1<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" 2 "http://www.w3.org/TR/html4/strict.dtd"> 3<html> 4<head> 5 <title>LLVM's Analysis and Transform Passes</title> 6 <link rel="stylesheet" href="llvm.css" type="text/css"> 7 <meta http-equiv="Content-Type" content="text/html; charset=UTF-8"> 8</head> 9<body> 10 11<!-- 12 13If Passes.html is up to date, the following "one-liner" should print 14an empty diff. 15 16egrep -e '^<tr><td><a href="#.*">-.*</a></td><td>.*</td></tr>$' \ 17 -e '^ <a name=".*">.*</a>$' < Passes.html >html; \ 18perl >help <<'EOT' && diff -u help html; rm -f help html 19open HTML, "<Passes.html" or die "open: Passes.html: $!\n"; 20while (<HTML>) { 21 m:^<tr><td><a href="#(.*)">-.*</a></td><td>.*</td></tr>$: or next; 22 $order{$1} = sprintf("%03d", 1 + int %order); 23} 24open HELP, "../Release/bin/opt -help|" or die "open: opt -help: $!\n"; 25while (<HELP>) { 26 m:^ -([^ ]+) +- (.*)$: or next; 27 my $o = $order{$1}; 28 $o = "000" unless defined $o; 29 push @x, "$o<tr><td><a href=\"#$1\">-$1</a></td><td>$2</td></tr>\n"; 30 push @y, "$o <a name=\"$1\">-$1: $2</a>\n"; 31} 32@x = map { s/^\d\d\d//; $_ } sort @x; 33@y = map { s/^\d\d\d//; $_ } sort @y; 34print @x, @y; 35EOT 36 37This (real) one-liner can also be helpful when converting comments to HTML: 38 39perl -e '$/ = undef; for (split(/\n/, <>)) { s:^ *///? ?::; print " <p>\n" if !$on && $_ =~ /\S/; print " </p>\n" if $on && $_ =~ /^\s*$/; print " $_\n"; $on = ($_ =~ /\S/); } print " </p>\n" if $on' 40 41 --> 42 43<h1>LLVM's Analysis and Transform Passes</h1> 44 45<ol> 46 <li><a href="#intro">Introduction</a></li> 47 <li><a href="#analyses">Analysis Passes</a> 48 <li><a href="#transforms">Transform Passes</a></li> 49 <li><a href="#utilities">Utility Passes</a></li> 50</ol> 51 52<div class="doc_author"> 53 <p>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a> 54 and Gordon Henriksen</p> 55</div> 56 57<!-- ======================================================================= --> 58<h2><a name="intro">Introduction</a></h2> 59<div> 60 <p>This document serves as a high level summary of the optimization features 61 that LLVM provides. Optimizations are implemented as Passes that traverse some 62 portion of a program to either collect information or transform the program. 63 The table below divides the passes that LLVM provides into three categories. 64 Analysis passes compute information that other passes can use or for debugging 65 or program visualization purposes. Transform passes can use (or invalidate) 66 the analysis passes. Transform passes all mutate the program in some way. 67 Utility passes provides some utility but don't otherwise fit categorization. 68 For example passes to extract functions to bitcode or write a module to 69 bitcode are neither analysis nor transform passes. 70 <p>The table below provides a quick summary of each pass and links to the more 71 complete pass description later in the document.</p> 72 73<table> 74<tr><th colspan="2"><b>ANALYSIS PASSES</b></th></tr> 75<tr><th>Option</th><th>Name</th></tr> 76<tr><td><a href="#aa-eval">-aa-eval</a></td><td>Exhaustive Alias Analysis Precision Evaluator</td></tr> 77<tr><td><a href="#basicaa">-basicaa</a></td><td>Basic Alias Analysis (stateless AA impl)</td></tr> 78<tr><td><a href="#basiccg">-basiccg</a></td><td>Basic CallGraph Construction</td></tr> 79<tr><td><a href="#count-aa">-count-aa</a></td><td>Count Alias Analysis Query Responses</td></tr> 80<tr><td><a href="#debug-aa">-debug-aa</a></td><td>AA use debugger</td></tr> 81<tr><td><a href="#domfrontier">-domfrontier</a></td><td>Dominance Frontier Construction</td></tr> 82<tr><td><a href="#domtree">-domtree</a></td><td>Dominator Tree Construction</td></tr> 83<tr><td><a href="#dot-callgraph">-dot-callgraph</a></td><td>Print Call Graph to 'dot' file</td></tr> 84<tr><td><a href="#dot-cfg">-dot-cfg</a></td><td>Print CFG of function to 'dot' file</td></tr> 85<tr><td><a href="#dot-cfg-only">-dot-cfg-only</a></td><td>Print CFG of function to 'dot' file (with no function bodies)</td></tr> 86<tr><td><a href="#dot-dom">-dot-dom</a></td><td>Print dominance tree of function to 'dot' file</td></tr> 87<tr><td><a href="#dot-dom-only">-dot-dom-only</a></td><td>Print dominance tree of function to 'dot' file (with no function bodies)</td></tr> 88<tr><td><a href="#dot-postdom">-dot-postdom</a></td><td>Print postdominance tree of function to 'dot' file</td></tr> 89<tr><td><a href="#dot-postdom-only">-dot-postdom-only</a></td><td>Print postdominance tree of function to 'dot' file (with no function bodies)</td></tr> 90<tr><td><a href="#globalsmodref-aa">-globalsmodref-aa</a></td><td>Simple mod/ref analysis for globals</td></tr> 91<tr><td><a href="#instcount">-instcount</a></td><td>Counts the various types of Instructions</td></tr> 92<tr><td><a href="#intervals">-intervals</a></td><td>Interval Partition Construction</td></tr> 93<tr><td><a href="#iv-users">-iv-users</a></td><td>Induction Variable Users</td></tr> 94<tr><td><a href="#lazy-value-info">-lazy-value-info</a></td><td>Lazy Value Information Analysis</td></tr> 95<tr><td><a href="#lda">-lda</a></td><td>Loop Dependence Analysis</td></tr> 96<tr><td><a href="#libcall-aa">-libcall-aa</a></td><td>LibCall Alias Analysis</td></tr> 97<tr><td><a href="#lint">-lint</a></td><td>Statically lint-checks LLVM IR</td></tr> 98<tr><td><a href="#loops">-loops</a></td><td>Natural Loop Information</td></tr> 99<tr><td><a href="#memdep">-memdep</a></td><td>Memory Dependence Analysis</td></tr> 100<tr><td><a href="#module-debuginfo">-module-debuginfo</a></td><td>Decodes module-level debug info</td></tr> 101<tr><td><a href="#no-aa">-no-aa</a></td><td>No Alias Analysis (always returns 'may' alias)</td></tr> 102<tr><td><a href="#no-profile">-no-profile</a></td><td>No Profile Information</td></tr> 103<tr><td><a href="#postdomfrontier">-postdomfrontier</a></td><td>Post-Dominance Frontier Construction</td></tr> 104<tr><td><a href="#postdomtree">-postdomtree</a></td><td>Post-Dominator Tree Construction</td></tr> 105<tr><td><a href="#print-alias-sets">-print-alias-sets</a></td><td>Alias Set Printer</td></tr> 106<tr><td><a href="#print-callgraph">-print-callgraph</a></td><td>Print a call graph</td></tr> 107<tr><td><a href="#print-callgraph-sccs">-print-callgraph-sccs</a></td><td>Print SCCs of the Call Graph</td></tr> 108<tr><td><a href="#print-cfg-sccs">-print-cfg-sccs</a></td><td>Print SCCs of each function CFG</td></tr> 109<tr><td><a href="#print-dbginfo">-print-dbginfo</a></td><td>Print debug info in human readable form</td></tr> 110<tr><td><a href="#print-dom-info">-print-dom-info</a></td><td>Dominator Info Printer</td></tr> 111<tr><td><a href="#print-externalfnconstants">-print-externalfnconstants</a></td><td>Print external fn callsites passed constants</td></tr> 112<tr><td><a href="#print-function">-print-function</a></td><td>Print function to stderr</td></tr> 113<tr><td><a href="#print-module">-print-module</a></td><td>Print module to stderr</td></tr> 114<tr><td><a href="#print-used-types">-print-used-types</a></td><td>Find Used Types</td></tr> 115<tr><td><a href="#profile-estimator">-profile-estimator</a></td><td>Estimate profiling information</td></tr> 116<tr><td><a href="#profile-loader">-profile-loader</a></td><td>Load profile information from llvmprof.out</td></tr> 117<tr><td><a href="#profile-verifier">-profile-verifier</a></td><td>Verify profiling information</td></tr> 118<tr><td><a href="#regions">-regions</a></td><td>Detect single entry single exit regions</td></tr> 119<tr><td><a href="#scalar-evolution">-scalar-evolution</a></td><td>Scalar Evolution Analysis</td></tr> 120<tr><td><a href="#scev-aa">-scev-aa</a></td><td>ScalarEvolution-based Alias Analysis</td></tr> 121<tr><td><a href="#targetdata">-targetdata</a></td><td>Target Data Layout</td></tr> 122 123 124<tr><th colspan="2"><b>TRANSFORM PASSES</b></th></tr> 125<tr><th>Option</th><th>Name</th></tr> 126<tr><td><a href="#adce">-adce</a></td><td>Aggressive Dead Code Elimination</td></tr> 127<tr><td><a href="#always-inline">-always-inline</a></td><td>Inliner for always_inline functions</td></tr> 128<tr><td><a href="#argpromotion">-argpromotion</a></td><td>Promote 'by reference' arguments to scalars</td></tr> 129<tr><td><a href="#block-placement">-block-placement</a></td><td>Profile Guided Basic Block Placement</td></tr> 130<tr><td><a href="#break-crit-edges">-break-crit-edges</a></td><td>Break critical edges in CFG</td></tr> 131<tr><td><a href="#codegenprepare">-codegenprepare</a></td><td>Optimize for code generation</td></tr> 132<tr><td><a href="#constmerge">-constmerge</a></td><td>Merge Duplicate Global Constants</td></tr> 133<tr><td><a href="#constprop">-constprop</a></td><td>Simple constant propagation</td></tr> 134<tr><td><a href="#dce">-dce</a></td><td>Dead Code Elimination</td></tr> 135<tr><td><a href="#deadargelim">-deadargelim</a></td><td>Dead Argument Elimination</td></tr> 136<tr><td><a href="#deadtypeelim">-deadtypeelim</a></td><td>Dead Type Elimination</td></tr> 137<tr><td><a href="#die">-die</a></td><td>Dead Instruction Elimination</td></tr> 138<tr><td><a href="#dse">-dse</a></td><td>Dead Store Elimination</td></tr> 139<tr><td><a href="#functionattrs">-functionattrs</a></td><td>Deduce function attributes</td></tr> 140<tr><td><a href="#globaldce">-globaldce</a></td><td>Dead Global Elimination</td></tr> 141<tr><td><a href="#globalopt">-globalopt</a></td><td>Global Variable Optimizer</td></tr> 142<tr><td><a href="#gvn">-gvn</a></td><td>Global Value Numbering</td></tr> 143<tr><td><a href="#indvars">-indvars</a></td><td>Canonicalize Induction Variables</td></tr> 144<tr><td><a href="#inline">-inline</a></td><td>Function Integration/Inlining</td></tr> 145<tr><td><a href="#insert-edge-profiling">-insert-edge-profiling</a></td><td>Insert instrumentation for edge profiling</td></tr> 146<tr><td><a href="#insert-optimal-edge-profiling">-insert-optimal-edge-profiling</a></td><td>Insert optimal instrumentation for edge profiling</td></tr> 147<tr><td><a href="#instcombine">-instcombine</a></td><td>Combine redundant instructions</td></tr> 148<tr><td><a href="#internalize">-internalize</a></td><td>Internalize Global Symbols</td></tr> 149<tr><td><a href="#ipconstprop">-ipconstprop</a></td><td>Interprocedural constant propagation</td></tr> 150<tr><td><a href="#ipsccp">-ipsccp</a></td><td>Interprocedural Sparse Conditional Constant Propagation</td></tr> 151<tr><td><a href="#jump-threading">-jump-threading</a></td><td>Jump Threading</td></tr> 152<tr><td><a href="#lcssa">-lcssa</a></td><td>Loop-Closed SSA Form Pass</td></tr> 