10821</div>
10822<div class="paragraph">
10823<p>Where <em>switch-statement-list</em> is a nested scope containing a list of zero or
10824more <em>switch-statement</em> and other statements defined by the language, where
10825<em>switch-statement</em> adds some forms of labels.
10826That is</p>
10827</div>
10828<div class="openblock bnf">
10829<div class="content">
10830<div class="dlist">
10831<dl>
10832<dt class="hdlist1"><em>switch-statement-list</em> : </dt>
10833<dd>
10834<p><em>switch-statement</em><br>
10835<em>switch-statement-list</em> <em>switch-statement</em></p>
10836</dd>
10837<dt class="hdlist1"><em>switch-statement</em> : </dt>
10838<dd>
10839<p><strong>case</strong> <em>constant-expression</em> <strong>:</strong><br>
10840<strong>default</strong> <strong>:</strong> <em>statement</em></p>
10841</dd>
10842</dl>
10843</div>
10844</div>
10845</div>
10846<div class="paragraph">
10847<p>Note the above grammar&#8217;s purpose is to aid discussion in this section; the
10848normative grammar is in &#8220;<a href="#shading-language-grammar">Shading Language
10849Grammar</a>&#8221;.</p>
10850</div>
10851<div class="paragraph">
10852<p>If an <strong>if</strong>-expression evaluates to <strong>true</strong>, then the first <em>statement</em> is
10853executed.
10854If it evaluates to <strong>false</strong> and there is an <strong>else</strong> part then the second
10855<em>statement</em> is executed.</p>
10856</div>
10857<div class="paragraph">
10858<p>Any expression whose type evaluates to a Boolean can be used as the
10859conditional expression <em>bool-expression</em>.
10860Vector types are not accepted as the expression to <strong>if</strong>.</p>
10861</div>
10862<div class="paragraph">
10863<p>Conditionals can be nested.</p>
10864</div>
10865<div class="paragraph">
10866<p>The type of <em>init-expression</em> in a <strong>switch</strong> statement must be a scalar
10867integer.
10868The type of <em>init-expression</em> must match the type of the <strong>case</strong> labels
10869within each <strong>switch</strong> statement.
10870Either signed integers or unsigned integers are allowed but there is no
10871implicit type conversion between the two.
10872If a <strong>case</strong> label has a <em>constant-expression</em> of equal value to
10873<em>init-expression</em>, execution will continue after that label.
10874Otherwise, if there is a <strong>default</strong> label, execution will continue after that
10875label.
10876Otherwise, execution skips the rest of the switch statement.
10877It is an error to have more than one <strong>default</strong> or a replicated
10878<em>constant-expression</em>.
10879A <strong>break</strong> statement not nested in a loop or other switch statement (either
10880not nested or nested only in <strong>if</strong> or <strong>if</strong>-<strong>else</strong> statements) will also skip
10881the rest of the switch statement.
10882Fall through labels are allowed, but it is an error to have no
10883statement between a label and the end of the switch statement.
10884No statements are allowed in a switch statement before the first <strong>case</strong>
10885statement.</p>
10886</div>
10887<div class="paragraph">
10888<p>The <strong>case</strong> and <strong>default</strong> labels can only appear within a <strong>switch</strong> statement.
10889No <strong>case</strong> or <strong>default</strong> labels can be nested inside other statements or
10890compound statements within their corresponding <strong>switch</strong>.</p>
10891</div>
10892</div>
10893<div class="sect2">
10894<h3 id="iteration">6.3. Iteration</h3>
10895<div class="paragraph">
10896<p>For, while, and do loops are allowed as follows:</p>
10897</div>
10898<div class="listingblock">
10899<div class="content">
10900<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">for</span> (init-expression; condition-expression; loop-expression)
10901    sub-statement
10902<span class="keyword">while</span> (condition-expression)
10903    sub-statement
10904<span class="keyword">do</span>
10905    statement
10906<span class="keyword">while</span> (condition-expression)</code></pre>
10907</div>
10908</div>
10909<div class="paragraph">
10910<p>See &#8220;<a href="#shading-language-grammar">Shading Language Grammar</a>&#8221; for the
10911definitive specification of loops.</p>
10912</div>
10913<div class="paragraph">
10914<p>The <strong>for</strong> loop first evaluates the <em>init-expression</em>, then the
10915<em>condition-expression</em>.
10916If the <em>condition-expression</em> evaluates to <strong>true</strong>, then the body of the loop
10917is executed.
10918After the body is executed, a <strong>for</strong> loop will then evaluate the
10919<em>loop-expression</em>, and then loop back to evaluate the
10920<em>condition-expression</em>, repeating until the <em>condition-expression</em> evaluates
10921to <strong>false</strong>.
10922The loop is then exited, skipping its body and skipping its
10923<em>loop-expression</em>.
10924Variables modified by the <em>loop-expression</em> maintain their value after the
10925loop is exited, provided they are still in scope.
10926Variables declared in <em>init-expression</em> or <em>condition-expression</em> are only
10927in scope until the end of the sub-statement of the <strong>for</strong> loop.</p>
10928</div>
10929<div class="paragraph">
10930<p>The <strong>while</strong> loop first evaluates the <em>condition-expression</em>.
10931If <strong>true</strong>, then the body is executed.
10932This is then repeated, until the <em>condition-expression</em> evaluates to
10933<strong>false</strong>, exiting the loop and skipping its body.
10934Variables declared in the <em>condition-expression</em> are only in scope until the
10935end of the sub-statement of the <strong>while</strong> loop.</p>
10936</div>
10937<div class="paragraph">
10938<p>For both <strong>for</strong> and <strong>while</strong> loops, the sub-statement does not introduce a new
10939scope for variable names, so the following has a redeclaration error:</p>
10940</div>
10941<div class="listingblock">
10942<div class="content">
10943<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">for</span> (<span class="predefined-type">int</span> i = <span class="integer">0</span>; i &lt; <span class="integer">10</span>; i++) +
10944{
10945    <span class="predefined-type">int</span> i; <span class="comment">// redeclaration error +</span>
10946}</code></pre>
10947</div>
10948</div>
10949<div class="paragraph">
10950<p>The <strong>do</strong>-<strong>while</strong> loop first executes the body, then executes the
10951<em>condition-expression</em>.
10952This is repeated until <em>condition-expression</em> evaluates to <strong>false</strong>, and then
10953the loop is exited.</p>
10954</div>
10955<div class="paragraph">
10956<p>Expressions for <em>condition-expression</em> must evaluate to a Boolean.</p>
10957</div>
10958<div class="paragraph">
10959<p>Both the <em>condition-expression</em> and the <em>init-expression</em> can declare and
10960initialize a variable, except in the <strong>do</strong>-<strong>while</strong> loop, which cannot declare
10961a variable in its <em>condition-expression</em>.
10962The variable&#8217;s scope lasts only until the end of the sub-statement that
10963forms the body of the loop.</p>
10964</div>
10965<div class="paragraph">
10966<p>Loops can be nested.</p>
10967</div>
10968<div class="paragraph">
10969<p>Non-terminating loops are allowed.
10970The consequences of very long or non-terminating loops are platform
10971dependent.</p>
10972</div>
10973</div>
10974<div class="sect2">
10975<h3 id="jumps">6.4. Jumps</h3>
10976<div class="paragraph">
10977<p>These are the jumps:</p>
10978</div>
10979<div class="openblock bnf">
10980<div class="content">
10981<div class="dlist">
10982<dl>
10983<dt class="hdlist1"><em>jump_statement</em> : </dt>
10984<dd>
10985<p><strong>continue</strong> <strong>;</strong><br>
10986<strong>break</strong> <strong>;</strong><br>
10987<strong>return</strong> <strong>;</strong><br>
10988<strong>return</strong> <em>expression</em> <strong>;</strong><br>
10989<strong>discard</strong> <strong>;</strong> // in the fragment shader language only</p>
10990</dd>
10991</dl>
10992</div>
10993</div>
10994</div>
10995<div class="paragraph">
10996<p>There is no &#8220;goto&#8221; or other non-structured flow of control.</p>
10997</div>
10998<div class="paragraph">
10999<p>The <strong>continue</strong> jump is used only in loops.
11000It skips the remainder of the body of the inner-most loop of which it is
11001inside.
11002For <strong>while</strong> and <strong>do</strong>-<strong>while</strong> loops, this jump is to the next evaluation of
11003the loop <em>condition-expression</em> from which the loop continues as previously
11004defined.
11005For <strong>for</strong> loops, the jump is to the <em>loop-expression</em>, followed by the
11006<em>condition-expression</em>.</p>
11007</div>
11008<div class="paragraph">
11009<p>The <strong>break</strong> jump can also be used only in loops and <strong>switch</strong> statements.
11010It is simply an immediate exit of the inner-most loop or <strong>switch</strong> statements
11011containing the <strong>break</strong>.
11012No further execution of <em>condition-expression</em>, <em>loop-expression</em>, or
11013<em>switch-statement</em> is done.</p>
11014</div>
11015<div class="paragraph">
11016<p>The <strong>discard</strong> keyword is only allowed within fragment shaders.
11017It can be used within a fragment shader to abandon the operation on the
11018current fragment.
11019This keyword causes the fragment to be discarded and no updates to any
11020buffers will occur.
11021Any prior writes to other buffers such as shader storage buffers are
11022unaffected.
11023Control flow exits the shader, and subsequent implicit or explicit
11024derivatives are undefined when this control flow is non-uniform (meaning
11025different fragments within the primitive take different control paths).
11026It would typically be used within a conditional statement, for example:</p>
11027</div>
11028<div class="listingblock">
11029<div class="content">
11030<pre class="CodeRay highlight"><code data-lang="c++"><span class="keyword">if</span> (intensity &lt; <span class="float">0</span><span class="float">.0</span>)
11031    discard;</code></pre>
11032</div>
11033</div>
11034<div class="paragraph">
11035<p>A fragment shader may test a fragment&#8217;s alpha value and discard the fragment
11036based on that test.
11037However, it should be noted that coverage testing occurs after the fragment
11038shader runs, and the coverage test can change the alpha value.</p>
11039</div>
11040<div class="paragraph">
11041<p>The <strong>return</strong> jump causes immediate exit of the current function.
11042If it has <em>expression</em> then that is the return value for the function.</p>
11043</div>
11044<div class="paragraph">
11045<p>The function <em>main</em> can use <strong>return</strong>.
11046This simply causes <em>main</em> to exit in the same way as when the end of the
11047function had been reached.
11048It does not imply a use of <strong>discard</strong> in a fragment shader.
11049Using <strong>return</strong> in <em>main</em> before defining outputs will have the same behavior
11050as reaching the end of <em>main</em> before defining outputs.</p>
11051</div>
11052</div>
11053</div>
11054</div>
11055<div class="sect1">
11056<h2 id="built-in-variables">7. Built-In Variables</h2>
11057<div class="sectionbody">
11058<div class="sect2">
11059<h3 id="built-in-language-variables">7.1. Built-In Language Variables</h3>
11060<div class="paragraph">
11061<p>Some operations occur outside shader functionality and need to provide
11062values to or receive values from shader executables.
11063Shaders communicate with fixed-function pipeline stages, and
11064optionally with other shader executables, through the use of built-in input
11065and output variables.</p>
11066</div>
11067<div class="sect3">
11068<h4 id="vertex-shader-special-variables">7.1.1. Vertex Shader Special Variables</h4>
11069<div class="paragraph">
11070<p>The built-in vertex shader variables are intrinsically declared as follows:</p>
11071</div>
11072<div class="listingblock">
11073<div class="content">
11074<pre class="CodeRay highlight"><code data-lang="c++">in highp <span class="predefined-type">int</span> gl_VertexID;      <span class="comment">// only present when not targeting Vulkan</span>
11075in highp <span class="predefined-type">int</span> gl_InstanceID;    <span class="comment">// only present when not targeting Vulkan</span>
11076in highp <span class="predefined-type">int</span> gl_VertexIndex;   <span class="comment">// only present when targeting Vulkan</span>
11077in highp <span class="predefined-type">int</span> gl_InstanceIndex; <span class="comment">// only present when targeting Vulkan</span>
11078
11079out gl_PerVertex {
11080    out highp vec4 gl_Position;
11081    out highp <span class="predefined-type">float</span> gl_PointSize;
11082};</code></pre>
11083</div>
11084</div>
11085<div class="paragraph">
11086<p>The variable <em>gl_Position</em> is intended for writing the homogeneous vertex
11087position.
11088It can be written at any time during shader execution.
11089This value will be used by primitive assembly, clipping, culling, and other
11090fixed functionality operations, if present, that operate on primitives after
11091vertex processing has occurred.
11092Its value is undefined after the vertex processing stage if the vertex
11093shader executable does not write <em>gl_Position</em>.</p>
11094</div>
11095<div class="paragraph">
11096<p>The variable <em>gl_PointSize</em> is intended for a shader to write the size of
11097the point to be rasterized.
11098It is measured in pixels.
11099If <em>gl_PointSize</em> is not written to, its value is undefined in subsequent
11100pipe stages.</p>
11101</div>
11102<div class="paragraph">
11103<p>The variable <em>gl_VertexID</em> is a vertex shader input variable that holds an
11104integer index for the vertex, as defined under &#8220;Shader Inputs&#8221; in section
1110511.1.3.9 &#8220;Shader Inputs&#8221; of the <a href="#references">OpenGL ES Specification</a>.
11106It is only present when not targeting Vulkan.
11107Even when present, the value of <em>gl_VertexID</em> is not always defined.</p>
11108</div>
11109<div class="paragraph">
11110<p>The variable <em>gl_InstanceID</em> is a vertex shader input variable that holds
11111the instance number of the current primitive in an instanced draw call (see
11112&#8220;Shader Inputs&#8221; in section 11.1.3.9 &#8220;Shader Inputs&#8221; of the
11113<a href="#references">OpenGL ES Specification</a>).
11114It is only present when not targeting Vulkan.
11115If the current primitive does not come from an instanced draw call, the
11116value of <em>gl_InstanceID</em> is zero.</p>
11117</div>
11118<div class="paragraph">
11119<p>The variable <em>gl_VertexIndex</em> is a vertex language input variable that
11120holds an integer index for the vertex, relative to a base.
11121It is only present when targeting Vulkan.
11122Even when present, the value of <em>gl_VertexIndex</em> is not always defined.</p>
11123</div>
11124<div class="paragraph">
11125<p>The variable <em>gl_InstanceIndex</em> is a vertex language input variable that
11126holds the instance number of the current primitive in an instanced draw
11127call, relative to a base.
11128It is only present when targeting Vulkan.
11129If the current primitive does not come from an instanced draw call,
11130the value of gl_InstanceIndex is zero.</p>
11131</div>
11132</div>
11133<div class="sect3">
11134<h4 id="tessellation-control-shader-special-variables">7.1.2. Tessellation Control Shader Special Variables</h4>
11135<div class="paragraph">
11136<p>In the tessellation control shader, built-in variables are intrinsically
11137declared as:</p>
11138</div>
11139<div class="listingblock">
11140<div class="content">
11141<pre class="CodeRay highlight"><code data-lang="c++">in gl_PerVertex {
11142    highp vec4 gl_Position;
11143} gl_in[gl_MaxPatchVertices];
11144
11145in highp <span class="predefined-type">int</span> gl_PatchVerticesIn;
11146in highp <span class="predefined-type">int</span> gl_PrimitiveID;
11147in highp <span class="predefined-type">int</span> gl_InvocationID;
11148
11149out gl_PerVertex {
11150    highp vec4 gl_Position;
11151} gl_out[];
11152
11153patch out highp <span class="predefined-type">float</span> gl_TessLevelOuter[<span class="integer">4</span>];
11154patch out highp <span class="predefined-type">float</span> gl_TessLevelInner[<span class="integer">2</span>];
11155patch out highp vec4 gl_BoundingBox[<span class="integer">2</span>];</code></pre>
11156</div>
11157</div>
11158<div class="sect4">
11159<h5 id="tessellation-control-input-variables">Tessellation Control Input Variables</h5>
11160<div class="paragraph">
11161<p><em>gl_Position</em> contains the output written in the previous shader stage to
11162<em>gl_Position</em>.</p>
11163</div>
11164<div class="paragraph">
11165<p><em>gl_PatchVerticesIn</em> contains the number of vertices in the input patch
11166being processed by the shader.
11167A single shader can read patches of differing sizes, so the value of
11168<em>gl_PatchVerticesIn</em> may differ between patches.</p>
11169</div>
11170<div class="paragraph">
11171<p><em>gl_PrimitiveID</em> contains the number of primitives processed by the shader
11172since the current set of rendering primitives was started.</p>
11173</div>
11174<div class="paragraph">
11175<p><em>gl_InvocationID</em> contains the number of the output patch vertex assigned to
11176the tessellation control shader invocation.
11177It is assigned integer values in the range [0, N-1], where N is the number
11178of output patch vertices per primitive.</p>
11179</div>
11180</div>
11181<div class="sect4">
11182<h5 id="tessellation-control-output-variables">Tessellation Control Output Variables</h5>
11183<div class="paragraph">
11184<p><em>gl_Position</em> is used in the same fashion as the corresponding output
11185variable in the vertex shader.</p>
11186</div>
11187<div class="paragraph">
11188<p>The values written to <em>gl_TessLevelOuter</em> and <em>gl_TessLevelInner</em> are
11189assigned to the corresponding outer and inner tessellation levels of the
11190output patch.
11191They are used by the tessellation primitive generator to control primitive
11192tessellation and may be read by tessellation evaluation shaders.</p>
11193</div>
11194<div class="paragraph">
11195<p>The values written to <em>gl_BoundingBox</em> specify the minimum and maximum
11196clip-space extents of a bounding box containing all primitives generated
11197from the patch by the primitive generator, geometry shader, and clipping.
11198Fragments may or may not be generated for portions of these primitives
11199that extend outside the window-coordinate projection of this bounding
11200box.</p>
11201</div>
11202</div>
11203</div>
11204<div class="sect3">
11205<h4 id="tessellation-evaluation-shader-special-variables">7.1.3. Tessellation Evaluation Shader Special Variables</h4>
11206<div class="paragraph">
11207<p>In the tessellation evaluation shader, built-in variables are intrinsically
11208declared as:</p>
11209</div>
11210<div class="listingblock">
11211<div class="content">
11212<pre class="CodeRay highlight"><code data-lang="c++">in gl_PerVertex {
11213    highp vec4 gl_Position;
11214} gl_in[gl_MaxPatchVertices];
11215
11216in highp <span class="predefined-type">int</span> gl_PatchVerticesIn;
11217in highp <span class="predefined-type">int</span> gl_PrimitiveID;
11218in highp vec3 gl_TessCoord;
11219patch in highp <span class="predefined-type">float</span> gl_TessLevelOuter[<span class="integer">4</span>];
11220patch in highp <span class="predefined-type">float</span> gl_TessLevelInner[<span class="integer">2</span>];
11221
11222out gl_PerVertex {
11223    highp vec4 gl_Position;
11224};</code></pre>
11225</div>
11226</div>
11227<div class="sect4">
11228<h5 id="tessellation-evaluation-input-variables">Tessellation Evaluation Input Variables</h5>
11229<div class="paragraph">
11230<p><em>gl_Position</em> contains the output written in the previous shader stage to
11231<em>gl_Position</em>.</p>
11232</div>
11233<div class="paragraph">
11234<p><em>gl_PatchVerticesIn</em> and <em>gl_PrimitiveID</em> are defined in the same fashion as
11235the corresponding input variables in the tessellation control shader.</p>
11236</div>
11237<div class="paragraph">
11238<p><em>gl_TessCoord</em> specifies a three-component <em>(u,v,w)</em> vector identifying the
11239position of the vertex being processed by the shader relative to the
11240primitive being tessellated.
11241Its values will obey the properties</p>
11242</div>
11243<div class="listingblock">
11244<div class="content">
11245<pre class="CodeRay highlight"><code data-lang="c++">gl_TessCoord.x == <span class="float">1</span><span class="float">.0</span> - (<span class="float">1</span><span class="float">.0</span> - gl_TessCoord.x) <span class="comment">// two operations performed</span>
11246gl_TessCoord.y == <span class="float">1</span><span class="float">.0</span> - (<span class="float">1</span><span class="float">.0</span> - gl_TessCoord.y) <span class="comment">// two operations performed</span>
11247gl_TessCoord.z == <span class="float">1</span><span class="float">.0</span> - (<span class="float">1</span><span class="float">.0</span> - gl_TessCoord.z) <span class="comment">// two operations performed</span></code></pre>
11248</div>
11249</div>
11250<div class="paragraph">
11251<p>to aid in replicating subdivision computations.</p>
11252</div>
11253<div class="paragraph">
11254<p><em>gl_TessLevelOuter</em> and <em>gl_TessLevelInner</em> are filled with the
11255corresponding outputs written by the active tessellation control shader.</p>
11256</div>
11257</div>
11258<div class="sect4">
11259<h5 id="tessellation-evaluation-output-variables">Tessellation Evaluation Output Variables</h5>
11260<div class="paragraph">
11261<p><em>gl_Position</em> is used in the same fashion as the corresponding output
11262variable in the vertex shader.</p>
11263</div>
11264</div>
11265</div>
11266<div class="sect3">
11267<h4 id="geometry-shader-special-variables">7.1.4. Geometry Shader Special Variables</h4>
11268<div class="paragraph">
11269<p>In the geometry shader, built-in variables are intrinsically declared as:</p>
11270</div>
11271<div class="listingblock">
11272<div class="content">
11273<pre class="CodeRay highlight"><code data-lang="c++">in gl_PerVertex {
11274    highp vec4 gl_Position;
11275} gl_in[];
11276
11277in highp <span class="predefined-type">int</span> gl_PrimitiveIDIn;
11278in highp <span class="predefined-type">int</span> gl_InvocationID;
11279
11280out gl_PerVertex {
11281    highp vec4 gl_Position;
11282};
11283
11284out highp <span class="predefined-type">int</span> gl_PrimitiveID;
11285out highp <span class="predefined-type">int</span> gl_Layer;</code></pre>
11286</div>
11287</div>
11288<div class="sect4">
11289<h5 id="geometry-shader-input-variables">Geometry Shader Input Variables</h5>
11290<div class="paragraph">
11291<p><em>gl_Position</em> contains the output written in the previous shader stage to
11292<em>gl_Position</em>.</p>
11293</div>
11294<div class="paragraph">
11295<p><em>gl_PrimitiveIDIn</em> contains the number of primitives processed by the shader
11296since the current set of rendering primitives was started.</p>
11297</div>
11298<div class="paragraph">
11299<p><em>gl_InvocationID</em> contains the invocation number assigned to the geometry
11300shader invocation.
11301It is assigned integer values in the range [0, N-1], where N is the number
11302of geometry shader invocations per primitive.</p>
11303</div>
11304</div>
11305<div class="sect4">
11306<h5 id="geometry-shader-output-variables">Geometry Shader Output Variables</h5>
11307<div class="paragraph">
11308<p><em>gl_Position</em> is used in the same fashion as the corresponding output
11309variable in the vertex shader.</p>
11310</div>
11311<div class="paragraph">
11312<p><em>gl_PrimitiveID</em> is filled with a single integer that serves as a primitive
11313identifier to the fragment shader.
11314This is then available to fragment shaders, which will select the written
11315primitive ID from the provoking vertex of the primitive being shaded.
11316If a fragment shader using <em>gl_PrimitiveID</em> is active and a geometry shader
11317is also active, the geometry shader must write to <em>gl_PrimitiveID</em> or the
11318fragment shader input <em>gl_PrimitiveID</em> is undefined.
11319See section 11.3.4.4 &#8220;Geometry Shader Outputs&#8221; of the
11320<a href="#references">OpenGL ES Specification</a> for more information.</p>
11321</div>
11322<div class="paragraph">
11323<p><em>gl_Layer</em> is used to select a specific layer (or face and layer of a cube
11324map) of a multi-layer framebuffer attachment.
11325The actual layer used will come from one of the vertices in the primitive
11326being shaded.
11327Which vertex the layer comes from is determined as discussed in section
1132811.3.4.4 of the <a href="#references">OpenGL ES Specification</a>
11329but may be undefined, so it is best to write the same layer value for all
11330vertices of a primitive.
11331If a shader statically assigns a value to <em>gl_Layer</em>, layered rendering mode
11332is enabled.
11333See section 11.3.4.4 &#8220;Geometry Shader Outputs&#8221; and section 9.8 &#8220;Layered
11334Framebuffers&#8221; of the <a href="#references">OpenGL ES Specification</a> for more information.
11335If a shader statically assigns a value to <em>gl_Layer</em>, and there is an
11336execution path through the shader that does not set <em>gl_Layer</em>, then the
11337value of <em>gl_Layer</em> is undefined for executions of the shader that take that
11338path.</p>
11339</div>
11340<div class="paragraph">
11341<p>The output variable <em>gl_Layer</em> takes on a special value when used with an
11342array of cube map textures.
11343Instead of only referring to the layer, it is used to select a cube map face
11344and a layer.
11345Setting <em>gl_Layer</em> to the value <em>layer*6+face</em> will render to face <em>face</em> of
11346the cube defined in layer <em>layer</em>.
11347The face values are defined in table 8.25 of the <a href="#references">OpenGL ES Specification</a>,
11348but repeated below for clarity.</p>
11349</div>
11350<table class="tableblock frame-all grid-all stretch">
11351<colgroup>
11352<col style="width: 50%;">
11353<col style="width: 50%;">
11354</colgroup>
11355<thead>
11356<tr>
11357<th class="tableblock halign-left valign-top">Face Value</th>
11358<th class="tableblock halign-left valign-top">Resulting Target</th>
11359</tr>
11360</thead>
11361<tbody>
11362<tr>
11363<td class="tableblock halign-left valign-top"><p class="tableblock">0</p></td>
11364<td class="tableblock halign-left valign-top"><p class="tableblock">TEXTURE_CUBE_MAP_POSITIVE_X</p></td>
11365</tr>
11366<tr>
11367<td class="tableblock halign-left valign-top"><p class="tableblock">1</p></td>
11368<td class="tableblock halign-left valign-top"><p class="tableblock">TEXTURE_CUBE_MAP_NEGATIVE_X</p></td>
11369</tr>
11370<tr>
11371<td class="tableblock halign-left valign-top"><p class="tableblock">2</p></td>
11372<td class="tableblock halign-left valign-top"><p class="tableblock">TEXTURE_CUBE_MAP_POSITIVE_Y</p></td>
11373</tr>
11374<tr>
11375<td class="tableblock halign-left valign-top"><p class="tableblock">3</p></td>
11376<td class="tableblock halign-left valign-top"><p class="tableblock">TEXTURE_CUBE_MAP_NEGATIVE_Y</p></td>
11377</tr>
11378<tr>
11379<td class="tableblock halign-left valign-top"><p class="tableblock">4</p></td>
11380<td class="tableblock halign-left valign-top"><p class="tableblock">TEXTURE_CUBE_MAP_POSITIVE_Z</p></td>
11381</tr>
11382<tr>
11383<td class="tableblock halign-left valign-top"><p class="tableblock">5</p></td>
11384<td class="tableblock halign-left valign-top"><p class="tableblock">TEXTURE_CUBE_MAP_NEGATIVE_Z</p></td>
11385</tr>
11386</tbody>
11387</table>
11388<div class="paragraph">
11389<p>For example, to render to the positive <em>y</em> cube map face located in the 5th
11390layer of the cube map array, <em>gl_Layer</em> should be set to <em>5 * 6 + 2</em>.</p>
11391</div>
11392</div>
11393</div>
11394<div class="sect3">
11395<h4 id="fragment-shader-special-variables">7.1.5. Fragment Shader Special Variables</h4>
11396<div class="paragraph">
11397<p>The built-in special variables that are accessible from a fragment shader
11398are intrinsically declared as follows:</p>
11399</div>
11400<div class="listingblock">
11401<div class="content">
11402<pre class="CodeRay highlight"><code data-lang="c++">in highp vec4 gl_FragCoord;
11403in <span class="predefined-type">bool</span> gl_FrontFacing;
11404out highp <span class="predefined-type">float</span> gl_FragDepth;
11405in mediump vec2 gl_PointCoord;
11406in <span class="predefined-type">bool</span> gl_HelperInvocation;
11407in highp <span class="predefined-type">int</span> gl_PrimitiveID;
11408in highp <span class="predefined-type">int</span> gl_Layer;
11409in lowp <span class="predefined-type">int</span> gl_SampleID;
11410in mediump vec2 gl_SamplePosition;
11411in highp <span class="predefined-type">int</span> gl_SampleMaskIn[(gl_MaxSamples+<span class="integer">31</span>)/<span class="integer">32</span>];
11412out highp <span class="predefined-type">int</span> gl_SampleMask[(gl_MaxSamples+<span class="integer">31</span>)/<span class="integer">32</span>];</code></pre>
11413</div>
11414</div>
11415<div class="paragraph">
11416<p>The output of the fragment shader executable is processed by the fixed
11417function operations at the back end of the API pipeline.</p>
11418</div>
11419<div class="paragraph">
11420<p>The fixed functionality computed depth for a fragment may be obtained by
11421reading <em>gl_FragCoord.z</em>, described below.</p>
11422</div>
11423<div class="paragraph">
11424<p>Writing to <em>gl_FragDepth</em> will establish the depth value for the fragment
11425being processed.
11426If depth buffering is enabled, and no shader writes <em>gl_FragDepth</em>, then the
11427fixed function value for depth will be used as the fragment&#8217;s depth value.
11428If a shader statically assigns a value to <em>gl_FragDepth</em>, and there is an
11429execution path through the shader that does not set <em>gl_FragDepth</em>, then the
11430value of the fragment&#8217;s depth may be undefined for executions of the shader
11431that take that path.
11432That is, if the set of linked fragment shaders statically contain a write to
11433<em>gl_FragDepth</em>, then it is responsible for always writing it.</p>
11434</div>
11435<div class="paragraph">
11436<p>If a shader executes the <strong>discard</strong> keyword, the fragment is discarded, and
11437the values of any user-defined fragment outputs, <em>gl_FragDepth</em>, and
11438<em>gl_SampleMask</em> become irrelevant.</p>
11439</div>
11440<div class="paragraph">
11441<p>The variable <em>gl_FragCoord</em> is available as an input variable from within
11442fragment shaders and it holds the window relative coordinates (<em>x</em>, <em>y</em>,
11443<em>z</em>, <em>1/w</em>) values for the fragment.
11444If multi-sampling, this value can be for any location within the pixel, or
11445one of the fragment samples.
11446The use of <strong>centroid</strong> does not further restrict this value to be inside the
11447current primitive.
11448This value is the result of the fixed functionality that interpolates
11449primitives after vertex processing to generate fragments.
11450The <em>z</em> component is the depth value that would be used for the fragment&#8217;s
11451depth if no shader contained any writes to <em>gl_FragDepth</em>.
11452This is useful for invariance if a shader conditionally computes
11453<em>gl_FragDepth</em> but otherwise wants the fixed functionality fragment depth.</p>
11454</div>
11455<div class="paragraph">
11456<p>Fragment shaders have access to the input built-in variable
11457<em>gl_FrontFacing</em>, whose value is <strong>true</strong> if the fragment belongs to a
11458front-facing primitive.
11459One use of this is to emulate two-sided lighting by selecting one of two
11460colors calculated by a vertex or geometry shader.</p>
11461</div>
11462<div class="paragraph">
11463<p>The values in <em>gl_PointCoord</em> are two-dimensional coordinates indicating
11464where within a point primitive the current fragment is located, when point
11465sprites are enabled.
11466They range from 0.0 to 1.0 across the point.
11467If the current primitive is not a point, or if point sprites are not
11468enabled, then the values read from <em>gl_PointCoord</em> are undefined.</p>
11469</div>
11470<div class="paragraph">
11471<p>For both the input array <em>gl_SampleMaskIn[]</em> and the output array
11472<em>gl_SampleMask[]</em>, bit <em>B</em> of mask <em>M</em> (<em>gl_SampleMaskIn[M]</em> or
11473<em>gl_SampleMask[M]</em>) corresponds to sample <em>32*M+B</em>.
11474These arrays have <strong>ceil</strong>(<em>s</em>/32) elements, where <em>s</em> is the maximum number
11475of color samples supported by the implementation.</p>
11476</div>
11477<div class="paragraph">
11478<p>The input variable <em>gl_SampleMaskIn</em> indicates the set of samples covered by
11479the primitive generating the fragment during multisample rasterization.
11480It has a sample bit set if and only if the sample is considered covered for
11481this fragment shader invocation.</p>
11482</div>
11483<div class="paragraph">
11484<p>The output array <em>gl_SampleMask[]</em> sets the sample mask for the fragment
11485being processed.
11486Coverage for the current fragment will become the logical AND of the
11487coverage mask and the output <em>gl_SampleMask</em>.
11488This array must be sized in the fragment shader either implicitly or
11489explicitly, to be no larger than the implementation-dependent maximum
11490sample-mask (as an array of 32bit elements), determined by the maximum
11491number of samples..
11492If the fragment shader statically assigns a value to <em>gl_SampleMask</em>, the
11493sample mask will be undefined for any array elements of any fragment shader
11494invocations that fail to assign a value.
11495If a shader does not statically assign a value to <em>gl_SampleMask</em>, the
11496sample mask has no effect on the processing of a fragment.</p>
11497</div>
11498<div class="paragraph">
11499<p>The input variable <em>gl_SampleID</em> is filled with the sample number of the
11500sample currently being processed.
11501This variable is in the range <em>0</em> to <em>gl_NumSamples-1</em>, where
11502<em>gl_NumSamples</em> is the total number of samples in the framebuffer, or 1 if
11503rendering to a non-multisample framebuffer.
11504Any static use of this variable in a fragment shader causes the entire
11505shader to be evaluated per-sample.</p>
11506</div>
11507<div class="paragraph">
11508<p>The input variable <em>gl_SamplePosition</em> contains the position of the current
11509sample within the multisample draw buffer.
11510The <em>x</em> and <em>y</em> components of <em>gl_SamplePosition</em> contain the sub-pixel
11511coordinate of the current sample and will have values in the range 0.0 to
115121.0.
11513Any static use of this variable in a fragment shader causes the entire
11514shader to be evaluated per sample.</p>
11515</div>
11516<div class="paragraph">
11517<p>The value <em>gl_HelperInvocation</em> is <strong>true</strong> if the fragment shader invocation
11518is considered a <em>helper invocation</em> and is <strong>false</strong> otherwise.
11519A helper invocation is a fragment shader invocation that is created solely
11520for the purposes of evaluating derivatives for use in non-helper fragment
11521shader invocations.
11522Such derivatives are computed implicitly in the built-in function
11523<strong>texture</strong>() (see &#8220;<a href="#texture-functions">Texture Functions</a>&#8221;), and
11524explicitly in the derivative functions in
11525&#8220;<a href="#derivative-functions">Derivative Functions</a>&#8221;, for example <strong>dFdx</strong>() and
11526<strong>dFdy</strong>().</p>
11527</div>
11528<div class="paragraph">
11529<p>Fragment shader helper invocations execute the same shader code as
11530non-helper invocations, but will not have side effects that modify the
11531framebuffer or other shader-accessible memory.
11532In particular:</p>
11533</div>
11534<div class="ulist">
11535<ul>
11536<li>
11537<p>Fragments corresponding to helper invocations are discarded when shader
11538execution is complete, without updating the framebuffer.</p>
11539</li>
11540<li>
11541<p>Stores to image and buffer variables performed by helper invocations
11542have no effect on the underlying image or buffer memory.</p>
11543</li>
11544<li>
11545<p>Atomic operations to image, buffer, or atomic counter variables
11546performed by helper invocations have no effect on the underlying image
11547or buffer memory.
11548The values returned by such atomic operations are undefined.</p>
11549</li>
11550</ul>
11551</div>
11552<div class="paragraph">
11553<p>Helper invocations may be generated for pixels not covered by a primitive
11554being rendered.
11555While fragment shader inputs qualified with <strong>centroid</strong> are normally required
11556to be sampled in the intersection of the pixel and the primitive, the
11557requirement is ignored for such pixels since there is no intersection
11558between the pixel and primitive.</p>
11559</div>
11560<div class="paragraph">
11561<p>Helper invocations may also be generated for fragments that are covered by a
11562primitive being rendered when the fragment is killed by early fragment tests
11563(using the <strong>early_fragment_tests</strong> qualifier) or where the implementation is
11564able to determine that executing the fragment shader would have no effect
11565other than assisting in computing derivatives for other fragment shader
11566invocations.</p>
11567</div>
11568<div class="paragraph">
11569<p>The set of helper invocations generated when processing any set of
11570primitives is implementation-dependent.</p>
11571</div>
11572<div class="paragraph">
11573<p>The input variable <em>gl_PrimitiveID</em> is filled with the value written to the
11574<em>gl_PrimitiveID</em> geometry shader output, if a geometry shader is present.
11575Otherwise, it is filled with the number of primitives processed by the
11576shader since the current set of rendering primitives was started.</p>
11577</div>
11578<div class="paragraph">
11579<p>The input variable <em>gl_Layer</em> is filled with the value written to the
11580<em>gl_Layer</em> geometry shader output, if a geometry shader is present.
11581If the geometry stage does not dynamically assign a value to <em>gl_Layer</em>, the
11582value of <em>gl_Layer</em> in the fragment stage will be undefined.
11583If the geometry stage makes no static assignment to <em>gl_Layer</em>, the input
11584value in the fragment stage will be zero.
11585Otherwise, the fragment stage will read the same value written by the
11586geometry stage, even if that value is out of range.
11587If a fragment shader contains a static access to <em>gl_Layer</em>, it will count
11588against the implementation defined limit for the maximum number of inputs to
11589the fragment stage.</p>
11590</div>
11591</div>
11592<div class="sect3">
11593<h4 id="compute-shader-special-variables">7.1.6. Compute Shader Special Variables</h4>
11594<div class="paragraph">
11595<p>In the compute shader, built-in variables are declared as follows:</p>
11596</div>
11597<div class="listingblock">
11598<div class="content">
11599<pre class="CodeRay highlight"><code data-lang="c++"><span class="comment">// workgroup dimensions</span>
11600in uvec3 gl_NumWorkGroups;
11601<span class="directive">const</span> uvec3 gl_WorkGroupSize;
11602
11603<span class="comment">// workgroup and invocation IDs</span>
11604in uvec3 gl_WorkGroupID;
11605in uvec3 gl_LocalInvocationID;
11606
11607<span class="comment">// derived variables</span>
11608in uvec3 gl_GlobalInvocationID;
11609in uint gl_LocalInvocationIndex;</code></pre>
11610</div>
11611</div>
11612<div class="paragraph">
11613<p>The built-in variable <em>gl_NumWorkGroups</em> is a compute-shader input variable
11614containing the number of workgroups in each dimension of the dispatch that
11615will execute the compute shader.
