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 (default AA impl)</td></tr>
79 <tr><td><a href="#basiccg">-basiccg</a></td><td>Basic CallGraph Construction</td></tr>
80 <tr><td><a href="#codegenprepare">-codegenprepare</a></td><td>Optimize for code generation</td></tr>
81 <tr><td><a href="#count-aa">-count-aa</a></td><td>Count Alias Analysis Query Responses</td></tr>
82 <tr><td><a href="#debug-aa">-debug-aa</a></td><td>AA use debugger</td></tr>
83 <tr><td><a href="#domfrontier">-domfrontier</a></td><td>Dominance Frontier Construction</td></tr>
84 <tr><td><a href="#domtree">-domtree</a></td><td>Dominator Tree Construction</td></tr>
85 <tr><td><a href="#dot-callgraph">-dot-callgraph</a></td><td>Print Call Graph to 'dot' file</td></tr>
86 <tr><td><a href="#dot-cfg">-dot-cfg</a></td><td>Print CFG of function to 'dot' file</td></tr>
87 <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>
88 <tr><td><a href="#dot-dom">-dot-dom</a></td><td>Print dominator tree of function to 'dot' file</td></tr>
89 <tr><td><a href="#dot-dom-only">-dot-dom-only</a></td><td>Print dominator tree of function to 'dot' file (with no function bodies)</td></tr>
90 <tr><td><a href="#dot-postdom">-dot-postdom</a></td><td>Print post dominator tree of function to 'dot' file</td></tr>
91 <tr><td><a href="#dot-postdom-only">-dot-postdom-only</a></td><td>Print post dominator tree of function to 'dot' file (with no function bodies)</td></tr>
92 <tr><td><a href="#globalsmodref-aa">-globalsmodref-aa</a></td><td>Simple mod/ref analysis for globals</td></tr>
93 <tr><td><a href="#instcount">-instcount</a></td><td>Counts the various types of Instructions</td></tr>
94 <tr><td><a href="#interprocedural-aa-eval">-interprocedural-aa-eval</a></td><td>Exhaustive Interprocedural Alias Analysis Precision Evaluator</td></tr>
95 <tr><td><a href="#interprocedural-basic-aa">-interprocedural-basic-aa</a></td><td>Interprocedural Basic Alias Analysis</td></tr>
96 <tr><td><a href="#intervals">-intervals</a></td><td>Interval Partition Construction</td></tr>
97 <tr><td><a href="#iv-users">-iv-users</a></td><td>Induction Variable Users</td></tr>
98 <tr><td><a href="#lazy-value-info">-lazy-value-info</a></td><td>Lazy Value Information Analysis</td></tr>
99 <tr><td><a href="#lda">-lda</a></td><td>Loop Dependence Analysis</td></tr>
100 <tr><td><a href="#libcall-aa">-libcall-aa</a></td><td>LibCall Alias Analysis</td></tr>
101 <tr><td><a href="#lint">-lint</a></td><td>Check for common errors in LLVM IR</td></tr>
102 <tr><td><a href="#live-values">-live-values</a></td><td>Value Liveness Analysis</td></tr>
103 <tr><td><a href="#loops">-loops</a></td><td>Natural Loop Information</td></tr>
104 <tr><td><a href="#memdep">-memdep</a></td><td>Memory Dependence Analysis</td></tr>
105 <tr><td><a href="#module-debuginfo">-module-debuginfo</a></td><td>Prints module debug info metadata</td></tr>
106 <tr><td><a href="#no-aa">-no-aa</a></td><td>No Alias Analysis (always returns 'may' alias)</td></tr>
107 <tr><td><a href="#no-profile">-no-profile</a></td><td>No Profile Information</td></tr>
108 <tr><td><a href="#pointertracking">-pointertracking</a></td><td>Track pointer bounds</td></tr>
109 <tr><td><a href="#postdomfrontier">-postdomfrontier</a></td><td>Post-Dominance Frontier Construction</td></tr>
110 <tr><td><a href="#postdomtree">-postdomtree</a></td><td>Post-Dominator Tree Construction</td></tr>
111 <tr><td><a href="#print-alias-sets">-print-alias-sets</a></td><td>Alias Set Printer</td></tr>
112 <tr><td><a href="#print-callgraph">-print-callgraph</a></td><td>Print a call graph</td></tr>
113 <tr><td><a href="#print-callgraph-sccs">-print-callgraph-sccs</a></td><td>Print SCCs of the Call Graph</td></tr>
114 <tr><td><a href="#print-cfg-sccs">-print-cfg-sccs</a></td><td>Print SCCs of each function CFG</td></tr>
115 <tr><td><a href="#print-dbginfo">-print-dbginfo</a></td><td>Print debug info in human readable form</td></tr>
116 <tr><td><a href="#print-dom-info">-print-dom-info</a></td><td>Dominator Info Printer</td></tr>
117 <tr><td><a href="#print-externalfnconstants">-print-externalfnconstants</a></td><td>Print external fn callsites passed constants</td></tr>
118 <tr><td><a href="#print-function">-print-function</a></td><td>Print function to stderr</td></tr>
119 <tr><td><a href="#print-module">-print-module</a></td><td>Print module to stderr</td></tr>
120 <tr><td><a href="#print-used-types">-print-used-types</a></td><td>Find Used Types</td></tr>
121 <tr><td><a href="#profile-estimator">-profile-estimator</a></td><td>Estimate profiling information</td></tr>
122 <tr><td><a href="#profile-loader">-profile-loader</a></td><td>Load profile information from llvmprof.out</td></tr>
123 <tr><td><a href="#regions">-regions</a></td><td>Detect single entry single exit regions in a function</td></tr>
124 <tr><td><a href="#profile-verifier">-profile-verifier</a></td><td>Verify profiling information</td></tr>
125 <tr><td><a href="#scalar-evolution">-scalar-evolution</a></td><td>Scalar Evolution Analysis</td></tr>
126 <tr><td><a href="#scev-aa">-scev-aa</a></td><td>ScalarEvolution-based Alias Analysis</td></tr>
127 <tr><td><a href="#targetdata">-targetdata</a></td><td>Target Data Layout</td></tr>
130 <tr><th colspan="2"><b>TRANSFORM PASSES</b></th></tr>
131 <tr><th>Option</th><th>Name</th></tr>
132 <tr><td><a href="#abcd">-abcd</a></td><td>Remove redundant conditional branches</td></tr>
133 <tr><td><a href="#adce">-adce</a></td><td>Aggressive Dead Code Elimination</td></tr>
134 <tr><td><a href="#always-inline">-always-inline</a></td><td>Inliner for always_inline functions</td></tr>
135 <tr><td><a href="#argpromotion">-argpromotion</a></td><td>Promote 'by reference' arguments to scalars</td></tr>
136 <tr><td><a href="#block-placement">-block-placement</a></td><td>Profile Guided Basic Block Placement</td></tr>
137 <tr><td><a href="#break-crit-edges">-break-crit-edges</a></td><td>Break critical edges in CFG</td></tr>
138 <tr><td><a href="#codegenprepare">-codegenprepare</a></td><td>Prepare a function for code generation </td></tr>
139 <tr><td><a href="#constmerge">-constmerge</a></td><td>Merge Duplicate Global Constants</td></tr>
140 <tr><td><a href="#constprop">-constprop</a></td><td>Simple constant propagation</td></tr>
141 <tr><td><a href="#dce">-dce</a></td><td>Dead Code Elimination</td></tr>
142 <tr><td><a href="#deadargelim">-deadargelim</a></td><td>Dead Argument Elimination</td></tr>
143 <tr><td><a href="#deadtypeelim">-deadtypeelim</a></td><td>Dead Type Elimination</td></tr>
144 <tr><td><a href="#die">-die</a></td><td>Dead Instruction Elimination</td></tr>
145 <tr><td><a href="#dse">-dse</a></td><td>Dead Store Elimination</td></tr>
146 <tr><td><a href="#functionattrs">-functionattrs</a></td><td>Deduce function attributes</td></tr>
147 <tr><td><a href="#globaldce">-globaldce</a></td><td>Dead Global Elimination</td></tr>
148 <tr><td><a href="#globalopt">-globalopt</a></td><td>Global Variable Optimizer</td></tr>
149 <tr><td><a href="#gvn">-gvn</a></td><td>Global Value Numbering</td></tr>
150 <tr><td><a href="#indvars">-indvars</a></td><td>Canonicalize Induction Variables</td></tr>
151 <tr><td><a href="#inline">-inline</a></td><td>Function Integration/Inlining</td></tr>
152 <tr><td><a href="#insert-edge-profiling">-insert-edge-profiling</a></td><td>Insert instrumentation for edge profiling</td></tr>
153 <tr><td><a href="#insert-optimal-edge-profiling">-insert-optimal-edge-profiling</a></td><td>Insert optimal instrumentation for edge profiling</td></tr>
154 <tr><td><a href="#instcombine">-instcombine</a></td><td>Combine redundant instructions</td></tr>
155 <tr><td><a href="#internalize">-internalize</a></td><td>Internalize Global Symbols</td></tr>
156 <tr><td><a href="#ipconstprop">-ipconstprop</a></td><td>Interprocedural constant propagation</td></tr>
157 <tr><td><a href="#ipsccp">-ipsccp</a></td><td>Interprocedural Sparse Conditional Constant Propagation</td></tr>