153<tr><td><a href="#licm">-licm</a></td><td>Loop Invariant Code Motion</td></tr> 154<tr><td><a href="#loop-deletion">-loop-deletion</a></td><td>Delete dead loops</td></tr> 155<tr><td><a href="#loop-extract">-loop-extract</a></td><td>Extract loops into new functions</td></tr> 156<tr><td><a href="#loop-extract-single">-loop-extract-single</a></td><td>Extract at most one loop into a new function</td></tr> 157<tr><td><a href="#loop-reduce">-loop-reduce</a></td><td>Loop Strength Reduction</td></tr> 158<tr><td><a href="#loop-rotate">-loop-rotate</a></td><td>Rotate Loops</td></tr> 159<tr><td><a href="#loop-simplify">-loop-simplify</a></td><td>Canonicalize natural loops</td></tr> 160<tr><td><a href="#loop-unroll">-loop-unroll</a></td><td>Unroll loops</td></tr> 161<tr><td><a href="#loop-unswitch">-loop-unswitch</a></td><td>Unswitch loops</td></tr> 162<tr><td><a href="#loweratomic">-loweratomic</a></td><td>Lower atomic intrinsics to non-atomic form</td></tr> 163<tr><td><a href="#lowerinvoke">-lowerinvoke</a></td><td>Lower invoke and unwind, for unwindless code generators</td></tr> 164<tr><td><a href="#lowersetjmp">-lowersetjmp</a></td><td>Lower Set Jump</td></tr> 165<tr><td><a href="#lowerswitch">-lowerswitch</a></td><td>Lower SwitchInst's to branches</td></tr> 166<tr><td><a href="#mem2reg">-mem2reg</a></td><td>Promote Memory to Register</td></tr> 167<tr><td><a href="#memcpyopt">-memcpyopt</a></td><td>MemCpy Optimization</td></tr> 168<tr><td><a href="#mergefunc">-mergefunc</a></td><td>Merge Functions</td></tr> 169<tr><td><a href="#mergereturn">-mergereturn</a></td><td>Unify function exit nodes</td></tr> 170<tr><td><a href="#partial-inliner">-partial-inliner</a></td><td>Partial Inliner</td></tr> 171<tr><td><a href="#prune-eh">-prune-eh</a></td><td>Remove unused exception handling info</td></tr> 172<tr><td><a href="#reassociate">-reassociate</a></td><td>Reassociate expressions</td></tr> 173<tr><td><a href="#reg2mem">-reg2mem</a></td><td>Demote all values to stack slots</td></tr> 174<tr><td><a href="#scalarrepl">-scalarrepl</a></td><td>Scalar Replacement of Aggregates (DT)</td></tr> 175<tr><td><a href="#sccp">-sccp</a></td><td>Sparse Conditional Constant Propagation</td></tr> 176<tr><td><a href="#simplify-libcalls">-simplify-libcalls</a></td><td>Simplify well-known library calls</td></tr> 177<tr><td><a href="#simplifycfg">-simplifycfg</a></td><td>Simplify the CFG</td></tr> 178<tr><td><a href="#sink">-sink</a></td><td>Code sinking</td></tr> 179<tr><td><a href="#sretpromotion">-sretpromotion</a></td><td>Promote sret arguments to multiple ret values</td></tr> 180<tr><td><a href="#strip">-strip</a></td><td>Strip all symbols from a module</td></tr> 181<tr><td><a href="#strip-dead-debug-info">-strip-dead-debug-info</a></td><td>Strip debug info for unused symbols</td></tr> 182<tr><td><a href="#strip-dead-prototypes">-strip-dead-prototypes</a></td><td>Strip Unused Function Prototypes</td></tr> 183<tr><td><a href="#strip-debug-declare">-strip-debug-declare</a></td><td>Strip all llvm.dbg.declare intrinsics</td></tr> 184<tr><td><a href="#strip-nondebug">-strip-nondebug</a></td><td>Strip all symbols, except dbg symbols, from a module</td></tr> 185<tr><td><a href="#tailcallelim">-tailcallelim</a></td><td>Tail Call Elimination</td></tr> 186<tr><td><a href="#tailduplicate">-tailduplicate</a></td><td>Tail Duplication</td></tr> 187 188 189<tr><th colspan="2"><b>UTILITY PASSES</b></th></tr> 190<tr><th>Option</th><th>Name</th></tr> 191<tr><td><a href="#deadarghaX0r">-deadarghaX0r</a></td><td>Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</td></tr> 192<tr><td><a href="#extract-blocks">-extract-blocks</a></td><td>Extract Basic Blocks From Module (for bugpoint use)</td></tr> 193<tr><td><a href="#instnamer">-instnamer</a></td><td>Assign names to anonymous instructions</td></tr> 194<tr><td><a href="#preverify">-preverify</a></td><td>Preliminary module verification</td></tr> 195<tr><td><a href="#verify">-verify</a></td><td>Module Verifier</td></tr> 196<tr><td><a href="#view-cfg">-view-cfg</a></td><td>View CFG of function</td></tr> 197<tr><td><a href="#view-cfg-only">-view-cfg-only</a></td><td>View CFG of function (with no function bodies)</td></tr> 198<tr><td><a href="#view-dom">-view-dom</a></td><td>View dominance tree of function</td></tr> 199<tr><td><a href="#view-dom-only">-view-dom-only</a></td><td>View dominance tree of function (with no function bodies)</td></tr> 200<tr><td><a href="#view-postdom">-view-postdom</a></td><td>View postdominance tree of function</td></tr> 201<tr><td><a href="#view-postdom-only">-view-postdom-only</a></td><td>View postdominance tree of function (with no function bodies)</td></tr> 202</table> 203 204</div> 205 206<!-- ======================================================================= --> 207<h2><a name="analyses">Analysis Passes</a></h2> 208<div> 209 <p>This section describes the LLVM Analysis Passes.</p> 210 211<!-------------------------------------------------------------------------- --> 212<h3> 213 <a name="aa-eval">-aa-eval: Exhaustive Alias Analysis Precision Evaluator</a> 214</h3> 215<div> 216 <p>This is a simple N^2 alias analysis accuracy evaluator. 217 Basically, for each function in the program, it simply queries to see how the 218 alias analysis implementation answers alias queries between each pair of 219 pointers in the function.</p> 220 221 <p>This is inspired and adapted from code by: Naveen Neelakantam, Francesco 222 Spadini, and Wojciech Stryjewski.</p> 223</div> 224 225<!-------------------------------------------------------------------------- --> 226<h3> 227 <a name="basicaa">-basicaa: Basic Alias Analysis (stateless AA impl)</a> 228</h3> 229<div> 230 <p> 231 This is the default implementation of the Alias Analysis interface 232 that simply implements a few identities (two different globals cannot alias, 233 etc), but otherwise does no analysis. 234 </p> 235</div> 236 237<!-------------------------------------------------------------------------- --> 238<h3> 239 <a name="basiccg">-basiccg: Basic CallGraph Construction</a> 240</h3> 241<div> 242 <p>Yet to be written.</p> 243</div> 244 245<!-------------------------------------------------------------------------- --> 246<h3> 247 <a name="count-aa">-count-aa: Count Alias Analysis Query Responses</a> 248</h3> 249<div> 250 <p> 251 A pass which can be used to count how many alias queries 252 are being made and how the alias analysis implementation being used responds. 253 </p> 254</div> 255 256<!-------------------------------------------------------------------------- --> 257<h3> 258 <a name="debug-aa">-debug-aa: AA use debugger</a> 259</h3> 260<div> 261 <p> 262 This simple pass checks alias analysis users to ensure that if they 263 create a new value, they do not query AA without informing it of the value. 264 It acts as a shim over any other AA pass you want. 265 </p> 266 267 <p> 268 Yes keeping track of every value in the program is expensive, but this is 269 a debugging pass. 270 </p> 271</div> 272 273<!-------------------------------------------------------------------------- --> 274<h3> 275 <a name="domfrontier">-domfrontier: Dominance Frontier Construction</a> 276</h3> 277<div> 278 <p> 279 This pass is a simple dominator construction algorithm for finding forward 280 dominator frontiers. 281 </p> 282</div> 283 284<!-------------------------------------------------------------------------- --> 285<h3> 286 <a name="domtree">-domtree: Dominator Tree Construction</a> 287</h3> 288<div> 289 <p> 290 This pass is a simple dominator construction algorithm for finding forward 291 dominators. 292 </p> 293</div> 294 295<!-------------------------------------------------------------------------- --> 296<h3> 297 <a name="dot-callgraph">-dot-callgraph: Print Call Graph to 'dot' file</a> 298</h3> 299<div> 300 <p> 301 This pass, only available in <code>opt</code>, prints the call graph into a 302 <code>.dot</code> graph. This graph can then be processed with the "dot" tool 303 to convert it to postscript or some other suitable format. 304 </p> 305</div> 306 307<!-------------------------------------------------------------------------- --> 308<h3> 309 <a name="dot-cfg">-dot-cfg: Print CFG of function to 'dot' file</a> 310</h3> 311<div> 312 <p> 313 This pass, only available in <code>opt</code>, prints the control flow graph 314 into a <code>.dot</code> graph. This graph can then be processed with the 315 "dot" tool to convert it to postscript or some other suitable format. 316 </p> 317</div> 318 319<!-------------------------------------------------------------------------- --> 320<h3> 321 <a name="dot-cfg-only">-dot-cfg-only: Print CFG of function to 'dot' file (with no function bodies)</a> 322</h3> 323<div> 324 <p> 325 This pass, only available in <code>opt</code>, prints the control flow graph 326 into a <code>.dot</code> graph, omitting the function bodies. This graph can 327 then be processed with the "dot" tool to convert it to postscript or some 328 other suitable format. 329 </p> 330</div> 331 332<!-------------------------------------------------------------------------- --> 333<h3> 334 <a name="dot-dom">-dot-dom: Print dominance tree of function to 'dot' file</a> 335</h3> 336<div> 337 <p> 338 This pass, only available in <code>opt</code>, prints the dominator tree 339 into a <code>.dot</code> graph. This graph can then be processed with the 340 "dot" tool to convert it to postscript or some other suitable format. 341 </p> 342</div> 343 344<!-------------------------------------------------------------------------- --> 345<h3> 346 <a name="dot-dom-only">-dot-dom-only: Print dominance tree of function to 'dot' file (with no function bodies)</a> 347</h3> 348<div> 349 <p> 350 This pass, only available in <code>opt</code>, prints the dominator tree 351 into a <code>.dot</code> graph, omitting the function bodies. This graph can 352 then be processed with the "dot" tool to convert it to postscript or some 353 other suitable format. 354 </p> 355</div> 356 357<!-------------------------------------------------------------------------- --> 358<h3> 359 <a name="dot-postdom">-dot-postdom: Print postdominance tree of function to 'dot' file</a> 360</h3> 361<div> 362 <p> 363 This pass, only available in <code>opt</code>, prints the post dominator tree 364 into a <code>.dot</code> graph. This graph can then be processed with the 365 "dot" tool to convert it to postscript or some other suitable format. 366 </p> 367</div> 368 369<!