11616Its content is equal to the values specified in the <em>num_groups_x</em>,
11617<em>num_groups_y</em>, and <em>num_groups_z</em> parameters passed to the
11618<em>DispatchCompute</em> API entry point.</p>
11619</div>
11620<div class="paragraph">
11621<p>The built-in constant <em>gl_WorkGroupSize</em> is a compute-shader constant
11622containing the workgroup size of the shader.
11623The size of the workgroup in the <em>X</em>, <em>Y</em>, and <em>Z</em> dimensions is stored in
11624the <em>x</em>, <em>y</em>, and <em>z</em> components.
11625The constants values in <em>gl_WorkGroupSize</em> will match those specified in the
11626required <strong>local_size_x</strong>, <strong>local_size_y</strong>, and <strong>local_size_z</strong> layout
11627qualifiers for the current shader.
11628This is a constant so that it can be used to size arrays of memory that can
11629be shared within the workgroup.
11630It is a compile-time error to use <em>gl_WorkGroupSize</em> in a shader that does
11631not declare a fixed workgroup size, or before that shader has declared a
11632fixed workgroup size, using <strong>local_size_x</strong>, <strong>local_size_y</strong>, and
11633<strong>local_size_z</strong>.</p>
11634</div>
11635<div class="paragraph">
11636<p>The built-in variable <em>gl_WorkGroupID</em> is a compute-shader input variable
11637containing the three-dimensional index of the workgroup that the
11638current invocation is executing in.
11639The possible values range across the parameters passed into
11640<em>DispatchCompute</em>, i.e., from (0, 0, 0) to (<em>gl_NumWorkGroups.x</em> - 1,
11641<em>gl_NumWorkGroups.y</em> - 1, <em>gl_NumWorkGroups.z</em> -1).</p>
11642</div>
11643<div class="paragraph">
11644<p>The built-in variable <em>gl_LocalInvocationID</em> is a compute-shader input
11645variable containing the three-dimensional index of the current work item
11646within the workgroup.
11647The possible values for this variable range across the workgroup
11648size, i.e., (0,0,0) to (<em>gl_WorkGroupSize.x</em> - 1, <em>gl_WorkGroupSize.y</em> - 1,
11649<em>gl_WorkGroupSize.z</em> - 1). Use of <em>gl_LocalInvocationID</em> is allowed
11650before declarations of <strong>local_size_x</strong>, <strong>local_size_y</strong>, and <strong>local_size_z</strong>.</p>
11651</div>
11652<div class="paragraph">
11653<p>The built-in variable <em>gl_GlobalInvocationID</em> is a compute shader input
11654variable containing the global index of the current work item.
11655This value uniquely identifies this invocation from all other invocations
11656across all workgroups initiated by the current <em>DispatchCompute</em> call.
11657This is computed as:</p>
11658</div>
11659<div class="listingblock">
11660<div class="content">
11661<pre class="CodeRay highlight"><code data-lang="c++">gl_GlobalInvocationID =
11662    gl_WorkGroupID * gl_WorkGroupSize + gl_LocalInvocationID;</code></pre>
11663</div>
11664</div>
11665<div class="paragraph">
11666<p>The built-in variable <em>gl_LocalInvocationIndex</em> is a compute shader input
11667variable that contains the one-dimensional representation of the
11668<em>gl_LocalInvocationID</em>.
11669This is computed as:</p>
11670</div>
11671<div class="listingblock">
11672<div class="content">
11673<pre class="CodeRay highlight"><code data-lang="c++">gl_LocalInvocationIndex =
11674    gl_LocalInvocationID.z * gl_WorkGroupSize.x * gl_WorkGroupSize.y +
11675    gl_LocalInvocationID.y * gl_WorkGroupSize.x +
11676    gl_LocalInvocationID.x;</code></pre>
11677</div>
11678</div>
11679<div class="paragraph">
11680<p>Use of <em>gl_LocalInvocationIndex</em> is allowed before declarations of
11681<strong>local_size_x</strong>, <strong>local_size_y</strong>, and <strong>local_size_z</strong>.</p>
11682</div>
11683</div>
11684</div>
11685<div class="sect2">
11686<h3 id="built-in-constants">7.2. Built-In Constants</h3>
11687<div class="paragraph">
11688<p>The following built-in constants are declared in all shaders.
11689The actual values used are implementation-dependent, but must be at least
11690the value shown.</p>
11691</div>
11692<div class="listingblock">
11693<div class="content">
11694<pre class="CodeRay highlight"><code data-lang="c++"><span class="comment">//</span>
11695<span class="comment">// Implementation-dependent constants. The example values below</span>
11696<span class="comment">// are the minimum values allowed for these maximums.</span>
11697<span class="comment">//</span></code></pre>
11698</div>
11699</div>
11700<div class="listingblock">
11701<div class="content">
11702<pre class="CodeRay highlight"><code data-lang="c++"><span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxVertexAttribs = <span class="integer">16</span>;
11703<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxVertexUniformVectors = <span class="integer">256</span>;
11704<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxVertexOutputVectors = <span class="integer">16</span>;
11705<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxVertexTextureImageUnits = <span class="integer">16</span>;
11706<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxVertexImageUniforms = <span class="integer">0</span>;
11707<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxVertexAtomicCounters = <span class="integer">0</span>;
11708<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxVertexAtomicCounterBuffers = <span class="integer">0</span>;
11709
11710<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessControlInputComponents = <span class="integer">64</span>;
11711<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessControlOutputComponents = <span class="integer">64</span>;
11712<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessControlTextureImageUnits = <span class="integer">16</span>;
11713<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessControlUniformComponents = <span class="integer">1024</span>;
11714<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessControlTotalOutputComponents = <span class="integer">2048</span>;
11715<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessControlImageUniforms = <span class="integer">0</span>;
11716<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessControlAtomicCounters = <span class="integer">0</span>;
11717<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessControlAtomicCounterBuffers = <span class="integer">0</span>;
11718
11719<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessPatchComponents = <span class="integer">120</span>;
11720<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxPatchVertices = <span class="integer">32</span>;
11721<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessGenLevel = <span class="integer">64</span>;
11722
11723<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessEvaluationInputComponents = <span class="integer">64</span>;
11724<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessEvaluationOutputComponents = <span class="integer">64</span>;
11725<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessEvaluationTextureImageUnits = <span class="integer">16</span>;
11726<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessEvaluationUniformComponents = <span class="integer">1024</span>;
11727<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessEvaluationImageUniforms = <span class="integer">0</span>;
11728<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessEvaluationAtomicCounters = <span class="integer">0</span>;
11729<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTessEvaluationAtomicCounterBuffers = <span class="integer">0</span>;
11730
11731<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxGeometryInputComponents = <span class="integer">64</span>;
11732<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxGeometryOutputComponents = <span class="integer">64</span>;
11733<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxGeometryImageUniforms = <span class="integer">0</span>;
11734<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxGeometryTextureImageUnits = <span class="integer">16</span>;
11735<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxGeometryOutputVertices = <span class="integer">256</span>;
11736<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxGeometryTotalOutputComponents = <span class="integer">1024</span>;
11737<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxGeometryUniformComponents = <span class="integer">1024</span>;
11738<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxGeometryAtomicCounters = <span class="integer">0</span>;
11739<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxGeometryAtomicCounterBuffers = <span class="integer">0</span>;
11740
11741<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxFragmentInputVectors = <span class="integer">15</span>;
11742<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxFragmentImageUniforms = <span class="integer">4</span>;
11743<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxFragmentUniformVectors = <span class="integer">256</span>;
11744<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxFragmentAtomicCounters = <span class="integer">8</span>;
11745<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxFragmentAtomicCounterBuffers = <span class="integer">1</span>;
11746
11747<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxDrawBuffers = <span class="integer">4</span>;
11748<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxTextureImageUnits = <span class="integer">16</span>;
11749<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MinProgramTexelOffset = -<span class="integer">8</span>;
11750<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxProgramTexelOffset = <span class="integer">7</span>;
11751<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxImageUnits = <span class="integer">4</span>;
11752<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxSamples = <span class="integer">4</span>;
11753
11754<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxComputeImageUniforms = <span class="integer">4</span>;
11755<span class="directive">const</span> highp ivec3 gl_MaxComputeWorkGroupCount = ivec3(<span class="integer">65535</span>, <span class="integer">65535</span>, <span class="integer">65535</span>);
11756<span class="directive">const</span> highp ivec3 gl_MaxComputeWorkGroupSize = ivec3(<span class="integer">128</span>, <span class="integer">128</span>, <span class="integer">64</span>);
11757<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxComputeUniformComponents = <span class="integer">1024</span>;
11758<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxComputeTextureImageUnits = <span class="integer">16</span>;
11759<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxComputeAtomicCounters = <span class="integer">8</span>;
11760<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxComputeAtomicCounterBuffers = <span class="integer">1</span>;
11761
11762<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxCombinedTextureImageUnits = <span class="integer">96</span>;
11763<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxCombinedImageUniforms = <span class="integer">4</span>;
11764<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxCombinedShaderOutputResources = <span class="integer">4</span>;
11765<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxCombinedAtomicCounters = <span class="integer">8</span>;
11766<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxCombinedAtomicCounterBuffers = <span class="integer">1</span>;
11767<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxAtomicCounterBindings = <span class="integer">1</span>;
11768<span class="directive">const</span> mediump <span class="predefined-type">int</span> gl_MaxAtomicCounterBufferSize = <span class="integer">32</span>;
11769
11770<span class="directive">const</span> highp <span class="predefined-type">int</span> gl_MaxInputAttachments = <span class="integer">1</span>;  <span class="comment">// only present when targeting Vulkan</span></code></pre>
11771</div>
11772</div>
11773</div>
11774<div class="sect2">
11775<h3 id="built-in-uniform-state">7.3. Built-In Uniform State</h3>
11776<div class="paragraph">
11777<p>Built-in uniform state is not available when generating SPIR-V.
11778Otherwise, as an aid to accessing OpenGL ES processing state, the following
11779uniform variables are built into the OpenGL ES Shading Language.</p>
11780</div>
11781<div class="listingblock">
11782<div class="content">
11783<pre class="CodeRay highlight"><code data-lang="c++"><span class="comment">//</span>
11784<span class="comment">// Depth range in window coordinates,</span>
11785<span class="comment">// section 12.5.1 &quot;Controlling the Viewport&quot; in the</span>
11786<span class="comment">// OpenGL ES Specification.</span>
11787<span class="comment">//</span>
11788<span class="keyword">struct</span> gl_DepthRangeParameters {
11789    highp <span class="predefined-type">float</span> near; <span class="comment">// n</span>
11790    highp <span class="predefined-type">float</span> far;  <span class="comment">// f</span>
11791    highp <span class="predefined-type">float</span> diff; <span class="comment">// f - n</span>
11792};
11793uniform gl_DepthRangeParameters gl_DepthRange;
11794uniform lowp <span class="predefined-type">int</span> gl_NumSamples;</code></pre>
11795</div>
11796</div>
11797<div class="paragraph">
11798<p>These variables are only guaranteed to be available in the fragment stage.
11799In other stages, their presence and function is implementation-defined.</p>
11800</div>
11801</div>
11802<div class="sect2">
11803<h3 id="redeclaring-built-in-blocks">7.4. Redeclaring Built-In Blocks</h3>
11804<div class="paragraph">
11805<p>The <em>gl_PerVertex</em> block can be redeclared in a shader to explicitly
11806indicate what subset of the fixed pipeline interface will be used.
11807This is necessary to establish the interface between multiple programs.</p>
11808</div>
11809<div class="paragraph">
11810<p>If the <em>gl_PerVertex</em> block is not redefined in a given program, the
11811intrinsically declared definition of that block is used for the program
11812interface.</p>
11813</div>
11814<div class="paragraph">
11815<p>For example:</p>
11816</div>
11817<div class="listingblock">
11818<div class="content">
11819<pre class="CodeRay highlight"><code data-lang="c++">out gl_PerVertex {
11820    highp vec4 gl_Position; <span class="comment">// will use gl_Position</span>
11821    highp vec4 t;           <span class="comment">// error, only gl_PerVertex members allowed</span>
11822}; <span class="comment">// no other members of gl_PerVertex will be used</span></code></pre>
11823</div>
11824</div>
11825<div class="paragraph">
11826<p>This establishes the output interface the shader will use with the
11827subsequent pipeline stage.
11828It must be a subset of the built-in members of <em>gl_PerVertex</em>.
11829Such a redeclaration can also add the <strong>invariant</strong> qualifier and
11830interpolation qualifiers.</p>
11831</div>
11832<div class="paragraph">
11833<p>Other layout qualifiers, like <strong>location</strong>, cannot be added to such a
11834redeclaration, unless specifically stated.</p>
11835</div>
11836<div class="paragraph">
11837<p>If a built-in interface block is redeclared, it must appear in the shader
11838before any use of any member included in the built-in declaration, or a
11839compile-time error will result.
11840It is also a compile-time error to redeclare the block more than once or to
11841redeclare a built-in block and then use a member from that built-in block
11842that was not included in the redeclaration.
11843Also, if a built-in interface block is redeclared, no member of the built-in
11844declaration can be redeclared outside the block redeclaration.
11845If multiple shaders using members of a built-in block belonging to the same
11846interface are linked together in the same program, they must all redeclare
11847the built-in block in the same way, as described in
11848&#8220;<a href="#interface-blocks">Interface Blocks</a>&#8221; for interface block matching, or a
11849link-time error will result.
11850It will also be a link-time error if some shaders in a program redeclare a
11851specific built-in interface block while another shader in that program does
11852not redeclare that interface block yet still uses a member of that interface
11853block.
11854If a built-in block interface is formed across shaders in different
11855programs, the shaders must all redeclare the built-in block in the same way
11856(as described for a single program), or the values passed along the
11857interface are undefined.</p>
11858</div>
11859</div>
11860</div>
11861</div>
11862<div class="sect1">
11863<h2 id="built-in-functions">8. Built-In Functions</h2>
11864<div class="sectionbody">
11865<div class="paragraph">
11866<p>The OpenGL ES Shading Language defines an assortment of built-in convenience functions for
11867scalar and vector operations.
11868Many of these built-in functions can be used in more than one type of
11869shader, but some are intended to provide a direct mapping to hardware and so
11870are available only for a specific type of shader.</p>
11871</div>
11872<div class="paragraph">
11873<p>The built-in functions basically fall into three categories:</p>
11874</div>
11875<div class="ulist">
11876<ul>
11877<li>
11878<p>They expose some necessary hardware functionality in a convenient way
11879such as accessing a texture map.
11880There is no way in the language for these functions to be emulated by a
11881shader.</p>
11882</li>
11883<li>
11884<p>They represent a trivial operation (clamp, mix, etc.) that is very
11885simple for the user to write, but they are very common and may have
11886direct hardware support.
11887It is a very hard problem for the compiler to map expressions to complex
11888assembler instructions.</p>
11889</li>
11890<li>
11891<p>They represent an operation graphics hardware is likely to accelerate at
11892some point.
11893The trigonometry functions fall into this category.</p>
11894</li>
11895</ul>
11896</div>
11897<div class="paragraph">
11898<p>Many of the functions are similar to the same named ones in common C
11899libraries, but they support vector input as well as the more traditional
11900scalar input.</p>
11901</div>
11902<div class="paragraph">
11903<p>Applications should be encouraged to use the built-in functions rather than
11904do the equivalent computations in their own shader code since the built-in
11905functions are assumed to be optimal (e.g. perhaps supported directly in
11906hardware).</p>
11907</div>
11908<div class="paragraph">
11909<p>When the built-in functions are specified below, where the input arguments
11910(and corresponding output) can be <strong>float</strong>, <strong>vec2</strong>, <strong>vec3</strong>, or <strong>vec4</strong>,
11911<em>genFType</em> is used as the argument.
11912Where the input arguments (and corresponding output) can be <strong>int</strong>, <strong>ivec2</strong>,
11913<strong>ivec3</strong>, or <strong>ivec4</strong>, <em>genIType</em> is used as the argument.
11914Where the input arguments (and corresponding output) can be <strong>uint</strong>, <strong>uvec2</strong>,
11915<strong>uvec3</strong>, or <strong>uvec4</strong>, <em>genUType</em> is used as the argument.
11916Where the input arguments (or corresponding output) can be <strong>bool</strong>, <strong>bvec2</strong>,
11917<strong>bvec3</strong>, or <strong>bvec4</strong>, <em>genBType</em> is used as the argument.
11918For any specific use of a function, the actual types substituted for
11919<em>genFType</em>, <em>genIType</em>, <em>genUType</em>, or <em>genBType</em> have to have the same
11920number of components for all arguments and for the return type.
11921Similarly, <em>mat</em> is used for any matrix basic
11922type.</p>
11923</div>
11924<div class="paragraph">
11925<p>Built-in functions have an effective precision qualification.
11926This qualification cannot be set explicitly and may be different from the
11927precision qualification of the result.</p>
11928</div>
11929<div class="paragraph">
11930<p>The precision qualification of the operation of a built-in function is based
11931on the precision qualification of its formal parameters and actual
11932parameters (input arguments): When a formal parameter specifies a precision
11933qualifier, that is used, otherwise, the precision qualification of the
11934actual (calling) argument is used.
11935The highest precision of these will be the precision of the operation of the
11936built-in function.
11937Generally, this is applied across all arguments to a built-in function, with
11938the exceptions being:</p>
11939</div>
11940<div class="ulist">
11941<ul>
11942<li>
11943<p><strong>bitfieldExtract</strong> and <strong>bitfieldInsert</strong> ignore the <em>offset</em> and <em>bits</em>
11944arguments.</p>
11945</li>
11946<li>
11947<p><strong>interpolateAt</strong> functions only look at the <em>interpolant</em> argument.</p>
11948</li>
11949</ul>
11950</div>
11951<div class="paragraph">
11952<p>The precision qualification of the result of a built-in function is
11953determined in one of the following ways:</p>
11954</div>
11955<div class="paragraph">
11956<p>For the texture sampling, image load and image store functions, the
11957precision of the return type matches the precision of the
11958texture-combined sampler type:</p>
11959</div>
11960<div class="listingblock">
11961<div class="content">
11962<pre class="CodeRay highlight"><code data-lang="c++">uniform lowp sampler2D texSampler;
11963highp vec2 coord;
11964...
11965lowp vec4 col = texture (texSampler, coord); <span class="comment">// texture() returns lowp</span></code></pre>
11966</div>
11967</div>
11968<div class="paragraph">
11969<p>Otherwise:</p>
11970</div>
11971<div class="ulist">
11972<ul>
11973<li>
11974<p>For prototypes that do not specify a resulting precision qualifier, the
11975precision will be the same as the precision of the operation (as defined
11976earlier).</p>
11977</li>
11978<li>
11979<p>For prototypes that do specify a resulting precision qualifier, the
11980specified precision qualifier is the precision qualification of the
11981result.</p>
11982</li>
11983</ul>
11984</div>
11985<div class="paragraph">
11986<p>Where the built-in functions in the following sections specify an equation,
11987the entire equation will be evaluated at the operation&#8217;s precision.
11988This may lead to underflow or overflow in the result, even when the correct
11989result could be represented in the operation precision.</p>
11990</div>
11991<div class="sect2">
11992<h3 id="angle-and-trigonometry-functions">8.1. Angle and Trigonometry Functions</h3>
11993<div class="paragraph">
11994<p>Function parameters specified as <em>angle</em> are assumed to be in units of
11995radians.
11996In no case will any of these functions result in a divide by zero error.
11997If the divisor of a ratio is 0, then results will be undefined.</p>
11998</div>
11999<div class="paragraph">
12000<p>These all operate component-wise.
12001The description is per component.</p>
12002</div>
12003<table class="tableblock frame-all grid-all stretch">
12004<colgroup>
12005<col style="width: 50%;">
12006<col style="width: 50%;">
12007</colgroup>
12008<thead>
12009<tr>
12010<th class="tableblock halign-left valign-top">Syntax</th>
12011<th class="tableblock halign-left valign-top">Description</th>
12012</tr>
12013</thead>
12014<tbody>
12015<tr>
12016<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>radians</strong>(genFType <em>degrees</em>)</p></td>
12017<td class="tableblock halign-left valign-top"><p class="tableblock">Converts <em>degrees</em> to radians, i.e.,
12018      <span class="eq">(π / 180) · degrees</span>.</p></td>
12019</tr>
12020<tr>
12021<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>degrees</strong>(genFType <em>radians</em>)</p></td>
12022<td class="tableblock halign-left valign-top"><p class="tableblock">Converts <em>radians</em> to degrees, i.e.,
12023      <span class="eq">(180 / π) · radians</span>.</p></td>
12024</tr>
12025<tr>
12026<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>sin</strong>(genFType <em>angle</em>)</p></td>
12027<td class="tableblock halign-left valign-top"><p class="tableblock">The standard trigonometric sine function.</p></td>
12028</tr>
12029<tr>
12030<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>cos</strong>(genFType <em>angle</em>)</p></td>
12031<td class="tableblock halign-left valign-top"><p class="tableblock">The standard trigonometric cosine function.</p></td>
12032</tr>
12033<tr>
12034<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>tan</strong>(genFType <em>angle</em>)</p></td>
12035<td class="tableblock halign-left valign-top"><p class="tableblock">The standard trigonometric tangent.</p></td>
12036</tr>
12037<tr>
12038<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>asin</strong>(genFType <em>x</em>)</p></td>
12039<td class="tableblock halign-left valign-top"><p class="tableblock">Arc sine.
12040      Returns an angle whose sine is <em>x</em>.
12041      The range of values returned by this function is
12042      <span class="eq">[-π / 2, π / 2]</span>.
12043      Results are undefined if <span class="eq">|x| &gt; 1</span>.</p></td>
12044</tr>
12045<tr>
12046<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>acos</strong>(genFType <em>x</em>)</p></td>
12047<td class="tableblock halign-left valign-top"><p class="tableblock">Arc cosine.
12048      Returns an angle whose cosine is <em>x</em>.
12049      The range of values returned by this function is <span class="eq">[0,π]</span>.
12050      Results are undefined if <span class="eq">|x| &gt; 1</span>.</p></td>
12051</tr>
12052<tr>
12053<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>atan</strong>(genFType <em>y</em>, genFType <em>x</em>)</p></td>
12054<td class="tableblock halign-left valign-top"><p class="tableblock">Arc tangent.
12055      Returns an angle whose tangent is <span class="eq">y / x</span>.
12056      The signs of <em>x</em> and <em>y</em> are used to determine what quadrant the angle
12057      is in.
12058      The range of values returned by this function is <span class="eq">[-π, π</span>.
12059      Results are undefined if <em>x</em> and <em>y</em> are both 0.</p></td>
12060</tr>
12061<tr>
12062<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>atan</strong>(genFType <em>y_over_x</em>)</p></td>
12063<td class="tableblock halign-left valign-top"><p class="tableblock">Arc tangent.
12064      Returns an angle whose tangent is <em>y_over_x</em>.
12065      The range of values returned by this function is
12066      <span class="eq">[-π / 2, π / 2]</span>.</p></td>
12067</tr>
12068<tr>
12069<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>sinh</strong>(genFType <em>x</em>)</p></td>
12070<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the hyperbolic sine function <span class="eq">(e<sup>x</sup> - e<sup>-x</sup>) / 2</span>.</p></td>
12071</tr>
12072<tr>
12073<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>cosh</strong>(genFType <em>x</em>)</p></td>
12074<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the hyperbolic cosine function <span class="eq">(e<sup>x</sup> + e<sup>-x</sup>) / 2</span>.</p></td>
12075</tr>
12076<tr>
12077<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>tanh</strong>(genFType <em>x</em>)</p></td>
12078<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the hyperbolic tangent function <span class="eq">sinh(x) / cosh(x)</span>.</p></td>
12079</tr>
12080<tr>
12081<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>asinh</strong>(genFType <em>x</em>)</p></td>
12082<td class="tableblock halign-left valign-top"><p class="tableblock">Arc hyperbolic sine; returns the inverse of <strong>sinh</strong>.</p></td>
12083</tr>
12084<tr>
12085<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>acosh</strong>(genFType <em>x</em>)</p></td>
12086<td class="tableblock halign-left valign-top"><p class="tableblock">Arc hyperbolic cosine; returns the non-negative inverse of <strong>cosh</strong>.
12087      Results are undefined if <span class="eq">x &lt; 1</span>.</p></td>
12088</tr>
12089<tr>
12090<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>atanh</strong>(genFType <em>x</em>)</p></td>
12091<td class="tableblock halign-left valign-top"><p class="tableblock">Arc hyperbolic tangent; returns the inverse of <strong>tanh</strong>.
12092      Results are undefined if <span class="eq">x ≥ 1</span>.</p></td>
12093</tr>
12094</tbody>
12095</table>
12096</div>
12097<div class="sect2">
12098<h3 id="exponential-functions">8.2. Exponential Functions</h3>
12099<div class="paragraph">
12100<p>These all operate component-wise.
12101The description is per component.</p>
12102</div>
12103<table class="tableblock frame-all grid-all stretch">
12104<colgroup>
12105<col style="width: 50%;">
12106<col style="width: 50%;">
12107</colgroup>
12108<thead>
12109<tr>
12110<th class="tableblock halign-left valign-top">Syntax</th>
12111<th class="tableblock halign-left valign-top">Description</th>
12112</tr>
12113</thead>
12114<tbody>
12115<tr>
12116<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>pow</strong>(genFType <em>x</em>, genFType <em>y</em>)</p></td>
12117<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <em>x</em> raised to the <em>y</em> power, i.e., <span class="eq">x<sup>y</sup></span>.
12118      Results are undefined if <span class="eq">x &lt; 0</span>.
12119      Results are undefined if <span class="eq">x = 0</span> and <span class="eq">y ≤ 0</span>.</p></td>
12120</tr>
12121<tr>
12122<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>exp</strong>(genFType <em>x</em>)</p></td>
12123<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the natural exponentiation of <em>x</em>, i.e., <span class="eq">e<sup>x</sup></span>.</p></td>
12124</tr>
12125<tr>
12126<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>log</strong>(genFType <em>x</em>)</p></td>
12127<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the natural logarithm of <em>x</em>, i.e., returns the value <em>y</em>
12128      which satisfies the equation <span class="eq">x = e<sup>y</sup></span>.
12129      Results are undefined if <span class="eq">x ≤ 0</span>.</p></td>
12130</tr>
12131<tr>
12132<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>exp2</strong>(genFType <em>x</em>)</p></td>
12133<td class="tableblock halign-left valign-top"><p class="tableblock">Returns 2 raised to the <em>x</em> power, i.e., <span class="eq">2<sup>x</sup></span>.</p></td>
12134</tr>
12135<tr>
12136<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>log2</strong>(genFType <em>x</em>)</p></td>
12137<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the base 2 logarithm of <em>x</em>, i.e., returns the value <em>y</em> which
12138      satisfies the equation <span class="eq">x = 2<sup>y</sup></span>.
12139      Results are undefined if <span class="eq">x ≤ 0</span>.</p></td>
12140</tr>
12141<tr>
12142<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>sqrt</strong>(genFType <em>x</em>)<br>
12143  genDType <strong>sqrt</strong>(genDType <em>x</em>)</p></td>
12144<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <span class="eq">sqrt(x)</span>.
12145      Results are undefined if <span class="eq">x &lt; 0</span>.</p></td>
12146</tr>
12147<tr>
12148<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>inversesqrt</strong>(genFType <em>x</em>)<br>
12149  genDType <strong>inversesqrt</strong>(genDType <em>x</em>)</p></td>
12150<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <span class="eq">1 / sqrt(x)</span>.
12151      Results are undefined if <span class="eq">x ≤ 0</span>.</p></td>
12152</tr>
12153</tbody>
12154</table>
12155</div>
12156<div class="sect2">
12157<h3 id="common-functions">8.3. Common Functions</h3>
12158<div class="paragraph">
12159<p>These all operate component-wise.
12160The description is per component.</p>
12161</div>
12162<table class="tableblock frame-all grid-all stretch">
12163<colgroup>
12164<col style="width: 50%;">
12165<col style="width: 50%;">
12166</colgroup>
12167<thead>
12168<tr>
12169<th class="tableblock halign-left valign-top">Syntax</th>
12170<th class="tableblock halign-left valign-top">Description</th>
12171</tr>
12172</thead>
12173<tbody>
12174<tr>
12175<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>abs</strong>(genFType <em>x</em>)<br>
12176  genIType <strong>abs</strong>(genIType <em>x</em>)</p></td>
12177<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <em>x</em> if <span class="eq">x ≥ 0</span>; otherwise it returns -<em>x</em>.</p></td>
12178</tr>
12179<tr>
12180<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>sign</strong>(genFType <em>x</em>)<br>
12181  genIType <strong>sign</strong>(genIType <em>x</em>)</p></td>
12182<td class="tableblock halign-left valign-top"><p class="tableblock">Returns 1.0 if <em>x</em> &gt; 0, 0.0 if <em>x</em> = 0, or -1.0 if <em>x</em> &lt; 0.</p></td>
12183</tr>
12184<tr>
12185<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>floor</strong>(genFType <em>x</em>)</p></td>
12186<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a value equal to the nearest integer that is less than or
12187      equal to <em>x</em>.</p></td>
12188</tr>
12189<tr>
12190<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>trunc</strong>(genFType <em>x</em>)</p></td>
12191<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a value equal to the nearest integer to <em>x</em> whose absolute
12192      value is not larger than the absolute value of <em>x</em>.</p></td>
12193</tr>
12194<tr>
12195<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>round</strong>(genFType <em>x</em>)</p></td>
12196<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a value equal to the nearest integer to <em>x</em>.
12197      The fraction 0.5 will round in a direction chosen by the
12198      implementation, presumably the direction that is fastest.
12199      This includes the possibility that <strong>round</strong>(<em>x</em>) returns the same value
12200      as <strong>roundEven</strong>(<em>x</em>) for all values of <em>x</em>.</p></td>
12201</tr>
12202<tr>
12203<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>roundEven</strong>(genFType <em>x</em>)</p></td>
12204<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a value equal to the nearest integer to <em>x</em>.
12205      A fractional part of 0.5 will round toward the nearest even integer.
12206      (Both 3.5 and 4.5 for x will return 4.0.)</p></td>
12207</tr>
12208<tr>
12209<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>ceil</strong>(genFType <em>x</em>)</p></td>
12210<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a value equal to the nearest integer that is greater than or
12211      equal to <em>x</em>.</p></td>
12212</tr>
12213<tr>
12214<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>fract</strong>(genFType <em>x</em>)</p></td>
12215<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <em>x</em> - <strong>floor</strong>(<em>x</em>).</p></td>
12216</tr>
12217<tr>
12218<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>mod</strong>(genFType <em>x</em>, float <em>y</em>)<br>
12219  genFType <strong>mod</strong>(genFType <em>x</em>, genFType <em>y</em>)</p></td>
12220<td class="tableblock halign-left valign-top"><p class="tableblock">Modulus.
12221      Returns <span class="eq">x - y · <strong>floor</strong>(x / y)</span>.</p></td>
12222</tr>
12223<tr>
12224<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>modf</strong>(genFType <em>x</em>, out genFType <em>i</em>)<br></p></td>
12225<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the fractional part of <em>x</em> and sets <em>i</em> to the integer part (as
12226      a whole number floating-point value).
12227      Both the return value and the output parameter will have the same sign
12228      as <em>x</em>.
12229      If <em>x</em> has the value +/- Inf, the return value should be NaN and must
12230      be either NaN or 0.0.
12231      For <strong>highp</strong> non-constant expressions, the value returned must be
12232      consistent.</p></td>
12233</tr>
12234<tr>
12235<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>min</strong>(genFType <em>x</em>, genFType <em>y</em>)<br>
12236  genFType <strong>min</strong>(genFType <em>x</em>, float <em>y</em>)<br>
12237  genIType <strong>min</strong>(genIType <em>x</em>, genIType <em>y</em>)<br>
12238  genIType <strong>min</strong>(genIType <em>x</em>, int <em>y</em>)<br>
12239  genUType <strong>min</strong>(genUType <em>x</em>, genUType <em>y</em>)<br>
12240  genUType <strong>min</strong>(genUType <em>x</em>, uint <em>y</em>)</p></td>
12241<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <em>y</em> if <em>y</em> &lt; <em>x;</em> otherwise it returns <em>x</em>.</p></td>
12242</tr>
12243<tr>
12244<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>max</strong>(genFType <em>x</em>, genFType <em>y</em>)<br>
12245  genFType <strong>max</strong>(genFType <em>x</em>, float <em>y</em>)<br>
12246  genIType <strong>max</strong>(genIType <em>x</em>, genIType <em>y</em>)<br>
12247  genIType <strong>max</strong>(genIType <em>x</em>, int <em>y</em>)<br>
12248  genUType <strong>max</strong>(genUType <em>x</em>, genUType <em>y</em>)<br>
12249  genUType <strong>max</strong>(genUType <em>x</em>, uint <em>y</em>)</p></td>
12250<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <em>y</em> if <em>x</em> &lt; <em>y;</em> otherwise it returns <em>x</em>.</p></td>
12251</tr>
12252<tr>
12253<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>clamp</strong>(genFType <em>x</em>, genFType <em>minVal</em>, genFType <em>maxVal</em>)<br>
12254  genFType <strong>clamp</strong>(genFType <em>x</em>, float <em>minVal</em>, float <em>maxVal</em>)<br>
12255  genIType <strong>clamp</strong>(genIType <em>x</em>, genIType <em>minVal</em>, genIType <em>maxVal</em>)<br>
12256  genIType <strong>clamp</strong>(genIType <em>x</em>, int <em>minVal</em>, int <em>maxVal</em>)<br>
12257  genUType <strong>clamp</strong>(genUType <em>x</em>, genUType <em>minVal</em>, genUType <em>maxVal</em>)<br>
12258  genUType <strong>clamp</strong>(genUType <em>x</em>, uint <em>minVal</em>, uint <em>maxVal</em>)</p></td>
12259<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <strong>min</strong>(<strong>max</strong>(<em>x</em>, <em>minVal</em>), <em>maxVal</em>).
12260      Results are undefined if <em>minVal</em> &gt; <em>maxVal</em>.</p></td>
12261</tr>
12262<tr>
12263<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>mix</strong>(genFType <em>x</em>, genFType <em>y</em>, genFType <em>a</em>)<br>
12264  genFType <strong>mix</strong>(genFType <em>x</em>, genFType <em>y</em>, float <em>a</em>)<br></p></td>
12265<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the linear blend of <em>x</em> and <em>y</em>, i.e.,
12266      <span class="eq">x · (1 - a) + y · a</span>.</p></td>
12267</tr>
12268<tr>
12269<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>mix</strong>(genFType <em>x</em>, genFType <em>y</em>,  genBType <em>a</em>)<br>
12270  genIType <strong>mix</strong>(genIType <em>x</em>, genIType <em>y</em>, genBType <em>a</em>)<br>
12271  genUType <strong>mix</strong>(genUType <em>x</em>, genUType <em>y</em>, genBType <em>a</em>)<br>
12272  genBType <strong>mix</strong>(genBType <em>x</em>, genBType <em>y</em>, genBType <em>a</em>)</p></td>
12273<td class="tableblock halign-left valign-top"><p class="tableblock">Selects which vector each returned component comes from.
12274      For a component of <em>a</em> that is <strong>false</strong>, the corresponding component of
12275      <em>x</em> is returned.
12276      For a component of <em>a</em> that is <strong>true</strong>, the corresponding component of
12277      <em>y</em> is returned.
12278      Components of <em>x</em> and <em>y</em> that are not selected are allowed to be
12279      invalid floating-point values and will have no effect on the results.
12280      Thus, this provides different functionality than, for example,<br>
12281      genFType <strong>mix</strong>(genFType <em>x</em>, genFType <em>y</em>, genFType(<em>a</em>))<br>
12282      where <em>a</em> is a Boolean vector.</p></td>
12283</tr>
12284<tr>
12285<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>step</strong>(genFType <em>edge</em>, genFType <em>x</em>)<br>
12286  genFType <strong>step</strong>(float <em>edge</em>, genFType <em>x</em>)</p></td>
12287<td class="tableblock halign-left valign-top"><p class="tableblock">Returns 0.0 if <em>x</em> &lt; <em>edge;</em> otherwise it returns 1.0.</p></td>
12288</tr>
12289<tr>
12290<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>smoothstep</strong>(genFType <em>edge0</em>, genFType <em>edge1</em>, genFType <em>x</em>)<br>
12291  genFType <strong>smoothstep</strong>(float <em>edge0</em>, float <em>edge1</em>, genFType <em>x</em>)<br></p></td>
12292<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
12293<p>Returns 0.0 if <span class="eq">x ≤ edge0</span> and 1.0 if <span class="eq">x ≥ edge1</span>, and
12294      performs smooth Hermite interpolation between 0 and 1 when <span class="eq">edge0
12295      &lt; x &lt; edge1</span>.
12296      This is useful in cases where you would want a threshold function with
12297      a smooth transition.
12298      This is equivalent to:</p>
12299</div>
12300<div class="openblock">
12301<div class="content">
12302<div class="listingblock">
12303<div class="content">
12304<pre class="CodeRay highlight"><code data-lang="c++">genFType t;
12305t = clamp ((x - edge0) / (edge1 - edge0), <span class="integer">0</span>, <span class="integer">1</span>);
12306<span class="keyword">return</span> t * t * (<span class="integer">3</span> - <span class="integer">2</span> * t);</code></pre>
12307</div>
12308</div>
12309<div class="paragraph">
12310<p>(And similarly for doubles.) Results are undefined if <span class="eq">edge0 ≥
12311edge1</span>.</p>
12312</div>
12313</div>
12314</div></div></td>
12315</tr>
12316<tr>
12317<td class="tableblock halign-left valign-top"><p class="tableblock">genBType <strong>isnan</strong>(genFType <em>x</em>)</p></td>
12318<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <strong>true</strong> if <em>x</em> holds a NaN.
12319      Returns <strong>false</strong> otherwise.
12320      Always returns <strong>false</strong> if NaNs are not implemented.</p></td>
12321</tr>
12322<tr>
12323<td class="tableblock halign-left valign-top"><p class="tableblock">genBType <strong>isinf</strong>(genFType <em>x</em>)</p></td>
12324<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <strong>true</strong> if <em>x</em> holds a positive infinity or negative infinity.
12325      Returns <strong>false</strong> otherwise.</p></td>
12326</tr>
12327<tr>
12328<td class="tableblock halign-left valign-top"><p class="tableblock">genIType <strong>floatBitsToInt</strong>(highp genFType <em>value</em>)<br>
12329  genUType <strong>floatBitsToUint</strong>(highp genFType <em>value</em>)</p></td>
12330<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a signed or unsigned integer value representing the encoding
12331      of a floating-point value.
