1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
2 "http://www.w3.org/TR/html4/strict.dtd">
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">
13 If Passes.html is up to date, the following "one-liner" should print
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: $!\n";
21 m:^<tr><td><a href="#(.*)">-.*</a></td><td>.*</td></tr>$: or next;
22 $order{$1} = sprintf("%03d", 1 + int %order);
24 open HELP, "../Release/bin/opt -help|" or die "open: opt -help: $!\n";
26 m:^ -([^ ]+) +- (.*)$: or next;
28 $o = "000" unless defined $o;
29 push @x, "$o<tr><td><a href=\"#$1\">-$1</a></td><td>$2</td></tr>\n";
30 push @y, "$o <a name=\"$1\">-$1: $2</a>\n";
32 @x = map { s/^\d\d\d//; $_ } sort @x;
33 @y = map { s/^\d\d\d//; $_ } sort @y;
37 This (real) one-liner can also be helpful when converting comments to HTML:
39 perl -e '$/ = 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'
43 <div class="doc_title">LLVM's Analysis and Transform Passes</div>
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>
52 <div class="doc_author">
53 <p>Written by <a href="mailto:rspencer@x10sys.com">Reid Spencer</a>
54 and Gordon Henriksen</p>
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>
73 <div class="doc_text" >
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="#basicaa">-basicaa</a></td><td>Basic Alias Analysis (stateless AA impl)</td></tr>
79 <tr><td><a href="#basiccg">-basiccg</a></td><td>Basic CallGraph Construction</td></tr>
80 <tr><td><a href="#count-aa">-count-aa</a></td><td>Count Alias Analysis Query Responses</td></tr>
81 <tr><td><a href="#debug-aa">-debug-aa</a></td><td>AA use debugger</td></tr>
82 <tr><td><a href="#domfrontier">-domfrontier</a></td><td>Dominance Frontier Construction</td></tr>
83 <tr><td><a href="#domtree">-domtree</a></td><td>Dominator Tree Construction</td></tr>
84 <tr><td><a href="#dot-callgraph">-dot-callgraph</a></td><td>Print Call Graph to 'dot' file</td></tr>
85 <tr><td><a href="#dot-cfg">-dot-cfg</a></td><td>Print CFG of function to 'dot' file</td></tr>
86 <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>
87 <tr><td><a href="#dot-dom">-dot-dom</a></td><td>Print dominance tree of function to 'dot' file</td></tr>
88 <tr><td><a href="#dot-dom-only">-dot-dom-only</a></td><td>Print dominance tree of function to 'dot' file (with no function bodies)</td></tr>
89 <tr><td><a href="#dot-postdom">-dot-postdom</a></td><td>Print postdominance tree of function to 'dot' file</td></tr>
90 <tr><td><a href="#dot-postdom-only">-dot-postdom-only</a></td><td>Print postdominance tree of function to 'dot' file (with no function bodies)</td></tr>
91 <tr><td><a href="#globalsmodref-aa">-globalsmodref-aa</a></td><td>Simple mod/ref analysis for globals</td></tr>
92 <tr><td><a href="#instcount">-instcount</a></td><td>Counts the various types of Instructions</td></tr>
93 <tr><td><a href="#intervals">-intervals</a></td><td>Interval Partition Construction</td></tr>
94 <tr><td><a href="#iv-users">-iv-users</a></td><td>Induction Variable Users</td></tr>
95 <tr><td><a href="#lazy-value-info">-lazy-value-info</a></td><td>Lazy Value Information Analysis</td></tr>
96 <tr><td><a href="#lda">-lda</a></td><td>Loop Dependence Analysis</td></tr>
97 <tr><td><a href="#libcall-aa">-libcall-aa</a></td><td>LibCall Alias Analysis</td></tr>
98 <tr><td><a href="#lint">-lint</a></td><td>Statically lint-checks LLVM IR</td></tr>
99 <tr><td><a href="#loops">-loops</a></td><td>Natural Loop Information</td></tr>
100 <tr><td><a href="#memdep">-memdep</a></td><td>Memory Dependence Analysis</td></tr>
101 <tr><td><a href="#module-debuginfo">-module-debuginfo</a></td><td>Decodes module-level debug info</td></tr>
102 <tr><td><a href="#no-aa">-no-aa</a></td><td>No Alias Analysis (always returns 'may' alias)</td></tr>
103 <tr><td><a href="#no-profile">-no-profile</a></td><td>No Profile Information</td></tr>
104 <tr><td><a href="#postdomfrontier">-postdomfrontier</a></td><td>Post-Dominance Frontier Construction</td></tr>
105 <tr><td><a href="#postdomtree">-postdomtree</a></td><td>Post-Dominator Tree Construction</td></tr>
106 <tr><td><a href="#print-alias-sets">-print-alias-sets</a></td><td>Alias Set Printer</td></tr>
107 <tr><td><a href="#print-callgraph">-print-callgraph</a></td><td>Print a call graph</td></tr>
108 <tr><td><a href="#print-callgraph-sccs">-print-callgraph-sccs</a></td><td>Print SCCs of the Call Graph</td></tr>
109 <tr><td><a href="#print-cfg-sccs">-print-cfg-sccs</a></td><td>Print SCCs of each function CFG</td></tr>
110 <tr><td><a href="#print-dbginfo">-print-dbginfo</a></td><td>Print debug info in human readable form</td></tr>
111 <tr><td><a href="#print-dom-info">-print-dom-info</a></td><td>Dominator Info Printer</td></tr>
112 <tr><td><a href="#print-externalfnconstants">-print-externalfnconstants</a></td><td>Print external fn callsites passed constants</td></tr>
113 <tr><td><a href="#print-function">-print-function</a></td><td>Print function to stderr</td></tr>
114 <tr><td><a href="#print-module">-print-module</a></td><td>Print module to stderr</td></tr>
115 <tr><td><a href="#print-used-types">-print-used-types</a></td><td>Find Used Types</td></tr>
116 <tr><td><a href="#profile-estimator">-profile-estimator</a></td><td>Estimate profiling information</td></tr>
117 <tr><td><a href="#profile-loader">-profile-loader</a></td><td>Load profile information from llvmprof.out</td></tr>
118 <tr><td><a href="#profile-verifier">-profile-verifier</a></td><td>Verify profiling information</td></tr>
119 <tr><td><a href="#regions">-regions</a></td><td>Detect single entry single exit regions</td></tr>
120 <tr><td><a href="#scalar-evolution">-scalar-evolution</a></td><td>Scalar Evolution Analysis</td></tr>
121 <tr><td><a href="#scev-aa">-scev-aa</a></td><td>ScalarEvolution-based Alias Analysis</td></tr>
122 <tr><td><a href="#targetdata">-targetdata</a></td><td>Target Data Layout</td></tr>
125 <tr><th colspan="2"><b>TRANSFORM PASSES</b></th></tr>
126 <tr><th>Option</th><th>Name</th></tr>
127 <tr><td><a href="#adce">-adce</a></td><td>Aggressive Dead Code Elimination</td></tr>
128 <tr><td><a href="#always-inline">-always-inline</a></td><td>Inliner for always_inline functions</td></tr>
129 <tr><td><a href="#argpromotion">-argpromotion</a></td><td>Promote 'by reference' arguments to scalars</td></tr>
130 <tr><td><a href="#block-placement">-block-placement</a></td><td>Profile Guided Basic Block Placement</td></tr>
131 <tr><td><a href="#break-crit-edges">-break-crit-edges</a></td><td>Break critical edges in CFG</td></tr>
132 <tr><td><a href="#codegenprepare">-codegenprepare</a></td><td>Optimize for code generation</td></tr>
133 <tr><td><a href="#constmerge">-constmerge</a></td><td>Merge Duplicate Global Constants</td></tr>
134 <tr><td><a href="#constprop">-constprop</a></td><td>Simple constant propagation</td></tr>
135 <tr><td><a href="#dce">-dce</a></td><td>Dead Code Elimination</td></tr>
136 <tr><td><a href="#deadargelim">-deadargelim</a></td><td>Dead Argument Elimination</td></tr>
137 <tr><td><a href="#deadtypeelim">-deadtypeelim</a></td><td>Dead Type Elimination</td></tr>
138 <tr><td><a href="#die">-die</a></td><td>Dead Instruction Elimination</td></tr>
139 <tr><td><a href="#dse">-dse</a></td><td>Dead Store Elimination</td></tr>
140 <tr><td><a href="#functionattrs">-functionattrs</a></td><td>Deduce function attributes</td></tr>
141 <tr><td><a href="#globaldce">-globaldce</a></td><td>Dead Global Elimination</td></tr>
142 <tr><td><a href="#globalopt">-globalopt</a></td><td>Global Variable Optimizer</td></tr>
143 <tr><td><a href="#gvn">-gvn</a></td><td>Global Value Numbering</td></tr>
144 <tr><td><a href="#indvars">-indvars</a></td><td>Canonicalize Induction Variables</td></tr>
145 <tr><td><a href="#inline">-inline</a></td><td>Function Integration/Inlining</td></tr>
146 <tr><td><a href="#insert-edge-profiling">-insert-edge-profiling</a></td><td>Insert instrumentation for edge profiling</td></tr>
147 <tr><td><a href="#insert-optimal-edge-profiling">-insert-optimal-edge-profiling</a></td><td>Insert optimal instrumentation for edge profiling</td></tr>
148 <tr><td><a href="#instcombine">-instcombine</a></td><td>Combine redundant instructions</td></tr>
149 <tr><td><a href="#internalize">-internalize</a></td><td>Internalize Global Symbols</td></tr>
150 <tr><td><a href="#ipconstprop">-ipconstprop</a></td><td>Interprocedural constant propagation</td></tr>
151 <tr><td><a href="#ipsccp">-ipsccp</a></td><td>Interprocedural Sparse Conditional Constant Propagation</td></tr>
152 <tr><td><a href="#jump-threading">-jump-threading</a></td><td>Jump Threading</td></tr>