158 <tr><td><a href="#jump-threading">-jump-threading</a></td><td>Thread control through conditional blocks </td></tr>
159 <tr><td><a href="#lcssa">-lcssa</a></td><td>Loop-Closed SSA Form Pass</td></tr>
160 <tr><td><a href="#licm">-licm</a></td><td>Loop Invariant Code Motion</td></tr>
161 <tr><td><a href="#loop-deletion">-loop-deletion</a></td><td>Dead Loop Deletion Pass </td></tr>
162 <tr><td><a href="#loop-extract">-loop-extract</a></td><td>Extract loops into new functions</td></tr>
163 <tr><td><a href="#loop-extract-single">-loop-extract-single</a></td><td>Extract at most one loop into a new function</td></tr>
164 <tr><td><a href="#loop-index-split">-loop-index-split</a></td><td>Index Split Loops</td></tr>
165 <tr><td><a href="#loop-reduce">-loop-reduce</a></td><td>Loop Strength Reduction</td></tr>
166 <tr><td><a href="#loop-rotate">-loop-rotate</a></td><td>Rotate Loops</td></tr>
167 <tr><td><a href="#loop-unroll">-loop-unroll</a></td><td>Unroll loops</td></tr>
168 <tr><td><a href="#loop-unswitch">-loop-unswitch</a></td><td>Unswitch loops</td></tr>
169 <tr><td><a href="#loopsimplify">-loopsimplify</a></td><td>Canonicalize natural loops</td></tr>
170 <tr><td><a href="#loweratomic">-loweratomic</a></td><td>Lower atomic intrinsics</td></tr>
171 <tr><td><a href="#lowerinvoke">-lowerinvoke</a></td><td>Lower invoke and unwind, for unwindless code generators</td></tr>
172 <tr><td><a href="#lowersetjmp">-lowersetjmp</a></td><td>Lower Set Jump</td></tr>
173 <tr><td><a href="#lowerswitch">-lowerswitch</a></td><td>Lower SwitchInst's to branches</td></tr>
174 <tr><td><a href="#mem2reg">-mem2reg</a></td><td>Promote Memory to Register</td></tr>
175 <tr><td><a href="#memcpyopt">-memcpyopt</a></td><td>Optimize use of memcpy and friends</td></tr>
176 <tr><td><a href="#mergefunc">-mergefunc</a></td><td>Merge Functions</td></tr>
177 <tr><td><a href="#mergereturn">-mergereturn</a></td><td>Unify function exit nodes</td></tr>
178 <tr><td><a href="#partial-inliner">-partial-inliner</a></td><td>Partial Inliner</td></tr>
179 <tr><td><a href="#partialspecialization">-partialspecialization</a></td><td>Partial Specialization</td></tr>
180 <tr><td><a href="#prune-eh">-prune-eh</a></td><td>Remove unused exception handling info</td></tr>
181 <tr><td><a href="#reassociate">-reassociate</a></td><td>Reassociate expressions</td></tr>
182 <tr><td><a href="#reg2mem">-reg2mem</a></td><td>Demote all values to stack slots</td></tr>
183 <tr><td><a href="#scalarrepl">-scalarrepl</a></td><td>Scalar Replacement of Aggregates</td></tr>
184 <tr><td><a href="#sccp">-sccp</a></td><td>Sparse Conditional Constant Propagation</td></tr>
185 <tr><td><a href="#sink">-sink</a></td><td>Code Sinking</td></tr>
186 <tr><td><a href="#simplify-libcalls">-simplify-libcalls</a></td><td>Simplify well-known library calls</td></tr>
187 <tr><td><a href="#simplify-libcalls-halfpowr">-simplify-libcalls-halfpowr</a></td><td>Simplify half_powr library calls</td></tr>
188 <tr><td><a href="#simplifycfg">-simplifycfg</a></td><td>Simplify the CFG</td></tr>
189 <tr><td><a href="#split-geps">-split-geps</a></td><td>Split complex GEPs into simple GEPs</td></tr>
190 <tr><td><a href="#ssi">-ssi</a></td><td>Static Single Information Construction</td></tr>
191 <tr><td><a href="#ssi-everything">-ssi-everything</a></td><td>Static Single Information Construction (everything, intended for debugging)</td></tr>
192 <tr><td><a href="#strip">-strip</a></td><td>Strip all symbols from a module</td></tr>
193 <tr><td><a href="#strip-dead-debug-info">-strip-dead-debug-info</a></td><td>Strip debug info for unused symbols</td></tr>
194 <tr><td><a href="#strip-dead-prototypes">-strip-dead-prototypes</a></td><td>Remove unused function declarations</td></tr>
195 <tr><td><a href="#strip-debug-declare">-strip-debug-declare</a></td><td>Strip all llvm.dbg.declare intrinsics</td></tr>
196 <tr><td><a href="#strip-nondebug">-strip-nondebug</a></td><td>Strip all symbols, except dbg symbols, from a module</td></tr>
197 <tr><td><a href="#sretpromotion">-sretpromotion</a></td><td>Promote sret arguments</td></tr>
198 <tr><td><a href="#tailcallelim">-tailcallelim</a></td><td>Tail Call Elimination</td></tr>
199 <tr><td><a href="#tailduplicate">-tailduplicate</a></td><td>Tail Duplication</td></tr>
202 <tr><th colspan="2"><b>UTILITY PASSES</b></th></tr>
203 <tr><th>Option</th><th>Name</th></tr>
204 <tr><td><a href="#deadarghaX0r">-deadarghaX0r</a></td><td>Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</td></tr>
205 <tr><td><a href="#extract-blocks">-extract-blocks</a></td><td>Extract Basic Blocks From Module (for bugpoint use)</td></tr>
206 <tr><td><a href="#instnamer">-instnamer</a></td><td>Assign names to anonymous instructions</td></tr>
207 <tr><td><a href="#preverify">-preverify</a></td><td>Preliminary module verification</td></tr>
208 <tr><td><a href="#verify">-verify</a></td><td>Module Verifier</td></tr>
209 <tr><td><a href="#view-cfg">-view-cfg</a></td><td>View CFG of function</td></tr>
210 <tr><td><a href="#view-cfg-only">-view-cfg-only</a></td><td>View CFG of function (with no function bodies)</td></tr>
211 <tr><td><a href="#view-dom">-view-dom</a></td><td>View dominator tree of function</td></tr>
212 <tr><td><a href="#view-dom-only">-view-dom-only</a></td><td>View dominator tree of function (with no function bodies)</td></tr>
213 <tr><td><a href="#view-postdom">-view-postdom</a></td><td>View post dominator tree of function</td></tr>
214 <tr><td><a href="#view-postdom-only">-view-postdom-only</a></td><td>View post dominator tree of function (with no function bodies)</td></tr>
218 <!-- ======================================================================= -->
219 <div class="doc_section"> <a name="example">Analysis Passes</a></div>
220 <div class="doc_text">
221 <p>This section describes the LLVM Analysis Passes.</p>
224 <!-------------------------------------------------------------------------- -->
225 <div class="doc_subsection">
226 <a name="aa-eval">-aa-eval: Exhaustive Alias Analysis Precision Evaluator</a>
228 <div class="doc_text">
229 <p>This is a simple N^2 alias analysis accuracy evaluator.
230 Basically, for each function in the program, it simply queries to see how the
231 alias analysis implementation answers alias queries between each pair of
232 pointers in the function.</p>
234 <p>This is inspired and adapted from code by: Naveen Neelakantam, Francesco
235 Spadini, and Wojciech Stryjewski.</p>
238 <!-------------------------------------------------------------------------- -->
239 <div class="doc_subsection">
240 <a name="basicaa">-basicaa: Basic Alias Analysis (default AA impl)</a>
242 <div class="doc_text">
244 This is the default implementation of the Alias Analysis interface
245 that simply implements a few identities (two different globals cannot alias,
246 etc), but otherwise does no analysis.
250 <!-------------------------------------------------------------------------- -->
251 <div class="doc_subsection">
252 <a name="basiccg">-basiccg: Basic CallGraph Construction</a>
254 <div class="doc_text">
255 <p>Yet to be written.</p>
258 <!-------------------------------------------------------------------------- -->
259 <div class="doc_subsection">
260 <a name="codegenprepare">-codegenprepare: Optimize for code generation</a>
262 <div class="doc_text">
264 This pass munges the code in the input function to better prepare it for
265 SelectionDAG-based code generation. This works around limitations in it's
266 basic-block-at-a-time approach. It should eventually be removed.
270 <!-------------------------------------------------------------------------- -->
271 <div class="doc_subsection">
272 <a name="count-aa">-count-aa: Count Alias Analysis Query Responses</a>
274 <div class="doc_text">
276 A pass which can be used to count how many alias queries
277 are being made and how the alias analysis implementation being used responds.