12332      The <strong>float</strong> value&#8217;s bit-level representation is preserved.</p></td>
12333</tr>
12334<tr>
12335<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>intBitsToFloat</strong>(highp genIType <em>value</em>)<br>
12336  genFType <strong>uintBitsToFloat</strong>(highp genUType <em>value</em>)</p></td>
12337<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a floating-point value corresponding to a signed or unsigned
12338      integer encoding of a floating-point value.
12339      If an Inf or NaN is passed in, it will not signal, and the resulting
12340      floating-point value is unspecified.
12341      Otherwise, the bit-level representation is preserved.</p></td>
12342</tr>
12343<tr>
12344<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>fma</strong>(genFType <em>a</em>, genFType <em>b</em>, genFType <em>c</em>)<br></p></td>
12345<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
12346<p>Computes and returns <code>a * b + c</code>.
12347      In uses where the return value is eventually consumed by a variable
12348      declared as <strong>precise</strong>:</p>
12349</div>
12350<div class="openblock">
12351<div class="content">
12352<div class="ulist">
12353<ul>
12354<li>
12355<p><strong>fma</strong>() is considered a single operation, whereas the expression <code>a * b
12356+ c</code> consumed by a variable declared <strong>precise</strong> is considered two
12357operations.</p>
12358</li>
12359<li>
12360<p>The precision of <strong>fma</strong>() can differ from the precision of the expression
12361<code>a * b + c</code>.</p>
12362</li>
12363<li>
12364<p><strong>fma</strong>() will be computed with the same precision as any other <strong>fma</strong>()
12365consumed by a precise variable, giving invariant results for the same
12366input values of <em>a</em>, <em>b</em>, and <em>c</em>.</p>
12367</li>
12368</ul>
12369</div>
12370<div class="paragraph">
12371<p>Otherwise, in the absence of <strong>precise</strong> consumption, there are no special
12372constraints on the number of operations or difference in precision between
12373<strong>fma</strong>() and the expression <code>a * b + c</code>.</p>
12374</div>
12375</div>
12376</div></div></td>
12377</tr>
12378<tr>
12379<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>frexp</strong>(highp genFType <em>x</em>, out highp genIType <em>exp</em>)</p></td>
12380<td class="tableblock halign-left valign-top"><p class="tableblock">Splits <em>x</em> into a floating-point significand in the range
12381      <span class="eq">[0.5,1.0]</span>, and an integral exponent of two, such that</p>
12382<p class="tableblock">      <span class="eq">x = significant · 2<sup>exponent</sup></span></p>
12383<p class="tableblock">      The significand is returned by the function and the exponent is
12384      returned in the parameter <em>exp</em>.
12385      For a floating-point value of zero, the significand and exponent are
12386      both zero.</p>
12387<p class="tableblock">      If an implementation supports signed zero, an input value of minus
12388      zero should return a significand of minus zero.
12389      For a floating-point value that is an infinity or is not a number, the
12390      results are undefined.</p>
12391<p class="tableblock">      If the input <em>x</em> is a vector, this operation is performed in a
12392      component-wise manner; the value returned by the function and the
12393      value written to <em>exp</em> are vectors with the same number of components
12394      as <em>x</em>.</p></td>
12395</tr>
12396<tr>
12397<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>ldexp</strong>(highp genFType <em>x</em>, highp genIType <em>exp</em>)<br></p></td>
12398<td class="tableblock halign-left valign-top"><p class="tableblock">Builds a floating-point number from <em>x</em> and the corresponding integral
12399      exponent of two in <em>exp</em>, returning:</p>
12400<p class="tableblock">      <span class="eq">significand · 2<sup>exponent</sup></span></p>
12401<p class="tableblock">      If this product is too large to be represented in the floating-point
12402      type, the result is undefined.</p>
12403<p class="tableblock">      If <em>exp</em> is greater than +128, the value returned is undefined.
12404      If <em>exp</em> is less than -126, the value returned may be flushed to zero.
12405      Additionally, splitting the value into a significand and exponent
12406      using <strong>frexp</strong>() and then reconstructing a floating-point value using
12407      <strong>ldexp</strong>() should yield the original input for zero and all finite
12408      non-denormalized values.<br>
12409      If the input <em>x</em> is a vector, this operation is performed in a
12410      component-wise manner; the value passed in <em>exp</em> and returned by the
12411      function are vectors with the same number of components as <em>x</em>.</p></td>
12412</tr>
12413</tbody>
12414</table>
12415</div>
12416<div class="sect2">
12417<h3 id="floating-point-pack-and-unpack-functions">8.4. Floating-Point Pack and Unpack Functions</h3>
12418<div class="paragraph">
12419<p>These functions do not operate component-wise, rather, as described in each
12420case.</p>
12421</div>
12422<table class="tableblock frame-all grid-all stretch">
12423<colgroup>
12424<col style="width: 50%;">
12425<col style="width: 50%;">
12426</colgroup>
12427<thead>
12428<tr>
12429<th class="tableblock halign-left valign-top">Syntax</th>
12430<th class="tableblock halign-left valign-top">Description</th>
12431</tr>
12432</thead>
12433<tbody>
12434<tr>
12435<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>packUnorm2x16</strong>(vec2 <em>v</em>)<br>
12436  highp uint <strong>packSnorm2x16</strong>(vec2 <em>v</em>)<br>
12437  highp uint <strong>packUnorm4x8</strong>(vec4 <em>v</em>)<br>
12438  highp uint <strong>packSnorm4x8</strong>(vec4 <em>v</em>)</p></td>
12439<td class="tableblock halign-left valign-top"><p class="tableblock">First, converts each component of the normalized floating-point value
12440      <em>v</em> into 16-bit (<strong>2x16</strong>) or 8-bit (<strong>4x8</strong>) integer values.
12441      Then, the results are packed into the returned 32-bit unsigned
12442      integer.</p>
12443<p class="tableblock">      The conversion for component <em>c</em> of <em>v</em> to fixed point is done as
12444      follows:</p>
12445<p class="tableblock">      <strong>packUnorm2x16</strong>: <strong>round</strong>(<strong>clamp</strong>(<em>c</em>, 0, +1) * 65535.0)<br>
12446      <strong>packSnorm2x16:</strong> <strong>round</strong>(<strong>clamp</strong>(<em>c</em>, -1, +1) * 32767.0)<br>
12447      <strong>packUnorm4x8</strong>: <strong>round</strong>(<strong>clamp</strong>(<em>c</em>, 0, +1) * 255.0)<br>
12448      <strong>packSnorm4x8</strong>: <strong>round</strong>(<strong>clamp</strong>(<em>c</em>, -1, +1) * 127.0)</p>
12449<p class="tableblock">      The first component of the vector will be written to the least
12450      significant bits of the output; the last component will be written to
12451      the most significant bits.</p></td>
12452</tr>
12453<tr>
12454<td class="tableblock halign-left valign-top"><p class="tableblock">highp vec2 <strong>unpackUnorm2x16</strong>(highp uint <em>p</em>)<br>
12455  highp vec2 <strong>unpackSnorm2x16</strong>(highp uint <em>p</em>)<br>
12456  mediump vec4 <strong>unpackUnorm4x8</strong>(highp uint <em>p</em>)<br>
12457  mediump vec4 <strong>unpackSnorm4x8</strong>(highp uint <em>p</em>)</p></td>
12458<td class="tableblock halign-left valign-top"><p class="tableblock">First, unpacks a single 32-bit unsigned integer <em>p</em> into a pair of
12459      16-bit unsigned integers, a pair of 16-bit signed integers, four 8-bit
12460      unsigned integers, or four 8-bit signed integers, respectively.
12461      Then, each component is converted to a normalized floating-point value
12462      to generate the returned two- or four-component vector.</p>
12463<p class="tableblock">      The conversion for unpacked fixed-point value <em>f</em> to floating-point is
12464      done as follows:</p>
12465<p class="tableblock">      <strong>unpackUnorm2x16</strong>: <em>f</em> / 65535.0<br>
12466      <strong>unpackSnorm2x16</strong>: <strong>clamp</strong>(<em>f</em> / 32767.0, -1, +1)<br>
12467      <strong>unpackUnorm4x8</strong>: <em>f</em> / 255.0<br>
12468      <strong>unpackSnorm4x8</strong>: <strong>clamp</strong>(<em>f</em> / 127.0, -1, +1)</p>
12469<p class="tableblock">      The first component of the returned vector will be extracted from the
12470      least significant bits of the input; the last component will be
12471      extracted from the most significant bits.</p></td>
12472</tr>
12473<tr>
12474<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>packHalf2x16</strong>(vec2 <em>v</em>)</p></td>
12475<td class="tableblock halign-left valign-top"><p class="tableblock">Returns an unsigned integer obtained by converting the components of a
12476      two-component floating-point vector to the 16-bit floating-point
12477      representation of the <a href="#references">API</a>, and
12478      then packing these two 16-bit integers into a 32-bit unsigned integer.</p>
12479<p class="tableblock">      The first vector component specifies the 16 least-significant bits of
12480      the result; the second component specifies the 16 most-significant
12481      bits.</p></td>
12482</tr>
12483<tr>
12484<td class="tableblock halign-left valign-top"><p class="tableblock">mediump vec2 <strong>unpackHalf2x16</strong>(highp uint <em>v</em>)</p></td>
12485<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a two-component floating-point vector with components obtained
12486      by unpacking a 32-bit unsigned integer into a pair of 16-bit values,
12487      interpreting those values as 16-bit floating-point numbers according
12488      to the <a href="#references">API</a>, and converting them to
12489      32-bit floating-point values.</p>
12490<p class="tableblock">      The first component of the vector is obtained from the 16
12491      least-significant bits of <em>v</em>; the second component is obtained from
12492      the 16 most-significant bits of <em>v</em>.</p></td>
12493</tr>
12494</tbody>
12495</table>
12496</div>
12497<div class="sect2">
12498<h3 id="geometric-functions">8.5. Geometric Functions</h3>
12499<div class="paragraph">
12500<p>These operate on vectors as vectors, not component-wise.</p>
12501</div>
12502<table class="tableblock frame-all grid-all stretch">
12503<colgroup>
12504<col style="width: 50%;">
12505<col style="width: 50%;">
12506</colgroup>
12507<thead>
12508<tr>
12509<th class="tableblock halign-left valign-top">Syntax</th>
12510<th class="tableblock halign-left valign-top">Description</th>
12511</tr>
12512</thead>
12513<tbody>
12514<tr>
12515<td class="tableblock halign-left valign-top"><p class="tableblock">float <strong>length</strong>(genFType <em>x</em>)</p></td>
12516<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
12517<p>Returns the length of vector <em>x</em>, i.e.,
12518      <span class="eq">sqrt( x<sub>0</sub><sup>2</sup> + x<sub>1</sub><sup>2</sup> + &#8230;&#8203; )</span>.</p>
12519</div></div></td>
12520</tr>
12521<tr>
12522<td class="tableblock halign-left valign-top"><p class="tableblock">float <strong>distance</strong>(genFType <em>p0</em>, genFType <em>p1</em>)</p></td>
12523<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the distance between <em>p0</em> and <em>p1</em>, i.e.,
12524      <strong>length</strong>(<em>p0</em> - <em>p1</em>)</p></td>
12525</tr>
12526<tr>
12527<td class="tableblock halign-left valign-top"><p class="tableblock">float <strong>dot</strong>(genFType <em>x</em>, genFType <em>y</em>)</p></td>
12528<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the dot product of <em>x</em> and <em>y</em>, i.e.,
12529      <span class="eq">x<sub>0</sub> · y<sub>0</sub> + x<sub>1</sub> · y<sub>1</sub> + &#8230;&#8203;</span></p></td>
12530</tr>
12531<tr>
12532<td class="tableblock halign-left valign-top"><p class="tableblock">vec3 <strong>cross</strong>(vec3 <em>x</em>, vec3 <em>y</em>)</p></td>
12533<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
12534<p>Returns the cross product of <em>x</em> and <em>y</em>, i.e.,
12535      <span class="eq">(x<sub>1</sub> · y<sub>2</sub> - y<sub>1</sub> · x<sub>2</sub>,
12536            x<sub>2</sub> · y<sub>0</sub> - y<sub>2</sub> · x<sub>0</sub>,
12537            x<sub>0</sub> · y<sub>1</sub> - y<sub>0</sub> · x<sub>1</sub>)</span>.</p>
12538</div></div></td>
12539</tr>
12540<tr>
12541<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>normalize</strong>(genFType <em>x</em>)</p></td>
12542<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a vector in the same direction as <em>x</em> but with a length of 1,
12543      i.e. <em>x</em> / <strong>length</strong>(x).</p></td>
12544</tr>
12545<tr>
12546<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>faceforward</strong>(genFType <em>N</em>, genFType <em>I</em>, genFType <em>Nref</em>)<br></p></td>
12547<td class="tableblock halign-left valign-top"><p class="tableblock">If <strong>dot</strong>(<em>Nref</em>, <em>I</em>) &lt; 0 return <em>N</em>, otherwise return -<em>N</em>.</p></td>
12548</tr>
12549<tr>
12550<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>reflect</strong>(genFType <em>I</em>, genFType <em>N</em>)</p></td>
12551<td class="tableblock halign-left valign-top"><p class="tableblock">For the incident vector <em>I</em> and surface orientation <em>N</em>, returns the
12552      reflection direction: <span class="eq">I - 2 · <strong>dot</strong>(N, I) · N</span>.
12553      <em>N</em> must already be normalized in order to achieve the desired result.</p></td>
12554</tr>
12555<tr>
12556<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>refract</strong>(genFType <em>I</em>, genFType <em>N</em>, float <em>eta</em>)<br></p></td>
12557<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
12558<p>For the incident vector <em>I</em> and surface normal <em>N</em>, and the ratio of
12559      indices of refraction <em>eta</em>, return the refraction vector.
12560      The result is computed by the <a href="#refraction-equation">refraction
12561      equation</a> shown below.</p>
12562</div>
12563<div class="literalblock">
12564<div class="content">
12565<pre>The input parameters for the incident vector _I_ and the surface
12566normal _N_ must already be normalized to get the desired results.</pre>
12567</div>
12568</div></div></td>
12569</tr>
12570</tbody>
12571</table>
12572<div class="sect3">
12573<h4 id="refraction-equation">8.5.1. Refraction Equation</h4>
12574<div class="stemblock">
12575<div class="content">
12576\[k = 1.0 - eta * eta * (1.0 - \textbf{dot}(N,I) \cdot \textbf{dot}(N,I))\]
12577</div>
12578</div>
12579<div class="stemblock">
12580<div class="content">
12581\[\begin{aligned}
12582result &amp;=
12583  \begin{cases}
12584    genFType(0.0), &amp; k &lt; 0.0 \\
12585    eta * I - (eta * \textbf{dot}(N,I) + \sqrt { k }) * N, &amp; \textbf{otherwise}
12586  \end{cases}
12587\end{aligned}\]
12588</div>
12589</div>
12590</div>
12591</div>
12592<div class="sect2">
12593<h3 id="matrix-functions">8.6. Matrix Functions</h3>
12594<div class="paragraph">
12595<p>For each of the following built-in matrix functions, there is both a
12596single-precision floating-point version, where all arguments and return
12597values are single precision, and a double-precision floating-point version,
12598where all arguments and return values are double precision.
12599Only the single-precision floating-point version is shown.</p>
12600</div>
12601<table class="tableblock frame-all grid-all stretch">
12602<colgroup>
12603<col style="width: 50%;">
12604<col style="width: 50%;">
12605</colgroup>
12606<thead>
12607<tr>
12608<th class="tableblock halign-left valign-top">Syntax</th>
12609<th class="tableblock halign-left valign-top">Description</th>
12610</tr>
12611</thead>
12612<tbody>
12613<tr>
12614<td class="tableblock halign-left valign-top"><p class="tableblock">mat <strong>matrixCompMult</strong>(mat <em>x</em>, mat <em>y</em>)</p></td>
12615<td class="tableblock halign-left valign-top"><p class="tableblock">Multiply matrix <em>x</em> by matrix <em>y</em> component-wise, i.e., result[i][j]
12616      is the scalar product of <em>x</em>[i][j] and <em>y</em>[i][j].<br></p>
12617<p class="tableblock">      Note: to get linear algebraic matrix multiplication, use the multiply
12618      operator (<strong>*</strong>).</p></td>
12619</tr>
12620<tr>
12621<td class="tableblock halign-left valign-top"><p class="tableblock">mat2 <strong>outerProduct</strong>(vec2 <em>c</em>, vec2 <em>r</em>)<br>
12622  mat3 <strong>outerProduct</strong>(vec3 <em>c</em>, vec3 <em>r</em>)<br>
12623  mat4 <strong>outerProduct</strong>(vec4 <em>c</em>, vec4 <em>r</em>)<br>
12624  mat2x3 <strong>outerProduct</strong>(vec3 <em>c</em>, vec2 <em>r</em>)<br>
12625  mat3x2 <strong>outerProduct</strong>(vec2 <em>c</em>, vec3 <em>r</em>)<br>
12626  mat2x4 <strong>outerProduct</strong>(vec4 <em>c</em>, vec2 <em>r</em>)<br>
12627  mat4x2 <strong>outerProduct</strong>(vec2 <em>c</em>, vec4 <em>r</em>)<br>
12628  mat3x4 <strong>outerProduct</strong>(vec4 <em>c</em>, vec3 <em>r</em>)<br>
12629  mat4x3 <strong>outerProduct</strong>(vec3 <em>c</em>, vec4 <em>r</em>)</p></td>
12630<td class="tableblock halign-left valign-top"><p class="tableblock">Treats the first parameter <em>c</em> as a column vector (matrix with one
12631      column) and the second parameter <em>r</em> as a row vector (matrix with one
12632      row) and does a linear algebraic matrix multiply <em>c</em> * <em>r</em>, yielding a
12633      matrix whose number of rows is the number of components in <em>c</em> and
12634      whose number of columns is the number of components in <em>r</em>.</p></td>
12635</tr>
12636<tr>
12637<td class="tableblock halign-left valign-top"><p class="tableblock">mat2 <strong>transpose</strong>(mat2 <em>m</em>)<br>
12638  mat3 <strong>transpose</strong>(mat3 <em>m</em>)<br>
12639  mat4 <strong>transpose</strong>(mat4 <em>m</em>)<br>
12640  mat2x3 <strong>transpose</strong>(mat3x2 <em>m</em>)<br>
12641  mat3x2 <strong>transpose</strong>(mat2x3 <em>m</em>)<br>
12642  mat2x4 <strong>transpose</strong>(mat4x2 <em>m</em>)<br>
12643  mat4x2 <strong>transpose</strong>(mat2x4 <em>m</em>)<br>
12644  mat3x4 <strong>transpose</strong>(mat4x3 <em>m</em>)<br>
12645  mat4x3 <strong>transpose</strong>(mat3x4 <em>m</em>)</p></td>
12646<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a matrix that is the transpose of <em>m</em>.
12647      The input matrix <em>m</em> is not modified.</p></td>
12648</tr>
12649<tr>
12650<td class="tableblock halign-left valign-top"><p class="tableblock">float <strong>determinant</strong>(mat2 <em>m</em>)<br>
12651  float <strong>determinant</strong>(mat3 <em>m</em>)<br>
12652  float <strong>determinant</strong>(mat4 <em>m</em>)</p></td>
12653<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the determinant of <em>m</em>.</p></td>
12654</tr>
12655<tr>
12656<td class="tableblock halign-left valign-top"><p class="tableblock">mat2 <strong>inverse</strong>(mat2 <em>m</em>)<br>
12657  mat3 <strong>inverse</strong>(mat3 <em>m</em>)<br>
12658  mat4 <strong>inverse</strong>(mat4 <em>m</em>)</p></td>
12659<td class="tableblock halign-left valign-top"><p class="tableblock">Returns a matrix that is the inverse of <em>m</em>.
12660      The input matrix <em>m</em> is not modified.
12661      The values in the returned matrix are undefined if <em>m</em> is singular or
12662      poorly-conditioned (nearly singular).</p></td>
12663</tr>
12664</tbody>
12665</table>
12666</div>
12667<div class="sect2">
12668<h3 id="vector-relational-functions">8.7. Vector Relational Functions</h3>
12669<div class="paragraph">
12670<p>Relational and equality operators (<strong>&lt;</strong>, <strong>&lt;=</strong>, <strong>&gt;</strong>, <strong>&gt;=</strong>, <strong>==</strong>, <strong>!=</strong>) are
12671defined to operate on scalars and produce scalar Boolean results.
12672For vector results, use the following built-in functions.
12673Below, the following placeholders are used for the listed specific types:</p>
12674</div>
12675<table class="tableblock frame-all grid-all stretch">
12676<colgroup>
12677<col style="width: 50%;">
12678<col style="width: 50%;">
12679</colgroup>
12680<thead>
12681<tr>
12682<th class="tableblock halign-left valign-top">Placeholder</th>
12683<th class="tableblock halign-left valign-top">Specific Types Allowed</th>
12684</tr>
12685</thead>
12686<tbody>
12687<tr>
12688<td class="tableblock halign-left valign-top"><p class="tableblock">bvec</p></td>
12689<td class="tableblock halign-left valign-top"><p class="tableblock">bvec2, bvec3, bvec4</p></td>
12690</tr>
12691<tr>
12692<td class="tableblock halign-left valign-top"><p class="tableblock">ivec</p></td>
12693<td class="tableblock halign-left valign-top"><p class="tableblock">ivec2, ivec3, ivec4</p></td>
12694</tr>
12695<tr>
12696<td class="tableblock halign-left valign-top"><p class="tableblock">uvec</p></td>
12697<td class="tableblock halign-left valign-top"><p class="tableblock">uvec2, uvec3, uvec4</p></td>
12698</tr>
12699<tr>
12700<td class="tableblock halign-left valign-top"><p class="tableblock">vec</p></td>
12701<td class="tableblock halign-left valign-top"><p class="tableblock">vec2, vec3, vec4</p></td>
12702</tr>
12703</tbody>
12704</table>
12705<div class="paragraph">
12706<p>In all cases, the sizes of all the input and return vectors for any
12707particular call must match.</p>
12708</div>
12709<table class="tableblock frame-all grid-all stretch">
12710<colgroup>
12711<col style="width: 50%;">
12712<col style="width: 50%;">
12713</colgroup>
12714<thead>
12715<tr>
12716<th class="tableblock halign-left valign-top">Syntax</th>
12717<th class="tableblock halign-left valign-top">Description</th>
12718</tr>
12719</thead>
12720<tbody>
12721<tr>
12722<td class="tableblock halign-left valign-top"><p class="tableblock">bvec <strong>lessThan</strong>(vec x, vec y)<br>
12723  bvec <strong>lessThan</strong>(ivec x, ivec y)<br>
12724  bvec <strong>lessThan</strong>(uvec x, uvec y)</p></td>
12725<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the component-wise compare of <span class="eq">x &lt; y</span>.</p></td>
12726</tr>
12727<tr>
12728<td class="tableblock halign-left valign-top"><p class="tableblock">bvec <strong>lessThanEqual</strong>(vec x, vec y)<br>
12729  bvec <strong>lessThanEqual</strong>(ivec x, ivec y)<br>
12730  bvec <strong>lessThanEqual</strong>(uvec x, uvec y)</p></td>
12731<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the component-wise compare of <span class="eq">x ≤ y</span>.</p></td>
12732</tr>
12733<tr>
12734<td class="tableblock halign-left valign-top"><p class="tableblock">bvec <strong>greaterThan</strong>(vec x, vec y)<br>
12735  bvec <strong>greaterThan</strong>(ivec x, ivec y)<br>
12736  bvec <strong>greaterThan</strong>(uvec x, uvec y)</p></td>
12737<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the component-wise compare of <span class="eq">x &gt; y</span>.</p></td>
12738</tr>
12739<tr>
12740<td class="tableblock halign-left valign-top"><p class="tableblock">bvec <strong>greaterThanEqual</strong>(vec x, vec y)<br>
12741  bvec <strong>greaterThanEqual</strong>(ivec x, ivec y)<br>
12742  bvec <strong>greaterThanEqual</strong>(uvec x, uvec y)</p></td>
12743<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the component-wise compare of <span class="eq">x ≥ y</span>.</p></td>
12744</tr>
12745<tr>
12746<td class="tableblock halign-left valign-top"><p class="tableblock">bvec <strong>equal</strong>(vec x, vec y)<br>
12747  bvec <strong>equal</strong>(ivec x, ivec y)<br>
12748  bvec <strong>equal</strong>(uvec x, uvec y)<br>
12749  bvec <strong>equal</strong>(bvec x, bvec y)</p></td>
12750<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the component-wise compare of <span class="eq">x == y</span>.</p></td>
12751</tr>
12752<tr>
12753<td class="tableblock halign-left valign-top"><p class="tableblock">bvec <strong>notEqual</strong>(vec x, vec y)<br>
12754  bvec <strong>notEqual</strong>(ivec x, ivec y)<br>
12755  bvec <strong>notEqual</strong>(uvec x, uvec y)<br>
12756  bvec <strong>notEqual</strong>(bvec x, bvec y)</p></td>
12757<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the component-wise compare of <span class="eq">x ≠ y</span>.</p></td>
12758</tr>
12759<tr>
12760<td class="tableblock halign-left valign-top"><p class="tableblock">bool <strong>any</strong>(bvec x)</p></td>
12761<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <strong>true</strong> if any component of <em>x</em> is <strong>true</strong>.</p></td>
12762</tr>
12763<tr>
12764<td class="tableblock halign-left valign-top"><p class="tableblock">bool <strong>all</strong>(bvec x)</p></td>
12765<td class="tableblock halign-left valign-top"><p class="tableblock">Returns <strong>true</strong> only if all components of <em>x</em> are <strong>true</strong>.</p></td>
12766</tr>
12767<tr>
12768<td class="tableblock halign-left valign-top"><p class="tableblock">bvec <strong>not</strong>(bvec x)</p></td>
12769<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the component-wise logical complement of <em>x</em>.</p></td>
12770</tr>
12771</tbody>
12772</table>
12773</div>
12774<div class="sect2">
12775<h3 id="integer-functions">8.8. Integer Functions</h3>
12776<div class="paragraph">
12777<p>These all operate component-wise.
12778The description is per component.
12779The notation [<em>a</em>, <em>b</em>] means the set of bits from bit-number <em>a</em> through
12780bit-number <em>b</em>, inclusive.
12781The lowest-order bit is bit 0.
12782&#8220;Bit number&#8221; will always refer to counting up from the lowest-order bit as
12783bit 0.</p>
12784</div>
12785<table class="tableblock frame-all grid-all stretch">
12786<colgroup>
12787<col style="width: 50%;">
12788<col style="width: 50%;">
12789</colgroup>
12790<thead>
12791<tr>
12792<th class="tableblock halign-left valign-top">Syntax</th>
12793<th class="tableblock halign-left valign-top">Description</th>
12794</tr>
12795</thead>
12796<tbody>
12797<tr>
12798<td class="tableblock halign-left valign-top"><p class="tableblock">genUType <strong>uaddCarry</strong>(highp genUType <em>x</em>, highp genUType <em>y</em>, out lowp genUType <em>carry</em>)</p></td>
12799<td class="tableblock halign-left valign-top"><p class="tableblock">Adds 32-bit unsigned integers <em>x</em> and <em>y</em>, returning the sum modulo
12800      2<sup>32</sup>.
12801      The value <em>carry</em> is set to zero if the sum was less than 2<sup>32</sup>, or
12802      one otherwise.</p></td>
12803</tr>
12804<tr>
12805<td class="tableblock halign-left valign-top"><p class="tableblock">genUType <strong>usubBorrow</strong>(highp genUType <em>x</em>, highp genUType <em>y</em>, out lowp genUType <em>borrow</em>)</p></td>
12806<td class="tableblock halign-left valign-top"><p class="tableblock">Subtracts the 32-bit unsigned integer <em>y</em> from <em>x</em>, returning the
12807      difference if non-negative, or 2<sup>32</sup> plus the difference otherwise.
12808      The value <em>borrow</em> is set to zero if <span class="eq">x ≥ y</span>, or one
12809      otherwise.</p></td>
12810</tr>
12811<tr>
12812<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>umulExtended</strong>(highp genUType <em>x</em>, highp genUType <em>y</em>, out highp genUType <em>msb</em>, out highp genUType <em>lsb</em>)<br>
12813  void <strong>imulExtended</strong>(highp genIType <em>x</em>, highp genIType <em>y</em>, out highp genIType <em>msb</em>, out highp genIType <em>lsb</em>)</p></td>
12814<td class="tableblock halign-left valign-top"><p class="tableblock">Multiplies 32-bit unsigned or signed integers <em>x</em> and <em>y</em>, producing a
12815      64-bit result.
12816      The 32 least-significant bits are returned in <em>lsb</em>.
12817      The 32 most-significant bits are returned in <em>msb</em>.</p></td>
12818</tr>
12819<tr>
12820<td class="tableblock halign-left valign-top"><p class="tableblock">genIType <strong>bitfieldExtract</strong>(genIType <em>value</em>, int <em>offset</em>, int <em>bits</em>)<br>
12821  genUType <strong>bitfieldExtract</strong>(genUType <em>value</em>, int <em>offset</em>, int <em>bits</em>)</p></td>
12822<td class="tableblock halign-left valign-top"><p class="tableblock">Extracts bits <span class="eq">[offset, offset + bits - 1]</span> from <em>value</em>,
12823      returning them in the least significant bits of the result.<br></p>
12824<p class="tableblock">      For unsigned data types, the most significant bits of the result will
12825      be set to zero.
12826      For signed data types, the most significant bits will be set to the
12827      value of bit <span class="eq">offset + bits - 1</span>.<br></p>
12828<p class="tableblock">      If <em>bits</em> is zero, the result will be zero.
12829      The result will be undefined if <em>offset</em> or <em>bits</em> is negative, or if
12830      the sum of <em>offset</em> and <em>bits</em> is greater than the number of bits used
12831      to store the operand.
12832      Note that for vector versions of <strong>bitfieldExtract</strong>(), a single pair of
12833      <em>offset</em> and <em>bits</em> values is shared for all components.</p></td>
12834</tr>
12835<tr>
12836<td class="tableblock halign-left valign-top"><p class="tableblock">genIType <strong>bitfieldInsert</strong>(genIType <em>base</em>, genIType <em>insert</em>, int <em>offset</em>, int <em>bits</em>)<br>
12837  genUType <strong>bitfieldInsert</strong>(genUType <em>base</em>, genUType <em>insert</em>, int <em>offset</em>, int <em>bits</em>)</p></td>
12838<td class="tableblock halign-left valign-top"><p class="tableblock">Inserts the <em>bits</em> least significant bits of <em>insert</em> into <em>base</em>.</p>
12839<p class="tableblock">      The result will have bits <span class="eq">[offset, offset + bits - 1]</span> taken from
12840      bits <span class="eq">[0, bits - 1]</span> of <em>insert</em>, and all other bits taken
12841      directly from the corresponding bits of <em>base</em>.
12842      If <em>bits</em> is zero, the result will simply be <em>base</em>.
12843      The result will be undefined if <em>offset</em> or <em>bits</em> is negative, or if
12844      the sum of <em>offset</em> and <em>bits</em> is greater than the number of bits used
12845      to store the operand.<br>
12846      Note that for vector versions of <strong>bitfieldInsert</strong>(), a single pair of
12847      <em>offset</em> and <em>bits</em> values is shared for all components.</p></td>
12848</tr>
12849<tr>
12850<td class="tableblock halign-left valign-top"><p class="tableblock">genIType <strong>bitfieldReverse</strong>(highp genIType <em>value</em>)<br>
12851  genUType <strong>bitfieldReverse</strong>(highp genUType <em>value</em>)</p></td>
12852<td class="tableblock halign-left valign-top"><p class="tableblock">Reverses the bits of <em>value</em>.
12853      The bit numbered <em>n</em> of the result will be taken from bit <span class="eq">(bits -
12854      1) - n</span> of <em>value</em>, where <em>bits</em> is the total number of bits used to
12855      represent <em>value</em>.</p></td>
12856</tr>
12857<tr>
12858<td class="tableblock halign-left valign-top"><p class="tableblock">lowp genIType <strong>bitCount</strong>(genIType <em>value</em>)<br>
12859  lowp genIType <strong>bitCount</strong>(genUType <em>value</em>)</p></td>
12860<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the number of one bits in the binary representation of
12861      <em>value</em>.</p></td>
12862</tr>
12863<tr>
12864<td class="tableblock halign-left valign-top"><p class="tableblock">lowp genIType <strong>findLSB</strong>(genIType <em>value</em>)<br>
12865  lowp genIType <strong>findLSB</strong>(genUType <em>value</em>)</p></td>
12866<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the bit number of the least significant one bit in the binary
12867      representation of <em>value</em>.
12868      If <em>value</em> is zero, -1 will be returned.</p></td>
12869</tr>
12870<tr>
12871<td class="tableblock halign-left valign-top"><p class="tableblock">lowp genIType <strong>findMSB</strong>(highp genIType <em>value</em>)<br>
12872  lowp genIType <strong>findMSB</strong>(highp genUType <em>value</em>)</p></td>
12873<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the bit number of the most significant bit in the binary
12874      representation of <em>value</em>.</p>
12875<p class="tableblock">      For positive integers, the result will be the bit number of the most
12876      significant one bit.
12877      For negative integers, the result will be the bit number of the most
12878      significant zero bit.
12879      For a <em>value</em> of zero or negative one, -1 will be returned.</p></td>
12880</tr>
12881</tbody>
12882</table>
12883</div>
12884<div class="sect2">
12885<h3 id="texture-functions">8.9. Texture Functions</h3>
12886<div class="paragraph">
12887<p>Texture lookup functions are available in all shading stages.
12888However, level-of-detail is implicitly computed only for fragment shaders.
12889Other shaders operate as though the base level-of-detail were computed as
12890zero.
12891The functions in the table below provide access to textures through
12892texture-combined samplers, as set up through the API.
12893Texture properties such as size, pixel format, number of dimensions,
12894filtering method, number of mipmap levels, depth comparison, and so on are
12895also defined by API calls.
12896Such properties are taken into account as the texture is accessed via the
12897built-in functions defined below.</p>
12898</div>
12899<div class="paragraph">
12900<p>Texture data can be stored by the GL as single-precision floating-point,
12901unsigned normalized integer, unsigned integer or signed integer data.
12902This is determined by the type of the internal format of the texture.</p>
12903</div>
12904<div class="paragraph">
12905<p>Texture lookup functions are provided that can return their result as
12906floating-point, unsigned integer or signed integer, depending on the sampler
12907type passed to the lookup function.
12908Care must be taken to use the right sampler type for texture access.
12909The following table lists the supported combinations of sampler types and
12910texture internal formats.
12911Blank entries are unsupported.
12912Doing a texture lookup will return undefined values for unsupported
12913combinations.</p>
12914</div>
12915<div class="paragraph">
12916<p>For depth/stencil textures, the internal texture format is determined by the
12917component being accessed as set through the API.
12918When the depth/stencil texture mode is set to DEPTH_COMPONENT, the internal
12919format of the depth component should be used.
12920When the depth/stencil texture mode is set to STENCIL_INDEX, the internal format
12921of the stencil component should be used.</p>
12922</div>
12923<table class="tableblock frame-all grid-all stretch">
12924<colgroup>
12925<col style="width: 25%;">
12926<col style="width: 25%;">
12927<col style="width: 25%;">
12928<col style="width: 25%;">
12929</colgroup>
12930<thead>
12931<tr>
12932<th class="tableblock halign-left valign-top">Internal Texture Format</th>
12933<th class="tableblock halign-left valign-top">Floating-Point Sampler Types</th>
12934<th class="tableblock halign-left valign-top">Signed Integer Sampler Types</th>
12935<th class="tableblock halign-left valign-top">Unsigned Integer Sampler Types</th>
12936</tr>
12937</thead>
12938<tbody>
12939<tr>
12940<td class="tableblock halign-left valign-top"><p class="tableblock">Floating-point</p></td>
12941<td class="tableblock halign-left valign-top"><p class="tableblock">Supported</p></td>
12942<td class="tableblock halign-left valign-top"></td>
12943<td class="tableblock halign-left valign-top"></td>
12944</tr>
12945<tr>
12946<td class="tableblock halign-left valign-top"><p class="tableblock">Normalized Integer</p></td>
12947<td class="tableblock halign-left valign-top"><p class="tableblock">Supported</p></td>
12948<td class="tableblock halign-left valign-top"></td>
12949<td class="tableblock halign-left valign-top"></td>
12950</tr>
12951<tr>
12952<td class="tableblock halign-left valign-top"><p class="tableblock">Signed Integer</p></td>
12953<td class="tableblock halign-left valign-top"></td>
12954<td class="tableblock halign-left valign-top"><p class="tableblock">Supported</p></td>
12955<td class="tableblock halign-left valign-top"></td>
12956</tr>
12957<tr>
12958<td class="tableblock halign-left valign-top"><p class="tableblock">Unsigned Integer</p></td>
12959<td class="tableblock halign-left valign-top"></td>
12960<td class="tableblock halign-left valign-top"></td>
12961<td class="tableblock halign-left valign-top"><p class="tableblock">Supported</p></td>
12962</tr>
12963</tbody>
12964</table>
12965<div class="paragraph">
12966<p>If an integer sampler type is used, the result of a texture lookup is an
12967<strong>ivec4</strong>.
12968If an unsigned integer sampler type is used, the result of a texture lookup
12969is a <strong>uvec4</strong>.
12970If a floating-point sampler type is used, the result of a texture lookup is
12971a <strong>vec4</strong>.</p>
12972</div>
12973<div class="paragraph">
12974<p>In the prototypes below, the <code>g</code> in the return type <code>gvec4</code> is used
12975as a placeholder for either nothing, <code>i</code>, or <code>u</code> making a return type of
12976<strong>vec4</strong>, <strong>ivec4</strong>, or <strong>uvec4</strong>.
12977In these cases, the sampler argument type also starts with <code>g</code>,
12978indicating the same substitution done on the return type; it is either a
12979floating-point, signed integer, or unsigned integer sampler, matching the
12980basic type of the return type, as described above.</p>
12981</div>
12982<div class="paragraph">
12983<p>For shadow forms (the sampler parameter is a shadow-type), a depth
12984comparison lookup on the depth texture bound to <em>sampler</em> is done as
12985described in section
129868.20
12987&#8220;Texture Comparison Modes&#8221; of the
12988<a href="#references">OpenGL ES Specification</a>.
12989See the table below for which component specifies <em>D<sub>ref</sub></em>.
12990The texture bound to <em>sampler</em> must be a depth texture, or results are
12991undefined.
12992If a non-shadow texture call is made to a sampler that represents a depth
12993texture with depth comparisons turned on, then results are undefined.
12994If a shadow texture call is made to a sampler that represents a depth
12995texture with depth comparisons turned off, then results are undefined.