153 <tr><td><a href="#lcssa">-lcssa</a></td><td>Loop-Closed SSA Form Pass</td></tr>
154 <tr><td><a href="#licm">-licm</a></td><td>Loop Invariant Code Motion</td></tr>
155 <tr><td><a href="#loop-deletion">-loop-deletion</a></td><td>Delete dead loops</td></tr>
156 <tr><td><a href="#loop-extract">-loop-extract</a></td><td>Extract loops into new functions</td></tr>
157 <tr><td><a href="#loop-extract-single">-loop-extract-single</a></td><td>Extract at most one loop into a new function</td></tr>
158 <tr><td><a href="#loop-reduce">-loop-reduce</a></td><td>Loop Strength Reduction</td></tr>
159 <tr><td><a href="#loop-rotate">-loop-rotate</a></td><td>Rotate Loops</td></tr>
160 <tr><td><a href="#loop-simplify">-loop-simplify</a></td><td>Canonicalize natural loops</td></tr>
161 <tr><td><a href="#loop-unroll">-loop-unroll</a></td><td>Unroll loops</td></tr>
162 <tr><td><a href="#loop-unswitch">-loop-unswitch</a></td><td>Unswitch loops</td></tr>
163 <tr><td><a href="#loweratomic">-loweratomic</a></td><td>Lower atomic intrinsics to non-atomic form</td></tr>
164 <tr><td><a href="#lowerinvoke">-lowerinvoke</a></td><td>Lower invoke and unwind, for unwindless code generators</td></tr>
165 <tr><td><a href="#lowersetjmp">-lowersetjmp</a></td><td>Lower Set Jump</td></tr>
166 <tr><td><a href="#lowerswitch">-lowerswitch</a></td><td>Lower SwitchInst's to branches</td></tr>
167 <tr><td><a href="#mem2reg">-mem2reg</a></td><td>Promote Memory to Register</td></tr>
168 <tr><td><a href="#memcpyopt">-memcpyopt</a></td><td>MemCpy Optimization</td></tr>
169 <tr><td><a href="#mergefunc">-mergefunc</a></td><td>Merge Functions</td></tr>
170 <tr><td><a href="#mergereturn">-mergereturn</a></td><td>Unify function exit nodes</td></tr>
171 <tr><td><a href="#partial-inliner">-partial-inliner</a></td><td>Partial Inliner</td></tr>
172 <tr><td><a href="#prune-eh">-prune-eh</a></td><td>Remove unused exception handling info</td></tr>
173 <tr><td><a href="#reassociate">-reassociate</a></td><td>Reassociate expressions</td></tr>
174 <tr><td><a href="#reg2mem">-reg2mem</a></td><td>Demote all values to stack slots</td></tr>
175 <tr><td><a href="#scalarrepl">-scalarrepl</a></td><td>Scalar Replacement of Aggregates (DT)</td></tr>
176 <tr><td><a href="#sccp">-sccp</a></td><td>Sparse Conditional Constant Propagation</td></tr>
177 <tr><td><a href="#simplify-libcalls">-simplify-libcalls</a></td><td>Simplify well-known library calls</td></tr>
178 <tr><td><a href="#simplifycfg">-simplifycfg</a></td><td>Simplify the CFG</td></tr>
179 <tr><td><a href="#sink">-sink</a></td><td>Code sinking</td></tr>
180 <tr><td><a href="#sretpromotion">-sretpromotion</a></td><td>Promote sret arguments to multiple ret values</td></tr>
181 <tr><td><a href="#strip">-strip</a></td><td>Strip all symbols from a module</td></tr>
182 <tr><td><a href="#strip-dead-debug-info">-strip-dead-debug-info</a></td><td>Strip debug info for unused symbols</td></tr>
183 <tr><td><a href="#strip-dead-prototypes">-strip-dead-prototypes</a></td><td>Strip Unused Function Prototypes</td></tr>
184 <tr><td><a href="#strip-debug-declare">-strip-debug-declare</a></td><td>Strip all llvm.dbg.declare intrinsics</td></tr>
185 <tr><td><a href="#strip-nondebug">-strip-nondebug</a></td><td>Strip all symbols, except dbg symbols, from a module</td></tr>
186 <tr><td><a href="#tailcallelim">-tailcallelim</a></td><td>Tail Call Elimination</td></tr>
187 <tr><td><a href="#tailduplicate">-tailduplicate</a></td><td>Tail Duplication</td></tr>
190 <tr><th colspan="2"><b>UTILITY PASSES</b></th></tr>
191 <tr><th>Option</th><th>Name</th></tr>
192 <tr><td><a href="#deadarghaX0r">-deadarghaX0r</a></td><td>Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</td></tr>
193 <tr><td><a href="#extract-blocks">-extract-blocks</a></td><td>Extract Basic Blocks From Module (for bugpoint use)</td></tr>
194 <tr><td><a href="#instnamer">-instnamer</a></td><td>Assign names to anonymous instructions</td></tr>
195 <tr><td><a href="#preverify">-preverify</a></td><td>Preliminary module verification</td></tr>
196 <tr><td><a href="#verify">-verify</a></td><td>Module Verifier</td></tr>
197 <tr><td><a href="#view-cfg">-view-cfg</a></td><td>View CFG of function</td></tr>
198 <tr><td><a href="#view-cfg-only">-view-cfg-only</a></td><td>View CFG of function (with no function bodies)</td></tr>
199 <tr><td><a href="#view-dom">-view-dom</a></td><td>View dominance tree of function</td></tr>
200 <tr><td><a href="#view-dom-only">-view-dom-only</a></td><td>View dominance tree of function (with no function bodies)</td></tr>
201 <tr><td><a href="#view-postdom">-view-postdom</a></td><td>View postdominance tree of function</td></tr>
202 <tr><td><a href="#view-postdom-only">-view-postdom-only</a></td><td>View postdominance tree of function (with no function bodies)</td></tr>
206 <!-- ======================================================================= -->
207 <div class="doc_section"> <a name="example">Analysis Passes</a></div>
208 <div class="doc_text">
209 <p>This section describes the LLVM Analysis Passes.</p>
212 <!-------------------------------------------------------------------------- -->
213 <div class="doc_subsection">
214 <a name="aa-eval">-aa-eval: Exhaustive Alias Analysis Precision Evaluator</a>
216 <div class="doc_text">
217 <p>This is a simple N^2 alias analysis accuracy evaluator.
218 Basically, for each function in the program, it simply queries to see how the
219 alias analysis implementation answers alias queries between each pair of
220 pointers in the function.</p>
222 <p>This is inspired and adapted from code by: Naveen Neelakantam, Francesco
223 Spadini, and Wojciech Stryjewski.</p>
226 <!-------------------------------------------------------------------------- -->
227 <div class="doc_subsection">
228 <a name="basicaa">-basicaa: Basic Alias Analysis (stateless AA impl)</a>
230 <div class="doc_text">
232 This is the default implementation of the Alias Analysis interface
233 that simply implements a few identities (two different globals cannot alias,
234 etc), but otherwise does no analysis.
238 <!-------------------------------------------------------------------------- -->
239 <div class="doc_subsection">
240 <a name="basiccg">-basiccg: Basic CallGraph Construction</a>
242 <div class="doc_text">
243 <p>Yet to be written.</p>
246 <!-------------------------------------------------------------------------- -->
247 <div class="doc_subsection">
248 <a name="codegenprepare">-codegenprepare: Optimize for code generation</a>
250 <div class="doc_text">
252 This pass munges the code in the input function to better prepare it for
253 SelectionDAG-based code generation. This works around limitations in it's
254 basic-block-at-a-time approach. It should eventually be removed.
258 <!-------------------------------------------------------------------------- -->
259 <div class="doc_subsection">
260 <a name="count-aa">-count-aa: Count Alias Analysis Query Responses</a>
262 <div class="doc_text">
264 A pass which can be used to count how many alias queries
265 are being made and how the alias analysis implementation being used responds.
269 <!-------------------------------------------------------------------------- -->
270 <div class="doc_subsection">
271 <a name="debug-aa">-debug-aa: AA use debugger</a>
273 <div class="doc_text">
275 This simple pass checks alias analysis users to ensure that if they
276 create a new value, they do not query AA without informing it of the value.
277 It acts as a shim over any other AA pass you want.
281 Yes keeping track of every value in the program is expensive, but this is
286 <!-------------------------------------------------------------------------- -->
287 <div class="doc_subsection">
288 <a name="domfrontier">-domfrontier: Dominance Frontier Construction</a>
290 <div class="doc_text">
292 This pass is a simple dominator construction algorithm for finding forward
297 <!-------------------------------------------------------------------------- -->
298 <div class="doc_subsection">
299 <a name="domtree">-domtree: Dominator Tree Construction</a>
301 <div class="doc_text">
303 This pass is a simple dominator construction algorithm for finding forward
308 <!-------------------------------------------------------------------------- -->
309 <div class="doc_subsection">
310 <a name="dot-callgraph">-dot-callgraph: Print Call Graph to 'dot' file</a>
312 <div class="doc_text">
314 This pass, only available in <code>opt</code>, prints the call graph into a
315 <code>.dot</code> graph. This graph can then be processed with the "dot" tool
316 to convert it to postscript or some other suitable format.