-------------------------------------------------------------------------- --> 370<h3> 371 <a name="dot-postdom-only">-dot-postdom-only: Print postdominance tree of function to 'dot' file (with no function bodies)</a> 372</h3> 373<div> 374 <p> 375 This pass, only available in <code>opt</code>, prints the post dominator tree 376 into a <code>.dot</code> graph, omitting the function bodies. This graph can 377 then be processed with the "dot" tool to convert it to postscript or some 378 other suitable format. 379 </p> 380</div> 381 382<!-------------------------------------------------------------------------- --> 383<h3> 384 <a name="globalsmodref-aa">-globalsmodref-aa: Simple mod/ref analysis for globals</a> 385</h3> 386<div> 387 <p> 388 This simple pass provides alias and mod/ref information for global values 389 that do not have their address taken, and keeps track of whether functions 390 read or write memory (are "pure"). For this simple (but very common) case, 391 we can provide pretty accurate and useful information. 392 </p> 393</div> 394 395<!-------------------------------------------------------------------------- --> 396<h3> 397 <a name="instcount">-instcount: Counts the various types of Instructions</a> 398</h3> 399<div> 400 <p> 401 This pass collects the count of all instructions and reports them 402 </p> 403</div> 404 405<!-------------------------------------------------------------------------- --> 406<h3> 407 <a name="intervals">-intervals: Interval Partition Construction</a> 408</h3> 409<div> 410 <p> 411 This analysis calculates and represents the interval partition of a function, 412 or a preexisting interval partition. 413 </p> 414 415 <p> 416 In this way, the interval partition may be used to reduce a flow graph down 417 to its degenerate single node interval partition (unless it is irreducible). 418 </p> 419</div> 420 421<!-------------------------------------------------------------------------- --> 422<h3> 423 <a name="iv-users">-iv-users: Induction Variable Users</a> 424</h3> 425<div> 426 <p>Bookkeeping for "interesting" users of expressions computed from 427 induction variables.</p> 428</div> 429 430<!-------------------------------------------------------------------------- --> 431<h3> 432 <a name="lazy-value-info">-lazy-value-info: Lazy Value Information Analysis</a> 433</h3> 434<div> 435 <p>Interface for lazy computation of value constraint information.</p> 436</div> 437 438<!-------------------------------------------------------------------------- --> 439<h3> 440 <a name="lda">-lda: Loop Dependence Analysis</a> 441</h3> 442<div> 443 <p>Loop dependence analysis framework, which is used to detect dependences in 444 memory accesses in loops.</p> 445</div> 446 447<!-------------------------------------------------------------------------- --> 448<h3> 449 <a name="libcall-aa">-libcall-aa: LibCall Alias Analysis</a> 450</h3> 451<div> 452 <p>LibCall Alias Analysis.</p> 453</div> 454 455<!-------------------------------------------------------------------------- --> 456<h3> 457 <a name="lint">-lint: Statically lint-checks LLVM IR</a> 458</h3> 459<div> 460 <p>This pass statically checks for common and easily-identified constructs 461 which produce undefined or likely unintended behavior in LLVM IR.</p> 462 463 <p>It is not a guarantee of correctness, in two ways. First, it isn't 464 comprehensive. There are checks which could be done statically which are 465 not yet implemented. Some of these are indicated by TODO comments, but 466 those aren't comprehensive either. Second, many conditions cannot be 467 checked statically. This pass does no dynamic instrumentation, so it 468 can't check for all possible problems.</p> 469 470 <p>Another limitation is that it assumes all code will be executed. A store 471 through a null pointer in a basic block which is never reached is harmless, 472 but this pass will warn about it anyway.</p> 473 474 <p>Optimization passes may make conditions that this pass checks for more or 475 less obvious. If an optimization pass appears to be introducing a warning, 476 it may be that the optimization pass is merely exposing an existing 477 condition in the code.</p> 478 479 <p>This code may be run before instcombine. In many cases, instcombine checks 480 for the same kinds of things and turns instructions with undefined behavior 481 into unreachable (or equivalent). Because of this, this pass makes some 482 effort to look through bitcasts and so on. 483 </p> 484</div> 485 486<!-------------------------------------------------------------------------- --> 487<h3> 488 <a name="loops">-loops: Natural Loop Information</a> 489</h3> 490<div> 491 <p> 492 This analysis is used to identify natural loops and determine the loop depth 493 of various nodes of the CFG. Note that the loops identified may actually be 494 several natural loops that share the same header node... not just a single 495 natural loop. 496 </p> 497</div> 498 499<!-------------------------------------------------------------------------- --> 500<h3> 501 <a name="memdep">-memdep: Memory Dependence Analysis</a> 502</h3> 503<div> 504 <p> 505 An analysis that determines, for a given memory operation, what preceding 506 memory operations it depends on. It builds on alias analysis information, and 507 tries to provide a lazy, caching interface to a common kind of alias 508 information query. 509 </p> 510</div> 511 512<!-------------------------------------------------------------------------- --> 513<h3> 514 <a name="module-debuginfo">-module-debuginfo: Decodes module-level debug info</a> 515</h3> 516<div> 517 <p>This pass decodes the debug info metadata in a module and prints in a 518 (sufficiently-prepared-) human-readable form. 519 520 For example, run this pass from opt along with the -analyze option, and 521 it'll print to standard output. 522 </p> 523</div> 524 525<!-------------------------------------------------------------------------- --> 526<h3> 527 <a name="no-aa">-no-aa: No Alias Analysis (always returns 'may' alias)</a> 528</h3> 529<div> 530 <p> 531 Always returns "I don't know" for alias queries. NoAA is unlike other alias 532 analysis implementations, in that it does not chain to a previous analysis. As 533 such it doesn't follow many of the rules that other alias analyses must. 534 </p> 535</div> 536 537<!-------------------------------------------------------------------------- --> 538<h3> 539 <a name="no-profile">-no-profile: No Profile Information</a> 540</h3> 541<div> 542 <p> 543 The default "no profile" implementation of the abstract 544 <code>ProfileInfo</code> interface. 545 </p> 546</div> 547 548<!-------------------------------------------------------------------------- --> 549<h3> 550 <a name="postdomfrontier">-postdomfrontier: Post-Dominance Frontier Construction</a> 551</h3> 552<div> 553 <p> 554 This pass is a simple post-dominator construction algorithm for finding 555 post-dominator frontiers. 556 </p> 557</div> 558 559<!-------------------------------------------------------------------------- --> 560<h3> 561 <a name="postdomtree">-postdomtree: Post-Dominator Tree Construction</a> 562</h3> 563<div> 564 <p> 565 This pass is a simple post-dominator construction algorithm for finding 566 post-dominators. 567 </p> 568</div> 569 570<!-------------------------------------------------------------------------- --> 571<h3> 572 <a name="print-alias-sets">-print-alias-sets: Alias Set Printer</a> 573</h3> 574<div> 575 <p>Yet to be written.</p> 576</div> 577 578<!-------------------------------------------------------------------------- --> 579<h3> 580 <a name="print-callgraph">-print-callgraph: Print a call graph</a> 581</h3> 582<div> 583 <p> 584 This pass, only available in <code>opt</code>, prints the call graph to 585 standard error in a human-readable form. 586 </p> 587</div> 588 589<!-------------------------------------------------------------------------- --> 590<h3> 591 <a name="print-callgraph-sccs">-print-callgraph-sccs: Print SCCs of the Call Graph</a> 592</h3> 593<div> 594 <p> 595 This pass, only available in <code>opt</code>, prints the SCCs of the call 596 graph to standard error in a human-readable form. 597 </p> 598</div> 599 600<!-------------------------------------------------------------------------- --> 601<h3> 602 <a name="print-cfg-sccs">-print-cfg-sccs: Print SCCs of each function CFG</a> 603</h3> 604<div> 605 <p> 606 This pass, only available in <code>opt</code>, prints the SCCs of each 607 function CFG to standard error in a human-readable form. 608 </p> 609</div> 610 611<!-------------------------------------------------------------------------- --> 612<h3> 613 <a name="print-dbginfo">-print-dbginfo: Print debug info in human readable form</a> 614</h3> 615<div> 616 <p>Pass that prints instructions, and associated debug info:</p> 617 <ul> 618 619 <li>source/line/col information</li> 620 <li>original variable name</li> 621 <li>original type name</li> 622 </ul> 623</div> 624 625<!-------------------------------------------------------------------------- --> 626<h3> 627 <a name="print-dom-info">-print-dom-info: Dominator Info Printer</a> 628</h3> 629<div> 630 <p>Dominator Info Printer.</p> 631</div> 632 633<!-------------------------------------------------------------------------- --> 634<h3> 635 <a name="print-externalfnconstants">-print-externalfnconstants: Print external fn callsites passed constants</a> 636</h3> 637<div> 638 <p> 639 This pass, only available in <code>opt</code>, prints out call sites to 640 external functions that are called with constant arguments. This can be 641 useful when looking for standard library functions we should constant fold 642 or handle in alias analyses. 643 </p> 644</div> 645 646<!-------------------------------------------------------------------------- --> 647<h3> 648 <a name="print-function">-print-function: Print function to stderr</a> 649</h3> 650<div> 651 <p> 652 The <code>PrintFunctionPass</code> class is designed to be pipelined with 653 other <code>FunctionPass</code>es, and prints out the functions of the module 654 as they are processed. 655 </p> 656</div> 657 658<!-------------------------------------------------------------------------- --> 659<h3> 660 <a name="print-module">-print-module: Print module to stderr</a> 661</h3> 662<div> 663 <p> 664 This pass simply prints out the entire module when it is executed. 665 </p> 666</div> 667 668<!-------------------------------------------------------------------------- --> 669<h3> 670 <a name="print-used-types">-print-used-types: Find Used Types</a> 671</h3> 672<div> 673 <p> 674 This pass is used to seek out all of the types in use by the program. Note 675 that this analysis explicitly does not include types only used by the symbol 676 table. 