12996If a shadow texture call is made to a sampler that does not represent a
12997depth texture, then results are undefined.</p>
12998</div>
12999<div class="paragraph">
13000<p>In all functions below, the <em>bias</em> parameter is optional for fragment
13001shaders.
13002The <em>bias</em> parameter is not accepted in any other shader stage.
13003For a fragment shader, if <em>bias</em> is present, it is added to the implicit
13004level-of-detail prior to performing the texture access operation.
13005No <em>bias</em> or <em>lod</em> parameters for
13006multisample textures, or texture buffers
13007are supported because mipmaps are not allowed for these types of textures.</p>
13008</div>
13009<div class="paragraph">
13010<p>The implicit level-of-detail is selected as follows: For a texture that is
13011not mipmapped, the texture is used directly.
13012If it is mipmapped and running in a fragment shader, the level-of-detail
13013computed by the implementation is used to do the texture lookup.
13014If it is mipmapped and running in a non-fragment shader, then the base
13015texture is used.</p>
13016</div>
13017<div class="paragraph">
13018<p>Some texture functions (non-&#8220;<strong>Lod</strong>&#8221; and non-&#8220;<strong>Grad</strong>&#8221; versions) may
13019require implicit derivatives.
13020Implicit derivatives are undefined within non-uniform control flow and for
13021non-fragment shader texture fetches.</p>
13022</div>
13023<div class="paragraph">
13024<p>For <strong>Cube</strong> forms, the direction of <em>P</em> is used to select which face to do a
130252-dimensional texture lookup in, as described in section 8.13 &#8220;Cube Map
13026Texture Selection&#8221; of the <a href="#references">OpenGL ES Specification</a>.</p>
13027</div>
13028<div class="paragraph">
13029<p>For <strong>Array</strong> forms, the array layer used will be</p>
13030</div>
13031<div class="paragraph">
13032<p>\(\max(0,\min(d-1,\left\lfloor layer + 0.5\right\rfloor))\)</p>
13033</div>
13034<div class="paragraph">
13035<p>where <em>d</em> is the depth of the texture array and <em>layer</em> comes from the
13036component indicated in the tables below.</p>
13037</div>
13038<div class="sect3">
13039<h4 id="texture-query-functions">8.9.1. Texture Query Functions</h4>
13040<div class="paragraph">
13041<p>The <strong>textureSize</strong> functions query the dimensions of a specific texture level
13042for a texture-combined sampler.</p>
13043</div>
13044<table class="tableblock frame-all grid-all stretch">
13045<colgroup>
13046<col style="width: 50%;">
13047<col style="width: 50%;">
13048</colgroup>
13049<thead>
13050<tr>
13051<th class="tableblock halign-left valign-top">Syntax</th>
13052<th class="tableblock halign-left valign-top">Description</th>
13053</tr>
13054</thead>
13055<tbody>
13056<tr>
13057<td class="tableblock halign-left valign-top"><p class="tableblock">highp ivec2 <strong>textureSize</strong>(gsampler2D <em>sampler</em>, int <em>lod</em>)<br>
13058  highp ivec3 <strong>textureSize</strong>(gsampler3D <em>sampler</em>, int <em>lod</em>)<br>
13059  highp ivec2 <strong>textureSize</strong>(gsamplerCube <em>sampler</em>, int <em>lod</em>)<br>
13060  highp ivec2 <strong>textureSize</strong>(sampler2DShadow <em>sampler</em>, int <em>lod</em>)<br>
13061  highp ivec2 <strong>textureSize</strong>(samplerCubeShadow <em>sampler</em>, int <em>lod</em>)
13062  highp ivec3 <strong>textureSize</strong>(gsamplerCubeArray <em>sampler</em>, int <em>lod</em>)<br>
13063  highp ivec3 <strong>textureSize</strong>(samplerCubeArrayShadow <em>sampler</em>, int <em>lod</em>)<br>
13064  highp ivec3 <strong>textureSize</strong>(gsampler2DArray <em>sampler</em>, int <em>lod</em>)<br>
13065  highp ivec3 <strong>textureSize</strong>(sampler2DArrayShadow <em>sampler</em>, int <em>lod</em>)<br>
13066  highp int <strong>textureSize</strong>(gsamplerBuffer <em>sampler</em>)<br>
13067  highp ivec2 <strong>textureSize</strong>(gsampler2DMS <em>sampler</em>)<br>
13068  highp ivec3 <strong>textureSize</strong>(gsampler2DMSArray <em>sampler</em>)</p></td>
13069<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the dimensions of level <em>lod</em> (if present) for the texture
13070      bound to <em>sampler</em>, as described in section
13071      8.11 &#8220;Texture Queries&#8221; of the <a href="#references">OpenGL ES Specification</a>.<br>
13072      The components in the return value are filled in, in order, with the
13073      width, height, and depth of the texture.</p>
13074<p class="tableblock">      For the array forms, the last component of the return value is the
13075      number of layers in the texture array, or the number of cubes in the
13076      texture cube map array.</p></td>
13077</tr>
13078</tbody>
13079</table>
13080</div>
13081<div class="sect3">
13082<h4 id="texel-lookup-functions">8.9.2. Texel Lookup Functions</h4>
13083<table class="tableblock frame-all grid-all stretch">
13084<colgroup>
13085<col style="width: 50%;">
13086<col style="width: 50%;">
13087</colgroup>
13088<thead>
13089<tr>
13090<th class="tableblock halign-left valign-top">Syntax</th>
13091<th class="tableblock halign-left valign-top">Description</th>
13092</tr>
13093</thead>
13094<tbody>
13095<tr>
13096<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>texture</strong>(gsampler2D <em>sampler</em>, vec2 <em>P</em> [, float <em>bias</em>] )<br>
13097  gvec4 <strong>texture</strong>(gsampler3D <em>sampler</em>, vec3 <em>P</em> [, float <em>bias</em>] )<br>
13098  gvec4 <strong>texture</strong>(gsamplerCube <em>sampler</em>, vec3 <em>P</em>[, float <em>bias</em>] )<br>
13099  float <strong>texture</strong>(sampler2DShadow <em>sampler</em>, vec3 <em>P</em> [, float <em>bias</em>])<br>
13100  float <strong>texture</strong>(samplerCubeShadow <em>sampler</em>, vec4 <em>P</em> [, float <em>bias</em>] )<br>
13101  gvec4 <strong>texture</strong>(gsampler2DArray <em>sampler</em>, vec3 <em>P</em> [, float <em>bias</em>] )<br>
13102  gvec4 <strong>texture</strong>(gsamplerCubeArray <em>sampler</em>, vec4 <em>P</em> [, float <em>bias</em>] )<br>
13103  float <strong>texture</strong>(sampler2DArrayShadow <em>sampler</em>, vec4 <em>P</em>)<br>
13104  float <strong>texture</strong>(samplerCubeArrayShadow <em>sampler</em>, vec4 <em>P</em>, float <em>compare</em>)</p></td>
13105<td class="tableblock halign-left valign-top"><p class="tableblock">Use the texture coordinate <em>P</em> to do a texture lookup in the texture
13106      currently bound to <em>sampler</em>.</p>
13107<p class="tableblock">      For shadow forms: When <em>compare</em> is present, it is used as <em>D<sub>ref</sub></em>
13108      and the array layer comes from the last component of <em>P</em>.
13109      When <em>compare</em> is not present, the last component of <em>P</em> is used as
13110      <em>D<sub>ref</sub></em> and the array layer comes from the second to last component
13111      of <em>P</em>.</p>
13112<p class="tableblock">      For non-shadow forms: the array layer comes from the last component of
13113      <em>P</em>.</p></td>
13114</tr>
13115<tr>
13116<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureProj</strong>(gsampler2D <em>sampler</em>, vec3 <em>P</em> [, float <em>bias</em>] )<br>
13117  gvec4 <strong>textureProj</strong>(gsampler2D <em>sampler</em>, vec4 <em>P</em> [, float <em>bias</em>] )<br>
13118  gvec4 <strong>textureProj</strong>(gsampler3D <em>sampler</em>, vec4 <em>P</em> [, float <em>bias</em>] )<br>
13119  float <strong>textureProj</strong>(sampler2DShadow <em>sampler</em>, vec4 <em>P</em> [, float <em>bias</em>] )<br></p></td>
13120<td class="tableblock halign-left valign-top"><p class="tableblock">Do a texture lookup with projection.
13121      The texture coordinates consumed from <em>P</em>, not including the last
13122      component of <em>P</em>, are divided by the last component of <em>P</em> to
13123      form projected coordinates <em>P'</em>.
13124      The resulting third component of <em>P</em> in the shadow forms is used as
13125      <em>D<sub>ref</sub></em>.
13126      The third component of <em>P</em> is ignored when <em>sampler</em> has type
13127      <strong>gsampler2D</strong> and <em>P</em> has type <strong>vec4</strong>.
13128      After these values are computed, texture lookup proceeds as in
13129      <strong>texture</strong>.</p></td>
13130</tr>
13131<tr>
13132<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureLod</strong>(gsampler2D <em>sampler</em>, vec2 <em>P</em>, float <em>lod</em>)<br>
13133  gvec4 <strong>textureLod</strong>(gsampler3D <em>sampler</em>, vec3 <em>P</em>, float <em>lod</em>)<br>
13134  gvec4 <strong>textureLod</strong>(gsamplerCube <em>sampler</em>, vec3 <em>P</em>, float <em>lod</em>)<br>
13135  float <strong>textureLod</strong>(sampler2DShadow <em>sampler</em>, vec3 <em>P</em>, float <em>lod</em>)<br>
13136  gvec4 <strong>textureLod</strong>(gsampler2DArray <em>sampler</em>, vec3 <em>P</em>, float <em>lod</em>)<br>
13137  gvec4 <strong>textureLod</strong>(gsamplerCubeArray <em>sampler</em>, vec4 <em>P</em>, float <em>lod</em>)</p></td>
13138<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
13139<p>Do a texture lookup as in <strong>texture</strong> but with explicit level-of-detail;
13140      <em>lod</em> specifies <span class="eq">λ<sub>base</sub>]</span> and sets the partial derivatives
13141      as follows:<br>
13142      (See section 8.14 &#8220;Texture Minification&#8221; and equations 8.4-8.6 of
13143      the <a href="#references">OpenGL ES Specification</a>.)<br>
13144     <br>
13145      <span class="eq">∂u / ∂x =
13146           ∂v / ∂x =
13147           ∂w / ∂x = 0</span>
13148     <br>
13149      <span class="eq">∂u / ∂y =
13150           ∂v / ∂y =
13151           ∂w / ∂y = 0</span></p>
13152</div></div></td>
13153</tr>
13154<tr>
13155<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureOffset</strong>(gsampler2D <em>sampler</em>, vec2 <em>P</em>, ivec2 <em>offset</em> [, float <em>bias</em>] )<br>
13156  gvec4 <strong>textureOffset</strong>(gsampler3D <em>sampler</em>, vec3 <em>P</em>, ivec3 <em>offset</em> [, float <em>bias</em>] )<br>
13157  float <strong>textureOffset</strong>(sampler2DShadow <em>sampler</em>, vec3 <em>P</em>, ivec2 <em>offset</em> [, float <em>bias</em>] )<br>
13158  gvec4 <strong>textureOffset</strong>(gsampler2DArray <em>sampler</em>, vec3 <em>P</em>, ivec2 <em>offset</em> [, float <em>bias</em>] )<br></p></td>
13159<td class="tableblock halign-left valign-top"><p class="tableblock">Do a texture lookup as in <strong>texture</strong> but with <em>offset</em> added to the
13160      <span class="eq">(u,v,w)</span> texel coordinates before looking up each texel.
13161      The offset value must be a constant expression.
13162      A limited range of offset values are supported; the minimum and
13163      maximum offset values are implementation-dependent and given by
13164      <em>gl_MinProgramTexelOffset</em> and <em>gl_MaxProgramTexelOffset</em>,
13165      respectively.</p>
13166<p class="tableblock">      Note that <em>offset</em> does not apply to the layer coordinate for texture
13167      arrays.
13168      This is explained in detail in section 8.14.2 &#8220;Coordinate Wrapping
13169      and Texel Selection&#8221; of the <a href="#references">OpenGL ES Specification</a>, where <em>offset</em>
13170      is <span class="eq">(δ<sub>u</sub>, δ<sub>v</sub>, δ<sub>w</sub>)</span>.<br>
13171      Note that texel offsets are also not supported for cube maps.</p></td>
13172</tr>
13173<tr>
13174<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>texelFetch</strong>(gsampler2D <em>sampler</em>, ivec2 <em>P</em>, int <em>lod</em>)<br>
13175  gvec4 <strong>texelFetch</strong>(gsampler3D <em>sampler</em>, ivec3 <em>P</em>, int <em>lod</em>)
13176  gvec4 <strong>texelFetch</strong>(gsampler2DArray <em>sampler</em>, ivec3 <em>P</em>, int <em>lod</em>)<br>
13177  gvec4 <strong>texelFetch</strong>(gsamplerBuffer <em>sampler</em>, int <em>P</em>)<br>
13178  gvec4 <strong>texelFetch</strong>(gsampler2DMS <em>sampler</em>, ivec2 <em>P</em>, int <em>sample</em>)<br>
13179  gvec4 <strong>texelFetch</strong>(gsampler2DMSArray <em>sampler</em>, ivec3 <em>P</em>, int <em>sample</em>)</p></td>
13180<td class="tableblock halign-left valign-top"><p class="tableblock">Use integer texture coordinate <em>P</em> to lookup a single texel from
13181      <em>sampler</em>.
13182      The array layer comes from the last component of <em>P</em> for the array
13183      forms.
13184      The level-of-detail <em>lod</em> (if present) is as described in sections
13185      11.1.3.2 &#8220;Texel Fetches&#8221; and 8.14.1 &#8220;Scale Factor and Level of
13186      Detail&#8221; of the <a href="#references">OpenGL ES Specification</a>.</p></td>
13187</tr>
13188<tr>
13189<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>texelFetchOffset</strong>(gsampler2D <em>sampler</em>, ivec2 <em>P</em>, int <em>lod</em>, ivec2 <em>offset</em>)<br>
13190  gvec4 <strong>texelFetchOffset</strong>(gsampler3D <em>sampler</em>, ivec3 <em>P</em>, int <em>lod</em>, ivec3 <em>offset</em>)<br>
13191  gvec4 <strong>texelFetchOffset</strong>(gsampler2DArray <em>sampler</em>, ivec3 <em>P</em>, int <em>lod</em>, ivec2 <em>offset</em>)</p></td>
13192<td class="tableblock halign-left valign-top"><p class="tableblock">Fetch a single texel as in <strong>texelFetch</strong>, offset by <em>offset</em> as
13193      described in <strong>textureOffset</strong>.</p></td>
13194</tr>
13195<tr>
13196<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureProjOffset</strong>(gsampler2D <em>sampler</em>, vec3 <em>P</em>, ivec2 <em>offset</em> [, float <em>bias</em>] )<br>
13197  gvec4 <strong>textureProjOffset</strong>(gsampler2D <em>sampler</em>, vec4 <em>P</em>, ivec2 <em>offset</em> [, float <em>bias</em>] )<br>
13198  gvec4 <strong>textureProjOffset</strong>(gsampler3D <em>sampler</em>, vec4 <em>P</em>, ivec3 <em>offset</em> [, float <em>bias</em>] )<br>
13199  float <strong>textureProjOffset</strong>(sampler2DShadow <em>sampler</em>, vec4 <em>P</em>, ivec2 <em>offset</em> [, float <em>bias</em>] )</p></td>
13200<td class="tableblock halign-left valign-top"><p class="tableblock">Do a projective texture lookup as described in <strong>textureProj</strong>, offset
13201      by <em>offset</em> as described in <strong>textureOffset</strong>.</p></td>
13202</tr>
13203<tr>
13204<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureLodOffset</strong>(gsampler2D <em>sampler</em>, vec2 <em>P</em>, float <em>lod</em>, ivec2 <em>offset</em>)<br>
13205  gvec4 <strong>textureLodOffset</strong>(gsampler3D <em>sampler</em>, vec3 <em>P</em>, float <em>lod</em>, ivec3 <em>offset</em>)<br>
13206  float <strong>textureLodOffset</strong>(sampler2DShadow <em>sampler</em>,  vec3 <em>P</em>, float <em>lod</em>, ivec2 <em>offset</em>)<br>
13207  gvec4 <strong>textureLodOffset</strong>(gsampler2DArray <em>sampler</em>, vec3 <em>P</em>, float <em>lod</em>, ivec2 <em>offset</em>)<br></p></td>
13208<td class="tableblock halign-left valign-top"><p class="tableblock">Do an offset texture lookup with explicit level-of-detail.
13209      See <strong>textureLod</strong> and <strong>textureOffset</strong>.</p></td>
13210</tr>
13211<tr>
13212<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureProjLod</strong>(gsampler2D <em>sampler</em>, vec3 <em>P</em>, float <em>lod</em>)<br>
13213  gvec4 <strong>textureProjLod</strong>(gsampler2D <em>sampler</em>, vec4 <em>P</em>, float <em>lod</em>)<br>
13214  gvec4 <strong>textureProjLod</strong>(gsampler3D <em>sampler</em>, vec4 <em>P</em>, float <em>lod</em>)<br>
13215  float <strong>textureProjLod</strong>(sampler2DShadow <em>sampler</em>, vec4 <em>P</em>, float <em>lod</em>)</p></td>
13216<td class="tableblock halign-left valign-top"><p class="tableblock">Do a projective texture lookup with explicit level-of-detail.
13217      See <strong>textureProj</strong> and <strong>textureLod</strong>.</p></td>
13218</tr>
13219<tr>
13220<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureProjLodOffset</strong>(gsampler2D <em>sampler</em>, vec3 <em>P</em>, float <em>lod</em>, ivec2 <em>offset</em>)<br>
13221  gvec4 <strong>textureProjLodOffset</strong>(gsampler2D <em>sampler</em>, vec4 <em>P</em>, float <em>lod</em>, ivec2 <em>offset</em>)<br>
13222  gvec4 <strong>textureProjLodOffset</strong>(gsampler3D <em>sampler</em>, vec4 <em>P</em>, float <em>lod</em>, ivec3 <em>offset</em>)<br>
13223  float <strong>textureProjLodOffset</strong>(sampler2DShadow <em>sampler</em>, vec4 <em>P</em>, float <em>lod</em>, ivec2 <em>offset</em>)</p></td>
13224<td class="tableblock halign-left valign-top"><p class="tableblock">Do an offset projective texture lookup with explicit level-of-detail.
13225      See <strong>textureProj</strong>, <strong>textureLod</strong>, and <strong>textureOffset</strong>.</p></td>
13226</tr>
13227<tr>
13228<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureGrad</strong>(gsampler2D <em>sampler</em>, vec2 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>)<br>
13229  gvec4 <strong>textureGrad</strong>(gsampler3D <em>sampler</em>,  <em>P</em>, vec3 <em>dPdx</em>, vec3 <em>dPdy</em>)<br>
13230  gvec4 <strong>textureGrad</strong>(gsamplerCube <em>sampler</em>, vec3 <em>P</em>, vec3 <em>dPdx</em>, vec3 <em>dPdy</em>)<br>
13231  float <strong>textureGrad</strong>(sampler2DShadow <em>sampler</em>, vec3 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>)<br>
13232  float <strong>textureGrad</strong>(samplerCubeShadow <em>sampler</em>, vec4 <em>P</em>, vec3 <em>dPdx</em>, vec3 <em>dPdy</em>)<br>
13233  float <strong>textureGrad</strong>(sampler2DArrayShadow <em>sampler</em>, vec4 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>)<br>
13234  gvec4 <strong>textureGrad</strong>(gsamplerCubeArray <em>sampler</em>, vec4 <em>P</em>, vec3 <em>dPdx</em>, vec3 <em>dPdy</em>)<br></p></td>
13235<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
13236<p>Do a texture lookup as in <strong>texture</strong> but with <a href="#explicit-gradients">explicit gradients</a> as shown below.
13237      The partial derivatives of <em>P</em> are with respect to window <em>x</em> and
13238      window <em>y</em>.
13239      For the cube version, the partial derivatives of <em>P</em> are assumed to be
13240      in the coordinate system used before texture coordinates are projected
13241      onto the appropriate cube face.</p>
13242</div></div></td>
13243</tr>
13244<tr>
13245<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureGradOffset</strong>(gsampler2D <em>sampler</em>, vec2 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>, ivec2 <em>offset</em>)<br>
13246  gvec4 <strong>textureGradOffset</strong>(gsampler3D <em>sampler</em>, vec3 <em>P</em>, vec3 <em>dPdx</em>, vec3 <em>dPdy</em>, ivec3 <em>offset</em>)<br>
13247  float <strong>textureGradOffset</strong>(sampler2DShadow <em>sampler</em>, vec3 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>, ivec2 <em>offset</em>)<br>
13248  gvec4 <strong>textureGradOffset</strong>(gsampler2DArray <em>sampler</em>, vec3 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>, ivec2 <em>offset</em>)<br>
13249  float <strong>textureGradOffset</strong>(sampler2DArrayShadow <em>sampler</em>, vec4 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>, ivec2 <em>offset</em>)<br></p></td>
13250<td class="tableblock halign-left valign-top"><p class="tableblock">Do a texture lookup with both explicit gradient and offset, as
13251      described in <strong>textureGrad</strong> and <strong>textureOffset</strong>.</p></td>
13252</tr>
13253<tr>
13254<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureProjGrad</strong>(gsampler2D <em>sampler</em>, vec3 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>)<br>
13255  gvec4 <strong>textureProjGrad</strong>(gsampler2D <em>sampler</em>, vec4 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>)<br>
13256  gvec4 <strong>textureProjGrad</strong>(gsampler3D <em>sampler</em>, vec4 <em>P</em>, vec3 <em>dPdx</em>, vec3 <em>dPdy</em>)<br>
13257  float <strong>textureProjGrad</strong>(sampler2DShadow <em>sampler</em>, vec4 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>)</p></td>
13258<td class="tableblock halign-left valign-top"><p class="tableblock">Do a texture lookup both projectively, as described in <strong>textureProj</strong>,
13259      and with explicit gradient as described in <strong>textureGrad</strong>.
13260      The partial derivatives <em>dPdx</em> and <em>dPdy</em> are assumed to be already
13261      projected.</p></td>
13262</tr>
13263<tr>
13264<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureProjGradOffset</strong>(gsampler2D <em>sampler</em>, vec3 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>, ivec2 <em>offset</em>)<br>
13265  gvec4 <strong>textureProjGradOffset</strong>(gsampler2D <em>sampler</em>, vec4 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>, ivec2 <em>offset</em>)<br>
13266  gvec4 <strong>textureProjGradOffset</strong>(gsampler3D <em>sampler</em>, vec4 <em>P</em>, vec3 <em>dPdx</em>, vec3 <em>dPdy</em>, ivec3 <em>offset</em>)<br>
13267  float <strong>textureProjGradOffset</strong>(sampler2DShadow <em>sampler</em>, vec4 <em>P</em>, vec2 <em>dPdx</em>, vec2 <em>dPdy</em>, ivec2 <em>offset</em>)</p></td>
13268<td class="tableblock halign-left valign-top"><p class="tableblock">Do a texture lookup projectively and with explicit gradient as
13269      described in <strong>textureProjGrad</strong>, as well as with offset, as described in
13270      <strong>textureOffset</strong>.</p></td>
13271</tr>
13272</tbody>
13273</table>
13274</div>
13275<div class="sect3">
13276<h4 id="explicit-gradients">8.9.3. Explicit Gradients</h4>
13277<div class="paragraph">
13278<p>In the <strong>textureGrad</strong> functions described above, explicit gradients control
13279texture lookups as follows:</p>
13280</div>
13281<div class="stemblock">
13282<div class="content">
13283\[\begin{aligned}
13284  \frac{\partial{s}}{\partial{x}} &amp; = \frac{\partial{P.s}}{\partial{x}} \\[0.8em]
13285  \frac{\partial{s}}{\partial{y}} &amp; = \frac{\partial{P.s}}{\partial{y}} \\[0.8em]
13286  \frac{\partial{t}}{\partial{x}} &amp; = \frac{\partial{P.t}}{\partial{x}} \\[0.8em]
13287  \frac{\partial{t}}{\partial{y}} &amp; = \frac{\partial{P.t}}{\partial{y}} \\[0.8em]
13288  \frac{\partial{r}}{\partial{x}} &amp; =
13289    \begin{cases}
13290      0.0,                               &amp; \text{for 2D} \\[0.8em]
13291      \frac{\partial{P.p}}{\partial{x}}, &amp; \text{cube, other}
13292    \end{cases} \\[2.5em]
13293  \frac{\partial{r}}{\partial{y}} &amp; =
13294    \begin{cases}
13295      0.0,                               &amp; \text{for 2D} \\[0.8em]
13296      \frac{\partial{P.p}}{\partial{y}}, &amp; \text{cube, other}
13297    \end{cases}
13298\end{aligned}\]
13299</div>
13300</div>
13301</div>
13302<div class="sect3">
13303<h4 id="texture-gather-functions">8.9.4. Texture Gather Functions</h4>
13304<div class="paragraph">
13305<p>The texture gather functions take components of a single floating-point
13306vector operand as a texture coordinate, determine a set of four texels to
13307sample from the base level-of-detail of the specified texture image, and
13308return one component from each texel in a four-component result vector.</p>
13309</div>
13310<div class="paragraph">
13311<p>When performing a texture gather operation, the minification and
13312magnification filters are ignored, and the rules for LINEAR filtering in the
13313<a href="#references">OpenGL ES Specification</a> are applied to the base level of the texture image
13314to identify the four texels <em>i<sub>0</sub> j<sub>1</sub></em>, <em>i<sub>1</sub> j<sub>1</sub></em>, <em>i<sub>1</sub> j<sub>0</sub></em>, and <em>i<sub>0</sub>
13315j<sub>0</sub></em>.
13316The texels are then converted to texture base colors (<em>R<sub>s</sub></em>, <em>G<sub>s</sub></em>,
13317<em>B<sub>s</sub></em>, <em>A<sub>s</sub></em>) according to table 15.1, followed by application of the
13318texture swizzle as described in section 15.2.1 &#8220;Texture Access&#8221; of the
13319<a href="#references">OpenGL ES Specification</a>.
13320A four-component vector is assembled by taking the selected component from
13321each of the post-swizzled texture source colors in the order (<em>i<sub>0</sub> j<sub>1</sub></em>,
13322<em>i<sub>1</sub> j<sub>1</sub></em>, <em>i<sub>1</sub> j<sub>0</sub></em>, <em>i<sub>0</sub> j<sub>0</sub></em>).</p>
13323</div>
13324<div class="paragraph">
13325<p>The selected component is identified by the optional <em>comp</em> argument, where
13326the values zero, one, two, and three identify the <em>R<sub>s</sub></em>, <em>G<sub>s</sub></em>, <em>B<sub>s</sub></em>, or
13327<em>A<sub>s</sub></em> component, respectively.
13328If <em>comp</em> is omitted, it is treated as identifying the <em>R<sub>s</sub></em> component.</p>
13329</div>
13330<div class="paragraph">
13331<p>Incomplete textures (see section 8.16 &#8220;Texture Completeness&#8221; of the
13332<a href="#references">OpenGL ES Specification</a>) return a texture source color of (0,0,0,1) for all
13333four source texels.</p>
13334</div>
13335<div class="paragraph">
13336<p>For texture gather functions using a texture-combined shadow sampler type,
13337each of the four
13338texel lookups perform a depth comparison against the depth reference value
13339passed in (<em>refZ</em>), and returns the result of that comparison in the
13340appropriate component of the result vector.</p>
13341</div>
13342<div class="paragraph">
13343<p>As with other texture lookup functions, the results of a texture gather are
13344undefined for shadow samplers if the texture referenced is not a depth
13345texture or has depth comparisons disabled; or for non-shadow samplers if the
13346texture referenced is a depth texture with depth comparisons enabled.</p>
13347</div>
13348<div class="paragraph">
13349<p>The <strong>textureGatherOffset</strong> built-in functions from the OpenGL ES Shading Language return a vector
13350derived from sampling four texels in the image array of level <em>level<sub>base</sub></em>.
13351For each of the four texel offsets specified by the <em>offsets</em> argument, the
13352rules for the LINEAR minification filter are applied to identify a 2 ×
133532 texel footprint, from which the single texel T<sub>i0j0</sub> is selected.
13354A four-component vector is then assembled by taking a single component from
13355each of the four T<sub>i0j0</sub> texels in the same manner as for the
13356<strong>textureGather</strong> function.</p>
13357</div>
13358<table class="tableblock frame-all grid-all stretch">
13359<colgroup>
13360<col style="width: 50%;">
13361<col style="width: 50%;">
13362</colgroup>
13363<thead>
13364<tr>
13365<th class="tableblock halign-left valign-top">Syntax</th>
13366<th class="tableblock halign-left valign-top">Description</th>
13367</tr>
13368</thead>
13369<tbody>
13370<tr>
13371<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureGather</strong>(gsampler2D <em>sampler</em>, vec2 <em>P</em> [, int <em>comp</em>])<br>
13372  gvec4 <strong>textureGather</strong>(gsampler2DArray <em>sampler</em>, vec3 <em>P</em> [, int <em>comp</em>])<br>
13373  gvec4 <strong>textureGather</strong>(gsamplerCube <em>sampler</em>, vec3 <em>P</em> [, int <em>comp</em>])<br>
13374  gvec4 <strong>textureGather</strong>(gsamplerCubeArray <em>sampler</em>, vec4 <em>P</em>[, int <em>comp</em>])<br>
13375  vec4 <strong>textureGather</strong>(sampler2DShadow <em>sampler</em>, vec2 <em>P</em>, float <em>refZ</em>)<br>
13376  vec4 <strong>textureGather</strong>(sampler2DArrayShadow <em>sampler</em>, vec3 <em>P</em>, float <em>refZ</em>)<br>
13377  vec4 <strong>textureGather</strong>(samplerCubeShadow <em>sampler</em>, vec3 <em>P</em>, float <em>refZ</em>)<br>
13378  vec4 <strong>textureGather</strong>(samplerCubeArrayShadow <em>sampler</em>, vec4 <em>P</em>, float <em>refZ</em>)<br></p></td>
13379<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
13380<p>Returns the value<br></p>
13381</div>
13382<div class="openblock">
13383<div class="content">
13384<div class="listingblock">
13385<div class="content">
13386<pre class="CodeRay highlight"><code data-lang="c++">vec4(Sample_i0_j1(P, base).comp,
13387     Sample_i1_j1(P, base).comp,
13388     Sample_i1_j0(P, base).comp,
13389     Sample_i0_j0(P, base).comp)</code></pre>
13390</div>
13391</div>
13392<div class="paragraph">
13393<p>If specified, the value of <em>comp</em> must be a constant integer expression with
13394a value of 0, 1, 2, or 3, identifying the <em>x</em>, <em>y</em>, <em>z</em>, or <em>w</em>
13395post-swizzled component of the four-component vector lookup result for each
13396texel, respectively.
13397If <em>comp</em> is not specified, it is treated as 0, selecting the <em>x</em> component
13398of each texel to generate the result.</p>
13399</div>
13400</div>
13401</div></div></td>
13402</tr>
13403<tr>
13404<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureGatherOffset</strong>(gsampler2D <em>sampler</em>, vec2 <em>P</em>, ivec2 <em>offset</em>, [ int <em>comp</em>])<br>
13405  gvec4 <strong>textureGatherOffset</strong>(gsampler2DArray <em>sampler</em>, vec3 <em>P</em>, ivec2 <em>offset</em> [ int <em>comp</em>])<br>
13406  vec4 <strong>textureGatherOffset</strong>(sampler2DShadow <em>sampler</em>, vec2 <em>P</em>, float <em>refZ</em>, ivec2 <em>offset</em>)<br>
13407  vec4 <strong>textureGatherOffset</strong>(sampler2DArrayShadow <em>sampler</em>, vec3 <em>P</em>, float <em>refZ</em>, ivec2 <em>offset</em>)<br></p></td>
13408<td class="tableblock halign-left valign-top"><p class="tableblock">Perform a texture gather operation as in <strong>textureGather</strong> by <em>offset</em>
13409      as described in <strong>textureOffset</strong> except that the <em>offset</em> can be
13410      variable (non constant) and the implementation-dependent minimum and
13411      maximum offset values are given by MIN_PROGRAM_TEXTURE_GATHER_OFFSET
13412      and MAX_PROGRAM_TEXTURE_GATHER_OFFSET, respectively.</p></td>
13413</tr>
13414<tr>
13415<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>textureGatherOffsets</strong>(gsampler2D <em>sampler</em>, vec2 <em>P</em>, ivec2 <em>offsets</em>[4] [, int <em>comp</em>])<br>
13416  gvec4 <strong>textureGatherOffsets</strong>(gsampler2DArray <em>sampler</em>, vec3 <em>P</em>, ivec2 <em>offsets</em>[4]   [, int <em>comp</em>])<br>
13417  vec4 <strong>textureGatherOffsets</strong>(sampler2DShadow <em>sampler</em>, vec2 <em>P</em>, float <em>refZ</em>, ivec2 <em>offsets</em>[4])<br>
13418  vec4 <strong>textureGatherOffsets</strong>(sampler2DArrayShadow <em>sampler</em>, vec3 <em>P</em>, float <em>refZ</em>, ivec2 <em>offsets</em>[4])<br></p></td>
13419<td class="tableblock halign-left valign-top"><p class="tableblock">Operate identically to <strong>textureGatherOffset</strong> except that <em>offsets</em> is
13420      used to determine the location of the four texels to sample.
13421      Each of the four texels is obtained by applying the corresponding
13422      offset in <em>offsets</em> as a (<em>u</em>, <em>v</em>) coordinate offset to <em>P</em>,
13423      identifying the four-texel LINEAR footprint, and then selecting the
13424      texel <em>i<sub>0</sub> j<sub>0</sub></em> of that footprint.
13425      The specified values in <em>offsets</em> must be constant integral
13426      expressions.</p></td>
13427</tr>
13428</tbody>
13429</table>
13430</div>
13431</div>
13432<div class="sect2">
13433<h3 id="atomic-counter-functions">8.10. Atomic Counter Functions</h3>
13434<div class="paragraph">
13435<p>The atomic-counter operations in this section operate atomically with
13436respect to each other.
13437They are atomic for any single counter, meaning any of these operations on a
13438specific counter in one shader instantiation will be indivisible by any of
13439these operations on the same counter from another shader instantiation.
13440There is no guarantee that these operations are atomic with respect to other
13441forms of access to the counter or that they are serialized when applied to
13442separate counters.
13443Such cases would require additional use of fences, barriers, or other forms
13444of synchronization, if atomicity or serialization is desired.</p>
13445</div>
13446<div class="paragraph">
13447<p>The underlying counter is a 32-bit unsigned integer.
13448The result of operations will wrap to [0, 2<sup>32</sup>-1].</p>
13449</div>
13450<table class="tableblock frame-all grid-all stretch">
13451<colgroup>
13452<col style="width: 50%;">
13453<col style="width: 50%;">
13454</colgroup>
13455<thead>
13456<tr>
13457<th class="tableblock halign-left valign-top">Syntax</th>
13458<th class="tableblock halign-left valign-top">Description</th>
13459</tr>
13460</thead>
13461<tbody>
13462<tr>
13463<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicCounterIncrement</strong>(atomic_uint <em>c</em>)</p></td>
13464<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
13465<p>Atomically<br></p>
13466</div>
13467<div class="openblock">
13468<div class="content">
13469<div class="olist arabic">
13470<ol class="arabic">
13471<li>
13472<p>increments the counter for <em>c</em>, and</p>
13473</li>
13474<li>
13475<p>returns its value prior to the increment operation.</p>
13476</li>
13477</ol>
13478</div>
13479<div class="paragraph">
13480<p>These two steps are done atomically with respect to the atomic counter
13481functions in this table.</p>
13482</div>
13483</div>
13484</div></div></td>
13485</tr>
13486<tr>
13487<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicCounterDecrement</strong>(atomic_uint <em>c</em>)</p></td>
13488<td class="tableblock halign-left valign-top"><div class="content"><div class="paragraph">
13489<p>Atomically<br></p>
13490</div>
13491<div class="openblock">
13492<div class="content">
13493<div class="olist arabic">
13494<ol class="arabic">
13495<li>
13496<p>decrements the counter for <em>c</em>, and</p>
13497</li>
13498<li>
13499<p>returns the value resulting from the decrement operation.</p>
13500</li>
13501</ol>
13502</div>
13503<div class="paragraph">
13504<p>These two steps are done atomically with respect to the atomic counter
13505functions in this table.</p>
13506</div>
13507</div>
13508</div></div></td>
13509</tr>
13510<tr>
13511<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicCounter</strong>(atomic_uint <em>c</em>)</p></td>
13512<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the counter value for <em>c</em>.</p></td>
13513</tr>
13514</tbody>
13515</table>
13516</div>
13517<div class="sect2">
13518<h3 id="atomic-memory-functions">8.11. Atomic Memory Functions</h3>
13519<div class="paragraph">
13520<p>Atomic memory functions perform atomic operations on an individual signed or
13521unsigned integer stored in buffer object or shared variable storage.
13522All of the atomic memory operations read a value from memory, compute a new
13523value using one of the operations described below, write the new value to
13524memory, and return the original value read.
13525The contents of the memory being updated by the atomic operation are
13526guaranteed not to be modified by any other assignment or atomic memory
13527function in any shader invocation between the time the original value is
13528read and the time the new value is written.</p>
13529</div>
13530<div class="paragraph">
13531<p>Atomic memory functions are supported only for a limited set of variables.
13532A shader will fail to compile if the value passed to the <em>mem</em> argument of
13533an atomic memory function does not correspond to a buffer or shared
13534variable.
13535It is acceptable to pass an element of an array or a single component of a
13536vector to the <em>mem</em> argument of an atomic memory function, as long as the
13537underlying array or vector is a buffer or shared variable.</p>
13538</div>
13539<div class="paragraph">
13540<p>All the built-in functions in this section accept arguments with
13541combinations of <strong>restrict</strong>, <strong>coherent</strong>, and <strong>volatile</strong> memory qualification,
13542despite not having them listed in the prototypes.
