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