677</div> 678 679<!-------------------------------------------------------------------------- --> 680<h3> 681 <a name="profile-estimator">-profile-estimator: Estimate profiling information</a> 682</h3> 683<div> 684 <p>Profiling information that estimates the profiling information 685 in a very crude and unimaginative way. 686 </p> 687</div> 688 689<!-------------------------------------------------------------------------- --> 690<h3> 691 <a name="profile-loader">-profile-loader: Load profile information from llvmprof.out</a> 692</h3> 693<div> 694 <p> 695 A concrete implementation of profiling information that loads the information 696 from a profile dump file. 697 </p> 698</div> 699 700<!-------------------------------------------------------------------------- --> 701<h3> 702 <a name="profile-verifier">-profile-verifier: Verify profiling information</a> 703</h3> 704<div> 705 <p>Pass that checks profiling information for plausibility.</p> 706</div> 707<h3> 708 <a name="regions">-regions: Detect single entry single exit regions</a> 709</h3> 710<div> 711 <p> 712 The <code>RegionInfo</code> pass detects single entry single exit regions in a 713 function, where a region is defined as any subgraph that is connected to the 714 remaining graph at only two spots. Furthermore, an hierarchical region tree is 715 built. 716 </p> 717</div> 718 719<!-------------------------------------------------------------------------- --> 720<h3> 721 <a name="scalar-evolution">-scalar-evolution: Scalar Evolution Analysis</a> 722</h3> 723<div> 724 <p> 725 The <code>ScalarEvolution</code> analysis can be used to analyze and 726 catagorize scalar expressions in loops. It specializes in recognizing general 727 induction variables, representing them with the abstract and opaque 728 <code>SCEV</code> class. Given this analysis, trip counts of loops and other 729 important properties can be obtained. 730 </p> 731 732 <p> 733 This analysis is primarily useful for induction variable substitution and 734 strength reduction. 735 </p> 736</div> 737 738<!-------------------------------------------------------------------------- --> 739<h3> 740 <a name="scev-aa">-scev-aa: ScalarEvolution-based Alias Analysis</a> 741</h3> 742<div> 743 <p>Simple alias analysis implemented in terms of ScalarEvolution queries. 744 745 This differs from traditional loop dependence analysis in that it tests 746 for dependencies within a single iteration of a loop, rather than 747 dependencies between different iterations. 748 749 ScalarEvolution has a more complete understanding of pointer arithmetic 750 than BasicAliasAnalysis' collection of ad-hoc analyses. 751 </p> 752</div> 753 754<!-------------------------------------------------------------------------- --> 755<h3> 756 <a name="targetdata">-targetdata: Target Data Layout</a> 757</h3> 758<div> 759 <p>Provides other passes access to information on how the size and alignment 760 required by the the target ABI for various data types.</p> 761</div> 762 763</div> 764 765<!-- ======================================================================= --> 766<h2><a name="transforms">Transform Passes</a></h2> 767<div> 768 <p>This section describes the LLVM Transform Passes.</p> 769 770<!-------------------------------------------------------------------------- --> 771<h3> 772 <a name="adce">-adce: Aggressive Dead Code Elimination</a> 773</h3> 774<div> 775 <p>ADCE aggressively tries to eliminate code. This pass is similar to 776 <a href="#dce">DCE</a> but it assumes that values are dead until proven 777 otherwise. This is similar to <a href="#sccp">SCCP</a>, except applied to 778 the liveness of values.</p> 779</div> 780 781<!-------------------------------------------------------------------------- --> 782<h3> 783 <a name="always-inline">-always-inline: Inliner for always_inline functions</a> 784</h3> 785<div> 786 <p>A custom inliner that handles only functions that are marked as 787 "always inline".</p> 788</div> 789 790<!-------------------------------------------------------------------------- --> 791<h3> 792 <a name="argpromotion">-argpromotion: Promote 'by reference' arguments to scalars</a> 793</h3> 794<div> 795 <p> 796 This pass promotes "by reference" arguments to be "by value" arguments. In 797 practice, this means looking for internal functions that have pointer 798 arguments. If it can prove, through the use of alias analysis, that an 799 argument is *only* loaded, then it can pass the value into the function 800 instead of the address of the value. This can cause recursive simplification 801 of code and lead to the elimination of allocas (especially in C++ template 802 code like the STL). 803 </p> 804 805 <p> 806 This pass also handles aggregate arguments that are passed into a function, 807 scalarizing them if the elements of the aggregate are only loaded. Note that 808 it refuses to scalarize aggregates which would require passing in more than 809 three operands to the function, because passing thousands of operands for a 810 large array or structure is unprofitable! 811 </p> 812 813 <p> 814 Note that this transformation could also be done for arguments that are only 815 stored to (returning the value instead), but does not currently. This case 816 would be best handled when and if LLVM starts supporting multiple return 817 values from functions. 818 </p> 819</div> 820 821<!-------------------------------------------------------------------------- --> 822<h3> 823 <a name="block-placement">-block-placement: Profile Guided Basic Block Placement</a> 824</h3> 825<div> 826 <p>This pass is a very simple profile guided basic block placement algorithm. 827 The idea is to put frequently executed blocks together at the start of the 828 function and hopefully increase the number of fall-through conditional 829 branches. If there is no profile information for a particular function, this 830 pass basically orders blocks in depth-first order.</p> 831</div> 832 833<!-------------------------------------------------------------------------- --> 834<h3> 835 <a name="break-crit-edges">-break-crit-edges: Break critical edges in CFG</a> 836</h3> 837<div> 838 <p> 839 Break all of the critical edges in the CFG by inserting a dummy basic block. 840 It may be "required" by passes that cannot deal with critical edges. This 841 transformation obviously invalidates the CFG, but can update forward dominator 842 (set, immediate dominators, tree, and frontier) information. 843 </p> 844</div> 845 846<!-------------------------------------------------------------------------- --> 847<h3> 848 <a name="codegenprepare">-codegenprepare: Optimize for code generation</a> 849</h3> 850<div> 851 This pass munges the code in the input function to better prepare it for 852 SelectionDAG-based code generation. This works around limitations in it's 853 basic-block-at-a-time approach. It should eventually be removed. 854</div> 855 856<!-------------------------------------------------------------------------- --> 857<h3> 858 <a name="constmerge">-constmerge: Merge Duplicate Global Constants</a> 859</h3> 860<div> 861 <p> 862 Merges duplicate global constants together into a single constant that is 863 shared. This is useful because some passes (ie TraceValues) insert a lot of 864 string constants into the program, regardless of whether or not an existing 865 string is available. 866 </p> 867</div> 868 869<!-------------------------------------------------------------------------- --> 870<h3> 871 <a name="constprop">-constprop: Simple constant propagation</a> 872</h3> 873<div> 874 <p>This file implements constant propagation and merging. It looks for 875 instructions involving only constant operands and replaces them with a 876 constant value instead of an instruction. For example:</p> 877 <blockquote><pre>add i32 1, 2</pre></blockquote> 878 <p>becomes</p> 879 <blockquote><pre>i32 3</pre></blockquote> 880 <p>NOTE: this pass has a habit of making definitions be dead. It is a good 881 idea to to run a <a href="#die">DIE</a> (Dead Instruction Elimination) pass 882 sometime after running this pass.</p> 883</div> 884 885<!-------------------------------------------------------------------------- --> 886<h3> 887 <a name="dce">-dce: Dead Code Elimination</a> 888</h3> 889<div> 890 <p> 891 Dead code elimination is similar to <a href="#die">dead instruction 892 elimination</a>, but it rechecks instructions that were used by removed 893 instructions to see if they are newly dead. 894 </p> 895</div> 896 897<!-------------------------------------------------------------------------- --> 898<h3> 899 <a name="deadargelim">-deadargelim: Dead Argument Elimination</a> 900</h3> 901<div> 902 <p> 903 This pass deletes dead arguments from internal functions. Dead argument 904 elimination removes arguments which are directly dead, as well as arguments 905 only passed into function calls as dead arguments of other functions. This 906 pass also deletes dead arguments in a similar way. 907 </p> 908 909 <p> 910 This pass is often useful as a cleanup pass to run after aggressive 911 interprocedural passes, which add possibly-dead arguments. 912 </p> 913</div> 914 915<!-------------------------------------------------------------------------- --> 916<h3> 917 <a name="deadtypeelim">-deadtypeelim: Dead Type Elimination</a> 918</h3> 919<div> 920 <p> 921 This pass is used to cleanup the output of GCC. It eliminate names for types 922 that are unused in the entire translation unit, using the <a 923 href="#findusedtypes">find used types</a> pass. 924 </p> 925</div> 926 927<!-------------------------------------------------------------------------- --> 928<h3> 929 <a name="die">-die: Dead Instruction Elimination</a> 930</h3> 931<div> 932 <p> 933 Dead instruction elimination performs a single pass over the function, 934 removing instructions that are obviously dead. 935 </p> 936</div> 937 938<!-------------------------------------------------------------------------- --> 939<h3> 940 <a name="dse">-dse: Dead Store Elimination</a> 941</h3> 942<div> 943 <p> 944 A trivial dead store elimination that only considers basic-block local 945 redundant stores. 946 </p> 947</div> 948 949<!