13543The atomic operation will operate as required by the calling argument&#8217;s
13544memory qualification, not by the built-in function&#8217;s formal parameter memory
13545qualification.</p>
13546</div>
13547<table class="tableblock frame-all grid-all stretch">
13548<colgroup>
13549<col style="width: 50%;">
13550<col style="width: 50%;">
13551</colgroup>
13552<thead>
13553<tr>
13554<th class="tableblock halign-left valign-top">Syntax</th>
13555<th class="tableblock halign-left valign-top">Description</th>
13556</tr>
13557</thead>
13558<tbody>
13559<tr>
13560<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicAdd</strong>(inout uint <em>mem</em>, uint <em>data</em>)<br>
13561  int <strong>atomicAdd</strong>(inout int <em>mem</em>, int <em>data</em>)</p></td>
13562<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by adding the value of <em>data</em> to the contents
13563      <em>mem</em>.</p></td>
13564</tr>
13565<tr>
13566<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicMin</strong>(inout uint <em>mem</em>, uint <em>data</em>)<br>
13567  int <strong>atomicMin</strong>(inout int <em>mem</em>, int <em>data</em>)</p></td>
13568<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by taking the minimum of the value of <em>data</em> and
13569      the contents of <em>mem</em>.</p></td>
13570</tr>
13571<tr>
13572<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicMax</strong>(inout uint <em>mem</em>, uint <em>data</em>)<br>
13573  int <strong>atomicMax</strong>(inout int <em>mem</em>, int <em>data</em>)</p></td>
13574<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by taking the maximum of the value of <em>data</em> and
13575      the contents of <em>mem</em>.</p></td>
13576</tr>
13577<tr>
13578<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicAnd</strong>(inout uint <em>mem</em>, uint <em>data</em>)<br>
13579  int <strong>atomicAnd</strong>(inout int <em>mem</em>, int <em>data</em>)</p></td>
13580<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by performing a bit-wise AND of the value of
13581      <em>data</em> and the contents of <em>mem</em>.</p></td>
13582</tr>
13583<tr>
13584<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicOr</strong>(inout uint <em>mem</em>, uint <em>data</em>)<br>
13585  int <strong>atomicOr</strong>(inout int <em>mem</em>, int <em>data</em>)</p></td>
13586<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by performing a bit-wise OR of the value of
13587      <em>data</em> and the contents of <em>mem</em>.</p></td>
13588</tr>
13589<tr>
13590<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicXor</strong>(inout uint <em>mem</em>, uint <em>data</em>)<br>
13591  int <strong>atomicXor</strong>(inout int <em>mem</em>, int <em>data</em>)</p></td>
13592<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by performing a bit-wise EXCLUSIVE OR of the
13593      value of <em>data</em> and the contents of <em>mem</em>.</p></td>
13594</tr>
13595<tr>
13596<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicExchange</strong>(inout uint <em>mem</em>, uint <em>data</em>)<br>
13597  int <strong>atomicExchange</strong>(inout int <em>mem</em>, int <em>data</em>)</p></td>
13598<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by simply copying the value of <em>data</em>.</p></td>
13599</tr>
13600<tr>
13601<td class="tableblock halign-left valign-top"><p class="tableblock">uint <strong>atomicCompSwap</strong>(inout uint <em>mem</em>, uint <em>compare</em>, uint <em>data</em>)<br>
13602  int <strong>atomicCompSwap</strong>(inout int <em>mem</em>, int <em>compare</em>, int <em>data</em>)</p></td>
13603<td class="tableblock halign-left valign-top"><p class="tableblock">Compares the value of <em>compare</em> and the contents of <em>mem</em>.
13604      If the values are equal, the new value is given by <em>data</em>; otherwise,
13605      it is taken from the original contents of <em>mem</em>.</p></td>
13606</tr>
13607</tbody>
13608</table>
13609</div>
13610<div class="sect2">
13611<h3 id="image-functions">8.12. Image Functions</h3>
13612<div class="paragraph">
13613<p>Variables using one of the image basic types may be used by the built-in
13614shader image memory functions defined in this section to read and write
13615individual texels of a texture.
13616Each image variable references an image unit, which has a texture image
13617attached.</p>
13618</div>
13619<div class="paragraph">
13620<p>When image memory functions below access memory, an individual texel in the
13621image is identified using an (<em>i</em>), (<em>i, j</em>), or (<em>i, j, k</em>) coordinate
13622corresponding to the values of <em>P</em>.
13623The coordinates
13624are used to select an individual texel in the manner described in section
136258.22
13626&#8220;Texture Image Loads and Stores&#8221; of the <a href="#references">OpenGL ES Specification</a>.</p>
13627</div>
13628<div class="paragraph">
13629<p>Loads and stores support float, integer, and unsigned integer types.
13630The data types below starting <code>gimage</code> serve as placeholders meaning
13631types starting either &#8220;<strong>image</strong>&#8221;, &#8220;<strong>iimage</strong>&#8221;, or &#8220;<strong>uimage</strong>&#8221; in the same
13632way as "<strong>gvec</strong>" or "<strong>gsampler</strong>" in earlier sections.</p>
13633</div>
13634<div class="paragraph">
13635<p>The <em>IMAGE_PARAMS</em> in the prototypes below is a placeholder representing
1363618
13637separate functions, each for a different type of image variable.
13638The <em>IMAGE_PARAMS</em> placeholder is replaced by one of the following parameter
13639lists:</p>
13640</div>
13641<div class="dlist">
13642<dl>
13643<dt class="hdlist1"></dt>
13644<dd>
13645<p>gimage2D <em>image</em>, ivec2 <em>P</em></p>
13646</dd>
13647<dt class="hdlist1"></dt>
13648<dd>
13649<p>gimage3D <em>image</em>, ivec3 <em>P</em></p>
13650</dd>
13651<dt class="hdlist1"></dt>
13652<dd>
13653<p>gimageCube <em>image</em>, ivec3 <em>P</em></p>
13654</dd>
13655<dt class="hdlist1"></dt>
13656<dd>
13657<p>gimageBuffer <em>image</em>, int <em>P</em></p>
13658</dd>
13659<dt class="hdlist1"></dt>
13660<dd>
13661<p>gimage2DArray <em>image</em>, ivec3 <em>P</em></p>
13662</dd>
13663<dt class="hdlist1"></dt>
13664<dd>
13665<p>gimageCubeArray <em>image</em>, ivec3 <em>P</em></p>
13666</dd>
13667</dl>
13668</div>
13669<div class="paragraph">
13670<p>where each of the lines represents one of three different image variable
13671types, and <em>image</em>,
13672<em>P</em>
13673specify the individual texel to operate on.
13674The method for identifying the individual texel operated on from <em>image</em>,
13675<em>P</em>,
13676and the method for reading and writing the texel are specified in section
136778.22
13678&#8220;Texture Image Loads and Stores&#8221; of the <a href="#references">OpenGL ES Specification</a>.</p>
13679</div>
13680<div class="paragraph">
13681<p>The atomic functions perform operations on individual texels or samples of
13682an image variable.
13683Atomic memory operations read a value from the selected texel, compute a new
13684value using one of the operations described below, write the new value to
13685the selected texel, and return the original value read.
13686The contents of the texel being updated by the atomic operation are
13687guaranteed not to be modified by any other image store or atomic function
13688between the time the original value is read and the time the new value is
13689written.</p>
13690</div>
13691<div class="paragraph">
13692<p>Atomic memory operations are supported on only a subset of all image
13693variable types; <em>image</em> must be either:</p>
13694</div>
13695<div class="ulist">
13696<ul>
13697<li>
13698<p>a signed integer image variable (type starts &#8220;<strong>iimage</strong>&#8221;) and a format
13699qualifier of <strong>r32i</strong>, used with a <em>data</em> argument of type <strong>int</strong>, or</p>
13700</li>
13701<li>
13702<p>an unsigned integer image variable (type starts &#8220;<strong>uimage</strong>&#8221;) and a
13703format qualifier of <strong>r32ui</strong>, used with a <em>data</em> argument of type <strong>uint</strong>,
13704or</p>
13705</li>
13706<li>
13707<p>a float image variable (type starts &#8220;<strong>image</strong>&#8221;) and a format qualifier
13708of <strong>r32f</strong>, used with a <em>data</em> argument of type <strong>float</strong>
13709(<strong>imageAtomicExchange</strong> only).</p>
13710</li>
13711</ul>
13712</div>
13713<div class="paragraph">
13714<p>All the built-in functions in this section accept arguments with
13715combinations of <strong>restrict</strong>, <strong>coherent</strong>, and <strong>volatile</strong> memory qualification,
13716despite not having them listed in the prototypes.
13717The image operation will operate as required by the calling argument&#8217;s
13718memory qualification, not by the built-in function&#8217;s formal parameter memory
13719qualification.</p>
13720</div>
13721<table class="tableblock frame-all grid-all stretch">
13722<colgroup>
13723<col style="width: 50%;">
13724<col style="width: 50%;">
13725</colgroup>
13726<thead>
13727<tr>
13728<th class="tableblock halign-left valign-top">Syntax</th>
13729<th class="tableblock halign-left valign-top">Description</th>
13730</tr>
13731</thead>
13732<tbody>
13733<tr>
13734<td class="tableblock halign-left valign-top"><p class="tableblock">highp ivec2 <strong>imageSize</strong>(readonly writeonly gimage2D <em>image</em>)<br>
13735  highp ivec3 <strong>imageSize</strong>(readonly writeonly gimage3D <em>image</em>)<br>
13736  highp ivec2 <strong>imageSize</strong>(readonly writeonly gimageCube <em>image</em>)<br>
13737  highp ivec3 <strong>imageSize</strong>(readonly writeonly gimageCubeArray <em>image</em>)<br>
13738  highp ivec3 <strong>imageSize</strong>(readonly writeonly gimage2DArray <em>image</em>)<br>
13739  highp   int <strong>imageSize</strong>(readonly writeonly gimageBuffer <em>image</em>)</p></td>
13740<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the dimensions of the image or images bound to <em>image</em>.
13741      For arrayed images, the last component of the return value will hold
13742      the size of the array.
13743      Cube images only return the dimensions of one face, and the number of
13744      cubes in the cube map array, if arrayed.<br>
13745      Note: The qualification <strong>readonly writeonly</strong> accepts a variable
13746      qualified with <strong>readonly</strong>, <strong>writeonly</strong>, both, or neither.
13747      It means the formal argument will be used for neither reading nor
13748      writing to the underlying memory.</p></td>
13749</tr>
13750<tr>
13751<td class="tableblock halign-left valign-top"><p class="tableblock">highp gvec4 <strong>imageLoad</strong>(readonly <em>IMAGE_PARAMS</em>)</p></td>
13752<td class="tableblock halign-left valign-top"><p class="tableblock">Loads the texel at the coordinate <em>P</em> from the image unit <em>image</em> (in
13753      <em>IMAGE_PARAMS</em>).
13754      When <em>image</em> and <em>P</em>
13755      identify a valid texel, the bits used to represent the selected texel in
13756      memory are converted to a <strong>vec4</strong>, <strong>ivec4</strong>, or <strong>uvec4</strong> in the manner
13757      described in section
137588.23
13759      &#8220;Texture Image Loads and Stores&#8221; of the
13760      <a href="#references">OpenGL ES Specification</a> and returned.</p></td>
13761</tr>
13762<tr>
13763<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>imageStore</strong>(writeonly <em>IMAGE_PARAMS</em>, gvec4 <em>data</em>)</p></td>
13764<td class="tableblock halign-left valign-top"><p class="tableblock">Stores <em>data</em> into the texel at the coordinate <em>P</em> from the image
13765      specified by <em>image</em>.
13766      When <em>image</em> and <em>P</em>
13767      identify a valid texel, the bits used to represent <em>data</em> are converted
13768      to the format of the image unit in the manner described in section
137698.23
13770      &#8220;Texture Image Loads and Stores&#8221; of the <a href="#references">OpenGL ES Specification</a>
13771      and stored to the specified texel.</p></td>
13772</tr>
13773<tr>
13774<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>imageAtomicAdd</strong>(<em>IMAGE_PARAMS</em>, uint <em>data</em>)<br>
13775  highp int <strong>imageAtomicAdd</strong>(<em>IMAGE_PARAMS</em>, int <em>data</em>)</p></td>
13776<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by adding the value of <em>data</em> to the contents of
13777      the selected texel.</p></td>
13778</tr>
13779<tr>
13780<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>imageAtomicMin</strong>(<em>IMAGE_PARAMS</em>, uint <em>data</em>)<br>
13781  highp int <strong>imageAtomicMin</strong>(<em>IMAGE_PARAMS</em>, int <em>data</em>)</p></td>
13782<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by taking the minimum of the value of <em>data</em> and
13783      the contents of the selected texel.</p></td>
13784</tr>
13785<tr>
13786<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>imageAtomicMax</strong>(<em>IMAGE_PARAMS</em>, uint <em>data</em>)<br>
13787  highp int <strong>imageAtomicMax</strong>(<em>IMAGE_PARAMS</em>, int <em>data</em>)</p></td>
13788<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by taking the maximum of the value <em>data</em> and the
13789      contents of the selected texel.</p></td>
13790</tr>
13791<tr>
13792<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>imageAtomicAnd</strong>(<em>IMAGE_PARAMS</em>, uint <em>data</em>)<br>
13793  highp int <strong>imageAtomicAnd</strong>(<em>IMAGE_PARAMS</em>, int <em>data</em>)</p></td>
13794<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by performing a bit-wise AND of the value of
13795      <em>data</em> and the contents of the selected texel.</p></td>
13796</tr>
13797<tr>
13798<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>imageAtomicOr</strong>(<em>IMAGE_PARAMS</em>, uint <em>data</em>)<br>
13799  highp int <strong>imageAtomicOr</strong>(<em>IMAGE_PARAMS</em>, int <em>data</em>)</p></td>
13800<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by performing a bit-wise OR of the value of
13801      <em>data</em> and the contents of the selected texel.</p></td>
13802</tr>
13803<tr>
13804<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>imageAtomicXor</strong>(<em>IMAGE_PARAMS</em>, uint <em>data</em>)<br>
13805  highp int <strong>imageAtomicXor</strong>(<em>IMAGE_PARAMS</em>, int <em>data</em>)</p></td>
13806<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by performing a bit-wise EXCLUSIVE OR of the
13807      value of <em>data</em> and the contents of the selected texel.</p></td>
13808</tr>
13809<tr>
13810<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>imageAtomicExchange</strong>(<em>IMAGE_PARAMS</em>, uint <em>data</em>)<br>
13811  highp int <strong>imageAtomicExchange</strong>(<em>IMAGE_PARAMS</em>, int <em>data</em>)<br>
13812  highp float <strong>imageAtomicExchange</strong>(<em>IMAGE_PARAMS</em>, float <em>data</em>)</p></td>
13813<td class="tableblock halign-left valign-top"><p class="tableblock">Computes a new value by simply copying the value of <em>data</em>.</p></td>
13814</tr>
13815<tr>
13816<td class="tableblock halign-left valign-top"><p class="tableblock">highp uint <strong>imageAtomicCompSwap</strong>(<em>IMAGE_PARAMS</em>, uint <em>compare</em>, uint <em>data</em>)<br>
13817  highp int <strong>imageAtomicCompSwap</strong>(<em>IMAGE_PARAMS</em>, int <em>compare</em>, int <em>data</em>)</p></td>
13818<td class="tableblock halign-left valign-top"><p class="tableblock">Compares the value of <em>compare</em> and the contents of the selected
13819      texel.
13820      If the values are equal, the new value is given by <em>data</em>; otherwise,
13821      it is taken from the original value loaded from the texel.</p></td>
13822</tr>
13823</tbody>
13824</table>
13825</div>
13826<div class="sect2">
13827<h3 id="geometry-shader-functions">8.13. Geometry Shader Functions</h3>
13828<div class="paragraph">
13829<p>These functions are only available in geometry shaders.
13830They are described in more depth following the table.</p>
13831</div>
13832<table class="tableblock frame-all grid-all stretch">
13833<colgroup>
13834<col style="width: 50%;">
13835<col style="width: 50%;">
13836</colgroup>
13837<thead>
13838<tr>
13839<th class="tableblock halign-left valign-top">Syntax</th>
13840<th class="tableblock halign-left valign-top">Description</th>
13841</tr>
13842</thead>
13843<tbody>
13844<tr>
13845<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>EmitVertex</strong>()<br></p></td>
13846<td class="tableblock halign-left valign-top"><p class="tableblock">Emits the current values of output variables to the current output
13847      primitive.
13848      On return from this call, the values of output variables are
13849      undefined.</p></td>
13850</tr>
13851<tr>
13852<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>EndPrimitive</strong>()</p></td>
13853<td class="tableblock halign-left valign-top"><p class="tableblock">Completes the current output primitive and starts a new one.
13854      No vertex is emitted.</p></td>
13855</tr>
13856</tbody>
13857</table>
13858<div class="paragraph">
13859<p>The function <strong>EmitVertex</strong>() specifies that a vertex is completed.
13860A vertex is added to the current output primitive using the current values
13861of all built-in and user-defined output variables.
13862The values of all output variables are undefined after a call to
13863<strong>EmitVertex</strong>().
13864If a geometry shader invocation has emitted more vertices than permitted by
13865the output layout qualifier <strong>max_vertices</strong>, the results of calling
13866<strong>EmitVertex</strong>() are undefined.</p>
13867</div>
13868<div class="paragraph">
13869<p>The function <strong>EndPrimitive</strong>() specifies that the current output primitive is
13870completed and a new output primitive (of the same type) will be started by
13871any subsequent <strong>EmitVertex</strong>().
13872This function does not emit a vertex.
13873If the output layout is declared to be <strong>points</strong>, calling <strong>EndPrimitive</strong>() is
13874optional.</p>
13875</div>
13876<div class="paragraph">
13877<p>A geometry shader starts with an output primitive containing no vertices.
13878When a geometry shader terminates, the current output primitive is
13879automatically completed.
13880It is not necessary to call <strong>EndPrimitive</strong>() if the geometry shader writes
13881only a single primitive.</p>
13882</div>
13883</div>
13884<div class="sect2">
13885<h3 id="fragment-processing-functions">8.14. Fragment Processing Functions</h3>
13886<div class="paragraph">
13887<p>Fragment processing functions are only available in fragment shaders.</p>
13888</div>
13889<div class="sect3">
13890<h4 id="derivative-functions">8.14.1. Derivative Functions</h4>
13891<div class="paragraph">
13892<p>Derivatives may be computationally expensive and/or numerically unstable.
13893Therefore, an implementation may approximate the true derivatives
13894by using a fast but not entirely accurate derivative computation.
13895Derivatives are undefined within non-uniform control flow.</p>
13896</div>
13897<div class="paragraph">
13898<p>The expected behavior of a derivative is specified using forward/backward
13899differencing.</p>
13900</div>
13901<div class="paragraph">
13902<p>Forward differencing:</p>
13903</div>
13904<div class="paragraph">
13905<p>\(F(x+dx) - F(x) \sim dFdx(x) \cdot dx (1a)\)</p>
13906</div>
13907<div class="paragraph">
13908<p>\(dFdx(x) \sim \frac{F(x+dx) - F(x)}{dx} (1b)\)</p>
13909</div>
13910<div class="paragraph">
13911<p>Backward differencing:</p>
13912</div>
13913<div class="paragraph">
13914<p>\(F(x-dx) - F(x) \sim -dFdx(x) \cdot dx (2a)\)</p>
13915</div>
13916<div class="paragraph">
13917<p>\(dFdx(x) \sim \frac{F(x) - F(x-dx)}{dx} (2b)\)</p>
13918</div>
13919<div class="paragraph">
13920<p>With single-sample rasterization, \(dx \leq 1.0\) in equations 1b
13921and 2b.
13922For multisample rasterization, \(dx &lt; 2.0\) in equations 1b and 2b.</p>
13923</div>
13924<div class="paragraph">
13925<p>\(dFdy\) is approximated similarly, with <em>y</em> replacing <em>x</em>.</p>
13926</div>
13927<div class="paragraph">
13928<p>An implementation may use the above or other methods to perform
13929the calculation, subject to the following conditions:</p>
13930</div>
13931<div class="olist arabic">
13932<ol class="arabic">
13933<li>
13934<p>The method may use piecewise linear approximations.
13935Such linear approximations imply that higher order derivatives,
13936<strong>dFdx</strong>(<strong>dFdx</strong>(<em>x</em>)) and above, are undefined.</p>
13937</li>
13938<li>
13939<p>The method may assume that the function evaluated is continuous.
13940Therefore derivatives within the body of a non-uniform conditional are
13941undefined.</p>
13942</li>
13943<li>
13944<p>The method may differ per fragment, subject to the constraint that the
13945method may vary by window coordinates, not screen coordinates.
13946The invariance requirement described in section 13.2 &#8220;Invariance&#8221; of
13947the <a href="#references">OpenGL ES Specification</a>, is relaxed for derivative calculations,
13948because the method may be a function of fragment location.</p>
13949</li>
13950</ol>
13951</div>
13952<div class="paragraph">
13953<p>Other properties that are desirable, but not required, are:</p>
13954</div>
13955<div class="olist arabic">
13956<ol class="arabic">
13957<li>
13958<p>Functions should be evaluated within the interior of a primitive
13959(interpolated, not extrapolated).</p>
13960</li>
13961<li>
13962<p>Functions for <strong>dFdx</strong> should be evaluated while holding <em>y</em> constant.
13963Functions for <strong>dFdy</strong> should be evaluated while holding <em>x</em> constant.
13964However, mixed higher order derivatives, like <strong>dFdx</strong>(<strong>dFdy</strong>(<em>y</em>)) and
13965<strong>dFdy</strong>(<strong>dFdx</strong>(<em>x</em>)) are undefined.</p>
13966</li>
13967<li>
13968<p>Derivatives of constant arguments should be 0.</p>
13969</li>
13970</ol>
13971</div>
13972<div class="paragraph">
13973<p>In some implementations, varying degrees of derivative accuracy may be
13974obtained by providing GL hints (see section 19.1 &#8220;Hints&#8221; of the
13975<a href="#references">OpenGL ES Specification</a>), allowing a user to make an image quality versus
13976speed trade off.</p>
13977</div>
13978<table class="tableblock frame-all grid-all stretch">
13979<colgroup>
13980<col style="width: 50%;">
13981<col style="width: 50%;">
13982</colgroup>
13983<thead>
13984<tr>
13985<th class="tableblock halign-left valign-top">Syntax</th>
13986<th class="tableblock halign-left valign-top">Description</th>
13987</tr>
13988</thead>
13989<tbody>
13990<tr>
13991<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>dFdx</strong>(genFType <em>p</em>)</p></td>
13992<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the derivative in x using local differencing for the input
13993      argument <em>p</em>.</p></td>
13994</tr>
13995<tr>
13996<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>dFdy</strong>(genFType <em>p</em>)</p></td>
13997<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the derivative in y using local differencing for the input
13998      argument <em>p</em>.<br>
13999      <br>
14000      These two functions are commonly used to estimate the filter width used
14001      to anti-alias procedural textures. We are assuming that the expression
14002      is being evaluated in parallel on a SIMD array so that at any given
14003      point in time the value of the function is known at the grid points
14004      represented by the SIMD array. Local differencing between SIMD array
14005      elements can therefore be used to derive <strong>dFdx</strong>, <strong>dFdy</strong>, etc.</p></td>
14006</tr>
14007<tr>
14008<td class="tableblock halign-left valign-top"><p class="tableblock">genFType <strong>fwidth</strong>(genFType <em>p</em>)</p></td>
14009<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the sum of the absolute derivative in x and y using local
14010      differencing for the input argument <em>p</em>, i.e., <strong>abs</strong>(<strong>dFdx</strong>(<em>p</em>))
14011      + <strong>abs</strong>(<strong>dFdy</strong>(<em>p</em>));</p></td>
14012</tr>
14013</tbody>
14014</table>
14015</div>
14016<div class="sect3">
14017<h4 id="interpolation-functions">8.14.2. Interpolation Functions</h4>
14018<div class="paragraph">
14019<p>Built-in interpolation functions are available to compute an interpolated
14020value of a fragment shader input variable at a shader-specified (<em>x</em>, <em>y</em>)
14021location.
14022A separate (<em>x</em>, <em>y</em>) location may be used for each invocation of the
14023built-in function, and those locations may differ from the default (<em>x</em>,
14024<em>y</em>) location used to produce the default value of the input.
14025For the <strong>interpolateAt*</strong> functions, the call will return a precision
14026qualification matching the precision of the <em>interpolant</em> argument to the
14027function call.</p>
14028</div>
14029<div class="paragraph">
14030<p>For all of the interpolation functions, <em>interpolant</em> must be an l-value
14031from an <strong>in</strong> declaration;
14032this can be either a variable, or an array element.
14033Component selection operators (e.g. <strong>.xy</strong>) may not be used when specifying <em>interpolant</em>.</p>
14034</div>
14035<div class="paragraph">
14036<p>If <em>interpolant</em> is declared with the <strong>flat</strong> qualifier, the interpolated
14037value will have the same value everywhere for a single primitive, so the
14038location used for interpolation has no effect and the functions just return
14039that same value.
14040If <em>interpolant</em> is declared with the <strong>centroid</strong> qualifier, the value
14041returned by <strong>interpolateAtSample</strong>() and <strong>interpolateAtOffset</strong>() will be
14042evaluated at the specified location, ignoring the location normally used
14043with the <strong>centroid</strong> qualifier.</p>
14044</div>
14045<table class="tableblock frame-all grid-all stretch">
14046<colgroup>
14047<col style="width: 50%;">
14048<col style="width: 50%;">
14049</colgroup>
14050<thead>
14051<tr>
14052<th class="tableblock halign-left valign-top">Syntax</th>
14053<th class="tableblock halign-left valign-top">Description</th>
14054</tr>
14055</thead>
14056<tbody>
14057<tr>
14058<td class="tableblock halign-left valign-top"><p class="tableblock">float <strong>interpolateAtCentroid</strong>(float <em>interpolant</em>)<br>
14059  vec2 <strong>interpolateAtCentroid</strong>(vec2 <em>interpolant</em>)<br>
14060  vec3 <strong>interpolateAtCentroid</strong>(vec3 <em>interpolant</em>)<br>
14061  vec4 <strong>interpolateAtCentroid</strong>(vec4 <em>interpolant</em>)</p></td>
14062<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the value of the input <em>interpolant</em> sampled at a location
14063      inside both the pixel and the primitive being processed.
14064      The value obtained would be the same value assigned to the input
14065      variable if declared with the <strong>centroid</strong> qualifier.</p></td>
14066</tr>
14067<tr>
14068<td class="tableblock halign-left valign-top"><p class="tableblock">float <strong>interpolateAtSample</strong>(float <em>interpolant</em>, int <em>sample</em>)<br>
14069  vec2 <strong>interpolateAtSample</strong>(vec2 <em>interpolant</em>, int <em>sample</em>)<br>
14070  vec3 <strong>interpolateAtSample</strong>(vec3 <em>interpolant</em>, int <em>sample</em>)<br>
14071  vec4 <strong>interpolateAtSample</strong>(vec4 <em>interpolant</em>, int <em>sample</em>)</p></td>
14072<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the value of the input <em>interpolant</em> variable at the location
14073      of sample number <em>sample</em>.
14074      If multisample buffers are not available, the input variable will be
14075      evaluated at the center of the pixel.
14076      If sample <em>sample</em> does not exist, the position used to interpolate
14077      the input variable is undefined.</p></td>
14078</tr>
14079<tr>
14080<td class="tableblock halign-left valign-top"><p class="tableblock">float <strong>interpolateAtOffset</strong>(float <em>interpolant</em>, vec2 offset)<br>
14081  vec2 <strong>interpolateAtOffset</strong>(vec2 <em>interpolant</em>, vec2 offset)<br>
14082  vec3 <strong>interpolateAtOffset</strong>(vec3 <em>interpolant</em>, vec2 offset)<br>
14083  vec4 <strong>interpolateAtOffset</strong>(vec4 <em>interpolant</em>, vec2 offset)</p></td>
14084<td class="tableblock halign-left valign-top"><p class="tableblock">Returns the value of the input <em>interpolant</em> variable sampled at an
14085      offset from the center of the pixel specified by <em>offset</em>.
14086      The two floating-point components of <em>offset</em>, give the offset in
14087      pixels in the <em>x</em> and <em>y</em> directions, respectively.<br>
14088      An offset of (0, 0) identifies the center of the pixel.
14089      The range and granularity of offsets supported by this function is
14090      implementation-dependent.</p></td>
14091</tr>
14092</tbody>
14093</table>
14094</div>
14095</div>
14096<div class="sect2">
14097<h3 id="shader-invocation-control-functions">8.15. Shader Invocation Control Functions</h3>
14098<div class="paragraph">
14099<p>The shader invocation control function is only available in tessellation
14100control and compute shaders.
14101It is used to control the relative execution order of multiple shader
14102invocations used to process a patch (in the case of tessellation control
14103shaders) or a workgroup (in the case of compute shaders), which are
14104otherwise executed with an undefined relative order.</p>
14105</div>
14106<table class="tableblock frame-all grid-all stretch">
14107<colgroup>
14108<col style="width: 50%;">
14109<col style="width: 50%;">
14110</colgroup>
14111<thead>
14112<tr>
14113<th class="tableblock halign-left valign-top">Syntax</th>
14114<th class="tableblock halign-left valign-top">Description</th>
14115</tr>
14116</thead>
14117<tbody>
14118<tr>
14119<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>barrier</strong>()</p></td>
14120<td class="tableblock halign-left valign-top"><p class="tableblock">For any given static instance of <strong>barrier</strong>(), all tessellation control
14121      shader invocations for a single input patch must enter it before any
14122      will be allowed to continue beyond it, or all compute shader
14123      invocations for a single workgroup must enter it before any will
14124      continue beyond it.</p></td>
14125</tr>
14126</tbody>
14127</table>
14128<div class="paragraph">
14129<p>The function <strong>barrier</strong>() provides a partially defined order of execution
14130between shader invocations.
14131The ensures that, for some types of memory accesses, values written by one
14132invocation prior to a given static instance of <strong>barrier</strong>() can be safely read
14133by other invocations after their call to the same static instance <strong>barrier</strong>().
14134Because invocations may execute in an undefined order between these barrier
14135calls, the values of a per-vertex or per-patch output variable for tessellation
14136control shaders, or the values of <strong>shared</strong> variables for compute shaders will be
14137undefined in a number of cases enumerated in
14138&#8220;<a href="#output-variables">Output Variables</a>&#8221; (for tessellation control shaders)
14139and &#8220;<a href="#shared-variables">Shared Variables</a>&#8221; (for compute shaders).</p>
14140</div>
14141<div class="paragraph">
14142<p>For tessellation control shaders, the <strong>barrier</strong>() function may only be
14143placed inside the function <strong>main</strong>() of the shader and may not be called
14144within any control flow.
14145Barriers are also disallowed after a return statement in the function
14146<strong>main</strong>().
14147Any such misplaced barriers result in a compile-time error.</p>
14148</div>
14149<div class="paragraph">
14150<p>A <strong>barrier</strong>() affects control flow but only synchronizes memory accesses
14151to <strong>shared</strong> variables and tessellation control output variables.
14152For other memory accesses, it does not ensure that values written by one invocation
14153prior to a given static instance of <strong>barrier</strong>() can be safely read by other
14154invocations after their call to the same static instance of <strong>barrier</strong>().
14155To achieve this requires the use of both <strong>barrier</strong>() and a memory barrier.</p>
14156</div>
14157<div class="paragraph">
14158<p>For compute shaders, the <strong>barrier</strong>() function may be placed within control
14159flow, but that control flow must be uniform control flow.
14160That is, all the controlling expressions that lead to execution of the
14161barrier must be dynamically uniform expressions.
14162This ensures that if any shader invocation enters a conditional statement,
14163then all invocations will enter it.
14164While compilers are encouraged to give warnings if they can detect this
14165might not happen, compilers cannot completely determine this.
14166Hence, it is the author&#8217;s responsibility to ensure <strong>barrier</strong>() only exists
14167inside uniform control flow.
14168Otherwise, some shader invocations will stall indefinitely, waiting for a
14169barrier that is never reached by other invocations.</p>
14170</div>
14171</div>
14172<div class="sect2">
14173<h3 id="shader-memory-control-functions">8.16. Shader Memory Control Functions</h3>
14174<div class="paragraph">
14175<p>Within a single shader invocation, the visibility and order of writes made
14176by that invocation are well-defined.
14177However, the relative order of reads and writes to a single shared memory
14178address from multiple separate shader invocations is largely undefined.
14179Additionally, the order of accesses to multiple memory addresses performed
14180by a single shader invocation, as observed by other shader invocations, is
14181also undefined.</p>
14182</div>
14183<div class="paragraph">
14184<p>The following built-in functions can be used to control the ordering of
14185reads and writes:</p>
14186</div>
14187<table class="tableblock frame-all grid-all stretch">
14188<colgroup>
14189<col style="width: 50%;">
14190<col style="width: 50%;">
14191</colgroup>
14192<thead>
14193<tr>
14194<th class="tableblock halign-left valign-top">Syntax</th>
14195<th class="tableblock halign-left valign-top">Description</th>
14196</tr>
14197</thead>
14198<tbody>
14199<tr>
14200<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>memoryBarrier</strong>()</p></td>
14201<td class="tableblock halign-left valign-top"><p class="tableblock">Control the ordering of memory transactions issued by a single shader
14202      invocation.</p></td>
14203</tr>
14204<tr>
14205<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>memoryBarrierAtomicCounter</strong>()</p></td>
14206<td class="tableblock halign-left valign-top"><p class="tableblock">Control the ordering of accesses to atomic-counter variables issued by
14207      a single shader invocation.</p></td>
14208</tr>
14209<tr>
14210<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>memoryBarrierBuffer</strong>()</p></td>
14211<td class="tableblock halign-left valign-top"><p class="tableblock">Control the ordering of memory transactions to buffer variables issued
14212      within a single shader invocation.</p></td>
14213</tr>
14214<tr>
14215<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>memoryBarrierShared</strong>()</p></td>
14216<td class="tableblock halign-left valign-top"><p class="tableblock">Control the ordering of memory transactions to shared variables issued
14217      within a single shader invocation, as viewed by other invocations in
14218      the same workgroup.<br>
14219      Only available in compute shaders.</p></td>
14220</tr>
14221<tr>
14222<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>memoryBarrierImage</strong>()</p></td>
14223<td class="tableblock halign-left valign-top"><p class="tableblock">Control the ordering of memory transactions to images issued within a
14224      single shader invocation.</p></td>
14225</tr>
14226<tr>
14227<td class="tableblock halign-left valign-top"><p class="tableblock">void <strong>groupMemoryBarrier</strong>()</p></td>
14228<td class="tableblock halign-left valign-top"><p class="tableblock">Control the ordering of all memory transactions issued within a single
14229      shader invocation, as viewed by other invocations in the same workgroup.<br>
14230      Only available in compute shaders.</p></td>
14231</tr>
14232</tbody>
14233</table>
14234<div class="paragraph">
14235<p>The memory barrier built-in functions can be used to order reads and writes
14236to variables stored in memory accessible to other shader invocations.
14237When called, these functions will wait for the completion of all reads and
14238writes previously performed by the caller that access selected variable
14239types, and then return with no other effect.
14240The built-in functions <strong>memoryBarrierAtomicCounter</strong>(),
14241<strong>memoryBarrierBuffer</strong>(), <strong>memoryBarrierImage</strong>(), and <strong>memoryBarrierShared</strong>()
14242wait for the completion of accesses to atomic counter, buffer, image, and
14243shared variables, respectively.
14244The built-in functions <strong>memoryBarrier</strong>() and <strong>groupMemoryBarrier</strong>() wait for
14245the completion of accesses to all of the above variable types.
14246The functions <strong>memoryBarrierShared</strong>() and <strong>groupMemoryBarrier</strong>() are
14247available only in compute shaders; the other functions are available in all
14248shader types.</p>
14249</div>
14250<div class="paragraph">
14251<p>When these functions return, the effects of any memory stores performed
14252using coherent variables prior to the call will be visible to any
14253future<sup>1</sup> coherent access to the same memory performed by any other shader
14254invocation.
14255In particular, the values written this way in one shader stage are
14256guaranteed to be visible to coherent memory accesses performed by shader
14257invocations in subsequent stages when those invocations were triggered by
14258the execution of the original shader invocation (e.g. fragment shader
14259invocations for a primitive resulting from a particular geometry
14260shader invocation).</p>
14261</div>
14262<div class="dlist">
14263<dl>
14264<dt class="hdlist1">1</dt>
14265<dd>
14266<p>An access is only a <em>future</em> access if a <em>happens-before</em> relation can
14267be established between the store and the load.</p>
14268</dd>
14269</dl>
14270</div>
14271<div class="paragraph">
14272<p>Additionally, memory barrier functions order stores performed by the calling
14273invocation, as observed by other shader invocations.
14274Without memory barriers, if one shader invocation performs two stores to
14275coherent variables, a second shader invocation might see the values written
14276by the second store prior to seeing those written by the first.
14277However, if the first shader invocation calls a memory barrier function
14278between the two stores, selected other shader invocations will never see the
14279results of the second store before seeing those of the first.
14280When using the functions <strong>groupMemoryBarrier</strong>() or <strong>memoryBarrierShared</strong>(),
14281this ordering guarantee applies only to other shader invocations in the same
14282compute shader workgroup; all other memory barrier functions provide the
14283guarantee to all other shader invocations.
14284No memory barrier is required to guarantee the order of memory stores as
14285observed by the invocation performing the stores; an invocation reading from
14286a variable that it previously wrote will always see the most recently
14287written value unless another shader invocation also wrote to the same
14288memory.</p>
14289</div>
14290</div>
14291<div class="sect2">
14292<h3 id="_subpass_input_functions">8.17. Subpass-Input Functions</h3>
14293<div class="paragraph">
14294<p>Subpass-input functions are only available when targeting a Vulkan fragment stage.</p>
14295</div>
14296<div class="paragraph">
14297<p>Subpass inputs are read through the built-in functions below. The <code>g</code> is again
14298a placeholder for either nothing, <code>i</code>, or <code>u</code>, indicating either a floating-point,
14299signed integer, or unsigned integer, and these must match between argument type
14300and return type.</p>
14301</div>
14302<table class="tableblock frame-all grid-all stretch">
14303<colgroup>
14304<col style="width: 50%;">
14305<col style="width: 50%;">
14306</colgroup>
14307<thead>
14308<tr>
14309<th class="tableblock halign-left valign-top">Syntax</th>
14310<th class="tableblock halign-left valign-top">Description</th>
14311</tr>
14312</thead>
14313<tbody>
14314<tr>
14315<td class="tableblock halign-left valign-top"><p class="tableblock">gvec4 <strong>subpassLoad</strong>(gsubpassInput subpass)<br>
14316  gvec4 <strong>subpassLoad</strong>(gsubpassInputMS subpass, int sample)</p></td>
14317<td class="tableblock halign-left valign-top"><p class="tableblock">Read from a subpass input, from the implicit location <em>(x, y, layer)</em>
14318      of the current fragment coordinate.</p></td>
14319</tr>
14320</tbody>
14321</table>
14322</div>
14323</div>
14324</div>
14325<div class="sect1">
14326<h2 id="shader-interface-matching">9. Shader Interface Matching</h2>
14327<div class="sectionbody">
14328<div class="paragraph">
14329<p>As described in chapter 7 of the <a href="#references">OpenGL ES Specification</a>, shaders may be
14330linked together to form a <em>program object</em> before being bound to the
14331pipeline or may be linked and bound individually as <em>separable program
14332objects</em>.</p>
14333</div>
14334<div class="admonitionblock note">
14335<table>
14336<tr>
14337<td class="icon">
14338<i class="fa icon-note" title="Note"></i>
14339</td>
14340<td class="content">
14341<div class="title">Note</div>
14342<div class="paragraph">
14343<p>These were previously known as separate shader objects (SSOs) but the
14344mechanism has been extended to support future versions of the specification
14345that have more than two shader stages.