281 <!-------------------------------------------------------------------------- -->
282 <div class="doc_subsection">
283 <a name="debug-aa">-debug-aa: AA use debugger</a>
285 <div class="doc_text">
287 This simple pass checks alias analysis users to ensure that if they
288 create a new value, they do not query AA without informing it of the value.
289 It acts as a shim over any other AA pass you want.
293 Yes keeping track of every value in the program is expensive, but this is
298 <!-------------------------------------------------------------------------- -->
299 <div class="doc_subsection">
300 <a name="domfrontier">-domfrontier: Dominance Frontier Construction</a>
302 <div class="doc_text">
304 This pass is a simple dominator construction algorithm for finding forward
309 <!-------------------------------------------------------------------------- -->
310 <div class="doc_subsection">
311 <a name="domtree">-domtree: Dominator Tree Construction</a>
313 <div class="doc_text">
315 This pass is a simple dominator construction algorithm for finding forward
320 <!-------------------------------------------------------------------------- -->
321 <div class="doc_subsection">
322 <a name="dot-callgraph">-dot-callgraph: Print Call Graph to 'dot' file</a>
324 <div class="doc_text">
326 This pass, only available in <code>opt</code>, prints the call graph into a
327 <code>.dot</code> graph. This graph can then be processed with the "dot" tool
328 to convert it to postscript or some other suitable format.
332 <!-------------------------------------------------------------------------- -->
333 <div class="doc_subsection">
334 <a name="dot-cfg">-dot-cfg: Print CFG of function to 'dot' file</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. This graph can then be processed with the
340 "dot" tool to convert it to postscript or some other suitable format.
344 <!-------------------------------------------------------------------------- -->
345 <div class="doc_subsection">
346 <a name="dot-cfg-only">-dot-cfg-only: Print CFG of function to 'dot' file (with no function bodies)</a>
348 <div class="doc_text">
350 This pass, only available in <code>opt</code>, prints the control flow graph
351 into a <code>.dot</code> graph, omitting the function bodies. This graph can
352 then be processed with the "dot" tool to convert it to postscript or some
353 other suitable format.
357 <!-------------------------------------------------------------------------- -->
358 <div class="doc_subsection">
359 <a name="dot-dom">-dot-dom: Print dominator tree of function to 'dot' file</a>
361 <div class="doc_text">
363 This pass, only available in <code>opt</code>, prints the dominator tree
364 into a <code>.dot</code> graph. This graph can then be processed with the
365 "dot" tool to convert it to postscript or some other suitable format.
369 <!-------------------------------------------------------------------------- -->
370 <div class="doc_subsection">
371 <a name="dot-dom-only">-dot-dom-only: Print dominator tree of function to 'dot' file (with no
374 <div class="doc_text">
376 This pass, only available in <code>opt</code>, prints the dominator tree
377 into a <code>.dot</code> graph, omitting the function bodies. This graph can
378 then be processed with the "dot" tool to convert it to postscript or some
379 other suitable format.
383 <!-------------------------------------------------------------------------- -->
384 <div class="doc_subsection">
385 <a name="dot-postdom">dot-postdom: Print post dominator tree of function to 'dot' file</a>
387 <div class="doc_text">
389 This pass, only available in <code>opt</code>, prints the post dominator tree
390 into a <code>.dot</code> graph. This graph can then be processed with the
391 "dot" tool to convert it to postscript or some other suitable format.
395 <!-------------------------------------------------------------------------- -->
396 <div class="doc_subsection">
397 <a name="dot-postdom-only">dot-postdom-only: Print post dominator tree of function to 'dot' file
398 (with no function bodies)</a>
400 <div class="doc_text">
402 This pass, only available in <code>opt</code>, prints the post dominator tree
403 into a <code>.dot</code> graph, omitting the function bodies. This graph can
404 then be processed with the "dot" tool to convert it to postscript or some
405 other suitable format.
409 <!-------------------------------------------------------------------------- -->
410 <div class="doc_subsection">
411 <a name="globalsmodref-aa">-globalsmodref-aa: Simple mod/ref analysis for globals</a>
413 <div class="doc_text">
415 This simple pass provides alias and mod/ref information for global values
416 that do not have their address taken, and keeps track of whether functions
417 read or write memory (are "pure"). For this simple (but very common) case,
418 we can provide pretty accurate and useful information.
422 <!-------------------------------------------------------------------------- -->
423 <div class="doc_subsection">
424 <a name="instcount">-instcount: Counts the various types of Instructions</a>
426 <div class="doc_text">
428 This pass collects the count of all instructions and reports them
432 <!-------------------------------------------------------------------------- -->
433 <div class="doc_subsection">
434 <a name="interprocedural-aa-eval">-interprocedural-aa-eval: Exhaustive Interprocedural Alias Analysis Precision Evaluator</a>
436 <div class="doc_text">
437 <p>This pass implements a simple N^2 alias analysis accuracy evaluator.
438 Basically, for each function in the program, it simply queries to see how the
439 alias analysis implementation answers alias queries between each pair of
440 pointers in the function.
444 <!-------------------------------------------------------------------------- -->
445 <div class="doc_subsection">
446 <a name="interprocedural-basic-aa">-interprocedural-basic-aa: Interprocedural Basic Alias Analysis</a>
448 <div class="doc_text">
449 <p>This pass defines the default implementation of the Alias Analysis interface
450 that simply implements a few identities (two different globals cannot alias,
451 etc), but otherwise does no analysis.
455 <!-------------------------------------------------------------------------- -->
456 <div class="doc_subsection">
457 <a name="intervals">-intervals: Interval Partition Construction</a>
459 <div class="doc_text">
461 This analysis calculates and represents the interval partition of a function,
462 or a preexisting interval partition.
466 In this way, the interval partition may be used to reduce a flow graph down
467 to its degenerate single node interval partition (unless it is irreducible).
471 <!-------------------------------------------------------------------------- -->
472 <div class="doc_subsection">
473 <a name="iv-users">-iv-users: Induction Variable Users</a>
475 <div class="doc_text">
476 <p>Bookkeeping for "interesting" users of expressions computed from
477 induction variables.</p>
480 <!-------------------------------------------------------------------------- -->
481 <div class="doc_subsection">
482 <a name="lazy-value-info">-lazy-value-info: Lazy Value Information Analysis</a>
484 <div class="doc_text">
485 <p>Interface for lazy computation of value constraint information.</p>
488 <!-------------------------------------------------------------------------- -->
489 <div class="doc_subsection">
490 <a name="lda">-lda: Loop Dependence Analysis</a>
492 <div class="doc_text">
493 <p>Loop dependence analysis framework, which is used to detect dependences in
494 memory accesses in loops.</p>
497 <!-------------------------------------------------------------------------- -->
498 <div class="doc_subsection">
499 <a name="libcall-aa">-libcall-aa: LibCall Alias Analysis</a>
501 <div class="doc_text">
502 <p>LibCall Alias Analysis.</p>
505 <!-------------------------------------------------------------------------- -->
506 <div class="doc_subsection">
507 <a name="lint">-lint: Check for common errors in LLVM IR</a>
509 <div class="doc_text">
510 <p>This pass statically checks for common and easily-identified constructs
511 which produce undefined or likely unintended behavior in LLVM IR.</p>
513 <p>It is not a guarantee of correctness, in two ways. First, it isn't
514 comprehensive. There are checks which could be done statically which are
515 not yet implemented. Some of these are indicated by TODO comments, but
516 those aren't comprehensive either. Second, many conditions cannot be
517 checked statically. This pass does no dynamic instrumentation, so it
518 can't check for all possible problems.</p>
520 <p>Another limitation is that it assumes all code will be executed. A store
521 through a null pointer in a basic block which is never reached is harmless,
522 but this pass will warn about it anyway.</p>
524 <p>Optimization passes may make conditions that this pass checks for more or
525 less obvious. If an optimization pass appears to be introducing a warning,
526 it may be that the optimization pass is merely exposing an existing
527 condition in the code.</p>
529 <p>This code may be run before instcombine. In many cases, instcombine checks
530 for the same kinds of things and turns instructions with undefined behavior
531 into unreachable (or equivalent). Because of this, this pass makes some
532 effort to look through bitcasts and so on.
536 <!-------------------------------------------------------------------------- -->
537 <div class="doc_subsection">
538 <a name="live-values">-live-values: Values Liveness Analysis</a>
540 <div class="doc_text">
541 <p>LLVM IR Value liveness analysis pass.</p>
544 <!-------------------------------------------------------------------------- -->
545 <div class="doc_subsection">
546 <a name="loops">-loops: Natural Loop Construction</a>
548 <div class="doc_text">
550 This analysis is used to identify natural loops and determine the loop depth
551 of various nodes of the CFG. Note that the loops identified may actually be
552 several natural loops that share the same header node... not just a single
557 <!-------------------------------------------------------------------------- -->
558 <div class="doc_subsection">
559 <a name="memdep">-memdep: Memory Dependence Analysis</a>
561 <div class="doc_text">
563 An analysis that determines, for a given memory operation, what preceding
564 memory operations it depends on. It builds on alias analysis information, and
565 tries to provide a lazy, caching interface to a common kind of alias
570 <!-------------------------------------------------------------------------- -->
571 <div class="doc_subsection">
572 <a name="module-debuginfo">-module-debuginfo: Prints module debug info metadata</a>
574 <div class="doc_text">
575 <p>This pass decodes the debug info metadata in a module and prints in a
576 (sufficiently-prepared-) human-readable form.
