1 /*
2 * Copyright © Microsoft Corporation
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * on the rights to use, copy, modify, merge, publish, distribute, sub
8 * license, and/or sell copies of the Software, and to permit persons to whom
9 * the Software is furnished to do so, subject to the following conditions:
10 *
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
13 * Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
19 * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
20 * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
21 * USE OR OTHER DEALINGS IN THE SOFTWARE.
22 */
23
24 #include "d3d12_nir_passes.h"
25
26 #include "nir_builder.h"
27 #include "nir_builtin_builder.h"
28
29 static enum pipe_format
get_input_target_format(nir_variable * var,const void * options)30 get_input_target_format(nir_variable *var, const void *options)
31 {
32 enum pipe_format *target_formats = (enum pipe_format *)options;
33 return target_formats[var->data.driver_location];
34 }
35
36 static bool
lower_vs_vertex_conversion_filter(const nir_instr * instr,const void * options)37 lower_vs_vertex_conversion_filter(const nir_instr *instr, const void *options)
38 {
39 if (instr->type != nir_instr_type_intrinsic)
40 return false;
41
42 nir_intrinsic_instr *inst = nir_instr_as_intrinsic(instr);
43 if (inst->intrinsic != nir_intrinsic_load_deref)
44 return false;
45
46 nir_variable *var = nir_intrinsic_get_var(inst, 0);
47 return (var->data.mode == nir_var_shader_in) &&
48 (get_input_target_format(var, options) != PIPE_FORMAT_NONE);
49 }
50
51 typedef nir_ssa_def *
52 (*shift_right_func)(nir_builder *build, nir_ssa_def *src0, nir_ssa_def *src1);
53
54 /* decoding the signed vs unsigned scaled format is handled
55 * by applying the signed or unsigned shift right function
56 * accordingly */
57 static nir_ssa_def *
from_10_10_10_2_scaled(nir_builder * b,nir_ssa_def * src,nir_ssa_def * lshift,shift_right_func shr)58 from_10_10_10_2_scaled(nir_builder *b, nir_ssa_def *src,
59 nir_ssa_def *lshift, shift_right_func shr)
60 {
61 nir_ssa_def *rshift = nir_imm_ivec4(b, 22, 22, 22, 30);
62 return nir_i2f32(b, shr(b, nir_ishl(b, src, lshift), rshift));
63 }
64
65 static nir_ssa_def *
from_10_10_10_2_snorm(nir_builder * b,nir_ssa_def * src,nir_ssa_def * lshift)66 from_10_10_10_2_snorm(nir_builder *b, nir_ssa_def *src, nir_ssa_def *lshift)
67 {
68 nir_ssa_def *split = from_10_10_10_2_scaled(b, src, lshift, nir_ishr);
69 nir_ssa_def *scale_rgb = nir_imm_vec4(b,
70 1.0f / 0x1ff,
71 1.0f / 0x1ff,
72 1.0f / 0x1ff,
73 1.0f);
74 return nir_fmul(b, split, scale_rgb);
75 }
76
77 static nir_ssa_def *
from_10_10_10_2_unorm(nir_builder * b,nir_ssa_def * src,nir_ssa_def * lshift)78 from_10_10_10_2_unorm(nir_builder *b, nir_ssa_def *src, nir_ssa_def *lshift)
79 {
80 nir_ssa_def *split = from_10_10_10_2_scaled(b, src, lshift, nir_ushr);
81 nir_ssa_def *scale_rgb = nir_imm_vec4(b,
82 1.0f / 0x3ff,
83 1.0f / 0x3ff,
84 1.0f / 0x3ff,
85 1.0f / 3.0f);
86 return nir_fmul(b, split, scale_rgb);
87 }
88
89 inline static nir_ssa_def *
lshift_rgba(nir_builder * b)90 lshift_rgba(nir_builder *b)
91 {
92 return nir_imm_ivec4(b, 22, 12, 2, 0);
93 }
94
95 inline static nir_ssa_def *
lshift_bgra(nir_builder * b)96 lshift_bgra(nir_builder *b)
97 {
98 return nir_imm_ivec4(b, 2, 12, 22, 0);
99 }
100
101 static nir_ssa_def *
lower_vs_vertex_conversion_impl(nir_builder * b,nir_instr * instr,void * options)102 lower_vs_vertex_conversion_impl(nir_builder *b, nir_instr *instr, void *options)
103 {
104 nir_intrinsic_instr *intr = nir_instr_as_intrinsic(instr);
105 nir_variable *var = nir_intrinsic_get_var(intr, 0);
106 enum pipe_format fmt = get_input_target_format(var, options);
107
108 if (!util_format_has_alpha(fmt)) {
109 /* these formats need the alpha channel replaced with 1: */
110 assert(fmt == PIPE_FORMAT_R8G8B8_SINT ||
111 fmt == PIPE_FORMAT_R8G8B8_UINT ||
112 fmt == PIPE_FORMAT_R16G16B16_SINT ||
113 fmt == PIPE_FORMAT_R16G16B16_UINT);
114 return nir_vector_insert_imm(b, &intr->dest.ssa, nir_imm_int(b, 1), 3);
115 } else {
116 nir_ssa_def *src = nir_channel(b, &intr->dest.ssa, 0);
117
118 switch (fmt) {
119 case PIPE_FORMAT_R10G10B10A2_SNORM:
120 return from_10_10_10_2_snorm(b, src, lshift_rgba(b));
121 case PIPE_FORMAT_B10G10R10A2_SNORM:
122 return from_10_10_10_2_snorm(b, src, lshift_bgra(b));
123 case PIPE_FORMAT_B10G10R10A2_UNORM:
124 return from_10_10_10_2_unorm(b, src, lshift_bgra(b));
125 case PIPE_FORMAT_R10G10B10A2_SSCALED:
126 return from_10_10_10_2_scaled(b, src, lshift_rgba(b), nir_ishr);
127 case PIPE_FORMAT_B10G10R10A2_SSCALED:
128 return from_10_10_10_2_scaled(b, src, lshift_bgra(b), nir_ishr);
129 case PIPE_FORMAT_R10G10B10A2_USCALED:
130 return from_10_10_10_2_scaled(b, src, lshift_rgba(b), nir_ushr);
131 case PIPE_FORMAT_B10G10R10A2_USCALED:
132 return from_10_10_10_2_scaled(b, src, lshift_bgra(b), nir_ushr);
133
134 default:
135 unreachable("Unsupported emulated vertex format");
136 }
137 }
138 }
139
140 /* Lower emulated vertex attribute input
141 * The vertex attributes are passed as R32_UINT that needs to be converted
142 * to one of the RGB10A2 formats that need to be emulated.
143 *
144 * @param target_formats contains the per attribute format to convert to
145 * or PIPE_FORMAT_NONE if no conversion is needed
146 */
147 bool
d3d12_nir_lower_vs_vertex_conversion(nir_shader * s,enum pipe_format target_formats[])148 d3d12_nir_lower_vs_vertex_conversion(nir_shader *s,
149 enum pipe_format target_formats[])
150 {
151 assert(s->info.stage == MESA_SHADER_VERTEX);
152
153 bool result =
154 nir_shader_lower_instructions(s,
155 lower_vs_vertex_conversion_filter,
156 lower_vs_vertex_conversion_impl,
157 target_formats);
158 return result;
159 }
160