14346It allows a subset of the shaders to be linked together.</p>
14347</div>
14348</td>
14349</tr>
14350</table>
14351</div>
14352<div class="paragraph">
14353<p>Within a <em>program object</em> or a <em>separable program object</em>, qualifiers for
14354matching variables must themselves match according to the rules specified in
14355this section.
14356There are also matching rules for qualifiers of matching variables between
14357separable program objects but only for variables across an input/output
14358boundary between shader stages.
14359For other shader interface variables such as uniforms, each program object
14360or separable program object has its own name space and so the same name can
14361refer to multiple independent variables.
14362Consequently, there are no matching rules for qualifiers in these cases.</p>
14363</div>
14364<div class="sect2">
14365<h3 id="input-output-matching-by-name-in-linked-programs">9.1. Input Output Matching by Name in Linked Programs</h3>
14366<div class="paragraph">
14367<p>When linking shaders, the type of declared vertex outputs and fragment
14368inputs with the same name must match, otherwise the link command will fail.
14369Only those fragment inputs statically used (i.e. read) in the fragment
14370shader must be declared as outputs in the vertex shader; declaring
14371superfluous vertex shader outputs is permissible.</p>
14372</div>
14373<div class="paragraph">
14374<p>The following table summarizes the rules for matching shader outputs to
14375shader inputs in consecutive stages when shaders are linked together.</p>
14376</div>
14377<table class="tableblock frame-all grid-all fit-content">
14378<colgroup>
14379<col>
14380<col>
14381<col>
14382<col>
14383<col>
14384</colgroup>
14385<tbody>
14386<tr>
14387<td class="tableblock halign-left valign-top" colspan="2" rowspan="2"><p class="tableblock">Treatment of Mismatched Input Variables</p></td>
14388<td class="tableblock halign-left valign-top" colspan="3"><p class="tableblock">Consuming Shader (input variables)</p></td>
14389</tr>
14390<tr>
14391<td class="tableblock halign-left valign-top"><p class="tableblock">No Declaration</p></td>
14392<td class="tableblock halign-left valign-top"><p class="tableblock">Declared but no Static Use</p></td>
14393<td class="tableblock halign-left valign-top"><p class="tableblock">Declared and Static Use</p></td>
14394</tr>
14395<tr>
14396<td class="tableblock halign-left valign-top" rowspan="3"><p class="tableblock">Generating Shader (output variables)</p></td>
14397<td class="tableblock halign-left valign-top"><p class="tableblock">No Declaration</p></td>
14398<td class="tableblock halign-left valign-top"><p class="tableblock">Allowed</p></td>
14399<td class="tableblock halign-left valign-top"><p class="tableblock">Allowed</p></td>
14400<td class="tableblock halign-left valign-top"><p class="tableblock">error</p></td>
14401</tr>
14402<tr>
14403<td class="tableblock halign-left valign-top"><p class="tableblock">Declares; no static Use</p></td>
14404<td class="tableblock halign-left valign-top"><p class="tableblock">Allowed</p></td>
14405<td class="tableblock halign-left valign-top"><p class="tableblock">Allowed</p></td>
14406<td class="tableblock halign-left valign-top"><p class="tableblock">Allowed (values are undefined)</p></td>
14407</tr>
14408<tr>
14409<td class="tableblock halign-left valign-top"><p class="tableblock">Declares and static Use</p></td>
14410<td class="tableblock halign-left valign-top"><p class="tableblock">Allowed</p></td>
14411<td class="tableblock halign-left valign-top"><p class="tableblock">Allowed</p></td>
14412<td class="tableblock halign-left valign-top"><p class="tableblock">Allowed (values are potentially undefined)</p></td>
14413</tr>
14414</tbody>
14415</table>
14416<div class="paragraph">
14417<p>See &#8220;<a href="#static-use">Static Use</a>&#8221; for the definition of <em>static use</em>.</p>
14418</div>
14419<div class="paragraph">
14420<p>The precision of a vertex output does not need to match the precision of the
14421corresponding fragment input.
14422The minimum precision at which vertex outputs are interpolated is the
14423minimum of the vertex output precision and the fragment input precision,
14424with the exception that for <strong>highp</strong>, implementations do not have to support
14425full IEEE 754 precision.</p>
14426</div>
14427<div class="paragraph">
14428<p>The precision of values exported to a transform feedback buffer is the
14429precision of the outputs of the vertex shader.
14430However, they are converted to <strong>highp</strong> format before being written.</p>
14431</div>
14432</div>
14433<div class="sect2">
14434<h3 id="matching-of-qualifiers">9.2. Matching of Qualifiers</h3>
14435<div class="paragraph">
14436<p>The following tables summarize the requirements for matching of qualifiers.
14437It applies whenever there are two or more matching variables in a shader
14438interface.</p>
14439</div>
14440<div class="paragraph">
14441<p>Notes:</p>
14442</div>
14443<div class="olist arabic">
14444<ol class="arabic">
14445<li>
14446<p><em>Yes</em> means the qualifiers must match.</p>
14447</li>
14448<li>
14449<p><em>No</em> means the qualifiers do not need to match.</p>
14450</li>
14451<li>
14452<p><em>Consistent</em> means qualifiers may be missing from a subset of
14453declarations but they cannot conflict</p>
14454</li>
14455<li>
14456<p>If there are conflicting qualifiers, only the last of these is
14457significant.</p>
14458</li>
14459<li>
14460<p>Matching is based only on the resulting qualification, not on the
14461presence or otherwise of qualifiers.</p>
14462</li>
14463<li>
14464<p>The rules apply to all declared variables, irrespective of whether they
14465are statically used, with the exception of inputs and outputs when
14466shaders are linked (see
14467&#8220;<a href="#input-output-matching-by-name-in-linked-programs">Input Output
14468Matching by Name in Linked Programs</a>&#8221;).</p>
14469</li>
14470<li>
14471<p>Errors are generated for any conflicts.</p>
14472</li>
14473</ol>
14474</div>
14475<div class="sect3">
14476<h4 id="linked-shaders">9.2.1. Linked Shaders</h4>
14477<table class="tableblock frame-all grid-all stretch">
14478<colgroup>
14479<col style="width: 16.6666%;">
14480<col style="width: 16.6666%;">
14481<col style="width: 16.6666%;">
14482<col style="width: 16.6666%;">
14483<col style="width: 16.6666%;">
14484<col style="width: 16.667%;">
14485</colgroup>
14486<thead>
14487<tr>
14488<th class="tableblock halign-left valign-top">Qualifier Class</th>
14489<th class="tableblock halign-left valign-top">Qualifier</th>
14490<th class="tableblock halign-left valign-top"><strong>in</strong>/<strong>out</strong></th>
14491<th class="tableblock halign-left valign-top">Default Uniforms</th>
14492<th class="tableblock halign-left valign-top"><strong>uniform</strong> Block</th>
14493<th class="tableblock halign-left valign-top"><strong>buffer</strong> Block</th>
14494</tr>
14495</thead>
14496<tbody>
14497<tr>
14498<td class="tableblock halign-left valign-top"><p class="tableblock">Storage<sup>1</sup></p></td>
14499<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>in</strong><br>
14500                    <strong>out</strong><br>
14501                    <strong>uniform</strong></p></td>
14502<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14503<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14504<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14505<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14506</tr>
14507<tr>
14508<td class="tableblock halign-left valign-top" rowspan="2"><p class="tableblock">Auxiliary</p></td>
14509<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>centroid</strong><br>
14510                    <strong>sample</strong></p></td>
14511<td class="tableblock halign-left valign-top"><p class="tableblock">No</p></td>
14512<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14513<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14514<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14515</tr>
14516<tr>
14517<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>patch</strong></p></td>
14518<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14519<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14520<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14521<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14522</tr>
14523<tr>
14524<td class="tableblock halign-left valign-top" rowspan="5"><p class="tableblock">Layout</p></td>
14525<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>location</strong></p></td>
14526<td class="tableblock halign-left valign-top"><p class="tableblock">Yes<sup>2</sup></p></td>
14527<td class="tableblock halign-left valign-top"><p class="tableblock">Consistent</p></td>
14528<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14529<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14530</tr>
14531<tr>
14532<td class="tableblock halign-left valign-top"><p class="tableblock">Block layout<sup>3,4</sup></p></td>
14533<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14534<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14535<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14536<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14537</tr>
14538<tr>
14539<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>binding</strong></p></td>
14540<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14541<td class="tableblock halign-left valign-top"><p class="tableblock">Consistent</p></td>
14542<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14543<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14544</tr>
14545<tr>
14546<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>offset</strong></p></td>
14547<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14548<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14549<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14550<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14551</tr>
14552<tr>
14553<td class="tableblock halign-left valign-top"><p class="tableblock">format</p></td>
14554<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14555<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14556<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14557<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14558</tr>
14559<tr>
14560<td class="tableblock halign-left valign-top"><p class="tableblock">Interpolation</p></td>
14561<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>smooth</strong><br>
14562                    <strong>flat</strong></p></td>
14563<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14564<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14565<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14566<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14567</tr>
14568<tr>
14569<td class="tableblock halign-left valign-top"><p class="tableblock">Precision</p></td>
14570<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>lowp</strong><br>
14571                    <strong>mediump</strong><br>
14572                    <strong>highp</strong></p></td>
14573<td class="tableblock halign-left valign-top"><p class="tableblock">No</p></td>
14574<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14575<td class="tableblock halign-left valign-top"><p class="tableblock">No</p></td>
14576<td class="tableblock halign-left valign-top"><p class="tableblock">No</p></td>
14577</tr>
14578<tr>
14579<td class="tableblock halign-left valign-top"><p class="tableblock">Variance</p></td>
14580<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>invariant</strong><br>
14581                    <strong>precise</strong></p></td>
14582<td class="tableblock halign-left valign-top"><p class="tableblock">No</p></td>
14583<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14584<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14585<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14586</tr>
14587<tr>
14588<td class="tableblock halign-left valign-top"><p class="tableblock">Memory</p></td>
14589<td class="tableblock halign-left valign-top"><p class="tableblock">all</p></td>
14590<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14591<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14592<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14593<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14594</tr>
14595</tbody>
14596</table>
14597<div class="dlist">
14598<dl>
14599<dt class="hdlist1">1</dt>
14600<dd>
14601<p>Storage qualifiers determine <em>when</em> variables match rather than being
14602<em>required</em> to match for matching variables.
14603Note also that each shader interface has a separate name space so for
14604example, it is possible to use the same name for a vertex output and
14605fragment uniform.</p>
14606</dd>
14607<dt class="hdlist1">2</dt>
14608<dd>
14609<p>If present, the <strong>location</strong> qualifier determines the matching of inputs
14610and outputs.
14611See section 7.4.1 &#8220;Shader interface Matching&#8221; of the
14612<a href="#references">OpenGL ES Specification</a> for details.]</p>
14613</dd>
14614<dt class="hdlist1">3</dt>
14615<dd>
14616<p>The <strong>row_major</strong> and <strong>column_major</strong> layout qualifiers do not need to
14617match when applied to non-matrix types.</p>
14618</dd>
14619<dt class="hdlist1">4</dt>
14620<dd>
14621<p>In cases where a layout qualifier overrides a previous layout qualifier
14622or a default, only the resulting qualification must match.</p>
14623</dd>
14624</dl>
14625</div>
14626</div>
14627<div class="sect3">
14628<h4 id="separable-programs">9.2.2. Separable Programs</h4>
14629<table class="tableblock frame-all grid-all stretch">
14630<colgroup>
14631<col style="width: 33.3333%;">
14632<col style="width: 33.3333%;">
14633<col style="width: 33.3334%;">
14634</colgroup>
14635<thead>
14636<tr>
14637<th class="tableblock halign-left valign-top">Qualifier Class</th>
14638<th class="tableblock halign-left valign-top">Qualifier</th>
14639<th class="tableblock halign-left valign-top"><strong>in</strong>/<strong>out</strong></th>
14640</tr>
14641</thead>
14642<tbody>
14643<tr>
14644<td class="tableblock halign-left valign-top"><p class="tableblock">Storage</p></td>
14645<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>in</strong><br>
14646                    <strong>out</strong><br>
14647                    <strong>uniform</strong></p></td>
14648<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14649</tr>
14650<tr>
14651<td class="tableblock halign-left valign-top" rowspan="2"><p class="tableblock">Auxiliary</p></td>
14652<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>centroid</strong><br>
14653                    <strong>sample</strong></p></td>
14654<td class="tableblock halign-left valign-top"><p class="tableblock">No</p></td>
14655</tr>
14656<tr>
14657<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>patch</strong></p></td>
14658<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14659</tr>
14660<tr>
14661<td class="tableblock halign-left valign-top" rowspan="5"><p class="tableblock">Layout</p></td>
14662<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>location</strong></p></td>
14663<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14664</tr>
14665<tr>
14666<td class="tableblock halign-left valign-top"><p class="tableblock">Block layout</p></td>
14667<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14668</tr>
14669<tr>
14670<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>binding</strong></p></td>
14671<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14672</tr>
14673<tr>
14674<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>offset</strong></p></td>
14675<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14676</tr>
14677<tr>
14678<td class="tableblock halign-left valign-top"><p class="tableblock">format</p></td>
14679<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14680</tr>
14681<tr>
14682<td class="tableblock halign-left valign-top"><p class="tableblock">Interpolation</p></td>
14683<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>smooth</strong><br>
14684                    <strong>flat</strong></p></td>
14685<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14686</tr>
14687<tr>
14688<td class="tableblock halign-left valign-top"><p class="tableblock">Precision</p></td>
14689<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>lowp</strong><br>
14690                    <strong>mediump</strong><br>
14691                    <strong>highp</strong></p></td>
14692<td class="tableblock halign-left valign-top"><p class="tableblock">Yes</p></td>
14693</tr>
14694<tr>
14695<td class="tableblock halign-left valign-top"><p class="tableblock">Variance</p></td>
14696<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>invariant</strong><br>
14697                    <strong>precise</strong></p></td>
14698<td class="tableblock halign-left valign-top"><p class="tableblock">No</p></td>
14699</tr>
14700<tr>
14701<td class="tableblock halign-left valign-top"><p class="tableblock">Memory</p></td>
14702<td class="tableblock halign-left valign-top"><p class="tableblock">all</p></td>
14703<td class="tableblock halign-left valign-top"><p class="tableblock">N/A</p></td>
14704</tr>
14705</tbody>
14706</table>
14707</div>
14708</div>
14709</div>
14710</div>
14711<div class="sect1">
14712<h2 id="shading-language-grammar">10. Shading Language Grammar</h2>
14713<div class="sectionbody">
14714<div class="paragraph">
14715<p>The grammar is fed from the output of lexical analysis.
14716The tokens returned from lexical analysis are</p>
14717</div>
14718<div class="openblock bnf">
14719<div class="content">
14720<div class="paragraph">
14721<p>CONST BOOL FLOAT INT UINT</p>
14722</div>
14723<div class="paragraph">
14724<p>BVEC2 BVEC3 BVEC4 IVEC2 IVEC3 IVEC4 UVEC2 UVEC3 UVEC4 VEC2 VEC3 VEC4</p>
14725</div>
14726<div class="paragraph">
14727<p>MAT2 MAT3 MAT4
14728MAT2X2 MAT2X3 MAT2X4
14729MAT3X2 MAT3X3 MAT3X4
14730MAT4X2 MAT4X3 MAT4X4</p>
14731</div>
14732<div class="paragraph">
14733<p>CENTROID IN OUT INOUT UNIFORM PATCH SAMPLE BUFFER SHARED
14734COHERENT VOLATILE RESTRICT READONLY WRITEONLY
14735FLAT SMOOTH LAYOUT</p>
14736</div>
14737<div class="paragraph">
14738<p>ATOMIC_UINT</p>
14739</div>
14740<div class="paragraph">
14741<p>SAMPLER2D SAMPLER3D SAMPLERCUBE SAMPLER2DSHADOW
14742SAMPLERCUBESHADOW SAMPLER2DARRAY SAMPLER2DARRAYSHADOW
14743ISAMPLER2D ISAMPLER3D ISAMPLERCUBE ISAMPLER2DARRAY
14744USAMPLER2D USAMPLER3D USAMPLERCUBE USAMPLER2DARRAY</p>
14745</div>
14746<div class="paragraph">
14747<p>SAMPLERBUFFER ISAMPLERBUFFER USAMPLERBUFFER
14748SAMPLERCUBEARRAY SAMPLERCUBEARRAYSHADOW
14749ISAMPLERCUBEARRAY USAMPLERCUBEARRAY
14750SAMPLER2DMS ISAMPLER2DMS USAMPLER2DMS
14751SAMPLER2DMSARRAY ISAMPLER2DMSARRAY USAMPLER2DMSARRAY
14752IMAGE2D IIMAGE2D UIMAGE2D
14753IMAGE3D IIMAGE3D UIMAGE3D
14754IMAGECUBE IIMAGECUBE UIMAGECUBE
14755IMAGEBUFFER IIMAGEBUFFER UIMAGEBUFFER
14756IMAGE2DARRAY IIMAGE2DARRAY UIMAGE2DARRAY
14757IMAGECUBEARRAY IIMAGECUBEARRAY UIMAGECUBEARRAY</p>
14758</div>
14759<div class="paragraph">
14760<p>STRUCT VOID</p>
14761</div>
14762<div class="paragraph">
14763<p>WHILE BREAK CONTINUE DO ELSE FOR IF DISCARD RETURN SWITCH CASE DEFAULT</p>
14764</div>
14765<div class="paragraph">
14766<p>IDENTIFIER TYPE_NAME
14767FLOATCONSTANT INTCONSTANT UINTCONSTANT BOOLCONSTANT
14768FIELD_SELECTION</p>
14769</div>
14770<div class="paragraph">
14771<p>LEFT_OP RIGHT_OP
14772INC_OP DEC_OP LE_OP GE_OP EQ_OP NE_OP
14773AND_OP OR_OP XOR_OP MUL_ASSIGN DIV_ASSIGN ADD_ASSIGN
14774MOD_ASSIGN LEFT_ASSIGN RIGHT_ASSIGN AND_ASSIGN XOR_ASSIGN OR_ASSIGN
14775SUB_ASSIGN
14776LEFT_PAREN RIGHT_PAREN LEFT_BRACKET RIGHT_BRACKET LEFT_BRACE RIGHT_BRACE DOT
14777COMMA COLON EQUAL SEMICOLON BANG DASH TILDE PLUS STAR SLASH PERCENT
14778LEFT_ANGLE RIGHT_ANGLE VERTICAL_BAR CARET AMPERSAND QUESTION</p>
14779</div>
14780<div class="paragraph">
14781<p>INVARIANT PRECISE
14782HIGH_PRECISION MEDIUM_PRECISION LOW_PRECISION PRECISION</p>
14783</div>
14784</div>
14785</div>
14786<div class="paragraph">
14787<p>The following describes the grammar for the OpenGL ES Shading Language in terms of the above
14788tokens.
14789The starting rule is <em>translation_unit</em>.</p>
14790</div>
14791<div class="openblock bnf">
14792<div class="content">
14793<div class="dlist">
14794<dl>
14795<dt class="hdlist1"><em>variable_identifier</em> : </dt>
14796<dd>
14797<p><em>IDENTIFIER</em></p>
14798</dd>
14799<dt class="hdlist1"><em>primary_expression</em> : </dt>
14800<dd>
14801<p><em>variable_identifier</em><br>
14802<em>INTCONSTANT</em><br>
14803<em>UINTCONSTANT</em><br>
14804<em>FLOATCONSTANT</em><br>
14805<em>BOOLCONSTANT</em><br>
14806<em>LEFT_PAREN</em> <em>expression</em> <em>RIGHT_PAREN</em></p>
14807</dd>
14808<dt class="hdlist1"><em>postfix_expression</em> : </dt>
14809<dd>
14810<p><em>primary_expression</em><br>
14811<em>postfix_expression</em> <em>LEFT_BRACKET</em> <em>integer_expression</em> <em>RIGHT_BRACKET</em><br>
14812<em>function_call</em><br>
14813<em>postfix_expression</em> <em>DOT</em> <em>FIELD_SELECTION</em><br>
14814<em>postfix_expression</em> <em>INC_OP</em><br>
14815<em>postfix_expression</em> <em>DEC_OP</em></p>
14816</dd>
14817</dl>
14818</div>
14819<div class="admonitionblock note">
14820<table>
14821<tr>
14822<td class="icon">
14823<i class="fa icon-note" title="Note"></i>
14824</td>
14825<td class="content">
14826<div class="paragraph">
14827<p>FIELD_SELECTION includes members in structures, component selection for
14828vectors and the 'length' identifier for the length() method</p>
14829</div>
14830</td>
14831</tr>
14832</table>
14833</div>
14834<div class="dlist">
14835<dl>
14836<dt class="hdlist1"><em>integer_expression</em> : </dt>
14837<dd>
14838<p><em>expression</em></p>
14839</dd>
14840<dt class="hdlist1"><em>function_call</em> : </dt>
14841<dd>
14842<p><em>function_call_or_method</em></p>
14843</dd>
14844<dt class="hdlist1"><em>function_call_or_method</em> : </dt>
14845<dd>
14846<p><em>function_call_generic</em></p>
14847</dd>
14848<dt class="hdlist1"><em>function_call_generic</em> : </dt>
14849<dd>
14850<p><em>function_call_header_with_parameters</em> <em>RIGHT_PAREN</em><br>
14851<em>function_call_header_no_parameters</em> <em>RIGHT_PAREN</em></p>
14852</dd>
14853<dt class="hdlist1"><em>function_call_header_no_parameters</em> : </dt>
14854<dd>
14855<p><em>function_call_header</em> <em>VOID</em><br>
14856<em>function_call_header</em></p>
14857</dd>
14858<dt class="hdlist1"><em>function_call_header_with_parameters</em> : </dt>
14859<dd>
14860<p><em>function_call_header</em> <em>assignment_expression</em><br>
14861<em>function_call_header_with_parameters</em> <em>COMMA</em> <em>assignment_expression</em></p>
14862</dd>
14863<dt class="hdlist1"><em>function_call_header</em> : </dt>
14864<dd>
14865<p><em>function_identifier</em> <em>LEFT_PAREN</em></p>
14866</dd>
14867</dl>
14868</div>
14869<div class="admonitionblock note">
14870<table>
14871<tr>
14872<td class="icon">
14873<i class="fa icon-note" title="Note"></i>
14874</td>
14875<td class="content">
14876<div class="paragraph">
14877<p>Grammar Note: Constructors look like functions, but lexical analysis
14878recognized most of them as keywords.
14879They are now recognized through <em>type_specifier</em>.</p>
14880</div>
14881</td>
14882</tr>
14883</table>
14884</div>
14885<div class="admonitionblock note">
14886<table>
14887<tr>
14888<td class="icon">
14889<i class="fa icon-note" title="Note"></i>
14890</td>
14891<td class="content">
14892<div class="paragraph">
14893<p>Methods (<strong>.length</strong>) and identifiers are recognized through
14894<em>postfix_expression</em>.</p>
14895</div>
14896</td>
14897</tr>
14898</table>
14899</div>
14900<div class="dlist">
14901<dl>
14902<dt class="hdlist1"><em>function_identifier</em> : </dt>
14903<dd>
14904<p><em>type_specifier</em><br>
14905<em>postfix_expression</em></p>
14906</dd>
14907<dt class="hdlist1"><em>unary_expression</em> : </dt>
14908<dd>
14909<p><em>postfix_expression</em><br>
14910<em>INC_OP</em> <em>unary_expression</em><br>
14911<em>DEC_OP</em> <em>unary_expression</em><br>
14912<em>unary_operator</em> <em>unary_expression</em></p>
14913</dd>
14914</dl>
14915</div>
14916<div class="admonitionblock note">
14917<table>
14918<tr>
14919<td class="icon">
14920<i class="fa icon-note" title="Note"></i>
14921</td>
14922<td class="content">
14923<div class="paragraph">
14924<p>Grammar Note: No traditional style type casts.</p>
14925</div>
14926</td>
14927</tr>
14928</table>
14929</div>
14930<div class="dlist">
14931<dl>
14932<dt class="hdlist1"><em>unary_operator</em> : </dt>
14933<dd>
14934<p><em>PLUS</em><br>
14935<em>DASH</em><br>
14936<em>BANG</em><br>
14937<em>TILDE</em></p>
14938</dd>
14939</dl>
14940</div>
14941<div class="admonitionblock note">
14942<table>
14943<tr>
14944<td class="icon">
14945<i class="fa icon-note" title="Note"></i>
14946</td>
14947<td class="content">
14948<div class="paragraph">
14949<p>Grammar Note: No '*' or '&amp;' unary ops.
14950Pointers are not supported.</p>
14951</div>
14952</td>
14953</tr>
14954</table>
14955</div>
14956<div class="dlist">
14957<dl>
14958<dt class="hdlist1"><em>multiplicative_expression</em> : </dt>
14959<dd>
14960<p><em>unary_expression</em><br>
14961<em>multiplicative_expression</em> <em>STAR</em> <em>unary_expression</em><br>
14962<em>multiplicative_expression</em> <em>SLASH</em> <em>unary_expression</em><br>
14963<em>multiplicative_expression</em> <em>PERCENT</em> <em>unary_expression</em></p>
14964</dd>
14965<dt class="hdlist1"><em>additive_expression</em> : </dt>
14966<dd>
14967<p><em>multiplicative_expression</em><br>
14968<em>additive_expression</em> <em>PLUS</em> <em>multiplicative_expression</em><br>
14969<em>additive_expression</em> <em>DASH</em> <em>multiplicative_expression</em></p>
14970</dd>
14971<dt class="hdlist1"><em>shift_expression</em> : </dt>
14972<dd>
14973<p><em>additive_expression</em><br>
14974<em>shift_expression</em> <em>LEFT_OP</em> <em>additive_expression</em><br>
14975<em>shift_expression</em> <em>RIGHT_OP</em> <em>additive_expression</em></p>
14976</dd>
14977<dt class="hdlist1"><em>relational_expression</em> : </dt>
14978<dd>
14979<p><em>shift_expression</em><br>
14980<em>relational_expression</em> <em>LEFT_ANGLE</em> <em>shift_expression</em><br>
14981<em>relational_expression</em> <em>RIGHT_ANGLE</em> <em>shift_expression</em><br>
14982<em>relational_expression</em> <em>LE_OP</em> <em>shift_expression</em><br>
14983<em>relational_expression</em> <em>GE_OP</em> <em>shift_expression</em></p>
14984</dd>
14985<dt class="hdlist1"><em>equality_expression</em> : </dt>
14986<dd>
14987<p><em>relational_expression</em><br>
14988<em>equality_expression</em> <em>EQ_OP</em> <em>relational_expression</em><br>
14989<em>equality_expression</em> <em>NE_OP</em> <em>relational_expression</em></p>
14990</dd>
14991<dt class="hdlist1"><em>and_expression</em> : </dt>
14992<dd>
14993<p><em>equality_expression</em><br>
14994<em>and_expression</em> <em>AMPERSAND</em> <em>equality_expression</em></p>
14995</dd>
14996<dt class="hdlist1"><em>exclusive_or_expression</em> : </dt>
14997<dd>
14998<p><em>and_expression</em><br>
14999<em>exclusive_or_expression</em> <em>CARET</em> <em>and_expression</em></p>
15000</dd>
15001<dt class="hdlist1"><em>inclusive_or_expression</em> : </dt>
15002<dd>
15003<p><em>exclusive_or_expression</em><br>
15004<em>inclusive_or_expression</em> <em>VERTICAL_BAR</em> <em>exclusive_or_expression</em></p>
15005</dd>
15006<dt class="hdlist1"><em>logical_and_expression</em> : </dt>
15007<dd>
15008<p><em>inclusive_or_expression</em><br>
15009<em>logical_and_expression</em> <em>AND_OP</em> <em>inclusive_or_expression</em></p>
15010</dd>
15011<dt class="hdlist1"><em>logical_xor_expression</em> : </dt>
15012<dd>
15013<p><em>logical_and_expression</em><br>
15014<em>logical_xor_expression</em> <em>XOR_OP</em> <em>logical_and_expression</em></p>
15015</dd>
15016<dt class="hdlist1"><em>logical_or_expression</em> : </dt>
15017<dd>
15018<p><em>logical_xor_expression</em><br>
15019<em>logical_or_expression</em> <em>OR_OP</em> <em>logical_xor_expression</em></p>
15020</dd>
15021<dt class="hdlist1"><em>conditional_expression</em> : </dt>
15022<dd>
15023<p><em>logical_or_expression</em><br>
15024<em>logical_or_expression</em> <em>QUESTION</em> <em>expression</em> <em>COLON</em>
15025<em>assignment_expression</em></p>
15026</dd>
15027<dt class="hdlist1"><em>assignment_expression</em> : </dt>
15028<dd>
15029<p><em>conditional_expression</em><br>
15030<em>unary_expression</em> <em>assignment_operator</em> <em>assignment_expression</em></p>
15031</dd>
15032<dt class="hdlist1"><em>assignment_operator</em> : </dt>
15033<dd>
15034<p><em>EQUAL</em><br>
15035<em>MUL_ASSIGN</em><br>
15036<em>DIV_ASSIGN</em><br>
15037<em>MOD_ASSIGN</em><br>
15038<em>ADD_ASSIGN</em><br>
15039<em>SUB_ASSIGN</em><br>
15040<em>LEFT_ASSIGN</em><br>
15041<em>RIGHT_ASSIGN</em><br>
15042<em>AND_ASSIGN</em><br>
15043<em>XOR_ASSIGN</em><br>
15044<em>OR_ASSIGN</em></p>
15045</dd>
15046<dt class="hdlist1"><em>expression</em> : </dt>
15047<dd>
15048<p><em>assignment_expression</em><br>
15049<em>expression</em> <em>COMMA</em> <em>assignment_expression</em></p>
15050</dd>
15051<dt class="hdlist1"><em>constant_expression</em> : </dt>
15052<dd>
15053<p><em>conditional_expression</em></p>
15054</dd>
15055<dt class="hdlist1"><em>declaration</em> : </dt>
15056<dd>
15057<p><em>function_prototype</em> <em>SEMICOLON</em><br>
15058<em>init_declarator_list</em> <em>SEMICOLON</em><br>
15059<em>PRECISION</em> <em>precision_qualifier</em> <em>type_specifier</em> <em>SEMICOLON</em><br>
15060<em>type_qualifier</em> <em>IDENTIFIER</em> <em>LEFT_BRACE</em> <em>struct_declaration_list</em>
15061<em>RIGHT_BRACE</em> <em>SEMICOLON</em><br>
15062<em>type_qualifier</em> <em>IDENTIFIER</em> <em>LEFT_BRACE</em> <em>struct_declaration_list</em>
15063<em>RIGHT_BRACE</em> <em>IDENTIFIER</em> <em>SEMICOLON</em><br>
15064<em>type_qualifier</em> <em>IDENTIFIER</em> <em>LEFT_BRACE</em> <em>struct_declaration_list</em>
15065<em>RIGHT_BRACE</em> <em>IDENTIFIER</em> <em>array_specifier</em> <em>SEMICOLON</em><br>
15066<em>type_qualifier</em> <em>SEMICOLON</em><br>
15067<em>type_qualifier</em> <em>IDENTIFIER</em> <em>SEMICOLON</em><br>
15068<em>type_qualifier</em> <em>IDENTIFIER</em> <em>identifier_list</em> <em>SEMICOLON</em></p>
15069</dd>
15070<dt class="hdlist1"><em>identifier_list</em> : </dt>
15071<dd>
15072<p><em>COMMA</em> <em>IDENTIFIER</em><br>
15073<em>identifier_list</em> <em>COMMA</em> <em>IDENTIFIER</em></p>
15074</dd>
15075<dt class="hdlist1"><em>function_prototype</em> : </dt>
15076<dd>
15077<p><em>function_declarator</em> <em>RIGHT_PAREN</em></p>
15078</dd>
15079<dt class="hdlist1"><em>function_declarator</em> : </dt>
15080<dd>
15081<p><em>function_header</em><br>
15082<em>function_header_with_parameters</em></p>
15083</dd>
15084<dt class="hdlist1"><em>function_header_with_parameters</em> : </dt>
15085<dd>
15086<p><em>function_header</em> <em>parameter_declaration</em><br>
15087<em>function_header_with_parameters</em> <em>COMMA</em> <em>parameter_declaration</em></p>
15088</dd>
15089<dt class="hdlist1"><em>function_header</em> : </dt>
15090<dd>
15091<p><em>fully_specified_type</em> <em>IDENTIFIER</em> <em>LEFT_PAREN</em></p>
15092</dd>
15093<dt class="hdlist1"><em>parameter_declarator</em> : </dt>
15094<dd>
15095<p><em>type_specifier</em> <em>IDENTIFIER</em><br>
15096<em>type_specifier</em> <em>IDENTIFIER</em> <em>array_specifier</em></p>
15097</dd>
15098<dt class="hdlist1"><em>parameter_declaration</em> : </dt>
15099<dd>
15100<p><em>type_qualifier</em> <em>parameter_declarator</em><br>
15101<em>parameter_declarator</em><br>
15102<em>type_qualifier</em> <em>parameter_type_specifier</em><br>
15103<em>parameter_type_specifier</em></p>
15104</dd>
15105<dt class="hdlist1"><em>parameter_type_specifier</em> : </dt>
15106<dd>
15107<p><em>type_specifier</em></p>
15108</dd>
15109<dt class="hdlist1"><em>init_declarator_list</em> : </dt>
15110<dd>
15111<p><em>single_declaration</em><br>
15112<em>init_declarator_list</em> <em>COMMA</em> <em>IDENTIFIER</em><br>
15113<em>init_declarator_list</em> <em>COMMA</em> <em>IDENTIFIER</em> <em>array_specifier</em><br>
15114<em>init_declarator_list</em> <em>COMMA</em> <em>IDENTIFIER</em> <em>array_specifier</em> <em>EQUAL</em>
15115<em>initializer</em><br>
15116<em>init_declarator_list</em> <em>COMMA</em> <em>IDENTIFIER</em> <em>EQUAL</em> <em>initializer</em></p>
15117</dd>
15118<dt class="hdlist1"><em>single_declaration</em> : </dt>
15119<dd>
15120<p><em>fully_specified_type</em><br>
15121<em>fully_specified_type</em> <em>IDENTIFIER</em><br>
15122<em>fully_specified_type</em> <em>IDENTIFIER</em> <em>array_specifier</em><br>
15123<em>fully_specified_type</em> <em>IDENTIFIER</em> <em>array_specifier</em> <em>EQUAL</em>
15124<em>initializer</em><br>
15125<em>fully_specified_type</em> <em>IDENTIFIER</em> <em>EQUAL</em> <em>initializer</em></p>
15126</dd>
15127</dl>
15128</div>
15129<div class="admonitionblock note">
15130<table>
15131<tr>
15132<td class="icon">
15133<i class="fa icon-note" title="Note"></i>
15134</td>
15135<td class="content">
15136<div class="paragraph">
15137<p>Grammar Note: No 'enum', or 'typedef'.</p>
15138</div>
15139</td>
15140</tr>
15141</table>
15142</div>
15143<div class="dlist">
15144<dl>
15145<dt class="hdlist1"><em>fully_specified_type</em> : </dt>
15146<dd>
15147<p><em>type_specifier</em><br>
15148<em>type_qualifier</em> <em>type_specifier</em></p>
15149</dd>
15150<dt class="hdlist1"><em>invariant_qualifier</em> : </dt>
15151<dd>
15152<p><em>INVARIANT</em></p>
15153</dd>
15154<dt class="hdlist1"><em>interpolation_qualifier</em> : </dt>
15155<dd>
15156<p><em>SMOOTH</em><br>
15157<em>FLAT</em></p>
15158</dd>
15159<dt class="hdlist1"><em>layout_qualifier</em> : </dt>
15160<dd>
15161<p><em>LAYOUT</em> <em>LEFT_PAREN</em> <em>layout_qualifier_id_list</em> <em>RIGHT_PAREN</em></p>
15162</dd>
15163<dt class="hdlist1"><em>layout_qualifier_id_list</em> : </dt>
15164<dd>
15165<p><em>layout_qualifier_id</em><br>