578 For example, run this pass from opt along with the -analyze option, and
579 it'll print to standard output.
583 <!-------------------------------------------------------------------------- -->
584 <div class="doc_subsection">
585 <a name="no-aa">-no-aa: No Alias Analysis (always returns 'may' alias)</a>
587 <div class="doc_text">
589 Always returns "I don't know" for alias queries. NoAA is unlike other alias
590 analysis implementations, in that it does not chain to a previous analysis. As
591 such it doesn't follow many of the rules that other alias analyses must.
595 <!-------------------------------------------------------------------------- -->
596 <div class="doc_subsection">
597 <a name="no-profile">-no-profile: No Profile Information</a>
599 <div class="doc_text">
601 The default "no profile" implementation of the abstract
602 <code>ProfileInfo</code> interface.
606 <!-------------------------------------------------------------------------- -->
607 <div class="doc_subsection">
608 <a name="pointertracking">-pointertracking: Track pointer bounds.</a>
610 <div class="doc_text">
611 <p>Tracking of pointer bounds.
615 <!-------------------------------------------------------------------------- -->
616 <div class="doc_subsection">
617 <a name="postdomfrontier">-postdomfrontier: Post-Dominance Frontier Construction</a>
619 <div class="doc_text">
621 This pass is a simple post-dominator construction algorithm for finding
622 post-dominator frontiers.
626 <!-------------------------------------------------------------------------- -->
627 <div class="doc_subsection">
628 <a name="postdomtree">-postdomtree: Post-Dominator Tree Construction</a>
630 <div class="doc_text">
632 This pass is a simple post-dominator construction algorithm for finding
637 <!-------------------------------------------------------------------------- -->
638 <div class="doc_subsection">
639 <a name="print-alias-sets">-print-alias-sets: Alias Set Printer</a>
641 <div class="doc_text">
642 <p>Yet to be written.</p>
645 <!-------------------------------------------------------------------------- -->
646 <div class="doc_subsection">
647 <a name="print-callgraph">-print-callgraph: Print a call graph</a>
649 <div class="doc_text">
651 This pass, only available in <code>opt</code>, prints the call graph to
652 standard output in a human-readable form.
656 <!-------------------------------------------------------------------------- -->
657 <div class="doc_subsection">
658 <a name="print-callgraph-sccs">-print-callgraph-sccs: Print SCCs of the Call Graph</a>
660 <div class="doc_text">
662 This pass, only available in <code>opt</code>, prints the SCCs of the call
663 graph to standard output in a human-readable form.
667 <!-------------------------------------------------------------------------- -->
668 <div class="doc_subsection">
669 <a name="print-cfg-sccs">-print-cfg-sccs: Print SCCs of each function CFG</a>
671 <div class="doc_text">
673 This pass, only available in <code>opt</code>, prints the SCCs of each
674 function CFG to standard output in a human-readable form.
678 <!-------------------------------------------------------------------------- -->
679 <div class="doc_subsection">
680 <a name="print-dbginfo">-print-dbginfo: Print debug info in human readable form</a>
682 <div class="doc_text">
683 <p>Pass that prints instructions, and associated debug info:
686 <li>source/line/col information</li>
687 <li>original variable name</li>
688 <li>original type name</li>
694 <!-------------------------------------------------------------------------- -->
695 <div class="doc_subsection">
696 <a name="print-dom-info">-print-dom-info: Dominator Info Printer</a>
698 <div class="doc_text">
699 <p>Dominator Info Printer.</p>
702 <!-------------------------------------------------------------------------- -->
703 <div class="doc_subsection">
704 <a name="print-externalfnconstants">-print-externalfnconstants: Print external fn callsites passed constants</a>
706 <div class="doc_text">
708 This pass, only available in <code>opt</code>, prints out call sites to
709 external functions that are called with constant arguments. This can be
710 useful when looking for standard library functions we should constant fold
711 or handle in alias analyses.
715 <!-------------------------------------------------------------------------- -->
716 <div class="doc_subsection">
717 <a name="print-function">-print-function: Print function to stderr</a>
719 <div class="doc_text">
721 The <code>PrintFunctionPass</code> class is designed to be pipelined with
722 other <code>FunctionPass</code>es, and prints out the functions of the module
723 as they are processed.
727 <!-------------------------------------------------------------------------- -->
728 <div class="doc_subsection">
729 <a name="print-module">-print-module: Print module to stderr</a>
731 <div class="doc_text">
733 This pass simply prints out the entire module when it is executed.
737 <!-------------------------------------------------------------------------- -->
738 <div class="doc_subsection">
739 <a name="print-used-types">-print-used-types: Find Used Types</a>
741 <div class="doc_text">
743 This pass is used to seek out all of the types in use by the program. Note
744 that this analysis explicitly does not include types only used by the symbol
748 <!-------------------------------------------------------------------------- -->
749 <div class="doc_subsection">
750 <a name="profile-estimator">-profile-estimator: Estimate profiling information</a>
752 <div class="doc_text">
753 <p>Profiling information that estimates the profiling information
754 in a very crude and unimaginative way.
758 <!-------------------------------------------------------------------------- -->
759 <div class="doc_subsection">
760 <a name="profile-loader">-profile-loader: Load profile information from llvmprof.out</a>
762 <div class="doc_text">
764 A concrete implementation of profiling information that loads the information
765 from a profile dump file.
769 <!-------------------------------------------------------------------------- -->
770 <div class="doc_subsection">
771 <a name="profile-verifier">-profile-verifier: Verify profiling information</a>
773 <div class="doc_text">
774 <p>Pass that checks profiling information for plausibility.</p>
776 <div class="doc_subsection">
777 <a name="regions">-regions: Detect single entry single exit regions in a function</a>
779 <div class="doc_text">
781 The <code>RegionInfo</code> pass detects single entry single exit regions in a
782 function, where a region is defined as any subgraph that is connected to the
783 remaining graph at only two spots. Furthermore, an hierarchical region tree is
788 <!-------------------------------------------------------------------------- -->
789 <div class="doc_subsection">
790 <a name="scalar-evolution">-scalar-evolution: Scalar Evolution Analysis</a>
792 <div class="doc_text">
794 The <code>ScalarEvolution</code> analysis can be used to analyze and
795 catagorize scalar expressions in loops. It specializes in recognizing general
796 induction variables, representing them with the abstract and opaque
797 <code>SCEV</code> class. Given this analysis, trip counts of loops and other
798 important properties can be obtained.
802 This analysis is primarily useful for induction variable substitution and
807 <!-------------------------------------------------------------------------- -->
808 <div class="doc_subsection">
809 <a name="scev-aa">-scev-aa: </a>
811 <div class="doc_text">
812 <p>Simple alias analysis implemented in terms of ScalarEvolution queries.
814 This differs from traditional loop dependence analysis in that it tests
815 for dependencies within a single iteration of a loop, rather than
816 dependencies between different iterations.
818 ScalarEvolution has a more complete understanding of pointer arithmetic
819 than BasicAliasAnalysis' collection of ad-hoc analyses.
823 <!-------------------------------------------------------------------------- -->
824 <div class="doc_subsection">
825 <a name="strip-dead-debug-info">-strip-dead-debug-info: Strip debug info for unused symbols</a>
827 <div class="doc_text">
829 performs code stripping. this transformation can delete:
833 <li>names for virtual registers</li>
834 <li>symbols for internal globals and functions</li>
835 <li>debug information</li>
839 note that this transformation makes code much less readable, so it should
840 only be used in situations where the <tt>strip</tt> utility would be used,
841 such as reducing code size or making it harder to reverse engineer code.
845 <!-------------------------------------------------------------------------- -->
846 <div class="doc_subsection">
847 <a name="targetdata">-targetdata: Target Data Layout</a>
849 <div class="doc_text">
850 <p>Provides other passes access to information on how the size and alignment
851 required by the the target ABI for various data types.</p>
854 <!-- ======================================================================= -->
855 <div class="doc_section"> <a name="transform">Transform Passes</a></div>
856 <div class="doc_text">
857 <p>This section describes the LLVM Transform Passes.</p>
860 <!-------------------------------------------------------------------------- -->
861 <div class="doc_subsection">
862 <a name="abcd">-abcd: Remove redundant conditional branches</a>
864 <div class="doc_text">
865 <p>ABCD removes conditional branch instructions that can be proved redundant.