-------------------------------------------------------------------------- --> 950<h3> 951 <a name="functionattrs">-functionattrs: Deduce function attributes</a> 952</h3> 953<div> 954 <p>A simple interprocedural pass which walks the call-graph, looking for 955 functions which do not access or only read non-local memory, and marking them 956 readnone/readonly. In addition, it marks function arguments (of pointer type) 957 'nocapture' if a call to the function does not create any copies of the pointer 958 value that outlive the call. This more or less means that the pointer is only 959 dereferenced, and not returned from the function or stored in a global. 960 This pass is implemented as a bottom-up traversal of the call-graph. 961 </p> 962</div> 963 964<!-------------------------------------------------------------------------- --> 965<h3> 966 <a name="globaldce">-globaldce: Dead Global Elimination</a> 967</h3> 968<div> 969 <p> 970 This transform is designed to eliminate unreachable internal globals from the 971 program. It uses an aggressive algorithm, searching out globals that are 972 known to be alive. After it finds all of the globals which are needed, it 973 deletes whatever is left over. This allows it to delete recursive chunks of 974 the program which are unreachable. 975 </p> 976</div> 977 978<!-------------------------------------------------------------------------- --> 979<h3> 980 <a name="globalopt">-globalopt: Global Variable Optimizer</a> 981</h3> 982<div> 983 <p> 984 This pass transforms simple global variables that never have their address 985 taken. If obviously true, it marks read/write globals as constant, deletes 986 variables only stored to, etc. 987 </p> 988</div> 989 990<!-------------------------------------------------------------------------- --> 991<h3> 992 <a name="gvn">-gvn: Global Value Numbering</a> 993</h3> 994<div> 995 <p> 996 This pass performs global value numbering to eliminate fully and partially 997 redundant instructions. It also performs redundant load elimination. 998 </p> 999</div> 1000 1001<!-------------------------------------------------------------------------- --> 1002<h3> 1003 <a name="indvars">-indvars: Canonicalize Induction Variables</a> 1004</h3> 1005<div> 1006 <p> 1007 This transformation analyzes and transforms the induction variables (and 1008 computations derived from them) into simpler forms suitable for subsequent 1009 analysis and transformation. 1010 </p> 1011 1012 <p> 1013 This transformation makes the following changes to each loop with an 1014 identifiable induction variable: 1015 </p> 1016 1017 <ol> 1018 <li>All loops are transformed to have a <em>single</em> canonical 1019 induction variable which starts at zero and steps by one.</li> 1020 <li>The canonical induction variable is guaranteed to be the first PHI node 1021 in the loop header block.</li> 1022 <li>Any pointer arithmetic recurrences are raised to use array 1023 subscripts.</li> 1024 </ol> 1025 1026 <p> 1027 If the trip count of a loop is computable, this pass also makes the following 1028 changes: 1029 </p> 1030 1031 <ol> 1032 <li>The exit condition for the loop is canonicalized to compare the 1033 induction value against the exit value. This turns loops like: 1034 <blockquote><pre>for (i = 7; i*i < 1000; ++i)</pre></blockquote> 1035 into 1036 <blockquote><pre>for (i = 0; i != 25; ++i)</pre></blockquote></li> 1037 <li>Any use outside of the loop of an expression derived from the indvar 1038 is changed to compute the derived value outside of the loop, eliminating 1039 the dependence on the exit value of the induction variable. If the only 1040 purpose of the loop is to compute the exit value of some derived 1041 expression, this transformation will make the loop dead.</li> 1042 </ol> 1043 1044 <p> 1045 This transformation should be followed by strength reduction after all of the 1046 desired loop transformations have been performed. Additionally, on targets 1047 where it is profitable, the loop could be transformed to count down to zero 1048 (the "do loop" optimization). 1049 </p> 1050</div> 1051 1052<!-------------------------------------------------------------------------- --> 1053<h3> 1054 <a name="inline">-inline: Function Integration/Inlining</a> 1055</h3> 1056<div> 1057 <p> 1058 Bottom-up inlining of functions into callees. 1059 </p> 1060</div> 1061 1062<!-------------------------------------------------------------------------- --> 1063<h3> 1064 <a name="insert-edge-profiling">-insert-edge-profiling: Insert instrumentation for edge profiling</a> 1065</h3> 1066<div> 1067 <p> 1068 This pass instruments the specified program with counters for edge profiling. 1069 Edge profiling can give a reasonable approximation of the hot paths through a 1070 program, and is used for a wide variety of program transformations. 1071 </p> 1072 1073 <p> 1074 Note that this implementation is very naïve. It inserts a counter for 1075 <em>every</em> edge in the program, instead of using control flow information 1076 to prune the number of counters inserted. 1077 </p> 1078</div> 1079 1080<!-------------------------------------------------------------------------- --> 1081<h3> 1082 <a name="insert-optimal-edge-profiling">-insert-optimal-edge-profiling: Insert optimal instrumentation for edge profiling</a> 1083</h3> 1084<div> 1085 <p>This pass instruments the specified program with counters for edge profiling. 1086 Edge profiling can give a reasonable approximation of the hot paths through a 1087 program, and is used for a wide variety of program transformations. 1088 </p> 1089</div> 1090 1091<!-------------------------------------------------------------------------- --> 1092<h3> 1093 <a name="instcombine">-instcombine: Combine redundant instructions</a> 1094</h3> 1095<div> 1096 <p> 1097 Combine instructions to form fewer, simple 1098 instructions. This pass does not modify the CFG This pass is where algebraic 1099 simplification happens. 1100 </p> 1101 1102 <p> 1103 This pass combines things like: 1104 </p> 1105 1106<blockquote><pre 1107>%Y = add i32 %X, 1 1108%Z = add i32 %Y, 1</pre></blockquote> 1109 1110 <p> 1111 into: 1112 </p> 1113 1114<blockquote><pre 1115>%Z = add i32 %X, 2</pre></blockquote> 1116 1117 <p> 1118 This is a simple worklist driven algorithm. 1119 </p> 1120 1121 <p> 1122 This pass guarantees that the following canonicalizations are performed on 1123 the program: 1124 </p> 1125 1126 <ul> 1127 <li>If a binary operator has a constant operand, it is moved to the right- 1128 hand side.</li> 1129 <li>Bitwise operators with constant operands are always grouped so that 1130 shifts are performed first, then <code>or</code>s, then 1131 <code>and</code>s, then <code>xor</code>s.</li> 1132 <li>Compare instructions are converted from <code><</code>, 1133 <code>></code>, <code>≤</code>, or <code>≥</code> to 1134 <code>=</code> or <code>≠</code> if possible.</li> 1135 <li>All <code>cmp</code> instructions on boolean values are replaced with 1136 logical operations.</li> 1137 <li><code>add <var>X</var>, <var>X</var></code> is represented as 1138 <code>mul <var>X</var>, 2</code> ⇒ <code>shl <var>X</var>, 1</code></li> 1139 <li>Multiplies with a constant power-of-two argument are transformed into 1140 shifts.</li> 1141 <li>… etc.</li> 1142 </ul> 1143</div> 1144 1145<!-------------------------------------------------------------------------- --> 1146<h3> 1147 <a name="internalize">-internalize: Internalize Global Symbols</a> 1148</h3> 1149<div> 1150 <p> 1151 This pass loops over all of the functions in the input module, looking for a 1152 main function. If a main function is found, all other functions and all 1153 global variables with initializers are marked as internal. 1154 </p> 1155</div> 1156 1157<!-------------------------------------------------------------------------- --> 1158<h3> 1159 <a name="ipconstprop">-ipconstprop: Interprocedural constant propagation</a> 1160</h3> 1161<div> 1162 <p> 1163 This pass implements an <em>extremely</em> simple interprocedural constant 1164 propagation pass. It could certainly be improved in many different ways, 1165 like using a worklist. This pass makes arguments dead, but does not remove 1166 them. The existing dead argument elimination pass should be run after this 1167 to clean up the mess. 1168 </p> 1169</div> 1170 1171<!-------------------------------------------------------------------------- --> 1172<h3> 1173 <a name="ipsccp">-ipsccp: Interprocedural Sparse Conditional Constant Propagation</a> 1174</h3> 1175<div> 1176 <p> 1177 An interprocedural variant of <a href="#sccp">Sparse Conditional Constant 1178 Propagation</a>. 1179 </p> 1180</div> 1181 1182<!-------------------------------------------------------------------------- --> 1183<h3> 1184 <a name="jump-threading">-jump-threading: Jump Threading</a> 1185</h3> 1186<div> 1187 <p> 1188 Jump threading tries to find distinct threads of control flow running through 1189 a basic block. This pass looks at blocks that have multiple predecessors and 1190 multiple successors. If one or more of the predecessors of the block can be 1191 proven to always cause a jump to one of the successors, we forward the edge 1192 from the predecessor to the successor by duplicating the contents of this 1193 block. 1194 </p> 1195 <p> 1196 An example of when this can occur is code like this: 1197 </p> 1198 1199 <pre 1200>if () { ... 1201 X = 4; 1202} 1203if (X < 3) {</pre> 1204 1205 <p> 1206 In this case, the unconditional branch at the end of the first if can be 1207 revectored to the false side of the second if. 1208 </p> 1209</div> 1210 1211<!-------------------------------------------------------------------------- --> 1212<h3> 1213 <a name="lcssa">-lcssa: Loop-Closed SSA Form Pass</a> 1214</h3> 1215<div> 1216 <p> 1217 This pass transforms loops by placing phi nodes at the end of the loops for 1218 all values that are live across the loop boundary. For example, it turns 1219 the left into the right code: 1220 </p> 1221 1222 <pre 1223>for (...) for (...) 1224 if (c) if (c) 1225 X1 = ... X1 = ... 1226 else else 1227 X2 = ... X2 = ... 1228 X3 = phi(X1, X2) X3 = phi(X1, X2) 1229... = X3 + 4 X4 = phi(X3) 1230 ... = X4 + 4</pre> 1231 1232 <p> 1233 This is still valid LLVM; the extra phi nodes are purely redundant, and will 1234 be trivially eliminated by <code>InstCombine</code>. The major benefit of 1235 this transformation is that it makes many other loop optimizations, such as 1236 LoopUnswitching, simpler. 1237 </p> 1238</div> 1239 1240<!-------------------------------------------------------------------------- --> 1241<h3> 1242 <a name="licm">-licm: Loop Invariant Code Motion</a> 1243</h3> 1244<div> 1245 <p> 1246 This pass performs loop invariant code motion, attempting to remove as much 1247 code from the body of a loop as possible. It does this by either hoisting 1248 code into the preheader block, or by sinking code to the exit blocks if it is 1249 safe. This pass also promotes must-aliased memory locations in the loop to 1250 live in registers, thus hoisting and sinking "invariant" loads and stores. 