15166<em>layout_qualifier_id_list</em> <em>COMMA</em> <em>layout_qualifier_id</em></p>
15167</dd>
15168<dt class="hdlist1"><em>layout_qualifier_id</em> : </dt>
15169<dd>
15170<p><em>IDENTIFIER</em><br>
15171<em>IDENTIFIER</em> <em>EQUAL</em> <em>INTCONSTANT</em><br>
15172<em>IDENTIFIER</em> <em>EQUAL</em> <em>UINTCONSTANT</em><br>
15173<em>SHARED</em></p>
15174</dd>
15175<dt class="hdlist1"><em>precise_qualifier</em> : </dt>
15176<dd>
15177<p><em>PRECISE</em></p>
15178</dd>
15179<dt class="hdlist1"><em>type_qualifier</em> : </dt>
15180<dd>
15181<p><em>single_type_qualifier</em><br>
15182<em>type_qualifier</em> <em>single_type_qualifier</em></p>
15183</dd>
15184<dt class="hdlist1"><em>single_type_qualifier</em> : </dt>
15185<dd>
15186<p><em>storage_qualifier</em><br>
15187<em>layout_qualifier</em><br>
15188<em>precision_qualifier</em><br>
15189<em>interpolation_qualifier</em><br>
15190<em>invariant_qualifier</em><br>
15191<em>precise_qualifier</em></p>
15192</dd>
15193<dt class="hdlist1"><em>storage_qualifier</em> : </dt>
15194<dd>
15195<p><em>CONST</em><br>
15196<em>IN</em><br>
15197<em>OUT</em><br>
15198<em>INOUT</em><br>
15199<em>CENTROID</em><br>
15200<em>PATCH</em><br>
15201<em>SAMPLE</em><br>
15202<em>UNIFORM</em><br>
15203<em>BUFFER</em><br>
15204<em>SHARED</em><br>
15205<em>COHERENT</em><br>
15206<em>VOLATILE</em><br>
15207<em>RESTRICT</em><br>
15208<em>READONLY</em><br>
15209<em>WRITEONLY</em></p>
15210</dd>
15211<dt class="hdlist1"><em>type_specifier</em> : </dt>
15212<dd>
15213<p><em>type_specifier_nonarray</em><br>
15214<em>type_specifier_nonarray</em> <em>array_specifier</em></p>
15215</dd>
15216<dt class="hdlist1"><em>array_specifier</em> : </dt>
15217<dd>
15218<p><em>LEFT_BRACKET</em> <em>RIGHT_BRACKET</em><br>
15219<em>LEFT_BRACKET</em> <em>conditional_expression</em> <em>RIGHT_BRACKET</em><br>
15220<em>array_specifier</em> <em>LEFT_BRACKET</em> <em>RIGHT_BRACKET</em><br>
15221<em>array_specifier</em> <em>LEFT_BRACKET</em> <em>conditional_expression</em> <em>RIGHT_BRACKET</em></p>
15222</dd>
15223<dt class="hdlist1"><em>type_specifier_nonarray</em> : </dt>
15224<dd>
15225<p><em>VOID</em><br>
15226<em>FLOAT</em><br>
15227<em>INT</em><br>
15228<em>UINT</em><br>
15229<em>BOOL</em><br>
15230<em>VEC2</em><br>
15231<em>VEC3</em><br>
15232<em>VEC4</em><br>
15233<em>BVEC2</em><br>
15234<em>BVEC3</em><br>
15235<em>BVEC4</em><br>
15236<em>IVEC2</em><br>
15237<em>IVEC3</em><br>
15238<em>IVEC4</em><br>
15239<em>UVEC2</em><br>
15240<em>UVEC3</em><br>
15241<em>UVEC4</em><br>
15242<em>MAT2</em><br>
15243<em>MAT3</em><br>
15244<em>MAT4</em><br>
15245<em>MAT2X2</em><br>
15246<em>MAT2X3</em><br>
15247<em>MAT2X4</em><br>
15248<em>MAT3X2</em><br>
15249<em>MAT3X3</em><br>
15250<em>MAT3X4</em><br>
15251<em>MAT4X2</em><br>
15252<em>MAT4X3</em><br>
15253<em>MAT4X4</em><br>
15254<em>ATOMIC_UINT</em><br>
15255<em>SAMPLER2D</em><br>
15256<em>SAMPLER3D</em><br>
15257<em>SAMPLERCUBE</em><br>
15258<em>SAMPLER2DSHADOW</em><br>
15259<em>SAMPLERCUBESHADOW</em><br>
15260<em>SAMPLER2DARRAY</em><br>
15261<em>SAMPLER2DARRAYSHADOW</em><br>
15262<em>SAMPLERCUBEARRAY</em><br>
15263<em>SAMPLERCUBEARRAYSHADOW</em><br>
15264<em>ISAMPLER2D</em><br>
15265<em>ISAMPLER3D</em><br>
15266<em>ISAMPLERCUBE</em><br>
15267<em>ISAMPLER2DARRAY</em><br>
15268<em>ISAMPLERCUBEARRAY</em><br>
15269<em>USAMPLER2D</em><br>
15270<em>USAMPLER3D</em><br>
15271<em>USAMPLERCUBE</em><br>
15272<em>USAMPLER2DARRAY</em><br>
15273<em>USAMPLERCUBEARRAY</em><br>
15274<em>SAMPLERBUFFER</em><br>
15275<em>ISAMPLERBUFFER</em><br>
15276<em>USAMPLERBUFFER</em><br>
15277<em>SAMPLER2DMS</em><br>
15278<em>ISAMPLER2DMS</em><br>
15279<em>USAMPLER2DMS</em><br>
15280<em>SAMPLER2DMSARRAY</em><br>
15281<em>ISAMPLER2DMSARRAY</em><br>
15282<em>USAMPLER2DMSARRAY</em><br>
15283<em>IMAGE2D</em><br>
15284<em>IIMAGE2D</em><br>
15285<em>UIMAGE2D</em><br>
15286<em>IMAGE3D</em><br>
15287<em>IIMAGE3D</em><br>
15288<em>UIMAGE3D</em><br>
15289<em>IMAGECUBE</em><br>
15290<em>IIMAGECUBE</em><br>
15291<em>UIMAGECUBE</em><br>
15292<em>IMAGEBUFFER</em><br>
15293<em>IIMAGEBUFFER</em><br>
15294<em>UIMAGEBUFFER</em><br>
15295<em>IMAGE2DARRAY</em><br>
15296<em>IIMAGE2DARRAY</em><br>
15297<em>UIMAGE2DARRAY</em><br>
15298<em>IMAGECUBEARRAY</em><br>
15299<em>IIMAGECUBEARRAY</em><br>
15300<em>UIMAGECUBEARRAY</em><br>
15301<em>struct_specifier</em><br>
15302<em>TYPE_NAME</em></p>
15303</dd>
15304<dt class="hdlist1"><em>precision_qualifier</em> : </dt>
15305<dd>
15306<p><em>HIGH_PRECISION</em><br>
15307<em>MEDIUM_PRECISION</em><br>
15308<em>LOW_PRECISION</em></p>
15309</dd>
15310<dt class="hdlist1"><em>struct_specifier</em> : </dt>
15311<dd>
15312<p><em>STRUCT</em> <em>IDENTIFIER</em> <em>LEFT_BRACE</em> <em>struct_declaration_list</em>
15313<em>RIGHT_BRACE</em><br>
15314<em>STRUCT</em> <em>LEFT_BRACE</em> <em>struct_declaration_list</em> <em>RIGHT_BRACE</em></p>
15315</dd>
15316<dt class="hdlist1"><em>struct_declaration_list</em> : </dt>
15317<dd>
15318<p><em>struct_declaration</em><br>
15319<em>struct_declaration_list</em> <em>struct_declaration</em></p>
15320</dd>
15321<dt class="hdlist1"><em>struct_declaration</em> : </dt>
15322<dd>
15323<p><em>type_specifier</em> <em>struct_declarator_list</em> <em>SEMICOLON</em><br>
15324<em>type_qualifier</em> <em>type_specifier</em> <em>struct_declarator_list</em> <em>SEMICOLON</em></p>
15325</dd>
15326<dt class="hdlist1"><em>struct_declarator_list</em> : </dt>
15327<dd>
15328<p><em>struct_declarator</em><br>
15329<em>struct_declarator_list</em> <em>COMMA</em> <em>struct_declarator</em></p>
15330</dd>
15331<dt class="hdlist1"><em>struct_declarator</em> : </dt>
15332<dd>
15333<p><em>IDENTIFIER</em><br>
15334<em>IDENTIFIER</em> <em>array_specifier</em></p>
15335</dd>
15336<dt class="hdlist1"><em>initializer</em> : </dt>
15337<dd>
15338<p><em>assignment_expression</em></p>
15339</dd>
15340<dt class="hdlist1"><em>declaration_statement</em> : </dt>
15341<dd>
15342<p><em>declaration</em></p>
15343</dd>
15344<dt class="hdlist1"><em>statement</em> : </dt>
15345<dd>
15346<p><em>compound_statement</em><br>
15347<em>simple_statement</em></p>
15348</dd>
15349</dl>
15350</div>
15351<div class="admonitionblock note">
15352<table>
15353<tr>
15354<td class="icon">
15355<i class="fa icon-note" title="Note"></i>
15356</td>
15357<td class="content">
15358<div class="paragraph">
15359<p>Grammar Note: labeled statements for SWITCH only; 'goto' is not supported.</p>
15360</div>
15361</td>
15362</tr>
15363</table>
15364</div>
15365<div class="dlist">
15366<dl>
15367<dt class="hdlist1"><em>simple_statement</em> : </dt>
15368<dd>
15369<p><em>declaration_statement</em><br>
15370<em>expression_statement</em><br>
15371<em>selection_statement</em><br>
15372<em>switch_statement</em><br>
15373<em>case_label</em><br>
15374<em>iteration_statement</em><br>
15375<em>jump_statement</em></p>
15376</dd>
15377<dt class="hdlist1"><em>compound_statement</em> : </dt>
15378<dd>
15379<p><em>LEFT_BRACE</em> <em>RIGHT_BRACE</em><br>
15380<em>LEFT_BRACE</em> <em>statement_list</em> <em>RIGHT_BRACE</em></p>
15381</dd>
15382<dt class="hdlist1"><em>statement_no_new_scope</em> : </dt>
15383<dd>
15384<p><em>compound_statement_no_new_scope</em><br>
15385<em>simple_statement</em></p>
15386</dd>
15387<dt class="hdlist1"><em>compound_statement_no_new_scope</em> : </dt>
15388<dd>
15389<p><em>LEFT_BRACE</em> <em>RIGHT_BRACE</em><br>
15390<em>LEFT_BRACE</em> <em>statement_list</em> <em>RIGHT_BRACE</em></p>
15391</dd>
15392<dt class="hdlist1"><em>statement_list</em> : </dt>
15393<dd>
15394<p><em>statement</em><br>
15395<em>statement_list</em> <em>statement</em></p>
15396</dd>
15397<dt class="hdlist1"><em>expression_statement</em> : </dt>
15398<dd>
15399<p><em>SEMICOLON</em><br>
15400<em>expression</em> <em>SEMICOLON</em></p>
15401</dd>
15402<dt class="hdlist1"><em>selection_statement</em> : </dt>
15403<dd>
15404<p><em>IF</em> <em>LEFT_PAREN</em> <em>expression</em> <em>RIGHT_PAREN</em> <em>selection_rest_statement</em></p>
15405</dd>
15406<dt class="hdlist1"><em>selection_rest_statement</em> : </dt>
15407<dd>
15408<p><em>statement</em> <em>ELSE</em> <em>statement</em><br>
15409<em>statement</em></p>
15410</dd>
15411<dt class="hdlist1"><em>condition</em> : </dt>
15412<dd>
15413<p><em>expression</em><br>
15414<em>fully_specified_type</em> <em>IDENTIFIER</em> <em>EQUAL</em> <em>initializer</em></p>
15415</dd>
15416<dt class="hdlist1"><em>switch_statement</em> : </dt>
15417<dd>
15418<p><em>SWITCH</em> <em>LEFT_PAREN</em> <em>expression</em> <em>RIGHT_PAREN</em> <em>LEFT_BRACE</em>
15419<em>switch_statement_list</em><br>
15420<em>RIGHT_BRACE</em></p>
15421</dd>
15422<dt class="hdlist1"><em>switch_statement_list</em> : </dt>
15423<dd>
15424<p>/* <em>nothing</em> */<br>
15425<em>statement_list</em></p>
15426</dd>
15427<dt class="hdlist1"><em>case_label</em> : </dt>
15428<dd>
15429<p><em>CASE</em> <em>expression</em> <em>COLON</em><br>
15430<em>DEFAULT</em> <em>COLON</em></p>
15431</dd>
15432<dt class="hdlist1"><em>iteration_statement</em> : </dt>
15433<dd>
15434<p><em>WHILE</em> <em>LEFT_PAREN</em> <em>condition</em> <em>RIGHT_PAREN</em> <em>statement_no_new_scope</em><br>
15435<em>DO</em> <em>statement</em> <em>WHILE</em> <em>LEFT_PAREN</em> <em>expression</em> <em>RIGHT_PAREN</em>
15436<em>SEMICOLON</em><br>
15437<em>FOR</em> <em>LEFT_PAREN</em> <em>for_init_statement</em> <em>for_rest_statement</em>
15438<em>RIGHT_PAREN</em> <em>statement_no_new_scope</em></p>
15439</dd>
15440<dt class="hdlist1"><em>for_init_statement</em> : </dt>
15441<dd>
15442<p><em>expression_statement</em><br>
15443<em>declaration_statement</em></p>
15444</dd>
15445<dt class="hdlist1"><em>conditionopt</em> : </dt>
15446<dd>
15447<p><em>condition</em><br>
15448/* <em>empty</em> */</p>
15449</dd>
15450<dt class="hdlist1"><em>for_rest_statement</em> : </dt>
15451<dd>
15452<p><em>conditionopt</em> <em>SEMICOLON</em><br>
15453<em>conditionopt</em> <em>SEMICOLON</em> <em>expression</em></p>
15454</dd>
15455<dt class="hdlist1"><em>jump_statement</em> : </dt>
15456<dd>
15457<p><em>CONTINUE</em> <em>SEMICOLON</em><br>
15458<em>BREAK</em> <em>SEMICOLON</em><br>
15459<em>RETURN</em> <em>SEMICOLON</em><br>
15460<em>RETURN</em> <em>expression</em> <em>SEMICOLON</em><br>
15461<em>DISCARD</em> <em>SEMICOLON</em> // Fragment shader only.</p>
15462</dd>
15463</dl>
15464</div>
15465<div class="admonitionblock note">
15466<table>
15467<tr>
15468<td class="icon">
15469<i class="fa icon-note" title="Note"></i>
15470</td>
15471<td class="content">
15472<div class="paragraph">
15473<p>Grammar Note: No 'goto'.
15474Gotos are not supported.</p>
15475</div>
15476</td>
15477</tr>
15478</table>
15479</div>
15480<div class="dlist">
15481<dl>
15482<dt class="hdlist1"><em>translation_unit</em> : </dt>
15483<dd>
15484<p><em>external_declaration</em><br>
15485<em>translation_unit</em> <em>external_declaration</em></p>
15486</dd>
15487<dt class="hdlist1"><em>external_declaration</em> : </dt>
15488<dd>
15489<p><em>function_definition</em><br>
15490<em>declaration</em></p>
15491</dd>
15492<dt class="hdlist1"><em>function_definition</em> : </dt>
15493<dd>
15494<p><em>function_prototype</em> <em>compound_statement_no_new_scope</em></p>
15495</dd>
15496</dl>
15497</div>
15498</div>
15499</div>
15500<div class="paragraph">
15501<p>In general the above grammar describes a super set of the OpenGL ES Shading Language.
15502Certain constructs that are valid purely in terms of the grammar are
15503disallowed by statements elsewhere in this specification.</p>
15504</div>
15505</div>
15506</div>
15507<div class="sect1">
15508<h2 id="counting-of-inputs-and-outputs">11. Counting of Inputs and Outputs</h2>
15509<div class="sectionbody">
15510<div class="paragraph">
15511<p>This section applies to outputs from the last active vertex processing stage
15512and inputs to the fragment stage.
15513GLSL ES 3.20 specifies the storage available for such variables in terms of
15514an array of 4-vectors.
15515The assumption is that variables will be packed into these arrays without
15516wasting space.
15517This places significant burden on implementations since optimal packing is
15518computationally intensive.
15519Implementations may have more internal resources than exposed to the
15520application and so avoid the need to perform packing but this is also
15521considered an expensive solution.</p>
15522</div>
15523<div class="paragraph">
15524<p>GLSL ES 3.20 therefore relaxes the requirements for packing by
15525specifying a simpler algorithm that may be used.
15526This algorithm specifies a minimum requirement for when a set of variables
15527must be supported by an implementation.
15528The implementation is allowed to support more than the minimum and so may
15529use a more efficient algorithm and/or may support more registers than the
15530virtual target machine.</p>
15531</div>
15532<div class="paragraph">
15533<p>Outputs from the last active vertex stage and inputs to the fragment stage are
15534counted separately.
15535If statically used in the fragment shader, the built-in special variables
15536(<em>gl_FragCoord</em>, <em>gl_FrontFacing</em> and <em>gl_PointCoord</em>) are included when
15537calculating the storage requirements of fragment inputs.</p>
15538</div>
15539<div class="paragraph">
15540<p>If the last active vertex-pipeline shader and fragment shader are linked
15541together, inputs and outputs are only counted if they are statically used
15542within the shader.
15543If the shaders are each compiled into a separable program, all declared inputs
15544and outputs are counted.</p>
15545</div>
15546<div class="admonitionblock note">
15547<table>
15548<tr>
15549<td class="icon">
15550<i class="fa icon-note" title="Note"></i>
15551</td>
15552<td class="content">
15553<div class="title">Note</div>
15554<div class="paragraph">
15555<p>GLSL ES 3.20 does not require the implementation to remove
15556outputs which are not statically used in the fragment shader.</p>
15557</div>
15558</td>
15559</tr>
15560</table>
15561</div>
15562<div class="paragraph">
15563<p>For the algorithm used, failing resource allocation for a variable must
15564result in an error.</p>
15565</div>
15566<div class="paragraph">
15567<p>The resource allocation of variables must succeed for all cases where the
15568following packing algorithm succeeds:</p>
15569</div>
15570<div class="ulist">
15571<ul>
15572<li>
15573<p>The target architecture consists of a grid of registers, 16 rows by 4
15574columns.
15575Each register can contain a scalar value, i.e. a float, int or uint.</p>
15576</li>
15577<li>
15578<p>Variables with an explicit location are allocated first.
15579When attempting to allocate a location for other variables, if there is
15580a conflict, the search moves to the next available free location.</p>
15581</li>
15582<li>
15583<p>Structures are assumed to be flattened.
15584Each data member is treated as if it were at global scope.</p>
15585</li>
15586<li>
15587<p>Variables are packed into the registers one at a time so that they each
15588occupy a contiguous sub-rectangle.
15589No splitting of variables is permitted.</p>
15590</li>
15591<li>
15592<p>The orientation of variables is fixed.
15593Vectors always occupy registers in a single row.
15594Elements of an array must be in different rows.
15595E.g. <strong>vec4</strong> will always occupy one row; float[16] will occupy one
15596column.
15597Since it is not permitted to split a variable, large arrays e.g.
15598float[32] will always fail with this algorithm.</p>
15599</li>
15600<li>
15601<p>Non-square matrices of type <strong>matCxR</strong> consume the same space as a square
15602matrix of type <strong>matN</strong> where N is the greater of C and R.
15603Variables of type <strong>mat2</strong> occupies 2 complete rows.
15604These rules allow implementations more flexibility in how variables are
15605stored.
15606+ Other variables consume only the minimum space required.</p>
15607</li>
15608<li>
15609<p>Arrays of size N are assumed to take N times the size of the base type.</p>
15610</li>
15611<li>
15612<p>Variables are packed in the following order:</p>
15613<div class="olist arabic">
15614<ol class="arabic">
15615<li>
15616<p><strong>mat4</strong> and arrays of <strong>mat4</strong>.</p>
15617</li>
15618<li>
15619<p><strong>mat2</strong> and arrays of <strong>mat2</strong> (since they occupy full rows)</p>
15620</li>
15621<li>
15622<p><strong>vec4</strong> and arrays of <strong>vec4</strong></p>
15623</li>
15624<li>
15625<p><strong>mat3</strong> and arrays of <strong>mat3</strong></p>
15626</li>
15627<li>
15628<p><strong>vec3</strong> and arrays of <strong>vec3</strong></p>
15629</li>
15630<li>
15631<p><strong>vec2</strong> and arrays of <strong>vec2</strong></p>
15632</li>
15633<li>
15634<p>Scalar types and arrays of scalar types</p>
15635</li>
15636</ol>
15637</div>
15638</li>
15639<li>
15640<p>For each of the above types, the arrays are processed in order of size,
15641largest first.
15642Arrays of size 1 and the base type are considered equivalent.
15643The first type to be packed will be mat4[4], mat4[3], mat[2] followed by
15644mat4, mat2[4]&#8230;&#8203;mat2[2], mat2, vec4[8], vec4[7],&#8230;&#8203;vec4[1], vec4,
15645mat3[2], mat3 and so on.
15646The last variables to be packed will be float (and float[1]).</p>
15647</li>
15648<li>
15649<p>For 2,3 and 4 component variables packing is started using the 1<sup>st</sup>
15650column of the 1<sup>st</sup> row.
15651Variables are then allocated to successive rows, aligning them to the
156521<sup>st</sup> column.</p>
15653</li>
15654<li>
15655<p>For 2 component variables, when there are no spare rows, the strategy is
15656switched to using the highest numbered row and the lowest numbered
15657column where the variable will fit.
15658(In practice, this means they will be aligned to the x or z component.)
15659Packing of any further 3 or 4 component variables will fail at this
15660point.</p>
15661</li>
15662<li>
15663<p>1 component variables (e.g. floats and arrays of floats) have their own
15664packing rule.
15665They are packed in order of size, largest first.
15666Each variable is placed in the column that leaves the least amount of
15667space in the column and aligned to the lowest available rows within that
15668column.
15669During this phase of packing, space will be available in up to 4
15670columns.
15671The space within each column is always contiguous in the case where no
15672variables have explicit locations.</p>
15673</li>
15674<li>
15675<p>For each type, variables with the 'smooth' property are packed first,
15676followed by variables with the 'flat' property.</p>
15677</li>
15678<li>
15679<p>Each row can contain either values with the 'smooth' property or the
15680'flat' property but not both.
15681If this situation is encountered during allocation, the algorithm skips
15682the component location and continues with the next available location.
15683These skipped locations may be used for other values later in the
15684allocation process.</p>
15685</li>
15686<li>
15687<p>There is no backtracking.
15688Once a value is assigned a location, it cannot be changed, even if such
15689a change is required for a successful allocation.</p>
15690</li>
15691</ul>
15692</div>
15693<div class="paragraph">
15694<p>Example: pack the following types:</p>
15695</div>
15696<div class="listingblock">
15697<div class="content">
15698<pre class="CodeRay highlight"><code data-lang="c++">out vec4 a;     <span class="comment">// top left</span>
15699out mat3 b;     <span class="comment">// align to left, lowest numbered rows</span>
15700out mat2x3 c;   <span class="comment">// same size as mat3, align to left</span>
15701out vec2 d[<span class="integer">6</span>];  <span class="comment">// align to left, lowest numbered rows</span>
15702out vec2 e[<span class="integer">4</span>];  <span class="comment">// Cannot align to left so align to z column, highest</span>
15703                <span class="comment">// numbered rows</span>
15704out vec2 f;     <span class="comment">// Align to left, lowest numbered rows.</span>
15705out <span class="predefined-type">float</span> g[<span class="integer">3</span>]  <span class="comment">// Column with minimum space</span>
15706out <span class="predefined-type">float</span> h[<span class="integer">2</span>]; <span class="comment">// Column with minimum space (choice of 3, any</span>
15707                <span class="comment">// can be used)</span>
15708out <span class="predefined-type">float</span> i;    <span class="comment">// Column with minimum space</span></code></pre>
15709</div>
15710</div>
15711<div class="paragraph">
15712<p>In this example, the variables happen to be listed in the order in which
15713they are packed.
15714Packing is independent of the order of declaration.</p>
15715</div>
15716<table class="tableblock frame-all grid-all stretch">
15717<colgroup>
15718<col style="width: 20%;">
15719<col style="width: 20%;">
15720<col style="width: 20%;">
15721<col style="width: 20%;">
15722<col style="width: 20%;">
15723</colgroup>
15724<tbody>
15725<tr>
15726<td class="tableblock halign-left valign-top"></td>
15727<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>x</strong></p></td>
15728<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>y</strong></p></td>
15729<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>z</strong></p></td>
15730<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>w</strong></p></td>
15731</tr>
15732<tr>
15733<td class="tableblock halign-left valign-top"><p class="tableblock">0</p></td>
15734<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>a</strong></p></td>
15735<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>a</strong></p></td>
15736<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>a</strong></p></td>
15737<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>a</strong></p></td>
15738</tr>
15739<tr>
15740<td class="tableblock halign-left valign-top"><p class="tableblock">1</p></td>
15741<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>b</strong></p></td>
15742<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>b</strong></p></td>
15743<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>b</strong></p></td>
15744<td class="tableblock halign-left valign-top"></td>
15745</tr>
15746<tr>
15747<td class="tableblock halign-left valign-top"><p class="tableblock">2</p></td>
15748<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>b</strong></p></td>
15749<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>b</strong></p></td>
15750<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>b</strong></p></td>
15751<td class="tableblock halign-left valign-top"></td>
15752</tr>
15753<tr>
15754<td class="tableblock halign-left valign-top"><p class="tableblock">3</p></td>
15755<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>b</strong></p></td>
15756<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>b</strong></p></td>
15757<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>b</strong></p></td>
15758<td class="tableblock halign-left valign-top"></td>
15759</tr>
15760<tr>
15761<td class="tableblock halign-left valign-top"><p class="tableblock">4</p></td>
15762<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>c</strong></p></td>
15763<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>c</strong></p></td>
15764<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>c</strong></p></td>
15765<td class="tableblock halign-left valign-top"></td>
15766</tr>
15767<tr>
15768<td class="tableblock halign-left valign-top"><p class="tableblock">5</p></td>
15769<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>c</strong></p></td>
15770<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>c</strong></p></td>
15771<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>c</strong></p></td>
15772<td class="tableblock halign-left valign-top"></td>
15773</tr>
15774<tr>
15775<td class="tableblock halign-left valign-top"><p class="tableblock">6</p></td>
15776<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>c</strong></p></td>
15777<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>c</strong></p></td>
15778<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>c</strong></p></td>
15779<td class="tableblock halign-left valign-top"></td>
15780</tr>
15781<tr>
15782<td class="tableblock halign-left valign-top"><p class="tableblock">7</p></td>
15783<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15784<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15785<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>g</strong></p></td>
15786<td class="tableblock halign-left valign-top"></td>
15787</tr>
15788<tr>
15789<td class="tableblock halign-left valign-top"><p class="tableblock">8</p></td>
15790<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15791<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15792<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>g</strong></p></td>
15793<td class="tableblock halign-left valign-top"></td>
15794</tr>
15795<tr>
15796<td class="tableblock halign-left valign-top"><p class="tableblock">9</p></td>
15797<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15798<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15799<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>g</strong></p></td>
15800<td class="tableblock halign-left valign-top"></td>
15801</tr>
15802<tr>
15803<td class="tableblock halign-left valign-top"><p class="tableblock">10</p></td>
15804<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15805<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15806<td class="tableblock halign-left valign-top"></td>
15807<td class="tableblock halign-left valign-top"></td>
15808</tr>
15809<tr>
15810<td class="tableblock halign-left valign-top"><p class="tableblock">11</p></td>
15811<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15812<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15813<td class="tableblock halign-left valign-top"></td>
15814<td class="tableblock halign-left valign-top"></td>
15815</tr>
15816<tr>
15817<td class="tableblock halign-left valign-top"><p class="tableblock">12</p></td>
15818<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15819<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>d</strong></p></td>
15820<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>e</strong></p></td>
15821<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>e</strong></p></td>
15822</tr>
15823<tr>
15824<td class="tableblock halign-left valign-top"><p class="tableblock">13</p></td>
15825<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>f</strong></p></td>
15826<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>f</strong></p></td>
15827<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>e</strong></p></td>
15828<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>e</strong></p></td>
15829</tr>
15830<tr>
15831<td class="tableblock halign-left valign-top"><p class="tableblock">14</p></td>
15832<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>h</strong></p></td>
15833<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>i</strong></p></td>
15834<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>e</strong></p></td>
15835<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>e</strong></p></td>
15836</tr>
15837<tr>
15838<td class="tableblock halign-left valign-top"><p class="tableblock">15</p></td>
15839<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>h</strong></p></td>
15840<td class="tableblock halign-left valign-top"></td>
15841<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>e</strong></p></td>
15842<td class="tableblock halign-left valign-top"><p class="tableblock"><strong>e</strong></p></td>
15843</tr>
15844</tbody>
15845</table>
15846<div class="paragraph">
15847<p>Some types e.g. mat4[8] will be too large to fit.
15848These always fail with this algorithm.</p>
15849</div>
15850</div>
15851</div>
15852<div class="sect1">
15853<h2 id="acknowledgments">12. Acknowledgments</h2>
15854<div class="sectionbody">
15855<div class="paragraph">
15856<p>This specification is based on the work of those who contributed to past
15857versions of the Open GL and Open GL ES Language Specifications and the
15858following contributors to this version:</p>
15859</div>
15860<div class="paragraph">
15861<p>Acorn Pooley, NVIDIA<br>
15862Alberto Moreira, Qualcomm<br>
15863Aleksandra Krstic, Qualcomm<br>
15864Alon Or-bach, Nokia &amp; Samsung<br>
15865Andrzej Kacprowski, Intel<br>
15866Arzhange Safdarzadeh, Intel<br>
15867Aske Simon Christensen, ARM<br>
15868Avi Shapira, Graphic Remedy<br>
15869Barthold Lichtenbelt, NVIDIA<br>
15870Ben Bowman, Imagination Technologies<br>
15871Ben Brierton, Broadcom<br>
15872Benj Lipchak, Apple<br>
15873Benson Tao, Vivante<br>
15874Bill Licea-Kane, Qualcomm<br>
15875Brent Insko, Intel<br>
15876Brian Murray, Freescale<br>
15877Bruce Merry, ARM<br>
15878Carlos Santa, TI<br>
15879Cass Everitt, Epic Games &amp; NVIDIA<br>
15880Cemil Azizoglu, TI<br>
15881Chang-Hyo Yu, Samsung<br>
15882Chris Dodd, NVIDIA<br>
15883Chris Knox, NVIDIA<br>
15884Chris Tserng, TI<br>
15885Clay Montgomery, TI<br>
15886Daniel Kartch, NVIDIA<br>
15887Daniel Koch, Transgaming&amp; NVIDIA<br>
15888Daoxiang Gong, Imagination Technologies<br>
15889Dave Shreiner, ARM<br>
15890David Garcia, AMD<br>
15891David Jarmon, Vivante<br>
15892Derek Cornish, Epic Games<br>
15893Dominick Witczak, Mobica<br>
15894Eben Upton, Broadcom<br>
15895Ed Plowman, ARM<br>
15896Eisaku Ohbuchi, DMP<br>
15897Elan Lennard, ARM<br>
15898Erik Faye-Lund, ARM<br>
15899Graham Connor, Imagination Technologies<br>
15900Graham Sellers, AMD<br>
15901Greg Roth, NVIDIA<br>
15902Guillaume Portier, Hi Corporation<br>
15903Guofang Jiao, Qualcomm<br>
15904Hans-Martin Will, Vincent<br>
15905Hwanyong Lee, Huone<br>
15906I-Gene Leong, NVIDIA<br>
15907Ian Romanick, Intel<br>
15908Ian South-Dickinson, NVIDIA<br>
15909Ilan Aelion-Exch, Samsung<br>
15910Inkyun Lee, Huone<br>
15911Jacob Strm, Ericsson<br>
15912James Adams, Broadcom<br>
15913James Jones, Imagination Technologies<br>
15914James McCombe, Imagination Technologies<br>
15915Jamie Gennis, Google<br>
15916Jan-Harald Fredriksen, ARM<br>
15917Jani Vaisanen, Nokia<br>
15918Jarkko Kemppainen, Symbio<br>
15919Jeff Bolz, NVIDIA<br>
15920Jeff Leger, Qualcomm<br>
15921Jeff Vigil, Qualcomm<br>
15922Jeremy Sandmel, Apple<br>
15923Jeremy Thorne, Broadcom<br>
15924Jim Hauxwell, Broadcom<br>
15925Jinsung Kim, Huone<br>
15926Jiyoung Yoon, Huone<br>
15927John Kessenich, Google<br>
15928Jon Kennedy, 3DLabs<br>
15929Jon Leech, Khronos<br>
15930Jonathan Putsman, Imagination Technologies<br>
15931Joohoon Lee, Samsung<br>
15932JoukoKylmäoja, Symbio<br>
15933Jrn Nystad, ARM<br>
15934Jussi Rasanen, NVIDIA<br>
15935Kalle Raita, drawElements<br>
15936Kari Pulli, Nokia<br>
15937Keith Whitwell, VMware<br>
15938Kent Miller, Netlogic Microsystems<br>
15939Kimmo Nikkanen, Nokia<br>
15940Konsta Karsisto, Nokia<br>
15941Krzysztof Kaminski, Intel<br>
15942Larry Seiler, Intel<br>
15943Lars Remes, Symbio<br>
15944Lee Thomason, Adobe<br>
15945Lefan Zhong, Vivante<br>
15946Marcus Lorentzon, Ericsson<br>
15947Mark Butler, Imagination Technologies<br>
15948Mark Callow, Hi Corporation<br>
15949Mark Cresswell, Broadcom<br>
15950Mark Snyder, Alt Software<br>
15951Mark Young, AMD<br>
15952Mathieu Robart, STM<br>
15953Matt Netsch, Qualcomm<br>
15954Matt Russo, Matrox<br>
15955Maurice Ribble, Qualcomm<br>
15956Max Kazakov, DMP<br>
15957Mika Pesonen, Nokia<br>
15958Mike Cai, Vivante<br>
15959Mike Weiblen, Zebra Imaging &amp; Qualcomm<br>
15960Mila Smith, AMD<br>
15961Nakhoon Baek, Kyungpook Univeristy<br>
15962Nate Huang, NVIDIA<br>
15963Neil Trevett, NVIDIA<br>
15964Nelson Kidd, Intel<br>
15965Nick Haemel, NVIDIA<br>
15966Nick Penwarden, Epic Games<br>
15967Niklas Smedberg, Epic Games<br>
15968Nizar Romdan, ARM<br>
15969Oliver Wohlmuth , Fujitsu<br>
15970Pat Brown, NVIDIA<br>
15971Paul Ruggieri, Qualcomm<br>
15972Per Wennersten, Ericsson<br>
15973Petri Talala, Symbio<br>
15974Phil Huxley, ZiiLabs<br>
15975Philip Hatcher, Freescale &amp; Intel<br>
15976Piers Daniell, NVIDIA<br>
15977Pyry Haulos, drawElements<br>
15978Piotr Tomaszewski, Ericsson<br>
15979Piotr Uminski, Intel<br>
15980Rami Mayer, Samsung<br>
15981Rauli Laatikainen, RightWare<br>
15982Rob Barris, NVIDIA<br>
15983Rob Simpson, Qualcomm<br>
15984Roj Langhi, Vivante<br>
15985Rune Holm, ARM<br>
15986Sami Kyostila, Nokia<br>
15987Sean Ellis, ARM<br>
15988Shereef Shehata, TI<br>
15989Sila Kayo, Nokia<br>
15990Slawomir Cygan, Intel<br>
15991Slawomir Grajewski, Intel<br>
15992Steve Hill, STM &amp; Broadcom<br>
15993Steven Olney, DMP<br>
15994Suman Sharma, Intel<br>
15995Tapani Palli, Nokia<br>
15996Teemu Laakso, Symbio<br>
15997Tero Karras, NVIDIA<br>
15998Timo Suoranta, Imagination Technologies &amp; Broadcom<br>
15999Tom Cooksey, ARM<br>
16000Tom McReynolds, NVIDIA<br>
16001Tom Olson, TI &amp; ARM<br>
16002Tomi Aarnio, Nokia<br>
16003Tommy Asano, Takumi<br>
16004Wes Bang, Nokia<br>
16005Yanjun Zhang, Vivante</p>
16006</div>
16007</div>
16008</div>
16009<div class="sect1">
16010<h2 id="references">13. Normative References</h2>
16011<div class="sectionbody">
16012<div class="olist arabic">
16013<ol class="arabic">
16014<li>
16015<p>International Standard ISO/IEC 14882:1998(E).
16016Programming Languages - C++. Referenced for preprocessor only</p>
16017</li>
16018<li>
16019<p>&#8220;OpenGL<sup>R</sup> ES, Version 3.2&#8221;,
16020<a href="https://www.khronos.org/registry/OpenGL/index_es.php" class="bare">https://www.khronos.org/registry/OpenGL/index_es.php</a>, November 3, 2016.</p>
16021</li>
16022<li>
16023<p>&#8220;The OpenGL<sup>R</sup> Graphics System: A Specification, Version 4.6 (Core
16024Profile)&#8221;, <a href="https://www.khronos.org/registry/OpenGL/index_gl.php" class="bare">https://www.khronos.org/registry/OpenGL/index_gl.php</a>, June
160251, 2016.</p>
16026</li>
16027<li>
16028<p>International Standard ISO/IEC 646:1991.
16029Information technology - ISO 7-bit coded character set for information
16030interchange</p>
16031</li>
16032<li>
16033<p>The Unicode Standard Version 6.0 - Core Specification.</p>
16034</li>
16035<li>
16036<p>IEEE 754-2008.