866 With the SSI representation, each variable has a constraint. By analyzing these
867 constraints we can prove that a branch is redundant. When a branch is proved
868 redundant it means that one direction will always be taken; thus, we can change
869 this branch into an unconditional jump.</p>
870 <p>It is advisable to run <a href="#simplifycfg">SimplifyCFG</a> and
871 <a href="#adce">Aggressive Dead Code Elimination</a> after ABCD
872 to clean up the code.</p>
875 <!-------------------------------------------------------------------------- -->
876 <div class="doc_subsection">
877 <a name="adce">-adce: Aggressive Dead Code Elimination</a>
879 <div class="doc_text">
880 <p>ADCE aggressively tries to eliminate code. This pass is similar to
881 <a href="#dce">DCE</a> but it assumes that values are dead until proven
882 otherwise. This is similar to <a href="#sccp">SCCP</a>, except applied to
883 the liveness of values.</p>
886 <!-------------------------------------------------------------------------- -->
887 <div class="doc_subsection">
888 <a name="always-inline">-always-inline: Inliner for always_inline functions</a>
890 <div class="doc_text">
891 <p>A custom inliner that handles only functions that are marked as
895 <!-------------------------------------------------------------------------- -->
896 <div class="doc_subsection">
897 <a name="argpromotion">-argpromotion: Promote 'by reference' arguments to scalars</a>
899 <div class="doc_text">
901 This pass promotes "by reference" arguments to be "by value" arguments. In
902 practice, this means looking for internal functions that have pointer
903 arguments. If it can prove, through the use of alias analysis, that an
904 argument is *only* loaded, then it can pass the value into the function
905 instead of the address of the value. This can cause recursive simplification
906 of code and lead to the elimination of allocas (especially in C++ template
911 This pass also handles aggregate arguments that are passed into a function,
912 scalarizing them if the elements of the aggregate are only loaded. Note that
913 it refuses to scalarize aggregates which would require passing in more than
914 three operands to the function, because passing thousands of operands for a
915 large array or structure is unprofitable!
919 Note that this transformation could also be done for arguments that are only
920 stored to (returning the value instead), but does not currently. This case
921 would be best handled when and if LLVM starts supporting multiple return
922 values from functions.
926 <!-------------------------------------------------------------------------- -->
927 <div class="doc_subsection">
928 <a name="block-placement">-block-placement: Profile Guided Basic Block Placement</a>
930 <div class="doc_text">
931 <p>This pass is a very simple profile guided basic block placement algorithm.
932 The idea is to put frequently executed blocks together at the start of the
933 function and hopefully increase the number of fall-through conditional
934 branches. If there is no profile information for a particular function, this
935 pass basically orders blocks in depth-first order.</p>
938 <!-------------------------------------------------------------------------- -->
939 <div class="doc_subsection">
940 <a name="break-crit-edges">-break-crit-edges: Break critical edges in CFG</a>
942 <div class="doc_text">
944 Break all of the critical edges in the CFG by inserting a dummy basic block.
945 It may be "required" by passes that cannot deal with critical edges. This
946 transformation obviously invalidates the CFG, but can update forward dominator
947 (set, immediate dominators, tree, and frontier) information.
951 <!-------------------------------------------------------------------------- -->
952 <div class="doc_subsection">
953 <a name="codegenprepare">-codegenprepare: Prepare a function for code generation</a>
955 <div class="doc_text">
956 This pass munges the code in the input function to better prepare it for
957 SelectionDAG-based code generation. This works around limitations in it's
958 basic-block-at-a-time approach. It should eventually be removed.
961 <!-------------------------------------------------------------------------- -->
962 <div class="doc_subsection">
963 <a name="constmerge">-constmerge: Merge Duplicate Global Constants</a>
965 <div class="doc_text">
967 Merges duplicate global constants together into a single constant that is
968 shared. This is useful because some passes (ie TraceValues) insert a lot of
969 string constants into the program, regardless of whether or not an existing
974 <!-------------------------------------------------------------------------- -->
975 <div class="doc_subsection">
976 <a name="constprop">-constprop: Simple constant propagation</a>
978 <div class="doc_text">
979 <p>This file implements constant propagation and merging. It looks for
980 instructions involving only constant operands and replaces them with a
981 constant value instead of an instruction. For example:</p>
982 <blockquote><pre>add i32 1, 2</pre></blockquote>
984 <blockquote><pre>i32 3</pre></blockquote>
985 <p>NOTE: this pass has a habit of making definitions be dead. It is a good
986 idea to to run a <a href="#die">DIE</a> (Dead Instruction Elimination) pass
987 sometime after running this pass.</p>
990 <!-------------------------------------------------------------------------- -->
991 <div class="doc_subsection">
992 <a name="dce">-dce: Dead Code Elimination</a>
994 <div class="doc_text">
996 Dead code elimination is similar to <a href="#die">dead instruction
997 elimination</a>, but it rechecks instructions that were used by removed
998 instructions to see if they are newly dead.
1002 <!-------------------------------------------------------------------------- -->
1003 <div class="doc_subsection">
1004 <a name="deadargelim">-deadargelim: Dead Argument Elimination</a>
1006 <div class="doc_text">
1008 This pass deletes dead arguments from internal functions. Dead argument
1009 elimination removes arguments which are directly dead, as well as arguments
1010 only passed into function calls as dead arguments of other functions. This
1011 pass also deletes dead arguments in a similar way.
1015 This pass is often useful as a cleanup pass to run after aggressive
1016 interprocedural passes, which add possibly-dead arguments.
1020 <!-------------------------------------------------------------------------- -->
1021 <div class="doc_subsection">
1022 <a name="deadtypeelim">-deadtypeelim: Dead Type Elimination</a>
1024 <div class="doc_text">
1026 This pass is used to cleanup the output of GCC. It eliminate names for types
1027 that are unused in the entire translation unit, using the <a
1028 href="#findusedtypes">find used types</a> pass.
1032 <!-------------------------------------------------------------------------- -->
1033 <div class="doc_subsection">
1034 <a name="die">-die: Dead Instruction Elimination</a>
1036 <div class="doc_text">
1038 Dead instruction elimination performs a single pass over the function,
1039 removing instructions that are obviously dead.
1043 <!-------------------------------------------------------------------------- -->
1044 <div class="doc_subsection">
1045 <a name="dse">-dse: Dead Store Elimination</a>
1047 <div class="doc_text">
1049 A trivial dead store elimination that only considers basic-block local
1054 <!-------------------------------------------------------------------------- -->
1055 <div class="doc_subsection">
1056 <a name="functionattrs">-functionattrs: Deduce function attributes</a>
1058 <div class="doc_text">
1059 <p>A simple interprocedural pass which walks the call-graph, looking for
1060 functions which do not access or only read non-local memory, and marking them
1061 readnone/readonly. In addition, it marks function arguments (of pointer type)
1062 'nocapture' if a call to the function does not create any copies of the pointer
1063 value that outlive the call. This more or less means that the pointer is only
1064 dereferenced, and not returned from the function or stored in a global.
1065 This pass is implemented as a bottom-up traversal of the call-graph.
1069 <!-------------------------------------------------------------------------- -->
1070 <div class="doc_subsection">
1071 <a name="globaldce">-globaldce: Dead Global Elimination</a>
1073 <div class="doc_text">
1075 This transform is designed to eliminate unreachable internal globals from the
1076 program. It uses an aggressive algorithm, searching out globals that are
1077 known to be alive. After it finds all of the globals which are needed, it
1078 deletes whatever is left over. This allows it to delete recursive chunks of
1079 the program which are unreachable.
1083 <!-------------------------------------------------------------------------- -->
1084 <div class="doc_subsection">
1085 <a name="globalopt">-globalopt: Global Variable Optimizer</a>
1087 <div class="doc_text">
1089 This pass transforms simple global variables that never have their address
1090 taken. If obviously true, it marks read/write globals as constant, deletes
1091 variables only stored to, etc.
1095 <!-------------------------------------------------------------------------- -->
1096 <div class="doc_subsection">
1097 <a name="gvn">-gvn: Global Value Numbering</a>
1099 <div class="doc_text">
1101 This pass performs global value numbering to eliminate fully and partially
1102 redundant instructions. It also performs redundant load elimination.
1106 <!-------------------------------------------------------------------------- -->
1107 <div class="doc_subsection">
1108 <a name="indvars">-indvars: Canonicalize Induction Variables</a>
1110 <div class="doc_text">
1112 This transformation analyzes and transforms the induction variables (and
1113 computations derived from them) into simpler forms suitable for subsequent
1114 analysis and transformation.
1118 This transformation makes the following changes to each loop with an
1119 identifiable induction variable:
1123 <li>All loops are transformed to have a <em>single</em> canonical
1124 induction variable which starts at zero and steps by one.</li>
1125 <li>The canonical induction variable is guaranteed to be the first PHI node
1126 in the loop header block.</li>
1127 <li>Any pointer arithmetic recurrences are raised to use array
1132 If the trip count of a loop is computable, this pass also makes the following
1137 <li>The exit condition for the loop is canonicalized to compare the
1138 induction value against the exit value. This turns loops like:
1139 <blockquote><pre>for (i = 7; i*i < 1000; ++i)</pre></blockquote>
1141 <blockquote><pre>for (i = 0; i != 25; ++i)</pre></blockquote></li>
1142 <li>Any use outside of the loop of an expression derived from the indvar
1143 is changed to compute the derived value outside of the loop, eliminating
1144 the dependence on the exit value of the induction variable. If the only
1145 purpose of the loop is to compute the exit value of some derived
1146 expression, this transformation will make the loop dead.</li>
1150 This transformation should be followed by strength reduction after all of the
1151 desired loop transformations have been performed. Additionally, on targets
1152 where it is profitable, the loop could be transformed to count down to zero
1153 (the "do loop" optimization).