320 <!-------------------------------------------------------------------------- -->
321 <div class="doc_subsection">
322 <a name="dot-cfg">-dot-cfg: Print CFG of function to 'dot' file</a>
324 <div class="doc_text">
326 This pass, only available in <code>opt</code>, prints the control flow graph
327 into a <code>.dot</code> graph. This graph can then be processed with the
328 "dot" tool to convert it to postscript or some other suitable format.
332 <!-------------------------------------------------------------------------- -->
333 <div class="doc_subsection">
334 <a name="dot-cfg-only">-dot-cfg-only: Print CFG of function to 'dot' file (with no function bodies)</a>
336 <div class="doc_text">
338 This pass, only available in <code>opt</code>, prints the control flow graph
339 into a <code>.dot</code> graph, omitting the function bodies. This graph can
340 then be processed with the "dot" tool to convert it to postscript or some
341 other suitable format.
345 <!-------------------------------------------------------------------------- -->
346 <div class="doc_subsection">
347 <a name="dot-dom">-dot-dom: Print dominator tree of function to 'dot' file</a>
349 <div class="doc_text">
351 This pass, only available in <code>opt</code>, prints the dominator tree
352 into a <code>.dot</code> graph. This graph can then be processed with the
353 "dot" tool to convert it to postscript or some other suitable format.
357 <!-------------------------------------------------------------------------- -->
358 <div class="doc_subsection">
359 <a name="dot-dom-only">-dot-dom-only: Print dominator tree of function to 'dot' file (with no
362 <div class="doc_text">
364 This pass, only available in <code>opt</code>, prints the dominator tree
365 into a <code>.dot</code> graph, omitting the function bodies. This graph can
366 then be processed with the "dot" tool to convert it to postscript or some
367 other suitable format.
371 <!-------------------------------------------------------------------------- -->
372 <div class="doc_subsection">
373 <a name="dot-postdom">-dot-postdom: Print post dominator tree of function to 'dot' file</a>
375 <div class="doc_text">
377 This pass, only available in <code>opt</code>, prints the post dominator tree
378 into a <code>.dot</code> graph. This graph can then be processed with the
379 "dot" tool to convert it to postscript or some other suitable format.
383 <!-------------------------------------------------------------------------- -->
384 <div class="doc_subsection">
385 <a name="dot-postdom-only">-dot-postdom-only: Print post dominator tree of function to 'dot' file
386 (with no function bodies)</a>
388 <div class="doc_text">
390 This pass, only available in <code>opt</code>, prints the post dominator tree
391 into a <code>.dot</code> graph, omitting the function bodies. This graph can
392 then be processed with the "dot" tool to convert it to postscript or some
393 other suitable format.
397 <!-------------------------------------------------------------------------- -->
398 <div class="doc_subsection">
399 <a name="globalsmodref-aa">-globalsmodref-aa: Simple mod/ref analysis for globals</a>
401 <div class="doc_text">
403 This simple pass provides alias and mod/ref information for global values
404 that do not have their address taken, and keeps track of whether functions
405 read or write memory (are "pure"). For this simple (but very common) case,
406 we can provide pretty accurate and useful information.
410 <!-------------------------------------------------------------------------- -->
411 <div class="doc_subsection">
412 <a name="instcount">-instcount: Counts the various types of Instructions</a>
414 <div class="doc_text">
416 This pass collects the count of all instructions and reports them
420 <!-------------------------------------------------------------------------- -->
421 <div class="doc_subsection">
422 <a name="interprocedural-aa-eval">-interprocedural-aa-eval: Exhaustive Interprocedural Alias Analysis Precision Evaluator</a>
424 <div class="doc_text">
425 <p>This pass implements a simple N^2 alias analysis accuracy evaluator.
426 Basically, for each function in the program, it simply queries to see how the
427 alias analysis implementation answers alias queries between each pair of
428 pointers in the function.
432 <!-------------------------------------------------------------------------- -->
433 <div class="doc_subsection">
434 <a name="interprocedural-basic-aa">-interprocedural-basic-aa: Interprocedural Basic Alias Analysis</a>
436 <div class="doc_text">
437 <p>This pass defines the default implementation of the Alias Analysis interface
438 that simply implements a few identities (two different globals cannot alias,
439 etc), but otherwise does no analysis.
443 <!-------------------------------------------------------------------------- -->
444 <div class="doc_subsection">
445 <a name="intervals">-intervals: Interval Partition Construction</a>
447 <div class="doc_text">
449 This analysis calculates and represents the interval partition of a function,
450 or a preexisting interval partition.
454 In this way, the interval partition may be used to reduce a flow graph down
455 to its degenerate single node interval partition (unless it is irreducible).
459 <!-------------------------------------------------------------------------- -->
460 <div class="doc_subsection">
461 <a name="iv-users">-iv-users: Induction Variable Users</a>
463 <div class="doc_text">
464 <p>Bookkeeping for "interesting" users of expressions computed from
465 induction variables.</p>
468 <!-------------------------------------------------------------------------- -->
469 <div class="doc_subsection">
470 <a name="lazy-value-info">-lazy-value-info: Lazy Value Information Analysis</a>
472 <div class="doc_text">
473 <p>Interface for lazy computation of value constraint information.</p>
476 <!-------------------------------------------------------------------------- -->
477 <div class="doc_subsection">
478 <a name="lda">-lda: Loop Dependence Analysis</a>
480 <div class="doc_text">
481 <p>Loop dependence analysis framework, which is used to detect dependences in
482 memory accesses in loops.</p>
485 <!-------------------------------------------------------------------------- -->
486 <div class="doc_subsection">
487 <a name="libcall-aa">-libcall-aa: LibCall Alias Analysis</a>
489 <div class="doc_text">
490 <p>LibCall Alias Analysis.</p>
493 <!-------------------------------------------------------------------------- -->
494 <div class="doc_subsection">
495 <a name="lint">-lint: Statically lint-checks LLVM IR</a>
497 <div class="doc_text">
498 <p>This pass statically checks for common and easily-identified constructs
499 which produce undefined or likely unintended behavior in LLVM IR.</p>
501 <p>It is not a guarantee of correctness, in two ways. First, it isn't
502 comprehensive. There are checks which could be done statically which are
503 not yet implemented. Some of these are indicated by TODO comments, but
504 those aren't comprehensive either. Second, many conditions cannot be
505 checked statically. This pass does no dynamic instrumentation, so it
506 can't check for all possible problems.</p>
508 <p>Another limitation is that it assumes all code will be executed. A store
509 through a null pointer in a basic block which is never reached is harmless,
510 but this pass will warn about it anyway.</p>
512 <p>Optimization passes may make conditions that this pass checks for more or
513 less obvious. If an optimization pass appears to be introducing a warning,
514 it may be that the optimization pass is merely exposing an existing
515 condition in the code.</p>
517 <p>This code may be run before instcombine. In many cases, instcombine checks
518 for the same kinds of things and turns instructions with undefined behavior
519 into unreachable (or equivalent). Because of this, this pass makes some
520 effort to look through bitcasts and so on.
524 <!-------------------------------------------------------------------------- -->
525 <div class="doc_subsection">
526 <a name="loops">-loops: Natural Loop Information</a>
528 <div class="doc_text">
530 This analysis is used to identify natural loops and determine the loop depth
531 of various nodes of the CFG. Note that the loops identified may actually be
532 several natural loops that share the same header node... not just a single
537 <!-------------------------------------------------------------------------- -->
538 <div class="doc_subsection">
539 <a name="memdep">-memdep: Memory Dependence Analysis</a>
541 <div class="doc_text">
543 An analysis that determines, for a given memory operation, what preceding
544 memory operations it depends on. It builds on alias analysis information, and
545 tries to provide a lazy, caching interface to a common kind of alias
550 <!-------------------------------------------------------------------------- -->
551 <div class="doc_subsection">
552 <a name="module-debuginfo">-module-debuginfo: Decodes module-level debug info</a>
554 <div class="doc_text">
555 <p>This pass decodes the debug info metadata in a module and prints in a
556 (sufficiently-prepared-) human-readable form.
558 For example, run this pass from opt along with the -analyze option, and
559 it'll print to standard output.
563 <!-------------------------------------------------------------------------- -->
564 <div class="doc_subsection">
565 <a name="no-aa">-no-aa: No Alias Analysis (always returns 'may' alias)</a>
567 <div class="doc_text">
569 Always returns "I don't know" for alias queries. NoAA is unlike other alias
570 analysis implementations, in that it does not chain to a previous analysis. As
571 such it doesn't follow many of the rules that other alias analyses must.
575 <!-------------------------------------------------------------------------- -->
576 <div class="doc_subsection">
577 <a name="no-profile">-no-profile: No Profile Information</a>
579 <div class="doc_text">
581 The default "no profile" implementation of the abstract
582 <code>ProfileInfo</code> interface.
586 <!-------------------------------------------------------------------------- -->
587 <div class="doc_subsection">
588 <a name="pointertracking">-pointertracking: Track pointer bounds.</a>
590 <div class="doc_text">
591 <p>Tracking of pointer bounds.
595 <!-------------------------------------------------------------------------- -->
596 <div class="doc_subsection">
597 <a name="postdomfrontier">-postdomfrontier: Post-Dominance Frontier Construction</a>
599 <div class="doc_text">
601 This pass is a simple post-dominator construction algorithm for finding
602 post-dominator frontiers.