16037<em>IEEE Standard for Floating-Point Arithmetic</em></p>
16038</li>
16039<li>
16040<p>&#8220;SPIR-V Specification, Version 1.3, Revision 7&#8221; ,
16041<a href="https://www.khronos.org/registry/spir-v/" class="bare">https://www.khronos.org/registry/spir-v/</a>.</p>
16042</li>
16043<li>
16044<p>&#8220;Vulkan<sup>R</sup> 1.1.105 - A Specification&#8221;,
16045<a href="https://www.khronos.org/registry/vulkan/" class="bare">https://www.khronos.org/registry/vulkan/</a>,
16046March 19, 2019.</p>
16047</li>
16048</ol>
16049</div>
16050</div>
16051</div>
16052<div class="sect1">
16053<h2 id="_non_normative_spir_v_mappings">14. Non-Normative SPIR-V Mappings</h2>
16054<div class="sectionbody">
16055<div class="paragraph">
16056<p>This appendix includes:</p>
16057</div>
16058<div class="ulist">
16059<ul>
16060<li>
16061<p>a comparision of feature differences with SPIR-V versus without, for both Vulkan and OpenGL</p>
16062</li>
16063<li>
16064<p>a discussion of how GLSL features logically map to SPIR-V features.</p>
16065</li>
16066</ul>
16067</div>
16068<div class="sect2">
16069<h3 id="_feature_comparisons">14.1. Feature Comparisons</h3>
16070<div class="paragraph">
16071<p>The following features are removed for both OpenGL and Vulkan:</p>
16072</div>
16073<div class="ulist">
16074<ul>
16075<li>
16076<p>subroutines</p>
16077</li>
16078<li>
16079<p>shared and packed block layouts</p>
16080</li>
16081<li>
16082<p>the already deprecated texturing functions (e.g., <code>texture2D()</code>)</p>
16083</li>
16084<li>
16085<p>the already deprecated noise functions (e.g., <code>noise1()</code>)</p>
16086</li>
16087<li>
16088<p>compatibility-profile features</p>
16089</li>
16090<li>
16091<p><em>gl_DepthRangeParameters</em> and <em>gl_NumSamples</em></p>
16092</li>
16093</ul>
16094</div>
16095<div class="paragraph">
16096<p>Vulkan removed the following features, which are still present for OpenGL:</p>
16097</div>
16098<div class="ulist">
16099<ul>
16100<li>
16101<p>Default uniforms, for non-opaque types:
16102The <strong>UniformConstant</strong> storage class can be used on individual
16103variables at global scope. (That is, uniforms don&#8217;t have to be in a
16104block, unless they are built-in members that are in block in GLSL
16105version 4.5 or above.)</p>
16106</li>
16107<li>
16108<p>GLSL atomic-counter bindings have the <em>offset</em> layout qualifier &#8594;
16109SPIR-V <strong>AtomicCounter</strong> storage class using the <strong>Offset</strong> decoration</p>
16110</li>
16111<li>
16112<p>GLSL <em>origin_lower_left</em> &#8594; SPIR-V <strong>OriginLowerLeft</strong></p>
16113</li>
16114<li>
16115<p>special rules for locations for input doubles in the vertex shader</p>
16116</li>
16117<li>
16118<p><em>gl_VertexID</em> and <em>gl_InstanceID</em> (more detail follows)</p>
16119</li>
16120</ul>
16121</div>
16122<div class="paragraph">
16123<p>The following features are added for both OpenGL and Vulkan:</p>
16124</div>
16125<div class="ulist">
16126<ul>
16127<li>
16128<p>specialization constants</p>
16129</li>
16130<li>
16131<p><em>offset</em> can organize members in a different order than declaration order</p>
16132</li>
16133<li>
16134<p><em>offset</em> and <em>align</em> layout qualifiers for uniform/buffer blocks for
16135versions that did not support them</p>
16136</li>
16137</ul>
16138</div>
16139<div class="paragraph">
16140<p>Vulkan Only: The following features are added:</p>
16141</div>
16142<div class="ulist">
16143<ul>
16144<li>
16145<p>push-constant buffers</p>
16146</li>
16147<li>
16148<p>shader combining of separate textures and samplers (SPIR-V <strong>OpTypeSampler</strong>)</p>
16149</li>
16150<li>
16151<p>descriptor sets (<strong>DescriptorSet</strong> must be 0, if present)</p>
16152</li>
16153<li>
16154<p><em>gl_VertexIndex</em> and <em>gl_InstanceIndex</em></p>
16155</li>
16156<li>
16157<p>subpass-input targets and input attachments (<em>input_attachment_index</em>)</p>
16158</li>
16159</ul>
16160</div>
16161<div class="paragraph">
16162<p>The following features are changed in both OpenGL and Vulkan:</p>
16163</div>
16164<div class="ulist">
16165<ul>
16166<li>
16167<p><em>gl_FragColor</em> will no longer indicate an implicit broadcast</p>
16168</li>
16169</ul>
16170</div>
16171<div class="paragraph">
16172<p>Vulkan Only: The following features are changed:</p>
16173</div>
16174<div class="ulist">
16175<ul>
16176<li>
16177<p>precision qualifiers (<strong>mediump</strong> and <strong>lowp</strong>) will be respected for all
16178versions, not dropped for desktop versions (default precision for
16179desktop versions is <strong>highp</strong> for all types)</p>
16180</li>
16181<li>
16182<p>arrays of uniforms and buffer blocks take only one binding number for
16183the entire object, not one per array element</p>
16184</li>
16185<li>
16186<p>the default origin is <em>origin_upper_left</em> instead of <em>origin_lower_left</em></p>
16187</li>
16188</ul>
16189</div>
16190<div class="paragraph">
16191<p>Vulkan does not allow multi-dimensional arrays of resources like
16192UBOs and SSBOs in its SPIR-V environment spec. SPIR-V supports
16193it and OpenGL already allows this for GLSL shaders. SPIR-V
16194for OpenGL also allows it.</p>
16195</div>
16196</div>
16197<div class="sect2">
16198<h3 id="_mapping_from_glsl_to_spir_v">14.2. Mapping from GLSL to SPIR-V</h3>
16199<div class="sect3">
16200<h4 id="_specialization_constants">14.2.1. Specialization Constants</h4>
16201<div class="paragraph">
16202<p>SPIR-V specialization constants, which can be set later by the client API,
16203can be declared using <code>layout(constant_id=&#8230;&#8203;)</code>. For example, to make a
16204specialization constant with a default value of 12:</p>
16205</div>
16206<div class="literalblock">
16207<div class="content">
16208<pre>layout(constant_id = 17) const int arraySize = 12;</pre>
16209</div>
16210</div>
16211<div class="paragraph">
16212<p>Above, <code>17</code> is the ID by which the API or other tools can later refer to
16213this specific specialization constant.  The API or an intermediate tool can
16214then change its value to another constant integer before it is fully
16215lowered to executable code.  If it is never changed before final lowering,
16216it will retain the value of 12.</p>
16217</div>
16218<div class="paragraph">
16219<p>Specialization constants have const semantics, except they don&#8217;t fold.
16220Hence, an array can be declared with <code>arraySize</code> from above:</p>
16221</div>
16222<div class="literalblock">
16223<div class="content">
16224<pre>vec4 data[arraySize];  // legal, even though arraySize might change</pre>
16225</div>
16226</div>
16227<div class="paragraph">
16228<p>Specialization constants can be in expressions:</p>
16229</div>
16230<div class="literalblock">
16231<div class="content">
16232<pre>vec4 data2[arraySize + 2];</pre>
16233</div>
16234</div>
16235<div class="paragraph">
16236<p>This will make <code>data2</code> be sized by 2 more than whatever constant value
16237<code>arraySize</code> has when it is time to lower the shader to executable code.</p>
16238</div>
16239<div class="paragraph">
16240<p>An expression formed with specialization constants also behaves in the
16241shader like a specialization constant, not a like a constant.</p>
16242</div>
16243<div class="literalblock">
16244<div class="content">
16245<pre>arraySize + 2       // a specialization constant (with no constant_id)</pre>
16246</div>
16247</div>
16248<div class="paragraph">
16249<p>Such expressions can be used in the same places as a constant.</p>
16250</div>
16251<div class="paragraph">
16252<p>The <em>constant_id</em> can only be applied to a scalar integer, a scalar floating-point
16253or a scalar Boolean.</p>
16254</div>
16255<div class="paragraph">
16256<p>Only basic operators and constructors can be applied to a specialization
16257constant and still result in a specialization constant:</p>
16258</div>
16259<div class="literalblock">
16260<div class="content">
16261<pre>layout(constant_id = 17) const int arraySize = 12;
16262sin(float(arraySize));    // result is not a specialization constant</pre>
16263</div>
16264</div>
16265<div class="paragraph">
16266<p>While SPIR-V specialization constants are only for scalars, a vector
16267can be made by operations on scalars:</p>
16268</div>
16269<div class="literalblock">
16270<div class="content">
16271<pre>layout(constant_id = 18) const int scX = 1;
16272layout(constant_id = 19) const int scZ = 1;
16273const vec3 scVec = vec3(scX, 1, scZ);  // partially specialized vector</pre>
16274</div>
16275</div>
16276<div class="paragraph">
16277<p>A built-in variable can have a <em>constant_id</em> attached to it:</p>
16278</div>
16279<div class="literalblock">
16280<div class="content">
16281<pre>layout(constant_id = 18) gl_MaxImageUnits;</pre>
16282</div>
16283</div>
16284<div class="paragraph">
16285<p>This makes it behave as a specialization constant.  It is not a full
16286redeclaration; all other characteristics are left intact from the
16287original built-in declaration.</p>
16288</div>
16289<div class="paragraph">
16290<p>The built-in vector <em>gl_WorkGroupSize</em> can be specialized using special
16291layout <code>local_size_{xyz}_id</code> applied to the <strong>in</strong> qualifier.  For example:</p>
16292</div>
16293<div class="literalblock">
16294<div class="content">
16295<pre>layout(local_size_x_id = 18, local_size_z_id = 19) in;</pre>
16296</div>
16297</div>
16298<div class="paragraph">
16299<p>This leaves <em>gl_WorkGroupSize.y</em> as a non-specialization constant, with
16300<em>gl_WorkGroupSize</em> being a partially specialized vector.  Its <em>x</em> and <em>z</em>
16301components can be later specialized using the ID&#8217;s 18 and 19.</p>
16302</div>
16303</div>
16304<div class="sect3">
16305<h4 id="_vulkan_only_push_constants">14.2.2. Vulkan Only: Push Constants</h4>
16306<div class="paragraph">
16307<p>Push constants reside in a uniform block declared using the new
16308layout-qualifier-id <em>push_constant</em> applied to a uniform-block declaration.
16309The API writes a set of constants to a push-constant buffer, and the shader
16310reads them from a <em>push_constant</em> block:</p>
16311</div>
16312<div class="literalblock">
16313<div class="content">
16314<pre>layout(push_constant) uniform BlockName {
16315    int member1;
16316    float member2;
16317    ...
16318} InstanceName; // optional instance name
16319... = InstanceName.member2; // read a push constant</pre>
16320</div>
16321</div>
16322<div class="paragraph">
16323<p>The memory accounting used for the <em>push_constant</em> uniform block is different
16324than for other uniform blocks:  There is a separate small pool of memory
16325it must fit within.  By default, a <em>push_constant</em> buffer follows the std430
16326packing rules.</p>
16327</div>
16328</div>
16329<div class="sect3">
16330<h4 id="_vulkan_only_descriptor_sets">14.2.3. Vulkan Only: Descriptor Sets</h4>
16331<div class="paragraph">
16332<p>Each shader resource in a descriptor set is assigned a tuple of (set
16333number, binding number, array element) that defines its location within
16334a descriptor set layout.
16335In GLSL, the set number and binding number are assigned via the <em>set</em>
16336and <em>binding</em> layout qualifiers respectively, and the array element is
16337implicitly assigned consecutively starting with index equal to zero for
16338the first element of an array (and array element is zero for non-array
16339variables):</p>
16340</div>
16341<div class="literalblock">
16342<div class="content">
16343<pre>// Assign set number = M, binding number = N, array element = 0
16344layout (set=M, binding=N) uniform sampler2D variableName;</pre>
16345</div>
16346</div>
16347<div class="literalblock">
16348<div class="content">
16349<pre>// Assign set number = M, binding number = N for all array elements,
16350// and array element = i for the ith member of an array of size I.
16351layout (set=M, binding=N) uniform sampler2D variableNameArray[I];</pre>
16352</div>
16353</div>
16354<div class="paragraph">
16355<p>For example, two combined texture/sampler objects can be declared in two
16356different descriptor sets as follows</p>
16357</div>
16358<div class="literalblock">
16359<div class="content">
16360<pre>layout(set = 0, binding = 0) uniform sampler2D ts3;
16361layout(set = 1, binding = 0) uniform sampler2D ts4;</pre>
16362</div>
16363</div>
16364<div class="paragraph">
16365<p>See the API documentation for more detail on the operation model of
16366descriptor sets.</p>
16367</div>
16368</div>
16369<div class="sect3">
16370<h4 id="_vulkan_only_samplers_images_textures_and_buffers">14.2.4. Vulkan Only: Samplers, Images, Textures, and Buffers</h4>
16371<div class="sect4">
16372<h5 id="_storage_images">Storage Images</h5>
16373<div class="paragraph">
16374<p>Storage images are declared in GLSL shader source using uniform image
16375variables of the appropriate dimensionality as well as a format layout
16376qualifier (if necessary):</p>
16377</div>
16378<div class="literalblock">
16379<div class="content">
16380<pre>layout (set=m, binding=n, r32f) uniform image2D myStorageImage;</pre>
16381</div>
16382</div>
16383<div class="paragraph">
16384<p>Which maps to the following SPIR-V:</p>
16385</div>
16386<div class="literalblock">
16387<div class="content">
16388<pre>        ...
16389%1 = OpExtInstImport "GLSL.std.450"
16390        ...
16391        OpName %9 "myStorageImage"
16392        OpDecorate %9 DescriptorSet m
16393        OpDecorate %9 Binding n
16394%2 = OpTypeVoid
16395%3 = OpTypeFunction %2
16396%6 = OpTypeFloat 32
16397%7 = OpTypeImage %6 2D 0 0 0 2 R32f
16398%8 = OpTypePointer UniformConstant %7
16399%9 = OpVariable %8 UniformConstant
16400        ...</pre>
16401</div>
16402</div>
16403</div>
16404<div class="sect4">
16405<h5 id="_samplers">Samplers</h5>
16406<div class="paragraph">
16407<p>SPIR-V samplers are declared in GLSL shader source using uniform <strong>sampler</strong> and
16408<strong>samplerShadow</strong> types:</p>
16409</div>
16410<div class="literalblock">
16411<div class="content">
16412<pre>layout (set=m, binding=n) uniform sampler mySampler;</pre>
16413</div>
16414</div>
16415<div class="paragraph">
16416<p>Which maps to the following SPIR-V:</p>
16417</div>
16418<div class="literalblock">
16419<div class="content">
16420<pre>        ...
16421%1 = OpExtInstImport "GLSL.std.450"
16422        ...
16423        OpName %8 "mySampler"
16424        OpDecorate %8 DescriptorSet m
16425        OpDecorate %8 Binding n
16426%2 = OpTypeVoid
16427%3 = OpTypeFunction %2
16428%6 = OpTypeSampler
16429%7 = OpTypePointer UniformConstant %6
16430%8 = OpVariable %7 UniformConstant
16431        ...</pre>
16432</div>
16433</div>
16434</div>
16435<div class="sect4">
16436<h5 id="_textures_sampled_images">Textures (Sampled Images)</h5>
16437<div class="paragraph">
16438<p>Textures are declared in GLSL shader source using uniform texture
16439variables of the appropriate dimensionality:</p>
16440</div>
16441<div class="literalblock">
16442<div class="content">
16443<pre>layout (set=m, binding=n) uniform texture2D mySampledImage;</pre>
16444</div>
16445</div>
16446<div class="paragraph">
16447<p>Which maps to the following SPIR-V:</p>
16448</div>
16449<div class="literalblock">
16450<div class="content">
16451<pre>        ...
16452%1 = OpExtInstImport "GLSL.std.450"
16453        ...
16454        OpName %9 "mySampledImage"
16455        OpDecorate %9 DescriptorSet m
16456        OpDecorate %9 Binding n
16457%2 = OpTypeVoid
16458%3 = OpTypeFunction %2
16459%6 = OpTypeFloat 32
16460%7 = OpTypeImage %6 2D 0 0 0 1 Unknown
16461%8 = OpTypePointer UniformConstant %7
16462%9 = OpVariable %8 UniformConstant
16463        ...</pre>
16464</div>
16465</div>
16466</div>
16467<div class="sect4">
16468<h5 id="_combined_texture_and_samplers">Combined Texture and Samplers</h5>
16469<div class="paragraph">
16470<p>Combined textures and samplers are declared in GLSL shader source using
16471uniform texture-combined sampler variables of the appropriate dimensionality:</p>
16472</div>
16473<div class="literalblock">
16474<div class="content">
16475<pre>layout (set=m, binding=n) uniform sampler2D myCombinedImageSampler;</pre>
16476</div>
16477</div>
16478<div class="paragraph">
16479<p>Which maps to the following SPIR-V:</p>
16480</div>
16481<div class="literalblock">
16482<div class="content">
16483<pre>        ...
16484%1 = OpExtInstImport "GLSL.std.450"
16485        ...
16486        OpName %10 "myCombinedImageSampler"
16487        OpDecorate %10 DescriptorSet m
16488        OpDecorate %10 Binding n
16489%2 = OpTypeVoid
16490%3 = OpTypeFunction %2
16491%6 = OpTypeFloat 32
16492%7 = OpTypeImage %6 2D 0 0 0 1 Unknown
16493%8 = OpTypeSampledImage %7
16494%9 = OpTypePointer UniformConstant %8
16495%10 = OpVariable %9 UniformConstant
16496        ...</pre>
16497</div>
16498</div>
16499<div class="paragraph">
16500<p>Note that a combined image sampler descriptor can be referred to as just
16501an image or sampler in the shader as per the above sections.</p>
16502</div>
16503</div>
16504<div class="sect4">
16505<h5 id="_combining_separate_samplers_and_textures">Combining Separate Samplers and Textures</h5>
16506<div class="paragraph">
16507<p>A sampler, declared with the keyword <strong>sampler</strong>, contains just filtering
16508information, containing neither a texture nor an image:</p>
16509</div>
16510<div class="literalblock">
16511<div class="content">
16512<pre>uniform sampler s;    // a handle to filtering information</pre>
16513</div>
16514</div>
16515<div class="paragraph">
16516<p>A texture, declared with keywords like <strong>texture2D</strong>, contains just image
16517information, not filtering information:</p>
16518</div>
16519<div class="literalblock">
16520<div class="content">
16521<pre>uniform texture2D t;  // a handle to a texture (an image in SPIR-V)</pre>
16522</div>
16523</div>
16524<div class="paragraph">
16525<p>Constructors can then be used to combine a sampler and a texture at the
16526point of making a texture lookup call:</p>
16527</div>
16528<div class="literalblock">
16529<div class="content">
16530<pre>texture(sampler2D(t, s), ...);</pre>
16531</div>
16532</div>
16533<div class="paragraph">
16534<p>Note, <code>layout()</code> information is omitted above for clarity of this feature.</p>
16535</div>
16536</div>
16537<div class="sect4">
16538<h5 id="_texture_buffers_uniform_texel_buffers">Texture Buffers (Uniform Texel Buffers)</h5>
16539<div class="paragraph">
16540<p>Texture buffers are declared in GLSL shader source using uniform
16541textureBuffer variables:</p>
16542</div>
16543<div class="literalblock">
16544<div class="content">
16545<pre>layout (set=m, binding=n) uniform textureBuffer myUniformTexelBuffer;</pre>
16546</div>
16547</div>
16548<div class="paragraph">
16549<p>Which maps to the following SPIR-V:</p>
16550</div>
16551<div class="literalblock">
16552<div class="content">
16553<pre>        ...
16554%1 = OpExtInstImport "GLSL.std.450"
16555        ...
16556        OpName %9 "myUniformTexelBuffer"
16557        OpDecorate %9 DescriptorSet m
16558        OpDecorate %9 Binding n
16559%2 = OpTypeVoid
16560%3 = OpTypeFunction %2
16561%6 = OpTypeFloat 32
16562%7 = OpTypeImage %6 Buffer 0 0 0 1 Unknown
16563%8 = OpTypePointer UniformConstant %7
16564%9 = OpVariable %8 UniformConstant
16565        ...</pre>
16566</div>
16567</div>
16568</div>
16569<div class="sect4">
16570<h5 id="_image_buffers_storage_texel_buffers">Image Buffers (Storage Texel Buffers)</h5>
16571<div class="paragraph">
16572<p>Image buffers are declared in GLSL shader source using uniform
16573imageBuffer variables:</p>
16574</div>
16575<div class="literalblock">
16576<div class="content">
16577<pre>layout (set=m, binding=n, r32f) uniform imageBuffer myStorageTexelBuffer;</pre>
16578</div>
16579</div>
16580<div class="paragraph">
16581<p>Which maps to the following SPIR-V:</p>
16582</div>
16583<div class="literalblock">
16584<div class="content">
16585<pre>        ...
16586%1 = OpExtInstImport "GLSL.std.450"
16587        ...
16588        OpName %9 "myStorageTexelBuffer"
16589        OpDecorate %9 DescriptorSet m
16590        OpDecorate %9 Binding n
16591%2 = OpTypeVoid
16592%3 = OpTypeFunction %2
16593%6 = OpTypeFloat 32
16594%7 = OpTypeImage %6 Buffer 0 0 0 2 R32f
16595%8 = OpTypePointer UniformConstant %7
16596%9 = OpVariable %8 UniformConstant
16597        ...</pre>
16598</div>
16599</div>
16600</div>
16601<div class="sect4">
16602<h5 id="_storage_buffers">Storage Buffers</h5>
16603<div class="paragraph">
16604<p>Storage buffers are declared in GLSL shader source using buffer storage
16605qualifier and block syntax:</p>
16606</div>
16607<div class="literalblock">
16608<div class="content">
16609<pre>layout (set=m, binding=n) buffer myStorageBuffer
16610{
16611    vec4 myElement[];
16612};</pre>
16613</div>
16614</div>
16615<div class="paragraph">
16616<p>Which maps to the following SPIR-V:</p>
16617</div>
16618<div class="literalblock">
16619<div class="content">
16620<pre>        ...
16621%1 = OpExtInstImport "GLSL.std.450"
16622        ...
16623        OpName %9 "myStorageBuffer"
16624        OpMemberName %9 0 "myElement"
16625        OpName %11 ""
16626        OpDecorate %8 ArrayStride 16
16627        OpMemberDecorate %9 0 Offset 0
16628        OpDecorate %9 BufferBlock
16629        OpDecorate %11 DescriptorSet m
16630        OpDecorate %11 Binding n
16631%2 = OpTypeVoid
16632%3 = OpTypeFunction %2
16633%6 = OpTypeFloat 32
16634%7 = OpTypeVector %6 4
16635%8 = OpTypeRuntimeArray %7
16636%9 = OpTypeStruct %8
16637%10 = OpTypePointer Uniform %9
16638%11 = OpVariable %10 Uniform
16639        ...</pre>
16640</div>
16641</div>
16642</div>
16643<div class="sect4">
16644<h5 id="_uniform_buffers">Uniform Buffers</h5>
16645<div class="paragraph">
16646<p>Uniform buffers are declared in GLSL shader source using the uniform storage
16647qualifier and block syntax:</p>
16648</div>
16649<div class="literalblock">
16650<div class="content">
16651<pre>layout (set=m, binding=n) uniform myUniformBuffer
16652{
16653    vec4 myElement[32];
16654};</pre>
16655</div>
16656</div>
16657<div class="paragraph">
16658<p>Which maps to the following SPIR-V:</p>
16659</div>
16660<div class="literalblock">
16661<div class="content">
16662<pre>        ...
16663%1 = OpExtInstImport "GLSL.std.450"
16664        ...
16665        OpName %11 "myUniformBuffer"
16666        OpMemberName %11 0 "myElement"
16667        OpName %13 ""
16668        OpDecorate %10 ArrayStride 16
16669        OpMemberDecorate %11 0 Offset 0
16670        OpDecorate %11 Block
16671        OpDecorate %13 DescriptorSet m
16672        OpDecorate %13 Binding n
16673%2 = OpTypeVoid
16674%3 = OpTypeFunction %2
16675%6 = OpTypeFloat 32
16676%7 = OpTypeVector %6 4
16677%8 = OpTypeInt 32 0
16678%9 = OpConstant %8 32
16679%10 = OpTypeArray %7 %9
16680%11 = OpTypeStruct %10
16681%12 = OpTypePointer Uniform %11
16682%13 = OpVariable %12 Uniform
16683        ...</pre>
16684</div>
16685</div>
16686</div>
16687<div class="sect4">
16688<h5 id="_subpass_inputs_2">Subpass Inputs</h5>
16689<div class="paragraph">
16690<p>Within a rendering pass, a subpass can write results to an output target
16691that can then be read by the next subpass as an input subpass.  The
16692"Subpass Input" feature regards the ability to read an output target.</p>
16693</div>
16694<div class="paragraph">
16695<p>Subpass inputs are read through a new set of types, available only
16696to fragment shaders:</p>
16697</div>
16698<div class="literalblock">
16699<div class="content">
16700<pre>subpassInput
16701subpassInputMS
16702isubpassInput
16703isubpassInputMS
16704usubpassInput
16705usubpassInputMS</pre>
16706</div>
16707</div>
16708<div class="paragraph">
16709<p>Unlike sampler and image objects, subpass inputs are implicitly addressed
16710by the fragment&#8217;s (<em>x</em>, <em>y</em>, <em>layer</em>) coordinate.</p>
16711</div>
16712<div class="paragraph">
16713<p>Input attachments are decorated with their input attachment index in
16714addition to descriptor set and binding numbers.</p>
16715</div>
16716<div class="literalblock">
16717<div class="content">
16718<pre>layout (input_attachment_index=i, set=m, binding=n) uniform subpassInput myInputAttachment;</pre>
16719</div>
16720</div>
16721<div class="paragraph">
16722<p>Which maps to the following SPIR-V:</p>
16723</div>
16724<div class="literalblock">
16725<div class="content">
16726<pre>        ...
16727%1 = OpExtInstImport "GLSL.std.450"
16728        ...
16729        OpName %9 "myInputAttachment"
16730        OpDecorate %9 DescriptorSet m
16731        OpDecorate %9 Binding n
16732        OpDecorate %9 InputAttachmentIndex i
16733%2 = OpTypeVoid
16734%3 = OpTypeFunction %2
16735%6 = OpTypeFloat 32
16736%7 = OpTypeImage %6 SubpassData 0 0 0 2 Unknown
16737%8 = OpTypePointer UniformConstant %7
16738%9 = OpVariable %8 UniformConstant
16739        ...</pre>
16740</div>
16741</div>
16742<div class="paragraph">
16743<p>An <em>input_attachment_index</em> of i selects the ith entry in the input pass
16744list. (See API specification for more information.)</p>
16745</div>
16746<div class="paragraph">
16747<p>These objects support reading the subpass input through the following
16748functions:</p>
16749</div>
16750<div class="literalblock">
16751<div class="content">
16752<pre>gvec4 subpassLoad(gsubpassInput   subpass);
16753gvec4 subpassLoad(gsubpassInputMS subpass, int sample);</pre>
16754</div>
16755</div>
16756</div>
16757</div>
16758<div class="sect3">
16759<h4 id="_mapping_variables">14.2.5. Mapping Variables</h4>
16760<div class="sect4">
16761<h5 id="_gl_fragcolor"><em>gl_FragColor</em></h5>
16762<div class="paragraph">
16763<p>The fragment-stage built-in <em>gl_FragColor</em>, which implies a broadcast to all
16764outputs, is not present in SPIR-V. Shaders where writing to <em>gl_FragColor</em>
16765is allowed can still write to it, but it only means to write to an output:</p>
16766</div>
16767<div class="ulist">
16768<ul>
16769<li>
16770<p>of the same type as <em>gl_FragColor</em></p>
16771</li>
16772<li>
16773<p>decorated with location 0</p>
16774</li>
16775<li>
16776<p>not decorated as a built-in variable.</p>
16777</li>
16778</ul>
16779</div>
16780<div class="paragraph">
16781<p>There is no implicit broadcast.</p>
16782</div>
16783</div>
16784<div class="sect4">
16785<h5 id="_vulkan_gl_vertexindex_and_gl_instanceindex">Vulkan <em>gl_VertexIndex</em> and <em>gl_InstanceIndex</em></h5>
16786<div class="paragraph">
16787<p>Adds two new built-in variables, <em>gl_VertexIndex</em> and <em>gl_InstanceIndex</em> to
16788replace the existing built-in variables <em>gl_VertexID</em> and <em>gl_InstanceID</em>.</p>
16789</div>
16790<div class="paragraph">
16791<p>In the situations where the indexing is relative to some base offset,
16792these built-in variables are defined, for Vulkan, to take on values as
16793follows:</p>
16794</div>
16795<div class="literalblock">
16796<div class="content">
16797<pre>gl_VertexIndex             base, base+1, base+2, ...
16798gl_InstanceIndex           base, base+1, base+2, ...</pre>
16799</div>
16800</div>
16801<div class="paragraph">
16802<p>Where it depends on the situation what the base actually is.</p>
16803</div>
16804</div>
16805<div class="sect4">
16806<h5 id="_storage_classes">Storage Classes:</h5>
16807<div class="literalblock">
16808<div class="content">
16809<pre>uniform sampler2D...;        -&gt; UniformConstant
16810uniform blockN { ... } ...;  -&gt; Uniform, with Block decoration
16811in / out variable            -&gt; Input/Output, possibly with block (below)
16812in / out block...            -&gt; Input/Output, with Block decoration
16813buffer  blockN { ... } ...;  -&gt; Uniform, with BufferBlock decoration
16814shared                       -&gt; Workgroup
16815&lt;normal global&gt;              -&gt; Private</pre>
16816</div>
16817</div>
16818<div class="literalblock">
16819<div class="content">
16820<pre>Vulkan Only: buffer  blockN { ... } ...;  -&gt; StorageBuffer, when requested
16821OpenGL Only: uniform variable (non-block) -&gt; UniformConstant
16822OpenGL Only: ... uniform atomic_uint ...  -&gt; AtomicCounter</pre>
16823</div>
16824</div>
16825</div>
16826<div class="sect4">
16827<h5 id="_inputoutput">Input/Output</h5>
16828<div class="paragraph">
16829<p>Mapping of input/output blocks or variables is the same for all versions
16830of GLSL or ESSL. To the extent variables or members are available in a
16831version, its location is as follows:</p>
16832</div>
16833<div class="paragraph">
16834<p>These are mapped to SPIR-V individual variables, with similarly spelled
16835built-in decorations (except as noted):</p>
16836</div>
16837<div class="paragraph">
16838<p>Any stage:</p>
16839</div>
16840<div class="literalblock">
16841<div class="content">
16842<pre>in gl_VertexIndex          (Vulkan only)
16843in gl_VertexID             (OpenGL only)
16844in gl_InstanceIndex        (Vulkan only)
16845in gl_InstanceID           (OpenGL only)
16846in gl_InvocationID
16847in gl_PatchVerticesIn      (PatchVertices)
16848in gl_PrimitiveIDIn        (PrimitiveID)
16849in/out gl_PrimitiveID      (in/out based only on storage qualifier)
16850in gl_TessCoord</pre>
16851</div>
16852</div>
16853<div class="literalblock">
16854<div class="content">
16855<pre>in/out gl_Layer
16856in/out gl_ViewportIndex</pre>
16857</div>
16858</div>
16859<div class="literalblock">
16860<div class="content">
16861<pre>patch in/out gl_TessLevelOuter  (uses Patch decoration)
16862patch in/out gl_TessLevelInner  (uses Patch decoration)</pre>
16863</div>
16864</div>
16865<div class="paragraph">
16866<p>Compute stage only:</p>
16867</div>
16868<div class="literalblock">
16869<div class="content">
16870<pre>in gl_NumWorkGroups
16871in gl_WorkGroupSize
16872in gl_WorkGroupID
16873in gl_LocalInvocationID
16874in gl_GlobalInvocationID
16875in gl_LocalInvocationIndex</pre>
16876</div>
16877</div>
16878<div class="paragraph">
16879<p>Fragment stage only:</p>
16880</div>
16881<div class="literalblock">
16882<div class="content">
16883<pre>in gl_FragCoord
16884in gl_FrontFacing
16885in gl_ClipDistance
16886in gl_CullDistance
16887in gl_PointCoord
16888in gl_SampleID
16889in gl_SamplePosition
16890in gl_HelperInvocation
16891out gl_FragDepth
16892in gl_SampleMaskIn        (SampleMask)
16893out gl_SampleMask         (in/out based only on storage qualifier)</pre>
16894</div>
16895</div>
16896<div class="paragraph">
16897<p>These are mapped to SPIR-V blocks, as implied by the pseudo code, with
16898the members decorated with similarly spelled built-in decorations:</p>
16899</div>
16900<div class="paragraph">
16901<p>Non-fragment stage:</p>
16902</div>
16903<div class="literalblock">
16904<div class="content">
16905<pre>in/out gl_PerVertex {   // some subset of these members will be used
16906    gl_Position
16907    gl_PointSize
16908    gl_ClipDistance
16909    gl_CullDistance
16910}                       // name of block is for debug only</pre>
16911</div>
16912</div>
16913<div class="paragraph">
16914<p>There is at most one input and one output block per stage in SPIR-V.
16915The subset and order of members will match between stages sharing an
16916interface.</p>
16917</div>
16918</div>
16919</div>
16920<div class="sect3">
16921<h4 id="_vulkan_only_mapping_of_precision_qualifiers">14.2.6. Vulkan Only: Mapping of Precision Qualifiers</h4>
16922<div class="literalblock">
16923<div class="content">
16924<pre>lowp     -&gt; RelaxedPrecision, on storage variable and operation
16925mediump  -&gt; RelaxedPrecision, on storage variable and operation
16926highp    -&gt; 32-bit, same as int or float</pre>
16927</div>
16928</div>
16929<div class="literalblock">
16930<div class="content">
16931<pre>portability tool/mode  -&gt; OpQuantizeToF16</pre>
16932</div>
16933</div>
16934</div>
16935<div class="sect3">
16936<h4 id="_mapping_of_precise">14.2.7. Mapping of <strong>precise</strong>:</h4>
16937<div class="literalblock">
16938<div class="content">
16939<pre>precise -&gt; NoContraction</pre>
16940</div>
16941</div>
16942</div>
16943<div class="sect3">
16944<h4 id="_opengl_mapping_of_atomic_uint_offset_layout_qualifier">14.2.8. OpenGL Mapping of <strong>atomic_uint</strong> <em>offset</em> layout qualifier</h4>
16945<div class="literalblock">
16946<div class="content">
16947<pre>offset         -&gt;  Offset (decoration)</pre>
16948</div>
16949</div>
16950</div>
16951<div class="sect3">
16952<h4 id="_mapping_of_images">14.2.9. Mapping of Images</h4>
16953<div class="literalblock">
16954<div class="content">
16955<pre>imageLoad()   -&gt; OpImageRead
16956imageStore()  -&gt; OpImageWrite
16957texelFetch()  -&gt; OpImageFetch
16958subpassInput  -&gt; OpTypeImage with Dim of SubpassData (Vulkan only)
16959subpassLoad() -&gt; OpImageRead                         (Vulkan only)</pre>
16960</div>
16961</div>
16962<div class="literalblock">
16963<div class="content">
16964<pre>imageAtomicXXX(params, data)  -&gt; %ptr = OpImageTexelPointer params
16965                                        OpAtomicXXX %ptr, data</pre>
16966</div>
16967</div>
16968<div class="literalblock">
16969<div class="content">
16970<pre>XXXQueryXXX(combined) -&gt; %image = OpImage combined
16971                                OpXXXQueryXXX %image</pre>
16972</div>
16973</div>
16974</div>
16975<div class="sect3">
16976<h4 id="_mapping_of_layouts">14.2.10. Mapping of Layouts</h4>
16977<div class="literalblock">
16978<div class="content">
16979<pre>std140/std430  -&gt;  explicit Offset, ArrayStride, and MatrixStride
16980                    Decoration on struct members
16981shared/packed  -&gt;  not allowed
16982&lt;default&gt;      -&gt;  not shared, but std140 or std430
16983xfb_offset     -&gt;  Offset Decoration on the object or struct member
16984xfb_buffer     -&gt;  XfbBuffer Decoration on the object
16985xfb_stride     -&gt;  XfbStride Decoration on the object
16986any xfb_*      -&gt;  the Xfb Execution Mode is set
16987captured XFB   -&gt;  has both XfbBuffer and Offset
16988non-captured   -&gt;  lacking XfbBuffer or Offset</pre>
16989</div>
16990</div>
16991<div class="literalblock">
16992<div class="content">
16993<pre>max_vertices   -&gt;  OutputVertices</pre>
16994</div>
16995</div>
16996</div>
16997<div class="sect3">
16998<h4 id="_mapping_of_barriers">14.2.11. Mapping of barriers</h4>
16999<div class="literalblock">
17000<div class="content">
17001<pre>barrier() (compute) -&gt; OpControlBarrier(/*Execution*/Workgroup,
17002                                        /*Memory*/Workgroup,
17003                                        /*Semantics*/AcquireRelease |
17004                                                    WorkgroupMemory)</pre>
17005</div>
17006</div>
17007<div class="literalblock">
17008<div class="content">
17009<pre>barrier() (tess control) -&gt; OpControlBarrier(/*Execution*/Workgroup,
17010                                            /*Memory*/Invocation,
17011                                            /*Semantics*/None)</pre>
17012</div>
17013</div>
17014<div class="literalblock">
17015<div class="content">
17016<pre>memoryBarrier() -&gt; OpMemoryBarrier(/*Memory*/Device,
17017                                    /*Semantics*/AcquireRelease |
17018                                                UniformMemory |
17019                                                WorkgroupMemory |
17020                                                ImageMemory)</pre>
17021</div>
17022</div>
17023<div class="literalblock">
17024<div class="content">
17025<pre>memoryBarrierBuffer() -&gt; OpMemoryBarrier(/*Memory*/Device,
17026                                        /*Semantics*/AcquireRelease |
17027                                                    UniformMemory)</pre>
17028</div>
17029</div>
17030<div class="literalblock">
17031<div class="content">
17032<pre>memoryBarrierShared() -&gt; OpMemoryBarrier(/*Memory*/Device,
17033                                        /*Semantics*/AcquireRelease |
17034                                                    WorkgroupMemory)</pre>
17035</div>
17036</div>
17037<div class="literalblock">
17038<div class="content">
17039<pre>memoryBarrierImage() -&gt; OpMemoryBarrier(/*Memory*/Device,
17040                                        /*Semantics*/AcquireRelease |
17041                                                    ImageMemory)</pre>
17042</div>
17043</div>
17044<div class="literalblock">
17045<div class="content">
17046<pre>groupMemoryBarrier() -&gt; OpMemoryBarrier(/*Memory*/Workgroup,
17047                                        /*Semantics*/AcquireRelease |
17048                                                    UniformMemory |
17049                                                    WorkgroupMemory |
17050                                                    ImageMemory)</pre>
17051</div>
17052</div>
17053</div>
17054<div class="sect3">
17055<h4 id="_mapping_of_atomics">14.2.12. Mapping of atomics</h4>
17056<div class="literalblock">
17057<div class="content">
17058<pre>all atomic builtin functions -&gt; Semantics = None(Relaxed)</pre>
17059</div>
17060</div>
17061<div class="literalblock">
17062<div class="content">
17063<pre>atomicExchange()             -&gt; OpAtomicExchange
17064imageAtomicExchange()        -&gt; OpAtomicExchange
17065atomicCompSwap()             -&gt; OpAtomicCompareExchange
17066imageAtomicCompSwap()        -&gt; OpAtomicCompareExchange
17067N/A                          -&gt; OpAtomicCompareExchangeWeak</pre>
17068</div>
17069</div>
17070</div>
17071<div class="sect3">
17072<h4 id="_opengl_only_mapping_of_atomics">14.2.13. OpenGL Only: Mapping of Atomics</h4>
17073<div class="literalblock">
17074<div class="content">
17075<pre>atomicCounterIncrement -&gt; OpAtomicIIncrement
17076atomicCounterDecrement -&gt; OpAtomicIDecrement
17077atomicCounter          -&gt; OpAtomicLoad</pre>
17078</div>
17079</div>
17080</div>
17081<div class="sect3">
17082<h4 id="_mapping_of_other_instructions">14.2.14. Mapping of other instructions</h4>
17083<div class="literalblock">
17084<div class="content">
17085<pre>%     -&gt; OpUMod/OpSMod
17086mod() -&gt; OpFMod
17087N/A   -&gt; OpSRem/OpFRem</pre>
17088</div>
17089</div>
17090<div class="literalblock">
17091<div class="content">
17092<pre>pack/unpack (conversion)    -&gt; pack/unpack in GLSL extended instructions
17093pack/unpack (no conversion) -&gt; OpBitcast</pre>
17094</div>
17095</div>
17096</div>
17097</div>
17098</div>
17099</div>
17100</div>
17101<div id="footer">
17102<div id="footer-text">
17103Version 3.20.6<br>
17104Last updated 2019-07-10 14:42:53 MDT
17105</div>
17106</div>
17107
17108<script src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.4/MathJax.js?config=TeX-MML-AM_HTMLorMML"></script>
17109</body>
17110</html>