docs/Passes.html

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