606 <!-------------------------------------------------------------------------- -->
607 <div class="doc_subsection">
608 <a name="postdomtree">-postdomtree: Post-Dominator Tree Construction</a>
610 <div class="doc_text">
612 This pass is a simple post-dominator construction algorithm for finding
617 <!-------------------------------------------------------------------------- -->
618 <div class="doc_subsection">
619 <a name="print-alias-sets">-print-alias-sets: Alias Set Printer</a>
621 <div class="doc_text">
622 <p>Yet to be written.</p>
625 <!-------------------------------------------------------------------------- -->
626 <div class="doc_subsection">
627 <a name="print-callgraph">-print-callgraph: Print a call graph</a>
629 <div class="doc_text">
631 This pass, only available in <code>opt</code>, prints the call graph to
632 standard error in a human-readable form.
636 <!-------------------------------------------------------------------------- -->
637 <div class="doc_subsection">
638 <a name="print-callgraph-sccs">-print-callgraph-sccs: Print SCCs of the Call Graph</a>
640 <div class="doc_text">
642 This pass, only available in <code>opt</code>, prints the SCCs of the call
643 graph to standard error in a human-readable form.
647 <!-------------------------------------------------------------------------- -->
648 <div class="doc_subsection">
649 <a name="print-cfg-sccs">-print-cfg-sccs: Print SCCs of each function CFG</a>
651 <div class="doc_text">
653 This pass, only available in <code>opt</code>, prints the SCCs of each
654 function CFG to standard error in a human-readable form.
658 <!-------------------------------------------------------------------------- -->
659 <div class="doc_subsection">
660 <a name="print-dbginfo">-print-dbginfo: Print debug info in human readable form</a>
662 <div class="doc_text">
663 <p>Pass that prints instructions, and associated debug info:</p>
666 <li>source/line/col information</li>
667 <li>original variable name</li>
668 <li>original type name</li>
672 <!-------------------------------------------------------------------------- -->
673 <div class="doc_subsection">
674 <a name="print-dom-info">-print-dom-info: Dominator Info Printer</a>
676 <div class="doc_text">
677 <p>Dominator Info Printer.</p>
680 <!-------------------------------------------------------------------------- -->
681 <div class="doc_subsection">
682 <a name="print-externalfnconstants">-print-externalfnconstants: Print external fn callsites passed constants</a>
684 <div class="doc_text">
686 This pass, only available in <code>opt</code>, prints out call sites to
687 external functions that are called with constant arguments. This can be
688 useful when looking for standard library functions we should constant fold
689 or handle in alias analyses.
693 <!-------------------------------------------------------------------------- -->
694 <div class="doc_subsection">
695 <a name="print-function">-print-function: Print function to stderr</a>
697 <div class="doc_text">
699 The <code>PrintFunctionPass</code> class is designed to be pipelined with
700 other <code>FunctionPass</code>es, and prints out the functions of the module
701 as they are processed.
705 <!-------------------------------------------------------------------------- -->
706 <div class="doc_subsection">
707 <a name="print-module">-print-module: Print module to stderr</a>
709 <div class="doc_text">
711 This pass simply prints out the entire module when it is executed.
715 <!-------------------------------------------------------------------------- -->
716 <div class="doc_subsection">
717 <a name="print-used-types">-print-used-types: Find Used Types</a>
719 <div class="doc_text">
721 This pass is used to seek out all of the types in use by the program. Note
722 that this analysis explicitly does not include types only used by the symbol
726 <!-------------------------------------------------------------------------- -->
727 <div class="doc_subsection">
728 <a name="profile-estimator">-profile-estimator: Estimate profiling information</a>
730 <div class="doc_text">
731 <p>Profiling information that estimates the profiling information
732 in a very crude and unimaginative way.
736 <!-------------------------------------------------------------------------- -->
737 <div class="doc_subsection">
738 <a name="profile-loader">-profile-loader: Load profile information from llvmprof.out</a>
740 <div class="doc_text">
742 A concrete implementation of profiling information that loads the information
743 from a profile dump file.
747 <!-------------------------------------------------------------------------- -->
748 <div class="doc_subsection">
749 <a name="profile-verifier">-profile-verifier: Verify profiling information</a>
751 <div class="doc_text">
752 <p>Pass that checks profiling information for plausibility.</p>
754 <div class="doc_subsection">
755 <a name="regions">-regions: Detect single entry single exit regions in a function</a>
757 <div class="doc_text">
759 The <code>RegionInfo</code> pass detects single entry single exit regions in a
760 function, where a region is defined as any subgraph that is connected to the
761 remaining graph at only two spots. Furthermore, an hierarchical region tree is
766 <!-------------------------------------------------------------------------- -->
767 <div class="doc_subsection">
768 <a name="scalar-evolution">-scalar-evolution: Scalar Evolution Analysis</a>
770 <div class="doc_text">
772 The <code>ScalarEvolution</code> analysis can be used to analyze and
773 catagorize scalar expressions in loops. It specializes in recognizing general
774 induction variables, representing them with the abstract and opaque
775 <code>SCEV</code> class. Given this analysis, trip counts of loops and other
776 important properties can be obtained.
780 This analysis is primarily useful for induction variable substitution and
785 <!-------------------------------------------------------------------------- -->
786 <div class="doc_subsection">
787 <a name="scev-aa">-scev-aa: </a>
789 <div class="doc_text">
790 <p>Simple alias analysis implemented in terms of ScalarEvolution queries.
792 This differs from traditional loop dependence analysis in that it tests
793 for dependencies within a single iteration of a loop, rather than
794 dependencies between different iterations.
796 ScalarEvolution has a more complete understanding of pointer arithmetic
797 than BasicAliasAnalysis' collection of ad-hoc analyses.
801 <!-------------------------------------------------------------------------- -->
802 <div class="doc_subsection">
803 <a name="strip-dead-debug-info">-strip-dead-debug-info: Strip debug info for unused symbols</a>
805 <div class="doc_text">
807 performs code stripping. this transformation can delete:
811 <li>names for virtual registers</li>
812 <li>symbols for internal globals and functions</li>
813 <li>debug information</li>
817 note that this transformation makes code much less readable, so it should
818 only be used in situations where the <tt>strip</tt> utility would be used,
819 such as reducing code size or making it harder to reverse engineer code.
823 <!-------------------------------------------------------------------------- -->
824 <div class="doc_subsection">
825 <a name="targetdata">-targetdata: Target Data Layout</a>
827 <div class="doc_text">
828 <p>Provides other passes access to information on how the size and alignment
829 required by the the target ABI for various data types.</p>
832 <!-- ======================================================================= -->
833 <div class="doc_section"> <a name="transform">Transform Passes</a></div>
834 <div class="doc_text">
835 <p>This section describes the LLVM Transform Passes.</p>
838 <!-------------------------------------------------------------------------- -->
839 <div class="doc_subsection">
840 <a name="adce">-adce: Aggressive Dead Code Elimination</a>
842 <div class="doc_text">
843 <p>ADCE aggressively tries to eliminate code. This pass is similar to
844 <a href="#dce">DCE</a> but it assumes that values are dead until proven
845 otherwise. This is similar to <a href="#sccp">SCCP</a>, except applied to
846 the liveness of values.</p>
849 <!-------------------------------------------------------------------------- -->
850 <div class="doc_subsection">
851 <a name="always-inline">-always-inline: Inliner for always_inline functions</a>
853 <div class="doc_text">
854 <p>A custom inliner that handles only functions that are marked as
858 <!-------------------------------------------------------------------------- -->
859 <div class="doc_subsection">
860 <a name="argpromotion">-argpromotion: Promote 'by reference' arguments to scalars</a>
862 <div class="doc_text">
864 This pass promotes "by reference" arguments to be "by value" arguments. In
865 practice, this means looking for internal functions that have pointer
866 arguments. If it can prove, through the use of alias analysis, that an
867 argument is *only* loaded, then it can pass the value into the function
868 instead of the address of the value. This can cause recursive simplification
869 of code and lead to the elimination of allocas (especially in C++ template
874 This pass also handles aggregate arguments that are passed into a function,
875 scalarizing them if the elements of the aggregate are only loaded. Note that
876 it refuses to scalarize aggregates which would require passing in more than
877 three operands to the function, because passing thousands of operands for a
878 large array or structure is unprofitable!
882 Note that this transformation could also be done for arguments that are only
883 stored to (returning the value instead), but does not currently. This case
884 would be best handled when and if LLVM starts supporting multiple return
885 values from functions.
889 <!-------------------------------------------------------------------------- -->
890 <div class="doc_subsection">
891 <a name="block-placement">-block-placement: Profile Guided Basic Block Placement</a>
893 <div class="doc_text">
894 <p>This pass is a very simple profile guided basic block placement algorithm.
895 The idea is to put frequently executed blocks together at the start of the
896 function and hopefully increase the number of fall-through conditional
897 branches. If there is no profile information for a particular function, this
898 pass basically orders blocks in depth-first order.</p>
901 <!-------------------------------------------------------------------------- -->
902 <div class="doc_subsection">
903 <a name="break-crit-edges">-break-crit-edges: Break critical edges in CFG</a>
905 <div class="doc_text">
907 Break all of the critical edges in the CFG by inserting a dummy basic block.
908 It may be "required" by passes that cannot deal with critical edges. This
909 transformation obviously invalidates the CFG, but can update forward dominator
910 (set, immediate dominators, tree, and frontier) information.
914 <!-------------------------------------------------------------------------- -->
915 <div class="doc_subsection">
916 <a name="codegenprepare">-codegenprepare: Prepare a function for code generation</a>
918 <div class="doc_text">
919 This pass munges the code in the input function to better prepare it for
920 SelectionDAG-based code generation. This works around limitations in it's
921 basic-block-at-a-time approach. It should eventually be removed.