1251 </p> 1252 1253 <p> 1254 This pass uses alias analysis for two purposes: 1255 </p> 1256 1257 <ul> 1258 <li>Moving loop invariant loads and calls out of loops. If we can determine 1259 that a load or call inside of a loop never aliases anything stored to, 1260 we can hoist it or sink it like any other instruction.</li> 1261 <li>Scalar Promotion of Memory - If there is a store instruction inside of 1262 the loop, we try to move the store to happen AFTER the loop instead of 1263 inside of the loop. This can only happen if a few conditions are true: 1264 <ul> 1265 <li>The pointer stored through is loop invariant.</li> 1266 <li>There are no stores or loads in the loop which <em>may</em> alias 1267 the pointer. There are no calls in the loop which mod/ref the 1268 pointer.</li> 1269 </ul> 1270 If these conditions are true, we can promote the loads and stores in the 1271 loop of the pointer to use a temporary alloca'd variable. We then use 1272 the mem2reg functionality to construct the appropriate SSA form for the 1273 variable.</li> 1274 </ul> 1275</div> 1276 1277<!-------------------------------------------------------------------------- --> 1278<h3> 1279 <a name="loop-deletion">-loop-deletion: Delete dead loops</a> 1280</h3> 1281<div> 1282 <p> 1283 This file implements the Dead Loop Deletion Pass. This pass is responsible 1284 for eliminating loops with non-infinite computable trip counts that have no 1285 side effects or volatile instructions, and do not contribute to the 1286 computation of the function's return value. 1287 </p> 1288</div> 1289 1290<!-------------------------------------------------------------------------- --> 1291<h3> 1292 <a name="loop-extract">-loop-extract: Extract loops into new functions</a> 1293</h3> 1294<div> 1295 <p> 1296 A pass wrapper around the <code>ExtractLoop()</code> scalar transformation to 1297 extract each top-level loop into its own new function. If the loop is the 1298 <em>only</em> loop in a given function, it is not touched. This is a pass most 1299 useful for debugging via bugpoint. 1300 </p> 1301</div> 1302 1303<!-------------------------------------------------------------------------- --> 1304<h3> 1305 <a name="loop-extract-single">-loop-extract-single: Extract at most one loop into a new function</a> 1306</h3> 1307<div> 1308 <p> 1309 Similar to <a href="#loop-extract">Extract loops into new functions</a>, 1310 this pass extracts one natural loop from the program into a function if it 1311 can. This is used by bugpoint. 1312 </p> 1313</div> 1314 1315<!-------------------------------------------------------------------------- --> 1316<h3> 1317 <a name="loop-reduce">-loop-reduce: Loop Strength Reduction</a> 1318</h3> 1319<div> 1320 <p> 1321 This pass performs a strength reduction on array references inside loops that 1322 have as one or more of their components the loop induction variable. This is 1323 accomplished by creating a new value to hold the initial value of the array 1324 access for the first iteration, and then creating a new GEP instruction in 1325 the loop to increment the value by the appropriate amount. 1326 </p> 1327</div> 1328 1329<!-------------------------------------------------------------------------- --> 1330<h3> 1331 <a name="loop-rotate">-loop-rotate: Rotate Loops</a> 1332</h3> 1333<div> 1334 <p>A simple loop rotation transformation.</p> 1335</div> 1336 1337<!-------------------------------------------------------------------------- --> 1338<h3> 1339 <a name="loop-simplify">-loop-simplify: Canonicalize natural loops</a> 1340</h3> 1341<div> 1342 <p> 1343 This pass performs several transformations to transform natural loops into a 1344 simpler form, which makes subsequent analyses and transformations simpler and 1345 more effective. 1346 </p> 1347 1348 <p> 1349 Loop pre-header insertion guarantees that there is a single, non-critical 1350 entry edge from outside of the loop to the loop header. This simplifies a 1351 number of analyses and transformations, such as LICM. 1352 </p> 1353 1354 <p> 1355 Loop exit-block insertion guarantees that all exit blocks from the loop 1356 (blocks which are outside of the loop that have predecessors inside of the 1357 loop) only have predecessors from inside of the loop (and are thus dominated 1358 by the loop header). This simplifies transformations such as store-sinking 1359 that are built into LICM. 1360 </p> 1361 1362 <p> 1363 This pass also guarantees that loops will have exactly one backedge. 1364 </p> 1365 1366 <p> 1367 Note that the simplifycfg pass will clean up blocks which are split out but 1368 end up being unnecessary, so usage of this pass should not pessimize 1369 generated code. 1370 </p> 1371 1372 <p> 1373 This pass obviously modifies the CFG, but updates loop information and 1374 dominator information. 1375 </p> 1376</div> 1377 1378<!-------------------------------------------------------------------------- --> 1379<h3> 1380 <a name="loop-unroll">-loop-unroll: Unroll loops</a> 1381</h3> 1382<div> 1383 <p> 1384 This pass implements a simple loop unroller. It works best when loops have 1385 been canonicalized by the <a href="#indvars"><tt>-indvars</tt></a> pass, 1386 allowing it to determine the trip counts of loops easily. 1387 </p> 1388</div> 1389 1390<!-------------------------------------------------------------------------- --> 1391<h3> 1392 <a name="loop-unswitch">-loop-unswitch: Unswitch loops</a> 1393</h3> 1394<div> 1395 <p> 1396 This pass transforms loops that contain branches on loop-invariant conditions 1397 to have multiple loops. For example, it turns the left into the right code: 1398 </p> 1399 1400 <pre 1401>for (...) if (lic) 1402 A for (...) 1403 if (lic) A; B; C 1404 B else 1405 C for (...) 1406 A; C</pre> 1407 1408 <p> 1409 This can increase the size of the code exponentially (doubling it every time 1410 a loop is unswitched) so we only unswitch if the resultant code will be 1411 smaller than a threshold. 1412 </p> 1413 1414 <p> 1415 This pass expects LICM to be run before it to hoist invariant conditions out 1416 of the loop, to make the unswitching opportunity obvious. 1417 </p> 1418</div> 1419 1420<!-------------------------------------------------------------------------- --> 1421<h3> 1422 <a name="loweratomic">-loweratomic: Lower atomic intrinsics to non-atomic form</a> 1423</h3> 1424<div> 1425 <p> 1426 This pass lowers atomic intrinsics to non-atomic form for use in a known 1427 non-preemptible environment. 1428 </p> 1429 1430 <p> 1431 The pass does not verify that the environment is non-preemptible (in 1432 general this would require knowledge of the entire call graph of the 1433 program including any libraries which may not be available in bitcode form); 1434 it simply lowers every atomic intrinsic. 1435 </p> 1436</div> 1437 1438<!-------------------------------------------------------------------------- --> 1439<h3> 1440 <a name="lowerinvoke">-lowerinvoke: Lower invoke and unwind, for unwindless code generators</a> 1441</h3> 1442<div> 1443 <p> 1444 This transformation is designed for use by code generators which do not yet 1445 support stack unwinding. This pass supports two models of exception handling 1446 lowering, the 'cheap' support and the 'expensive' support. 1447 </p> 1448 1449 <p> 1450 'Cheap' exception handling support gives the program the ability to execute 1451 any program which does not "throw an exception", by turning 'invoke' 1452 instructions into calls and by turning 'unwind' instructions into calls to 1453 abort(). If the program does dynamically use the unwind instruction, the 1454 program will print a message then abort. 1455 </p> 1456 1457 <p> 1458 'Expensive' exception handling support gives the full exception handling 1459 support to the program at the cost of making the 'invoke' instruction 1460 really expensive. It basically inserts setjmp/longjmp calls to emulate the 1461 exception handling as necessary. 1462 </p> 1463 1464 <p> 1465 Because the 'expensive' support slows down programs a lot, and EH is only 1466 used for a subset of the programs, it must be specifically enabled by the 1467 <tt>-enable-correct-eh-support</tt> option. 1468 </p> 1469 1470 <p> 1471 Note that after this pass runs the CFG is not entirely accurate (exceptional 1472 control flow edges are not correct anymore) so only very simple things should 1473 be done after the lowerinvoke pass has run (like generation of native code). 1474 This should not be used as a general purpose "my LLVM-to-LLVM pass doesn't 1475 support the invoke instruction yet" lowering pass. 1476 </p> 1477</div> 1478 1479<!-------------------------------------------------------------------------- --> 1480<h3> 1481 <a name="lowersetjmp">-lowersetjmp: Lower Set Jump</a> 1482</h3> 1483<div> 1484 <p> 1485 Lowers <tt>setjmp</tt> and <tt>longjmp</tt> to use the LLVM invoke and unwind 1486 instructions as necessary. 1487 </p> 1488 1489 <p> 1490 Lowering of <tt>longjmp</tt> is fairly trivial. We replace the call with a 1491 call to the LLVM library function <tt>__llvm_sjljeh_throw_longjmp()</tt>. 1492 This unwinds the stack for us calling all of the destructors for 1493 objects allocated on the stack. 1494 </p> 1495 1496 <p> 1497 At a <tt>setjmp</tt> call, the basic block is split and the <tt>setjmp</tt> 1498 removed. The calls in a function that have a <tt>setjmp</tt> are converted to 1499 invoke where the except part checks to see if it's a <tt>longjmp</tt> 1500 exception and, if so, if it's handled in the function. If it is, then it gets 1501 the value returned by the <tt>longjmp</tt> and goes to where the basic block 1502 was split. <tt>invoke</tt> instructions are handled in a similar fashion with 1503 the original except block being executed if it isn't a <tt>longjmp</tt> 1504 except that is handled by that function. 1505 </p> 1506</div> 1507 1508<!-------------------------------------------------------------------------- --> 1509<h3> 1510 <a name="lowerswitch">-lowerswitch: Lower SwitchInst's to branches</a> 1511</h3> 1512<div> 1513 <p> 1514 Rewrites <tt>switch</tt> instructions with a sequence of branches, which 1515 allows targets to get away with not implementing the switch instruction until 1516 it is convenient. 1517 </p> 1518</div> 1519 1520<!