1157 <!-------------------------------------------------------------------------- -->
1158 <div class="doc_subsection">
1159 <a name="inline">-inline: Function Integration/Inlining</a>
1161 <div class="doc_text">
1163 Bottom-up inlining of functions into callees.
1167 <!-------------------------------------------------------------------------- -->
1168 <div class="doc_subsection">
1169 <a name="insert-edge-profiling">-insert-edge-profiling: Insert instrumentation for edge profiling</a>
1171 <div class="doc_text">
1173 This pass instruments the specified program with counters for edge profiling.
1174 Edge profiling can give a reasonable approximation of the hot paths through a
1175 program, and is used for a wide variety of program transformations.
1179 Note that this implementation is very naïve. It inserts a counter for
1180 <em>every</em> edge in the program, instead of using control flow information
1181 to prune the number of counters inserted.
1185 <!-------------------------------------------------------------------------- -->
1186 <div class="doc_subsection">
1187 <a name="insert-optimal-edge-profiling">-insert-optimal-edge-profiling: Insert optimal instrumentation for edge profiling</a>
1189 <div class="doc_text">
1190 <p>This pass instruments the specified program with counters for edge profiling.
1191 Edge profiling can give a reasonable approximation of the hot paths through a
1192 program, and is used for a wide variety of program transformations.
1196 <!-------------------------------------------------------------------------- -->
1197 <div class="doc_subsection">
1198 <a name="instcombine">-instcombine: Combine redundant instructions</a>
1200 <div class="doc_text">
1202 Combine instructions to form fewer, simple
1203 instructions. This pass does not modify the CFG This pass is where algebraic
1204 simplification happens.
1208 This pass combines things like:
1213 %Z = add i32 %Y, 1</pre></blockquote>
1220 >%Z = add i32 %X, 2</pre></blockquote>
1223 This is a simple worklist driven algorithm.
1227 This pass guarantees that the following canonicalizations are performed on
1232 <li>If a binary operator has a constant operand, it is moved to the right-
1234 <li>Bitwise operators with constant operands are always grouped so that
1235 shifts are performed first, then <code>or</code>s, then
1236 <code>and</code>s, then <code>xor</code>s.</li>
1237 <li>Compare instructions are converted from <code><</code>,
1238 <code>></code>, <code>≤</code>, or <code>≥</code> to
1239 <code>=</code> or <code>≠</code> if possible.</li>
1240 <li>All <code>cmp</code> instructions on boolean values are replaced with
1241 logical operations.</li>
1242 <li><code>add <var>X</var>, <var>X</var></code> is represented as
1243 <code>mul <var>X</var>, 2</code> ⇒ <code>shl <var>X</var>, 1</code></li>
1244 <li>Multiplies with a constant power-of-two argument are transformed into
1250 <!-------------------------------------------------------------------------- -->
1251 <div class="doc_subsection">
1252 <a name="internalize">-internalize: Internalize Global Symbols</a>
1254 <div class="doc_text">
1256 This pass loops over all of the functions in the input module, looking for a
1257 main function. If a main function is found, all other functions and all
1258 global variables with initializers are marked as internal.
1262 <!-------------------------------------------------------------------------- -->
1263 <div class="doc_subsection">
1264 <a name="ipconstprop">-ipconstprop: Interprocedural constant propagation</a>
1266 <div class="doc_text">
1268 This pass implements an <em>extremely</em> simple interprocedural constant
1269 propagation pass. It could certainly be improved in many different ways,
1270 like using a worklist. This pass makes arguments dead, but does not remove
1271 them. The existing dead argument elimination pass should be run after this
1272 to clean up the mess.
1276 <!-------------------------------------------------------------------------- -->
1277 <div class="doc_subsection">
1278 <a name="ipsccp">-ipsccp: Interprocedural Sparse Conditional Constant Propagation</a>
1280 <div class="doc_text">
1282 An interprocedural variant of <a href="#sccp">Sparse Conditional Constant
1287 <!-------------------------------------------------------------------------- -->
1288 <div class="doc_subsection">
1289 <a name="jump-threading">-jump-threading: Thread control through conditional blocks</a>
1291 <div class="doc_text">
1293 Jump threading tries to find distinct threads of control flow running through
1294 a basic block. This pass looks at blocks that have multiple predecessors and
1295 multiple successors. If one or more of the predecessors of the block can be
1296 proven to always cause a jump to one of the successors, we forward the edge
1297 from the predecessor to the successor by duplicating the contents of this
1301 An example of when this can occur is code like this:
1308 if (X < 3) {</pre>
1311 In this case, the unconditional branch at the end of the first if can be
1312 revectored to the false side of the second if.
1316 <!-------------------------------------------------------------------------- -->
1317 <div class="doc_subsection">
1318 <a name="lcssa">-lcssa: Loop-Closed SSA Form Pass</a>
1320 <div class="doc_text">
1322 This pass transforms loops by placing phi nodes at the end of the loops for
1323 all values that are live across the loop boundary. For example, it turns
1324 the left into the right code:
1328 >for (...) for (...)
1333 X3 = phi(X1, X2) X3 = phi(X1, X2)
1334 ... = X3 + 4 X4 = phi(X3)
1338 This is still valid LLVM; the extra phi nodes are purely redundant, and will
1339 be trivially eliminated by <code>InstCombine</code>. The major benefit of
1340 this transformation is that it makes many other loop optimizations, such as
1341 LoopUnswitching, simpler.
1345 <!-------------------------------------------------------------------------- -->
1346 <div class="doc_subsection">
1347 <a name="licm">-licm: Loop Invariant Code Motion</a>
1349 <div class="doc_text">
1351 This pass performs loop invariant code motion, attempting to remove as much
1352 code from the body of a loop as possible. It does this by either hoisting
1353 code into the preheader block, or by sinking code to the exit blocks if it is
1354 safe. This pass also promotes must-aliased memory locations in the loop to
1355 live in registers, thus hoisting and sinking "invariant" loads and stores.
1359 This pass uses alias analysis for two purposes:
1363 <li>Moving loop invariant loads and calls out of loops. If we can determine
1364 that a load or call inside of a loop never aliases anything stored to,
1365 we can hoist it or sink it like any other instruction.</li>
1366 <li>Scalar Promotion of Memory - If there is a store instruction inside of
1367 the loop, we try to move the store to happen AFTER the loop instead of
1368 inside of the loop. This can only happen if a few conditions are true:
1370 <li>The pointer stored through is loop invariant.</li>
1371 <li>There are no stores or loads in the loop which <em>may</em> alias
1372 the pointer. There are no calls in the loop which mod/ref the
1375 If these conditions are true, we can promote the loads and stores in the
1376 loop of the pointer to use a temporary alloca'd variable. We then use
1377 the mem2reg functionality to construct the appropriate SSA form for the
1381 <!-------------------------------------------------------------------------- -->
1382 <div class="doc_subsection">
1383 <a name="loop-deletion">-loop-deletion: Dead Loop Deletion Pass</a>
1385 <div class="doc_text">
1387 This file implements the Dead Loop Deletion Pass. This pass is responsible
1388 for eliminating loops with non-infinite computable trip counts that have no
1389 side effects or volatile instructions, and do not contribute to the
1390 computation of the function's return value.
1394 <!-------------------------------------------------------------------------- -->
1395 <div class="doc_subsection">
1396 <a name="loop-extract">-loop-extract: Extract loops into new functions</a>
1398 <div class="doc_text">
1400 A pass wrapper around the <code>ExtractLoop()</code> scalar transformation to
1401 extract each top-level loop into its own new function. If the loop is the
1402 <em>only</em> loop in a given function, it is not touched. This is a pass most
1403 useful for debugging via bugpoint.
1407 <!-------------------------------------------------------------------------- -->
1408 <div class="doc_subsection">
1409 <a name="loop-extract-single">-loop-extract-single: Extract at most one loop into a new function</a>
1411 <div class="doc_text">
1413 Similar to <a href="#loop-extract">Extract loops into new functions</a>,
1414 this pass extracts one natural loop from the program into a function if it
1415 can. This is used by bugpoint.
1419 <!-------------------------------------------------------------------------- -->
1420 <div class="doc_subsection">
1421 <a name="loop-index-split">-loop-index-split: Index Split Loops</a>
1423 <div class="doc_text">
1425 This pass divides loop's iteration range by spliting loop such that each
1426 individual loop is executed efficiently.
1430 <!-------------------------------------------------------------------------- -->
1431 <div class="doc_subsection">
1432 <a name="loop-reduce">-loop-reduce: Loop Strength Reduction</a>
1434 <div class="doc_text">
1436 This pass performs a strength reduction on array references inside loops that
1437 have as one or more of their components the loop induction variable. This is
1438 accomplished by creating a new value to hold the initial value of the array
1439 access for the first iteration, and then creating a new GEP instruction in
1440 the loop to increment the value by the appropriate amount.