924 <!-------------------------------------------------------------------------- -->
925 <div class="doc_subsection">
926 <a name="constmerge">-constmerge: Merge Duplicate Global Constants</a>
928 <div class="doc_text">
930 Merges duplicate global constants together into a single constant that is
931 shared. This is useful because some passes (ie TraceValues) insert a lot of
932 string constants into the program, regardless of whether or not an existing
937 <!-------------------------------------------------------------------------- -->
938 <div class="doc_subsection">
939 <a name="constprop">-constprop: Simple constant propagation</a>
941 <div class="doc_text">
942 <p>This file implements constant propagation and merging. It looks for
943 instructions involving only constant operands and replaces them with a
944 constant value instead of an instruction. For example:</p>
945 <blockquote><pre>add i32 1, 2</pre></blockquote>
947 <blockquote><pre>i32 3</pre></blockquote>
948 <p>NOTE: this pass has a habit of making definitions be dead. It is a good
949 idea to to run a <a href="#die">DIE</a> (Dead Instruction Elimination) pass
950 sometime after running this pass.</p>
953 <!-------------------------------------------------------------------------- -->
954 <div class="doc_subsection">
955 <a name="dce">-dce: Dead Code Elimination</a>
957 <div class="doc_text">
959 Dead code elimination is similar to <a href="#die">dead instruction
960 elimination</a>, but it rechecks instructions that were used by removed
961 instructions to see if they are newly dead.
965 <!-------------------------------------------------------------------------- -->
966 <div class="doc_subsection">
967 <a name="deadargelim">-deadargelim: Dead Argument Elimination</a>
969 <div class="doc_text">
971 This pass deletes dead arguments from internal functions. Dead argument
972 elimination removes arguments which are directly dead, as well as arguments
973 only passed into function calls as dead arguments of other functions. This
974 pass also deletes dead arguments in a similar way.
978 This pass is often useful as a cleanup pass to run after aggressive
979 interprocedural passes, which add possibly-dead arguments.
983 <!-------------------------------------------------------------------------- -->
984 <div class="doc_subsection">
985 <a name="deadtypeelim">-deadtypeelim: Dead Type Elimination</a>
987 <div class="doc_text">
989 This pass is used to cleanup the output of GCC. It eliminate names for types
990 that are unused in the entire translation unit, using the <a
991 href="#findusedtypes">find used types</a> pass.
995 <!-------------------------------------------------------------------------- -->
996 <div class="doc_subsection">
997 <a name="die">-die: Dead Instruction Elimination</a>
999 <div class="doc_text">
1001 Dead instruction elimination performs a single pass over the function,
1002 removing instructions that are obviously dead.
1006 <!-------------------------------------------------------------------------- -->
1007 <div class="doc_subsection">
1008 <a name="dse">-dse: Dead Store Elimination</a>
1010 <div class="doc_text">
1012 A trivial dead store elimination that only considers basic-block local
1017 <!-------------------------------------------------------------------------- -->
1018 <div class="doc_subsection">
1019 <a name="functionattrs">-functionattrs: Deduce function attributes</a>
1021 <div class="doc_text">
1022 <p>A simple interprocedural pass which walks the call-graph, looking for
1023 functions which do not access or only read non-local memory, and marking them
1024 readnone/readonly. In addition, it marks function arguments (of pointer type)
1025 'nocapture' if a call to the function does not create any copies of the pointer
1026 value that outlive the call. This more or less means that the pointer is only
1027 dereferenced, and not returned from the function or stored in a global.
1028 This pass is implemented as a bottom-up traversal of the call-graph.
1032 <!-------------------------------------------------------------------------- -->
1033 <div class="doc_subsection">
1034 <a name="globaldce">-globaldce: Dead Global Elimination</a>
1036 <div class="doc_text">
1038 This transform is designed to eliminate unreachable internal globals from the
1039 program. It uses an aggressive algorithm, searching out globals that are
1040 known to be alive. After it finds all of the globals which are needed, it
1041 deletes whatever is left over. This allows it to delete recursive chunks of
1042 the program which are unreachable.
1046 <!-------------------------------------------------------------------------- -->
1047 <div class="doc_subsection">
1048 <a name="globalopt">-globalopt: Global Variable Optimizer</a>
1050 <div class="doc_text">
1052 This pass transforms simple global variables that never have their address
1053 taken. If obviously true, it marks read/write globals as constant, deletes
1054 variables only stored to, etc.
1058 <!-------------------------------------------------------------------------- -->
1059 <div class="doc_subsection">
1060 <a name="gvn">-gvn: Global Value Numbering</a>
1062 <div class="doc_text">
1064 This pass performs global value numbering to eliminate fully and partially
1065 redundant instructions. It also performs redundant load elimination.
1069 <!-------------------------------------------------------------------------- -->
1070 <div class="doc_subsection">
1071 <a name="indvars">-indvars: Canonicalize Induction Variables</a>
1073 <div class="doc_text">
1075 This transformation analyzes and transforms the induction variables (and
1076 computations derived from them) into simpler forms suitable for subsequent
1077 analysis and transformation.
1081 This transformation makes the following changes to each loop with an
1082 identifiable induction variable:
1086 <li>All loops are transformed to have a <em>single</em> canonical
1087 induction variable which starts at zero and steps by one.</li>
1088 <li>The canonical induction variable is guaranteed to be the first PHI node
1089 in the loop header block.</li>
1090 <li>Any pointer arithmetic recurrences are raised to use array
1095 If the trip count of a loop is computable, this pass also makes the following
1100 <li>The exit condition for the loop is canonicalized to compare the
1101 induction value against the exit value. This turns loops like:
1102 <blockquote><pre>for (i = 7; i*i < 1000; ++i)</pre></blockquote>
1104 <blockquote><pre>for (i = 0; i != 25; ++i)</pre></blockquote></li>
1105 <li>Any use outside of the loop of an expression derived from the indvar
1106 is changed to compute the derived value outside of the loop, eliminating
1107 the dependence on the exit value of the induction variable. If the only
1108 purpose of the loop is to compute the exit value of some derived
1109 expression, this transformation will make the loop dead.</li>
1113 This transformation should be followed by strength reduction after all of the
1114 desired loop transformations have been performed. Additionally, on targets
1115 where it is profitable, the loop could be transformed to count down to zero
1116 (the "do loop" optimization).
1120 <!-------------------------------------------------------------------------- -->
1121 <div class="doc_subsection">
1122 <a name="inline">-inline: Function Integration/Inlining</a>
1124 <div class="doc_text">
1126 Bottom-up inlining of functions into callees.
1130 <!-------------------------------------------------------------------------- -->
1131 <div class="doc_subsection">
1132 <a name="insert-edge-profiling">-insert-edge-profiling: Insert instrumentation for edge profiling</a>
1134 <div class="doc_text">
1136 This pass instruments the specified program with counters for edge profiling.
1137 Edge profiling can give a reasonable approximation of the hot paths through a
1138 program, and is used for a wide variety of program transformations.
1142 Note that this implementation is very naïve. It inserts a counter for
1143 <em>every</em> edge in the program, instead of using control flow information
1144 to prune the number of counters inserted.
1148 <!-------------------------------------------------------------------------- -->
1149 <div class="doc_subsection">
1150 <a name="insert-optimal-edge-profiling">-insert-optimal-edge-profiling: Insert optimal instrumentation for edge profiling</a>
1152 <div class="doc_text">
1153 <p>This pass instruments the specified program with counters for edge profiling.
1154 Edge profiling can give a reasonable approximation of the hot paths through a
1155 program, and is used for a wide variety of program transformations.
1159 <!-------------------------------------------------------------------------- -->
1160 <div class="doc_subsection">
1161 <a name="instcombine">-instcombine: Combine redundant instructions</a>
1163 <div class="doc_text">
1165 Combine instructions to form fewer, simple
1166 instructions. This pass does not modify the CFG This pass is where algebraic
1167 simplification happens.
1171 This pass combines things like:
1176 %Z = add i32 %Y, 1</pre></blockquote>
1183 >%Z = add i32 %X, 2</pre></blockquote>
1186 This is a simple worklist driven algorithm.
1190 This pass guarantees that the following canonicalizations are performed on
1195 <li>If a binary operator has a constant operand, it is moved to the right-
1197 <li>Bitwise operators with constant operands are always grouped so that
1198 shifts are performed first, then <code>or</code>s, then
1199 <code>and</code>s, then <code>xor</code>s.</li>
1200 <li>Compare instructions are converted from <code><</code>,
1201 <code>></code>, <code>≤</code>, or <code>≥</code> to
1202 <code>=</code> or <code>≠</code> if possible.</li>
1203 <li>All <code>cmp</code> instructions on boolean values are replaced with
1204 logical operations.</li>
1205 <li><code>add <var>X</var>, <var>X</var></code> is represented as
1206 <code>mul <var>X</var>, 2</code> ⇒ <code>shl <var>X</var>, 1</code></li>
1207 <li>Multiplies with a constant power-of-two argument are transformed into
1213 <!-------------------------------------------------------------------------- -->
1214 <div class="doc_subsection">
1215 <a name="internalize">-internalize: Internalize Global Symbols</a>
1217 <div class="doc_text">
1219 This pass loops over all of the functions in the input module, looking for a
1220 main function. If a main function is found, all other functions and all
1221 global variables with initializers are marked as internal.
1225 <!-------------------------------------------------------------------------- -->
1226 <div class="doc_subsection">
1227 <a name="ipconstprop">-ipconstprop: Interprocedural constant propagation</a>
1229 <div class="doc_text">
1231 This pass implements an <em>extremely</em> simple interprocedural constant
1232 propagation pass. It could certainly be improved in many different ways,
1233 like using a worklist. This pass makes arguments dead, but does not remove
1234 them. The existing dead argument elimination pass should be run after this
1235 to clean up the mess.