-------------------------------------------------------------------------- --> 1521<h3> 1522 <a name="mem2reg">-mem2reg: Promote Memory to Register</a> 1523</h3> 1524<div> 1525 <p> 1526 This file promotes memory references to be register references. It promotes 1527 <tt>alloca</tt> instructions which only have <tt>load</tt>s and 1528 <tt>store</tt>s as uses. An <tt>alloca</tt> is transformed by using dominator 1529 frontiers to place <tt>phi</tt> nodes, then traversing the function in 1530 depth-first order to rewrite <tt>load</tt>s and <tt>store</tt>s as 1531 appropriate. This is just the standard SSA construction algorithm to construct 1532 "pruned" SSA form. 1533 </p> 1534</div> 1535 1536<!-------------------------------------------------------------------------- --> 1537<h3> 1538 <a name="memcpyopt">-memcpyopt: MemCpy Optimization</a> 1539</h3> 1540<div> 1541 <p> 1542 This pass performs various transformations related to eliminating memcpy 1543 calls, or transforming sets of stores into memset's. 1544 </p> 1545</div> 1546 1547<!-------------------------------------------------------------------------- --> 1548<h3> 1549 <a name="mergefunc">-mergefunc: Merge Functions</a> 1550</h3> 1551<div> 1552 <p>This pass looks for equivalent functions that are mergable and folds them. 1553 1554 A hash is computed from the function, based on its type and number of 1555 basic blocks. 1556 1557 Once all hashes are computed, we perform an expensive equality comparison 1558 on each function pair. This takes n^2/2 comparisons per bucket, so it's 1559 important that the hash function be high quality. The equality comparison 1560 iterates through each instruction in each basic block. 1561 1562 When a match is found the functions are folded. If both functions are 1563 overridable, we move the functionality into a new internal function and 1564 leave two overridable thunks to it. 1565 </p> 1566</div> 1567 1568<!-------------------------------------------------------------------------- --> 1569<h3> 1570 <a name="mergereturn">-mergereturn: Unify function exit nodes</a> 1571</h3> 1572<div> 1573 <p> 1574 Ensure that functions have at most one <tt>ret</tt> instruction in them. 1575 Additionally, it keeps track of which node is the new exit node of the CFG. 1576 </p> 1577</div> 1578 1579<!-------------------------------------------------------------------------- --> 1580<h3> 1581 <a name="partial-inliner">-partial-inliner: Partial Inliner</a> 1582</h3> 1583<div> 1584 <p>This pass performs partial inlining, typically by inlining an if 1585 statement that surrounds the body of the function. 1586 </p> 1587</div> 1588 1589<!-------------------------------------------------------------------------- --> 1590<h3> 1591 <a name="prune-eh">-prune-eh: Remove unused exception handling info</a> 1592</h3> 1593<div> 1594 <p> 1595 This file implements a simple interprocedural pass which walks the call-graph, 1596 turning <tt>invoke</tt> instructions into <tt>call</tt> instructions if and 1597 only if the callee cannot throw an exception. It implements this as a 1598 bottom-up traversal of the call-graph. 1599 </p> 1600</div> 1601 1602<!-------------------------------------------------------------------------- --> 1603<h3> 1604 <a name="reassociate">-reassociate: Reassociate expressions</a> 1605</h3> 1606<div> 1607 <p> 1608 This pass reassociates commutative expressions in an order that is designed 1609 to promote better constant propagation, GCSE, LICM, PRE, etc. 1610 </p> 1611 1612 <p> 1613 For example: 4 + (<var>x</var> + 5) ⇒ <var>x</var> + (4 + 5) 1614 </p> 1615 1616 <p> 1617 In the implementation of this algorithm, constants are assigned rank = 0, 1618 function arguments are rank = 1, and other values are assigned ranks 1619 corresponding to the reverse post order traversal of current function 1620 (starting at 2), which effectively gives values in deep loops higher rank 1621 than values not in loops. 1622 </p> 1623</div> 1624 1625<!-------------------------------------------------------------------------- --> 1626<h3> 1627 <a name="reg2mem">-reg2mem: Demote all values to stack slots</a> 1628</h3> 1629<div> 1630 <p> 1631 This file demotes all registers to memory references. It is intented to be 1632 the inverse of <a href="#mem2reg"><tt>-mem2reg</tt></a>. By converting to 1633 <tt>load</tt> instructions, the only values live across basic blocks are 1634 <tt>alloca</tt> instructions and <tt>load</tt> instructions before 1635 <tt>phi</tt> nodes. It is intended that this should make CFG hacking much 1636 easier. To make later hacking easier, the entry block is split into two, such 1637 that all introduced <tt>alloca</tt> instructions (and nothing else) are in the 1638 entry block. 1639 </p> 1640</div> 1641 1642<!-------------------------------------------------------------------------- --> 1643<h3> 1644 <a name="scalarrepl">-scalarrepl: Scalar Replacement of Aggregates (DT)</a> 1645</h3> 1646<div> 1647 <p> 1648 The well-known scalar replacement of aggregates transformation. This 1649 transform breaks up <tt>alloca</tt> instructions of aggregate type (structure 1650 or array) into individual <tt>alloca</tt> instructions for each member if 1651 possible. Then, if possible, it transforms the individual <tt>alloca</tt> 1652 instructions into nice clean scalar SSA form. 1653 </p> 1654 1655 <p> 1656 This combines a simple scalar replacement of aggregates algorithm with the <a 1657 href="#mem2reg"><tt>mem2reg</tt></a> algorithm because often interact, 1658 especially for C++ programs. As such, iterating between <tt>scalarrepl</tt>, 1659 then <a href="#mem2reg"><tt>mem2reg</tt></a> until we run out of things to 1660 promote works well. 1661 </p> 1662</div> 1663 1664<!-------------------------------------------------------------------------- --> 1665<h3> 1666 <a name="sccp">-sccp: Sparse Conditional Constant Propagation</a> 1667</h3> 1668<div> 1669 <p> 1670 Sparse conditional constant propagation and merging, which can be summarized 1671 as: 1672 </p> 1673 1674 <ol> 1675 <li>Assumes values are constant unless proven otherwise</li> 1676 <li>Assumes BasicBlocks are dead unless proven otherwise</li> 1677 <li>Proves values to be constant, and replaces them with constants</li> 1678 <li>Proves conditional branches to be unconditional</li> 1679 </ol> 1680 1681 <p> 1682 Note that this pass has a habit of making definitions be dead. It is a good 1683 idea to to run a DCE pass sometime after running this pass. 1684 </p> 1685</div> 1686 1687<!-------------------------------------------------------------------------- --> 1688<h3> 1689 <a name="simplify-libcalls">-simplify-libcalls: Simplify well-known library calls</a> 1690</h3> 1691<div> 1692 <p> 1693 Applies a variety of small optimizations for calls to specific well-known 1694 function calls (e.g. runtime library functions). For example, a call 1695 <tt>exit(3)</tt> that occurs within the <tt>main()</tt> function can be 1696 transformed into simply <tt>return 3</tt>. 1697 </p> 1698</div> 1699 1700<!-------------------------------------------------------------------------- --> 1701<h3> 1702 <a name="simplifycfg">-simplifycfg: Simplify the CFG</a> 1703</h3> 1704<div> 1705 <p> 1706 Performs dead code elimination and basic block merging. Specifically: 1707 </p> 1708 1709 <ol> 1710 <li>Removes basic blocks with no predecessors.</li> 1711 <li>Merges a basic block into its predecessor if there is only one and the 1712 predecessor only has one successor.</li> 1713 <li>Eliminates PHI nodes for basic blocks with a single predecessor.</li> 1714 <li>Eliminates a basic block that only contains an unconditional 1715 branch.</li> 1716 </ol> 1717</div> 1718 1719<!-------------------------------------------------------------------------- --> 1720<h3> 1721 <a name="sink">-sink: Code sinking</a> 1722</h3> 1723<div> 1724 <p>This pass moves instructions into successor blocks, when possible, so that 1725 they aren't executed on paths where their results aren't needed. 1726 </p> 1727</div> 1728 1729<!-------------------------------------------------------------------------- --> 1730<h3> 1731 <a name="sretpromotion">-sretpromotion: Promote sret arguments to multiple ret values</a> 1732</h3> 1733<div> 1734 <p> 1735 This pass finds functions that return a struct (using a pointer to the struct 1736 as the first argument of the function, marked with the '<tt>sret</tt>' attribute) and 1737 replaces them with a new function that simply returns each of the elements of 1738 that struct (using multiple return values). 1739 </p> 1740 1741 <p> 1742 This pass works under a number of conditions: 1743 </p> 1744 1745 <ul> 1746 <li>The returned struct must not contain other structs</li> 1747 <li>The returned struct must only be used to load values from</li> 1748 <li>The placeholder struct passed in is the result of an <tt>alloca</tt></li> 1749 </ul> 1750</div> 1751 1752<!-------------------------------------------------------------------------- --> 1753<h3> 1754 <a name="strip">-strip: Strip all symbols from a module</a> 1755</h3> 1756<div> 1757 <p> 1758 performs code stripping. this transformation can delete: 1759 </p> 1760 1761 <ol> 1762 <li>names for virtual registers</li> 1763 <li>symbols for internal globals and functions</li> 1764 <li>debug information</li> 1765 </ol> 1766 1767 <p> 1768 note that this transformation makes code much less readable, so it should 1769 only be used in situations where the <tt>strip</tt> utility would be used, 1770 such as reducing code size or making it harder to reverse engineer code. 1771 </p> 1772</div> 1773 1774<!-------------------------------------------------------------------------- --> 1775<h3> 1776 <a name="strip-dead-debug-info">-strip-dead-debug-info: Strip debug info for unused symbols</a> 1777</h3> 1778<div> 1779 <p> 1780 performs code stripping. this transformation can delete: 1781 </p> 1782 1783 <ol> 1784 <li>names for virtual registers</li> 1785 <li>symbols for internal globals and functions</li> 1786 <li>debug information</li> 1787 </ol> 1788 1789 <p> 1790 note that this transformation makes code much less readable, so it should 1791 only be used in situations where the <tt>strip</tt> utility would be used, 1792 such as reducing code size or making it harder to reverse engineer code. 1793 </p> 1794</div> 1795 1796<!-------------------------------------------------------------------------- --> 1797<h3> 1798 <a name="strip-dead-prototypes">-strip-dead-prototypes: Strip Unused Function Prototypes</a> 1799</h3> 1800<div> 1801 <p> 1802 This pass loops over all of the functions in the input module, looking for 1803 dead declarations and removes them. Dead declarations are declarations of 1804 functions for which no implementation is available (i.e., declarations for 1805 unused library functions). 