1444 <!-------------------------------------------------------------------------- -->
1445 <div class="doc_subsection">
1446 <a name="loop-rotate">-loop-rotate: Rotate Loops</a>
1448 <div class="doc_text">
1449 <p>A simple loop rotation transformation.</p>
1452 <!-------------------------------------------------------------------------- -->
1453 <div class="doc_subsection">
1454 <a name="loop-unroll">-loop-unroll: Unroll loops</a>
1456 <div class="doc_text">
1458 This pass implements a simple loop unroller. It works best when loops have
1459 been canonicalized by the <a href="#indvars"><tt>-indvars</tt></a> pass,
1460 allowing it to determine the trip counts of loops easily.
1464 <!-------------------------------------------------------------------------- -->
1465 <div class="doc_subsection">
1466 <a name="loop-unswitch">-loop-unswitch: Unswitch loops</a>
1468 <div class="doc_text">
1470 This pass transforms loops that contain branches on loop-invariant conditions
1471 to have multiple loops. For example, it turns the left into the right code:
1483 This can increase the size of the code exponentially (doubling it every time
1484 a loop is unswitched) so we only unswitch if the resultant code will be
1485 smaller than a threshold.
1489 This pass expects LICM to be run before it to hoist invariant conditions out
1490 of the loop, to make the unswitching opportunity obvious.
1494 <!-------------------------------------------------------------------------- -->
1495 <div class="doc_subsection">
1496 <a name="loopsimplify">-loopsimplify: Canonicalize natural loops</a>
1498 <div class="doc_text">
1500 This pass performs several transformations to transform natural loops into a
1501 simpler form, which makes subsequent analyses and transformations simpler and
1506 Loop pre-header insertion guarantees that there is a single, non-critical
1507 entry edge from outside of the loop to the loop header. This simplifies a
1508 number of analyses and transformations, such as LICM.
1512 Loop exit-block insertion guarantees that all exit blocks from the loop
1513 (blocks which are outside of the loop that have predecessors inside of the
1514 loop) only have predecessors from inside of the loop (and are thus dominated
1515 by the loop header). This simplifies transformations such as store-sinking
1516 that are built into LICM.
1520 This pass also guarantees that loops will have exactly one backedge.
1524 Note that the simplifycfg pass will clean up blocks which are split out but
1525 end up being unnecessary, so usage of this pass should not pessimize
1530 This pass obviously modifies the CFG, but updates loop information and
1531 dominator information.
1535 <!-------------------------------------------------------------------------- -->
1536 <div class="doc_subsection">
1537 <a name="lowerallocs">-lowerallocs: Lower allocations from instructions to calls</a>
1539 <div class="doc_text">
1541 Turn <tt>malloc</tt> and <tt>free</tt> instructions into <tt>@malloc</tt> and
1542 <tt>@free</tt> calls.
1546 This is a target-dependent tranformation because it depends on the size of
1547 data types and alignment constraints.
1551 <!-------------------------------------------------------------------------- -->
1552 <div class="doc_subsection">
1553 <a name="loweratomic">-loweratomic: Lower atomic intrinsics
1555 <div class="doc_text">
1557 This pass lowers atomic intrinsics to non-atomic form for use in a known
1558 non-preemptible environment.
1562 The pass does not verify that the environment is non-preemptible (in
1563 general this would require knowledge of the entire call graph of the
1564 program including any libraries which may not be available in bitcode form);
1565 it simply lowers every atomic intrinsic.
1569 <!-------------------------------------------------------------------------- -->
1570 <div class="doc_subsection">
1571 <a name="lowerinvoke">-lowerinvoke: Lower invoke and unwind, for unwindless code generators</a>
1573 <div class="doc_text">
1575 This transformation is designed for use by code generators which do not yet
1576 support stack unwinding. This pass supports two models of exception handling
1577 lowering, the 'cheap' support and the 'expensive' support.
1581 'Cheap' exception handling support gives the program the ability to execute
1582 any program which does not "throw an exception", by turning 'invoke'
1583 instructions into calls and by turning 'unwind' instructions into calls to
1584 abort(). If the program does dynamically use the unwind instruction, the
1585 program will print a message then abort.
1589 'Expensive' exception handling support gives the full exception handling
1590 support to the program at the cost of making the 'invoke' instruction
1591 really expensive. It basically inserts setjmp/longjmp calls to emulate the
1592 exception handling as necessary.
1596 Because the 'expensive' support slows down programs a lot, and EH is only
1597 used for a subset of the programs, it must be specifically enabled by the
1598 <tt>-enable-correct-eh-support</tt> option.
1602 Note that after this pass runs the CFG is not entirely accurate (exceptional
1603 control flow edges are not correct anymore) so only very simple things should
1604 be done after the lowerinvoke pass has run (like generation of native code).
1605 This should not be used as a general purpose "my LLVM-to-LLVM pass doesn't
1606 support the invoke instruction yet" lowering pass.
1610 <!-------------------------------------------------------------------------- -->
1611 <div class="doc_subsection">
1612 <a name="lowersetjmp">-lowersetjmp: Lower Set Jump</a>
1614 <div class="doc_text">
1616 Lowers <tt>setjmp</tt> and <tt>longjmp</tt> to use the LLVM invoke and unwind
1617 instructions as necessary.
1621 Lowering of <tt>longjmp</tt> is fairly trivial. We replace the call with a
1622 call to the LLVM library function <tt>__llvm_sjljeh_throw_longjmp()</tt>.
1623 This unwinds the stack for us calling all of the destructors for
1624 objects allocated on the stack.
1628 At a <tt>setjmp</tt> call, the basic block is split and the <tt>setjmp</tt>
1629 removed. The calls in a function that have a <tt>setjmp</tt> are converted to
1630 invoke where the except part checks to see if it's a <tt>longjmp</tt>
1631 exception and, if so, if it's handled in the function. If it is, then it gets
1632 the value returned by the <tt>longjmp</tt> and goes to where the basic block
1633 was split. <tt>invoke</tt> instructions are handled in a similar fashion with
1634 the original except block being executed if it isn't a <tt>longjmp</tt>
1635 except that is handled by that function.
1639 <!-------------------------------------------------------------------------- -->
1640 <div class="doc_subsection">
1641 <a name="lowerswitch">-lowerswitch: Lower SwitchInst's to branches</a>
1643 <div class="doc_text">
1645 Rewrites <tt>switch</tt> instructions with a sequence of branches, which
1646 allows targets to get away with not implementing the switch instruction until
1651 <!-------------------------------------------------------------------------- -->
1652 <div class="doc_subsection">
1653 <a name="mem2reg">-mem2reg: Promote Memory to Register</a>
1655 <div class="doc_text">
1657 This file promotes memory references to be register references. It promotes
1658 <tt>alloca</tt> instructions which only have <tt>load</tt>s and
1659 <tt>store</tt>s as uses. An <tt>alloca</tt> is transformed by using dominator
1660 frontiers to place <tt>phi</tt> nodes, then traversing the function in
1661 depth-first order to rewrite <tt>load</tt>s and <tt>store</tt>s as
1662 appropriate. This is just the standard SSA construction algorithm to construct
1667 <!-------------------------------------------------------------------------- -->
1668 <div class="doc_subsection">
1669 <a name="memcpyopt">-memcpyopt: Optimize use of memcpy and friend</a>
1671 <div class="doc_text">
1673 This pass performs various transformations related to eliminating memcpy
1674 calls, or transforming sets of stores into memset's.
1678 <!-------------------------------------------------------------------------- -->
1679 <div class="doc_subsection">
1680 <a name="mergefunc">-mergefunc: Merge Functions</a>
1682 <div class="doc_text">
1683 <p>This pass looks for equivalent functions that are mergable and folds them.
1685 A hash is computed from the function, based on its type and number of
1688 Once all hashes are computed, we perform an expensive equality comparison
1689 on each function pair. This takes n^2/2 comparisons per bucket, so it's
1690 important that the hash function be high quality. The equality comparison
1691 iterates through each instruction in each basic block.
1693 When a match is found the functions are folded. If both functions are
1694 overridable, we move the functionality into a new internal function and
1695 leave two overridable thunks to it.
1699 <!-------------------------------------------------------------------------- -->
1700 <div class="doc_subsection">
1701 <a name="mergereturn">-mergereturn: Unify function exit nodes</a>
1703 <div class="doc_text">
1705 Ensure that functions have at most one <tt>ret</tt> instruction in them.
1706 Additionally, it keeps track of which node is the new exit node of the CFG.
1710 <!-------------------------------------------------------------------------- -->
1711 <div class="doc_subsection">
1712 <a name="partial-inliner">-partial-inliner: Partial Inliner</a>
1714 <div class="doc_text">
1715 <p>This pass performs partial inlining, typically by inlining an if
1716 statement that surrounds the body of the function.
1720 <!-------------------------------------------------------------------------- -->
1721 <div class="doc_subsection">
1722 <a name="partialspecialization">-partialspecialization: Partial Specialization</a>
1724 <div class="doc_text">
1725 <p>This pass finds function arguments that are often a common constant and
1726 specializes a version of the called function for that constant.
1728 This pass simply does the cloning for functions it specializes. It depends
1729 on <a href="#ipsccp">IPSCCP</a> and <a href="#deadargelim">DAE</a> to clean up the results.