1239 <!-------------------------------------------------------------------------- -->
1240 <div class="doc_subsection">
1241 <a name="ipsccp">-ipsccp: Interprocedural Sparse Conditional Constant Propagation</a>
1243 <div class="doc_text">
1245 An interprocedural variant of <a href="#sccp">Sparse Conditional Constant
1250 <!-------------------------------------------------------------------------- -->
1251 <div class="doc_subsection">
1252 <a name="jump-threading">-jump-threading: Jump Threading</a>
1254 <div class="doc_text">
1256 Jump threading tries to find distinct threads of control flow running through
1257 a basic block. This pass looks at blocks that have multiple predecessors and
1258 multiple successors. If one or more of the predecessors of the block can be
1259 proven to always cause a jump to one of the successors, we forward the edge
1260 from the predecessor to the successor by duplicating the contents of this
1264 An example of when this can occur is code like this:
1271 if (X < 3) {</pre>
1274 In this case, the unconditional branch at the end of the first if can be
1275 revectored to the false side of the second if.
1279 <!-------------------------------------------------------------------------- -->
1280 <div class="doc_subsection">
1281 <a name="lcssa">-lcssa: Loop-Closed SSA Form Pass</a>
1283 <div class="doc_text">
1285 This pass transforms loops by placing phi nodes at the end of the loops for
1286 all values that are live across the loop boundary. For example, it turns
1287 the left into the right code:
1291 >for (...) for (...)
1296 X3 = phi(X1, X2) X3 = phi(X1, X2)
1297 ... = X3 + 4 X4 = phi(X3)
1301 This is still valid LLVM; the extra phi nodes are purely redundant, and will
1302 be trivially eliminated by <code>InstCombine</code>. The major benefit of
1303 this transformation is that it makes many other loop optimizations, such as
1304 LoopUnswitching, simpler.
1308 <!-------------------------------------------------------------------------- -->
1309 <div class="doc_subsection">
1310 <a name="licm">-licm: Loop Invariant Code Motion</a>
1312 <div class="doc_text">
1314 This pass performs loop invariant code motion, attempting to remove as much
1315 code from the body of a loop as possible. It does this by either hoisting
1316 code into the preheader block, or by sinking code to the exit blocks if it is
1317 safe. This pass also promotes must-aliased memory locations in the loop to
1318 live in registers, thus hoisting and sinking "invariant" loads and stores.
1322 This pass uses alias analysis for two purposes:
1326 <li>Moving loop invariant loads and calls out of loops. If we can determine
1327 that a load or call inside of a loop never aliases anything stored to,
1328 we can hoist it or sink it like any other instruction.</li>
1329 <li>Scalar Promotion of Memory - If there is a store instruction inside of
1330 the loop, we try to move the store to happen AFTER the loop instead of
1331 inside of the loop. This can only happen if a few conditions are true:
1333 <li>The pointer stored through is loop invariant.</li>
1334 <li>There are no stores or loads in the loop which <em>may</em> alias
1335 the pointer. There are no calls in the loop which mod/ref the
1338 If these conditions are true, we can promote the loads and stores in the
1339 loop of the pointer to use a temporary alloca'd variable. We then use
1340 the mem2reg functionality to construct the appropriate SSA form for the
1344 <!-------------------------------------------------------------------------- -->
1345 <div class="doc_subsection">
1346 <a name="loop-deletion">-loop-deletion: Delete dead loops</a>
1348 <div class="doc_text">
1350 This file implements the Dead Loop Deletion Pass. This pass is responsible
1351 for eliminating loops with non-infinite computable trip counts that have no
1352 side effects or volatile instructions, and do not contribute to the
1353 computation of the function's return value.
1357 <!-------------------------------------------------------------------------- -->
1358 <div class="doc_subsection">
1359 <a name="loop-extract">-loop-extract: Extract loops into new functions</a>
1361 <div class="doc_text">
1363 A pass wrapper around the <code>ExtractLoop()</code> scalar transformation to
1364 extract each top-level loop into its own new function. If the loop is the
1365 <em>only</em> loop in a given function, it is not touched. This is a pass most
1366 useful for debugging via bugpoint.
1370 <!-------------------------------------------------------------------------- -->
1371 <div class="doc_subsection">
1372 <a name="loop-extract-single">-loop-extract-single: Extract at most one loop into a new function</a>
1374 <div class="doc_text">
1376 Similar to <a href="#loop-extract">Extract loops into new functions</a>,
1377 this pass extracts one natural loop from the program into a function if it
1378 can. This is used by bugpoint.
1382 <!-------------------------------------------------------------------------- -->
1383 <div class="doc_subsection">
1384 <a name="loop-reduce">-loop-reduce: Loop Strength Reduction</a>
1386 <div class="doc_text">
1388 This pass performs a strength reduction on array references inside loops that
1389 have as one or more of their components the loop induction variable. This is
1390 accomplished by creating a new value to hold the initial value of the array
1391 access for the first iteration, and then creating a new GEP instruction in
1392 the loop to increment the value by the appropriate amount.
1396 <!-------------------------------------------------------------------------- -->
1397 <div class="doc_subsection">
1398 <a name="loop-rotate">-loop-rotate: Rotate Loops</a>
1400 <div class="doc_text">
1401 <p>A simple loop rotation transformation.</p>
1404 <!-------------------------------------------------------------------------- -->
1405 <div class="doc_subsection">
1406 <a name="loop-unroll">-loop-unroll: Unroll loops</a>
1408 <div class="doc_text">
1410 This pass implements a simple loop unroller. It works best when loops have
1411 been canonicalized by the <a href="#indvars"><tt>-indvars</tt></a> pass,
1412 allowing it to determine the trip counts of loops easily.
1416 <!-------------------------------------------------------------------------- -->
1417 <div class="doc_subsection">
1418 <a name="loop-unswitch">-loop-unswitch: Unswitch loops</a>
1420 <div class="doc_text">
1422 This pass transforms loops that contain branches on loop-invariant conditions
1423 to have multiple loops. For example, it turns the left into the right code:
1435 This can increase the size of the code exponentially (doubling it every time
1436 a loop is unswitched) so we only unswitch if the resultant code will be
1437 smaller than a threshold.
1441 This pass expects LICM to be run before it to hoist invariant conditions out
1442 of the loop, to make the unswitching opportunity obvious.
1446 <!-------------------------------------------------------------------------- -->
1447 <div class="doc_subsection">
1448 <a name="loop-simplify">-loop-simplify: Canonicalize natural loops</a>
1450 <div class="doc_text">
1452 This pass performs several transformations to transform natural loops into a
1453 simpler form, which makes subsequent analyses and transformations simpler and
1458 Loop pre-header insertion guarantees that there is a single, non-critical
1459 entry edge from outside of the loop to the loop header. This simplifies a
1460 number of analyses and transformations, such as LICM.
1464 Loop exit-block insertion guarantees that all exit blocks from the loop
1465 (blocks which are outside of the loop that have predecessors inside of the
1466 loop) only have predecessors from inside of the loop (and are thus dominated
1467 by the loop header). This simplifies transformations such as store-sinking
1468 that are built into LICM.
1472 This pass also guarantees that loops will have exactly one backedge.
1476 Note that the simplifycfg pass will clean up blocks which are split out but
1477 end up being unnecessary, so usage of this pass should not pessimize
1482 This pass obviously modifies the CFG, but updates loop information and
1483 dominator information.
1487 <!-------------------------------------------------------------------------- -->
1488 <div class="doc_subsection">
1489 <a name="lowerallocs">-lowerallocs: Lower allocations from instructions to calls</a>
1491 <div class="doc_text">
1493 Turn <tt>malloc</tt> and <tt>free</tt> instructions into <tt>@malloc</tt> and
1494 <tt>@free</tt> calls.
1498 This is a target-dependent tranformation because it depends on the size of
1499 data types and alignment constraints.
1503 <!-------------------------------------------------------------------------- -->
1504 <div class="doc_subsection">
1505 <a name="loweratomic">-loweratomic: Lower atomic intrinsics</a>
1507 <div class="doc_text">
1509 This pass lowers atomic intrinsics to non-atomic form for use in a known
1510 non-preemptible environment.
1514 The pass does not verify that the environment is non-preemptible (in
1515 general this would require knowledge of the entire call graph of the
1516 program including any libraries which may not be available in bitcode form);
1517 it simply lowers every atomic intrinsic.
1521 <!-------------------------------------------------------------------------- -->
1522 <div class="doc_subsection">
1523 <a name="lowerinvoke">-lowerinvoke: Lower invoke and unwind, for unwindless code generators</a>
1525 <div class="doc_text">
1527 This transformation is designed for use by code generators which do not yet
1528 support stack unwinding. This pass supports two models of exception handling
1529 lowering, the 'cheap' support and the 'expensive' support.
1533 'Cheap' exception handling support gives the program the ability to execute
1534 any program which does not "throw an exception", by turning 'invoke'
1535 instructions into calls and by turning 'unwind' instructions into calls to
1536 abort(). If the program does dynamically use the unwind instruction, the
1537 program will print a message then abort.
1541 'Expensive' exception handling support gives the full exception handling
1542 support to the program at the cost of making the 'invoke' instruction
1543 really expensive. It basically inserts setjmp/longjmp calls to emulate the
1544 exception handling as necessary.
1548 Because the 'expensive' support slows down programs a lot, and EH is only
1549 used for a subset of the programs, it must be specifically enabled by the
1550 <tt>-enable-correct-eh-support</tt> option.