1806 </p> 1807</div> 1808 1809<!-------------------------------------------------------------------------- --> 1810<h3> 1811 <a name="strip-debug-declare">-strip-debug-declare: Strip all llvm.dbg.declare intrinsics</a> 1812</h3> 1813<div> 1814 <p>This pass implements code stripping. Specifically, it can delete:</p> 1815 <ul> 1816 <li>names for virtual registers</li> 1817 <li>symbols for internal globals and functions</li> 1818 <li>debug information</li> 1819 </ul> 1820 <p> 1821 Note that this transformation makes code much less readable, so it should 1822 only be used in situations where the 'strip' utility would be used, such as 1823 reducing code size or making it harder to reverse engineer code. 1824 </p> 1825</div> 1826 1827<!-------------------------------------------------------------------------- --> 1828<h3> 1829 <a name="strip-nondebug">-strip-nondebug: Strip all symbols, except dbg symbols, from a module</a> 1830</h3> 1831<div> 1832 <p>This pass implements code stripping. Specifically, it can delete:</p> 1833 <ul> 1834 <li>names for virtual registers</li> 1835 <li>symbols for internal globals and functions</li> 1836 <li>debug information</li> 1837 </ul> 1838 <p> 1839 Note that this transformation makes code much less readable, so it should 1840 only be used in situations where the 'strip' utility would be used, such as 1841 reducing code size or making it harder to reverse engineer code. 1842 </p> 1843</div> 1844 1845<!-------------------------------------------------------------------------- --> 1846<h3> 1847 <a name="tailcallelim">-tailcallelim: Tail Call Elimination</a> 1848</h3> 1849<div> 1850 <p> 1851 This file transforms calls of the current function (self recursion) followed 1852 by a return instruction with a branch to the entry of the function, creating 1853 a loop. This pass also implements the following extensions to the basic 1854 algorithm: 1855 </p> 1856 1857 <ul> 1858 <li>Trivial instructions between the call and return do not prevent the 1859 transformation from taking place, though currently the analysis cannot 1860 support moving any really useful instructions (only dead ones). 1861 <li>This pass transforms functions that are prevented from being tail 1862 recursive by an associative expression to use an accumulator variable, 1863 thus compiling the typical naive factorial or <tt>fib</tt> implementation 1864 into efficient code. 1865 <li>TRE is performed if the function returns void, if the return 1866 returns the result returned by the call, or if the function returns a 1867 run-time constant on all exits from the function. It is possible, though 1868 unlikely, that the return returns something else (like constant 0), and 1869 can still be TRE'd. It can be TRE'd if <em>all other</em> return 1870 instructions in the function return the exact same value. 1871 <li>If it can prove that callees do not access theier caller stack frame, 1872 they are marked as eligible for tail call elimination (by the code 1873 generator). 1874 </ul> 1875</div> 1876 1877<!-------------------------------------------------------------------------- --> 1878<h3> 1879 <a name="tailduplicate">-tailduplicate: Tail Duplication</a> 1880</h3> 1881<div> 1882 <p> 1883 This pass performs a limited form of tail duplication, intended to simplify 1884 CFGs by removing some unconditional branches. This pass is necessary to 1885 straighten out loops created by the C front-end, but also is capable of 1886 making other code nicer. After this pass is run, the CFG simplify pass 1887 should be run to clean up the mess. 1888 </p> 1889</div> 1890 1891</div> 1892 1893<!-- ======================================================================= --> 1894<h2><a name="utilities">Utility Passes</a></h2> 1895<div> 1896 <p>This section describes the LLVM Utility Passes.</p> 1897 1898<!-------------------------------------------------------------------------- --> 1899<h3> 1900 <a name="deadarghaX0r">-deadarghaX0r: Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</a> 1901</h3> 1902<div> 1903 <p> 1904 Same as dead argument elimination, but deletes arguments to functions which 1905 are external. This is only for use by <a 1906 href="Bugpoint.html">bugpoint</a>.</p> 1907</div> 1908 1909<!-------------------------------------------------------------------------- --> 1910<h3> 1911 <a name="extract-blocks">-extract-blocks: Extract Basic Blocks From Module (for bugpoint use)</a> 1912</h3> 1913<div> 1914 <p> 1915 This pass is used by bugpoint to extract all blocks from the module into their 1916 own functions.</p> 1917</div> 1918 1919<!-------------------------------------------------------------------------- --> 1920<h3> 1921 <a name="instnamer">-instnamer: Assign names to anonymous instructions</a> 1922</h3> 1923<div> 1924 <p>This is a little utility pass that gives instructions names, this is mostly 1925 useful when diffing the effect of an optimization because deleting an 1926 unnamed instruction can change all other instruction numbering, making the 1927 diff very noisy. 1928 </p> 1929</div> 1930 1931<!-------------------------------------------------------------------------- --> 1932<h3> 1933 <a name="preverify">-preverify: Preliminary module verification</a> 1934</h3> 1935<div> 1936 <p> 1937 Ensures that the module is in the form required by the <a 1938 href="#verifier">Module Verifier</a> pass. 1939 </p> 1940 1941 <p> 1942 Running the verifier runs this pass automatically, so there should be no need 1943 to use it directly. 1944 </p> 1945</div> 1946 1947<!-------------------------------------------------------------------------- --> 1948<h3> 1949 <a name="verify">-verify: Module Verifier</a> 1950</h3> 1951<div> 1952 <p> 1953 Verifies an LLVM IR code. This is useful to run after an optimization which is 1954 undergoing testing. Note that <tt>llvm-as</tt> verifies its input before 1955 emitting bitcode, and also that malformed bitcode is likely to make LLVM 1956 crash. All language front-ends are therefore encouraged to verify their output 1957 before performing optimizing transformations. 1958 </p> 1959 1960 <ul> 1961 <li>Both of a binary operator's parameters are of the same type.</li> 1962 <li>Verify that the indices of mem access instructions match other 1963 operands.</li> 1964 <li>Verify that arithmetic and other things are only performed on 1965 first-class types. Verify that shifts and logicals only happen on 1966 integrals f.e.</li> 1967 <li>All of the constants in a switch statement are of the correct type.</li> 1968 <li>The code is in valid SSA form.</li> 1969 <li>It is illegal to put a label into any other type (like a structure) or 1970 to return one.</li> 1971 <li>Only phi nodes can be self referential: <tt>%x = add i32 %x, %x</tt> is 1972 invalid.</li> 1973 <li>PHI nodes must have an entry for each predecessor, with no extras.</li> 1974 <li>PHI nodes must be the first thing in a basic block, all grouped 1975 together.</li> 1976 <li>PHI nodes must have at least one entry.</li> 1977 <li>All basic blocks should only end with terminator insts, not contain 1978 them.</li> 1979 <li>The entry node to a function must not have predecessors.</li> 1980 <li>All Instructions must be embedded into a basic block.</li> 1981 <li>Functions cannot take a void-typed parameter.</li> 1982 <li>Verify that a function's argument list agrees with its declared 1983 type.</li> 1984 <li>It is illegal to specify a name for a void value.</li> 1985 <li>It is illegal to have a internal global value with no initializer.</li> 1986 <li>It is illegal to have a ret instruction that returns a value that does 1987 not agree with the function return value type.</li> 1988 <li>Function call argument types match the function prototype.</li> 1989 <li>All other things that are tested by asserts spread about the code.</li> 1990 </ul> 1991 1992 <p> 1993 Note that this does not provide full security verification (like Java), but 1994 instead just tries to ensure that code is well-formed. 1995 </p> 1996</div> 1997 1998<!-------------------------------------------------------------------------- --> 1999<h3> 2000 <a name="view-cfg">-view-cfg: View CFG of function</a> 2001</h3> 2002<div> 2003 <p> 2004 Displays the control flow graph using the GraphViz tool. 2005 </p> 2006</div> 2007 2008<!-------------------------------------------------------------------------- --> 2009<h3> 2010 <a name="view-cfg-only">-view-cfg-only: View CFG of function (with no function bodies)</a> 2011</h3> 2012<div> 2013 <p> 2014 Displays the control flow graph using the GraphViz tool, but omitting function 2015 bodies. 2016 </p> 2017</div> 2018 2019<!-------------------------------------------------------------------------- --> 2020<h3> 2021 <a name="view-dom">-view-dom: View dominance tree of function</a> 2022</h3> 2023<div> 2024 <p> 2025 Displays the dominator tree using the GraphViz tool. 2026 </p> 2027</div> 2028 2029<!-------------------------------------------------------------------------- --> 2030<h3> 2031 <a name="view-dom-only">-view-dom-only: View dominance tree of function (with no function bodies)</a> 2032</h3> 2033<div> 2034 <p> 2035 Displays the dominator tree using the GraphViz tool, but omitting function 2036 bodies. 2037 </p> 2038</div> 2039 2040<!-------------------------------------------------------------------------- --> 2041<h3> 2042 <a name="view-postdom">-view-postdom: View postdominance tree of function</a> 2043</h3> 2044<div> 2045 <p> 2046 Displays the post dominator tree using the GraphViz tool. 2047 </p> 2048</div> 2049 2050<!-------------------------------------------------------------------------- --> 2051<h3> 2052 <a name="view-postdom-only">-view-postdom-only: View postdominance tree of function (with no function bodies)</a> 2053</h3> 2054<div> 2055 <p> 2056 Displays the post dominator tree using the GraphViz tool, but omitting 2057 function bodies. 2058 </p> 2059</div> 2060 2061</div> 2062 2063<!-- *********************************************************************** --> 2064 2065<hr> 2066<address> 2067 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img 2068 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a> 2069 <a href="http://validator.w3.org/check/referer"><img 2070 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a> 2071 2072 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br> 2073 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br> 2074 Last modified: $Date$ 2075</address> 2076 2077</body> 2078</html> 2079