1554 Note that after this pass runs the CFG is not entirely accurate (exceptional
1555 control flow edges are not correct anymore) so only very simple things should
1556 be done after the lowerinvoke pass has run (like generation of native code).
1557 This should not be used as a general purpose "my LLVM-to-LLVM pass doesn't
1558 support the invoke instruction yet" lowering pass.
1562 <!-------------------------------------------------------------------------- -->
1563 <div class="doc_subsection">
1564 <a name="lowersetjmp">-lowersetjmp: Lower Set Jump</a>
1566 <div class="doc_text">
1568 Lowers <tt>setjmp</tt> and <tt>longjmp</tt> to use the LLVM invoke and unwind
1569 instructions as necessary.
1573 Lowering of <tt>longjmp</tt> is fairly trivial. We replace the call with a
1574 call to the LLVM library function <tt>__llvm_sjljeh_throw_longjmp()</tt>.
1575 This unwinds the stack for us calling all of the destructors for
1576 objects allocated on the stack.
1580 At a <tt>setjmp</tt> call, the basic block is split and the <tt>setjmp</tt>
1581 removed. The calls in a function that have a <tt>setjmp</tt> are converted to
1582 invoke where the except part checks to see if it's a <tt>longjmp</tt>
1583 exception and, if so, if it's handled in the function. If it is, then it gets
1584 the value returned by the <tt>longjmp</tt> and goes to where the basic block
1585 was split. <tt>invoke</tt> instructions are handled in a similar fashion with
1586 the original except block being executed if it isn't a <tt>longjmp</tt>
1587 except that is handled by that function.
1591 <!-------------------------------------------------------------------------- -->
1592 <div class="doc_subsection">
1593 <a name="lowerswitch">-lowerswitch: Lower SwitchInst's to branches</a>
1595 <div class="doc_text">
1597 Rewrites <tt>switch</tt> instructions with a sequence of branches, which
1598 allows targets to get away with not implementing the switch instruction until
1603 <!-------------------------------------------------------------------------- -->
1604 <div class="doc_subsection">
1605 <a name="mem2reg">-mem2reg: Promote Memory to Register</a>
1607 <div class="doc_text">
1609 This file promotes memory references to be register references. It promotes
1610 <tt>alloca</tt> instructions which only have <tt>load</tt>s and
1611 <tt>store</tt>s as uses. An <tt>alloca</tt> is transformed by using dominator
1612 frontiers to place <tt>phi</tt> nodes, then traversing the function in
1613 depth-first order to rewrite <tt>load</tt>s and <tt>store</tt>s as
1614 appropriate. This is just the standard SSA construction algorithm to construct
1619 <!-------------------------------------------------------------------------- -->
1620 <div class="doc_subsection">
1621 <a name="memcpyopt">-memcpyopt: MemCpy Optimization</a>
1623 <div class="doc_text">
1625 This pass performs various transformations related to eliminating memcpy
1626 calls, or transforming sets of stores into memset's.
1630 <!-------------------------------------------------------------------------- -->
1631 <div class="doc_subsection">
1632 <a name="mergefunc">-mergefunc: Merge Functions</a>
1634 <div class="doc_text">
1635 <p>This pass looks for equivalent functions that are mergable and folds them.
1637 A hash is computed from the function, based on its type and number of
1640 Once all hashes are computed, we perform an expensive equality comparison
1641 on each function pair. This takes n^2/2 comparisons per bucket, so it's
1642 important that the hash function be high quality. The equality comparison
1643 iterates through each instruction in each basic block.
1645 When a match is found the functions are folded. If both functions are
1646 overridable, we move the functionality into a new internal function and
1647 leave two overridable thunks to it.
1651 <!-------------------------------------------------------------------------- -->
1652 <div class="doc_subsection">
1653 <a name="mergereturn">-mergereturn: Unify function exit nodes</a>
1655 <div class="doc_text">
1657 Ensure that functions have at most one <tt>ret</tt> instruction in them.
1658 Additionally, it keeps track of which node is the new exit node of the CFG.
1662 <!-------------------------------------------------------------------------- -->
1663 <div class="doc_subsection">
1664 <a name="partial-inliner">-partial-inliner: Partial Inliner</a>
1666 <div class="doc_text">
1667 <p>This pass performs partial inlining, typically by inlining an if
1668 statement that surrounds the body of the function.
1672 <!-------------------------------------------------------------------------- -->
1673 <div class="doc_subsection">
1674 <a name="prune-eh">-prune-eh: Remove unused exception handling info</a>
1676 <div class="doc_text">
1678 This file implements a simple interprocedural pass which walks the call-graph,
1679 turning <tt>invoke</tt> instructions into <tt>call</tt> instructions if and
1680 only if the callee cannot throw an exception. It implements this as a
1681 bottom-up traversal of the call-graph.
1685 <!-------------------------------------------------------------------------- -->
1686 <div class="doc_subsection">
1687 <a name="reassociate">-reassociate: Reassociate expressions</a>
1689 <div class="doc_text">
1691 This pass reassociates commutative expressions in an order that is designed
1692 to promote better constant propagation, GCSE, LICM, PRE, etc.
1696 For example: 4 + (<var>x</var> + 5) ⇒ <var>x</var> + (4 + 5)
1700 In the implementation of this algorithm, constants are assigned rank = 0,
1701 function arguments are rank = 1, and other values are assigned ranks
1702 corresponding to the reverse post order traversal of current function
1703 (starting at 2), which effectively gives values in deep loops higher rank
1704 than values not in loops.
1708 <!-------------------------------------------------------------------------- -->
1709 <div class="doc_subsection">
1710 <a name="reg2mem">-reg2mem: Demote all values to stack slots</a>
1712 <div class="doc_text">
1714 This file demotes all registers to memory references. It is intented to be
1715 the inverse of <a href="#mem2reg"><tt>-mem2reg</tt></a>. By converting to
1716 <tt>load</tt> instructions, the only values live across basic blocks are
1717 <tt>alloca</tt> instructions and <tt>load</tt> instructions before
1718 <tt>phi</tt> nodes. It is intended that this should make CFG hacking much
1719 easier. To make later hacking easier, the entry block is split into two, such
1720 that all introduced <tt>alloca</tt> instructions (and nothing else) are in the
1725 <!-------------------------------------------------------------------------- -->
1726 <div class="doc_subsection">
1727 <a name="scalarrepl">-scalarrepl: Scalar Replacement of Aggregates (DT)</a>
1729 <div class="doc_text">
1731 The well-known scalar replacement of aggregates transformation. This
1732 transform breaks up <tt>alloca</tt> instructions of aggregate type (structure
1733 or array) into individual <tt>alloca</tt> instructions for each member if
1734 possible. Then, if possible, it transforms the individual <tt>alloca</tt>
1735 instructions into nice clean scalar SSA form.
1739 This combines a simple scalar replacement of aggregates algorithm with the <a
1740 href="#mem2reg"><tt>mem2reg</tt></a> algorithm because often interact,
1741 especially for C++ programs. As such, iterating between <tt>scalarrepl</tt>,
1742 then <a href="#mem2reg"><tt>mem2reg</tt></a> until we run out of things to
1747 <!-------------------------------------------------------------------------- -->
1748 <div class="doc_subsection">
1749 <a name="sccp">-sccp: Sparse Conditional Constant Propagation</a>
1751 <div class="doc_text">
1753 Sparse conditional constant propagation and merging, which can be summarized
1758 <li>Assumes values are constant unless proven otherwise</li>
1759 <li>Assumes BasicBlocks are dead unless proven otherwise</li>
1760 <li>Proves values to be constant, and replaces them with constants</li>
1761 <li>Proves conditional branches to be unconditional</li>
1765 Note that this pass has a habit of making definitions be dead. It is a good
1766 idea to to run a DCE pass sometime after running this pass.
1770 <!-------------------------------------------------------------------------- -->
1771 <div class="doc_subsection">
1772 <a name="sink">-sink: Code Sinking</a>
1774 <div class="doc_text">
1775 <p>This pass moves instructions into successor blocks, when possible, so that
1776 they aren't executed on paths where their results aren't needed.
1780 <!-------------------------------------------------------------------------- -->
1781 <div class="doc_subsection">
1782 <a name="simplify-libcalls">-simplify-libcalls: Simplify well-known library calls</a>
1784 <div class="doc_text">
1786 Applies a variety of small optimizations for calls to specific well-known
1787 function calls (e.g. runtime library functions). For example, a call
1788 <tt>exit(3)</tt> that occurs within the <tt>main()</tt> function can be
1789 transformed into simply <tt>return 3</tt>.
1793 <!-------------------------------------------------------------------------- -->
1794 <div class="doc_subsection">
1795 <a name="simplifycfg">-simplifycfg: Simplify the CFG</a>
1797 <div class="doc_text">
1799 Performs dead code elimination and basic block merging. Specifically:
1803 <li>Removes basic blocks with no predecessors.</li>
1804 <li>Merges a basic block into its predecessor if there is only one and the
1805 predecessor only has one successor.</li>
1806 <li>Eliminates PHI nodes for basic blocks with a single predecessor.</li>
1807 <li>Eliminates a basic block that only contains an unconditional
1812 <!-------------------------------------------------------------------------- -->
1813 <div class="doc_subsection">
1814 <a name="strip">-strip: Strip all symbols from a module</a>
1816 <div class="doc_text">
1818 performs code stripping. this transformation can delete:
1822 <li>names for virtual registers</li>
1823 <li>symbols for internal globals and functions</li>
1824 <li>debug information</li>
1828 note that this transformation makes code much less readable, so it should
1829 only be used in situations where the <tt>strip</tt> utility would be used,
1830 such as reducing code size or making it harder to reverse engineer code.
1834 <!-------------------------------------------------------------------------- -->
1835 <div class="doc_subsection">
1836 <a name="strip-dead-prototypes">-strip-dead-prototypes: Strip Unused Function Prototypes</a>
1838 <div class="doc_text">
1840 This pass loops over all of the functions in the input module, looking for
1841 dead declarations and removes them. Dead declarations are declarations of
1842 functions for which no implementation is available (i.e., declarations for
1843 unused library functions).
1847 <!-------------------------------------------------------------------------- -->
1848 <div class="doc_subsection">
1849 <a name="strip-debug-declare">-strip-debug-declare: Strip all llvm.dbg.declare intrinsics</a>
1851 <div class="doc_text">
1852 <p>This pass implements code stripping. Specifically, it can delete:</p>
1854 <li>names for virtual registers</li>
1855 <li>symbols for internal globals and functions</li>
1856 <li>debug information</li>
1859 Note that this transformation makes code much less readable, so it should
1860 only be used in situations where the 'strip' utility would be used, such as
1861 reducing code size or making it harder to reverse engineer code.
1865 <!-------------------------------------------------------------------------- -->
1866 <div class="doc_subsection">
1867 <a name="strip-nondebug">-strip-nondebug: Strip all symbols, except dbg symbols, from a module</a>
1869 <div class="doc_text">
1870 <p>This pass implements code stripping. Specifically, it can delete:</p>
1872 <li>names for virtual registers</li>
1873 <li>symbols for internal globals and functions</li>
1874 <li>debug information</li>
1877 Note that this transformation makes code much less readable, so it should
1878 only be used in situations where the 'strip' utility would be used, such as
1879 reducing code size or making it harder to reverse engineer code.
1883 <!-------------------------------------------------------------------------- -->
1884 <div class="doc_subsection">
1885 <a name="sretpromotion">-sretpromotion: Promote sret arguments</a>
1887 <div class="doc_text">
1889 This pass finds functions that return a struct (using a pointer to the struct
1890 as the first argument of the function, marked with the '<tt>sret</tt>' attribute) and
1891 replaces them with a new function that simply returns each of the elements of
1892 that struct (using multiple return values).
1896 This pass works under a number of conditions:
1900 <li>The returned struct must not contain other structs</li>
1901 <li>The returned struct must only be used to load values from</li>
1902 <li>The placeholder struct passed in is the result of an <tt>alloca</tt></li>
1906 <!-------------------------------------------------------------------------- -->
1907 <div class="doc_subsection">
1908 <a name="tailcallelim">-tailcallelim: Tail Call Elimination</a>
1910 <div class="doc_text">
1912 This file transforms calls of the current function (self recursion) followed
1913 by a return instruction with a branch to the entry of the function, creating
1914 a loop. This pass also implements the following extensions to the basic
1919 <li>Trivial instructions between the call and return do not prevent the
1920 transformation from taking place, though currently the analysis cannot
1921 support moving any really useful instructions (only dead ones).
1922 <li>This pass transforms functions that are prevented from being tail
1923 recursive by an associative expression to use an accumulator variable,
1924 thus compiling the typical naive factorial or <tt>fib</tt> implementation
1925 into efficient code.
1926 <li>TRE is performed if the function returns void, if the return
1927 returns the result returned by the call, or if the function returns a
1928 run-time constant on all exits from the function. It is possible, though
1929 unlikely, that the return returns something else (like constant 0), and
1930 can still be TRE'd. It can be TRE'd if <em>all other</em> return
1931 instructions in the function return the exact same value.
1932 <li>If it can prove that callees do not access theier caller stack frame,
1933 they are marked as eligible for tail call elimination (by the code
1938 <!-------------------------------------------------------------------------- -->
1939 <div class="doc_subsection">
1940 <a name="tailduplicate">-tailduplicate: Tail Duplication</a>
1942 <div class="doc_text">
1944 This pass performs a limited form of tail duplication, intended to simplify
1945 CFGs by removing some unconditional branches. This pass is necessary to
1946 straighten out loops created by the C front-end, but also is capable of
1947 making other code nicer. After this pass is run, the CFG simplify pass
1948 should be run to clean up the mess.
1952 <!-- ======================================================================= -->
1953 <div class="doc_section"> <a name="transform">Utility Passes</a></div>
1954 <div class="doc_text">
1955 <p>This section describes the LLVM Utility Passes.</p>
1958 <!-------------------------------------------------------------------------- -->
1959 <div class="doc_subsection">
1960 <a name="deadarghaX0r">-deadarghaX0r: Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</a>
1962 <div class="doc_text">
1964 Same as dead argument elimination, but deletes arguments to functions which
1965 are external. This is only for use by <a
1966 href="Bugpoint.html">bugpoint</a>.</p>
1969 <!-------------------------------------------------------------------------- -->
1970 <div class="doc_subsection">
1971 <a name="extract-blocks">-extract-blocks: Extract Basic Blocks From Module (for bugpoint use)</a>
1973 <div class="doc_text">
1975 This pass is used by bugpoint to extract all blocks from the module into their
1979 <!-------------------------------------------------------------------------- -->
1980 <div class="doc_subsection">
1981 <a name="instnamer">-instnamer: Assign names to anonymous instructions</a>
1983 <div class="doc_text">
1984 <p>This is a little utility pass that gives instructions names, this is mostly
1985 useful when diffing the effect of an optimization because deleting an
1986 unnamed instruction can change all other instruction numbering, making the
1991 <!-------------------------------------------------------------------------- -->
1992 <div class="doc_subsection">
1993 <a name="preverify">-preverify: Preliminary module verification</a>
1995 <div class="doc_text">
1997 Ensures that the module is in the form required by the <a
1998 href="#verifier">Module Verifier</a> pass.
2002 Running the verifier runs this pass automatically, so there should be no need
2007 <!-------------------------------------------------------------------------- -->
2008 <div class="doc_subsection">
2009 <a name="verify">-verify: Module Verifier</a>
2011 <div class="doc_text">
2013 Verifies an LLVM IR code. This is useful to run after an optimization which is
2014 undergoing testing. Note that <tt>llvm-as</tt> verifies its input before
2015 emitting bitcode, and also that malformed bitcode is likely to make LLVM
2016 crash. All language front-ends are therefore encouraged to verify their output
2017 before performing optimizing transformations.
2021 <li>Both of a binary operator's parameters are of the same type.</li>
2022 <li>Verify that the indices of mem access instructions match other
2024 <li>Verify that arithmetic and other things are only performed on
2025 first-class types. Verify that shifts and logicals only happen on
2027 <li>All of the constants in a switch statement are of the correct type.</li>
2028 <li>The code is in valid SSA form.</li>
2029 <li>It is illegal to put a label into any other type (like a structure) or
2031 <li>Only phi nodes can be self referential: <tt>%x = add i32 %x, %x</tt> is
2033 <li>PHI nodes must have an entry for each predecessor, with no extras.</li>
2034 <li>PHI nodes must be the first thing in a basic block, all grouped
2036 <li>PHI nodes must have at least one entry.</li>
2037 <li>All basic blocks should only end with terminator insts, not contain
2039 <li>The entry node to a function must not have predecessors.</li>
2040 <li>All Instructions must be embedded into a basic block.</li>
2041 <li>Functions cannot take a void-typed parameter.</li>
2042 <li>Verify that a function's argument list agrees with its declared
2044 <li>It is illegal to specify a name for a void value.</li>
2045 <li>It is illegal to have a internal global value with no initializer.</li>
2046 <li>It is illegal to have a ret instruction that returns a value that does
2047 not agree with the function return value type.</li>
2048 <li>Function call argument types match the function prototype.</li>
2049 <li>All other things that are tested by asserts spread about the code.</li>
2053 Note that this does not provide full security verification (like Java), but
2054 instead just tries to ensure that code is well-formed.
2058 <!-------------------------------------------------------------------------- -->
2059 <div class="doc_subsection">
2060 <a name="view-cfg">-view-cfg: View CFG of function</a>
2062 <div class="doc_text">
2064 Displays the control flow graph using the GraphViz tool.
2068 <!-------------------------------------------------------------------------- -->
2069 <div class="doc_subsection">
2070 <a name="view-cfg-only">-view-cfg-only: View CFG of function (with no function bodies)</a>
2072 <div class="doc_text">
2074 Displays the control flow graph using the GraphViz tool, but omitting function
2079 <!-------------------------------------------------------------------------- -->
2080 <div class="doc_subsection">
2081 <a name="view-dom">-view-dom: View dominance tree of function</a>
2083 <div class="doc_text">
2085 Displays the dominator tree using the GraphViz tool.
2089 <!-------------------------------------------------------------------------- -->
2090 <div class="doc_subsection">
2091 <a name="view-dom-only">-view-dom-only: View dominance tree of function (with no function
2095 <div class="doc_text">
2097 Displays the dominator tree using the GraphViz tool, but omitting function
2102 <!-------------------------------------------------------------------------- -->
2103 <div class="doc_subsection">
2104 <a name="view-postdom">-view-postdom: View postdominance tree of function</a>
2106 <div class="doc_text">
2108 Displays the post dominator tree using the GraphViz tool.
2112 <!-------------------------------------------------------------------------- -->
2113 <div class="doc_subsection">
2114 <a name="view-postdom-only">-view-postdom-only: View postdominance tree of function (with no
2118 <div class="doc_text">
2120 Displays the post dominator tree using the GraphViz tool, but omitting
2125 <!-- *********************************************************************** -->
2129 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2130 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
2131 <a href="http://validator.w3.org/check/referer"><img
2132 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
2134 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br>
2135 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
2136 Last modified: $Date$