1731 The initial heuristic favors constant arguments that are used in control
1736 <!-------------------------------------------------------------------------- -->
1737 <div class="doc_subsection">
1738 <a name="prune-eh">-prune-eh: Remove unused exception handling info</a>
1740 <div class="doc_text">
1742 This file implements a simple interprocedural pass which walks the call-graph,
1743 turning <tt>invoke</tt> instructions into <tt>call</tt> instructions if and
1744 only if the callee cannot throw an exception. It implements this as a
1745 bottom-up traversal of the call-graph.
1749 <!-------------------------------------------------------------------------- -->
1750 <div class="doc_subsection">
1751 <a name="reassociate">-reassociate: Reassociate expressions</a>
1753 <div class="doc_text">
1755 This pass reassociates commutative expressions in an order that is designed
1756 to promote better constant propagation, GCSE, LICM, PRE, etc.
1760 For example: 4 + (<var>x</var> + 5) ⇒ <var>x</var> + (4 + 5)
1764 In the implementation of this algorithm, constants are assigned rank = 0,
1765 function arguments are rank = 1, and other values are assigned ranks
1766 corresponding to the reverse post order traversal of current function
1767 (starting at 2), which effectively gives values in deep loops higher rank
1768 than values not in loops.
1772 <!-------------------------------------------------------------------------- -->
1773 <div class="doc_subsection">
1774 <a name="reg2mem">-reg2mem: Demote all values to stack slots</a>
1776 <div class="doc_text">
1778 This file demotes all registers to memory references. It is intented to be
1779 the inverse of <a href="#mem2reg"><tt>-mem2reg</tt></a>. By converting to
1780 <tt>load</tt> instructions, the only values live across basic blocks are
1781 <tt>alloca</tt> instructions and <tt>load</tt> instructions before
1782 <tt>phi</tt> nodes. It is intended that this should make CFG hacking much
1783 easier. To make later hacking easier, the entry block is split into two, such
1784 that all introduced <tt>alloca</tt> instructions (and nothing else) are in the
1789 <!-------------------------------------------------------------------------- -->
1790 <div class="doc_subsection">
1791 <a name="scalarrepl">-scalarrepl: Scalar Replacement of Aggregates</a>
1793 <div class="doc_text">
1795 The well-known scalar replacement of aggregates transformation. This
1796 transform breaks up <tt>alloca</tt> instructions of aggregate type (structure
1797 or array) into individual <tt>alloca</tt> instructions for each member if
1798 possible. Then, if possible, it transforms the individual <tt>alloca</tt>
1799 instructions into nice clean scalar SSA form.
1803 This combines a simple scalar replacement of aggregates algorithm with the <a
1804 href="#mem2reg"><tt>mem2reg</tt></a> algorithm because often interact,
1805 especially for C++ programs. As such, iterating between <tt>scalarrepl</tt>,
1806 then <a href="#mem2reg"><tt>mem2reg</tt></a> until we run out of things to
1811 <!-------------------------------------------------------------------------- -->
1812 <div class="doc_subsection">
1813 <a name="sccp">-sccp: Sparse Conditional Constant Propagation</a>
1815 <div class="doc_text">
1817 Sparse conditional constant propagation and merging, which can be summarized
1822 <li>Assumes values are constant unless proven otherwise</li>
1823 <li>Assumes BasicBlocks are dead unless proven otherwise</li>
1824 <li>Proves values to be constant, and replaces them with constants</li>
1825 <li>Proves conditional branches to be unconditional</li>
1829 Note that this pass has a habit of making definitions be dead. It is a good
1830 idea to to run a DCE pass sometime after running this pass.
1834 <!-------------------------------------------------------------------------- -->
1835 <div class="doc_subsection">
1836 <a name="sink">-sink: Code Sinking</a>
1838 <div class="doc_text">
1839 <p>This pass moves instructions into successor blocks, when possible, so that
1840 they aren't executed on paths where their results aren't needed.
1844 <!-------------------------------------------------------------------------- -->
1845 <div class="doc_subsection">
1846 <a name="simplify-libcalls">-simplify-libcalls: Simplify well-known library calls</a>
1848 <div class="doc_text">
1850 Applies a variety of small optimizations for calls to specific well-known
1851 function calls (e.g. runtime library functions). For example, a call
1852 <tt>exit(3)</tt> that occurs within the <tt>main()</tt> function can be
1853 transformed into simply <tt>return 3</tt>.
1857 <!-------------------------------------------------------------------------- -->
1858 <div class="doc_subsection">
1859 <a name="simplify-libcalls-halfpowr">-simplify-libcalls-halfpowr: Simplify half_powr library calls</a>
1861 <div class="doc_text">
1862 <p>Simple pass that applies an experimental transformation on calls
1863 to specific functions.
1867 <!-------------------------------------------------------------------------- -->
1868 <div class="doc_subsection">
1869 <a name="simplifycfg">-simplifycfg: Simplify the CFG</a>
1871 <div class="doc_text">
1873 Performs dead code elimination and basic block merging. Specifically:
1877 <li>Removes basic blocks with no predecessors.</li>
1878 <li>Merges a basic block into its predecessor if there is only one and the
1879 predecessor only has one successor.</li>
1880 <li>Eliminates PHI nodes for basic blocks with a single predecessor.</li>
1881 <li>Eliminates a basic block that only contains an unconditional
1886 <!-------------------------------------------------------------------------- -->
1887 <div class="doc_subsection">
1888 <a name="split-geps">-split-geps: Split complex GEPs into simple GEPs</a>
1890 <div class="doc_text">
1891 <p>This function breaks GEPs with more than 2 non-zero operands into smaller
1892 GEPs each with no more than 2 non-zero operands. This exposes redundancy
1893 between GEPs with common initial operand sequences.
1897 <!-------------------------------------------------------------------------- -->
1898 <div class="doc_subsection">
1899 <a name="ssi">-ssi: Static Single Information Construction</a>
1901 <div class="doc_text">
1902 <p>This pass converts a list of variables to the Static Single Information
1905 We are building an on-demand representation, that is, we do not convert
1906 every single variable in the target function to SSI form. Rather, we receive
1907 a list of target variables that must be converted. We also do not
1908 completely convert a target variable to the SSI format. Instead, we only
1909 change the variable in the points where new information can be attached
1910 to its live range, that is, at branch points.
1914 <!-------------------------------------------------------------------------- -->
1915 <div class="doc_subsection">
1916 <a name="ssi-everything">-ssi-everything: Static Single Information Construction (everything, intended for debugging)</a>
1918 <div class="doc_text">
1919 <p>A pass that runs <a href="#ssi">SSI</a> on every non-void variable, intended for debugging.
1923 <!-------------------------------------------------------------------------- -->
1924 <div class="doc_subsection">
1925 <a name="strip">-strip: Strip all symbols from a module</a>
1927 <div class="doc_text">
1929 performs code stripping. this transformation can delete:
1933 <li>names for virtual registers</li>
1934 <li>symbols for internal globals and functions</li>
1935 <li>debug information</li>
1939 note that this transformation makes code much less readable, so it should
1940 only be used in situations where the <tt>strip</tt> utility would be used,
1941 such as reducing code size or making it harder to reverse engineer code.
1945 <!-------------------------------------------------------------------------- -->
1946 <div class="doc_subsection">
1947 <a name="strip-dead-prototypes">-strip-dead-prototypes: Remove unused function declarations</a>
1949 <div class="doc_text">
1951 This pass loops over all of the functions in the input module, looking for
1952 dead declarations and removes them. Dead declarations are declarations of
1953 functions for which no implementation is available (i.e., declarations for
1954 unused library functions).
1958 <!-------------------------------------------------------------------------- -->
1959 <div class="doc_subsection">
1960 <a name="strip-debug-declare">-strip-debug-declare: Strip all llvm.dbg.declare intrinsics</a>
1962 <div class="doc_text">
1963 <p>This pass implements code stripping. Specifically, it can delete:
1965 <li>names for virtual registers</li>
1966 <li>symbols for internal globals and functions</li>
1967 <li>debug information</li>
1969 Note that this transformation makes code much less readable, so it should
1970 only be used in situations where the 'strip' utility would be used, such as
1971 reducing code size or making it harder to reverse engineer code.
1975 <!-------------------------------------------------------------------------- -->
1976 <div class="doc_subsection">
1977 <a name="strip-nondebug">-strip-nondebug: Strip all symbols, except dbg symbols, from a module</a>
1979 <div class="doc_text">
1980 <p>This pass implements code stripping. Specifically, it can delete:
1982 <li>names for virtual registers</li>
1983 <li>symbols for internal globals and functions</li>
1984 <li>debug information</li>
1986 Note that this transformation makes code much less readable, so it should
1987 only be used in situations where the 'strip' utility would be used, such as
1988 reducing code size or making it harder to reverse engineer code.
1992 <!-------------------------------------------------------------------------- -->
1993 <div class="doc_subsection">
1994 <a name="sretpromotion">-sretpromotion: Promote sret arguments</a>
1996 <div class="doc_text">
1998 This pass finds functions that return a struct (using a pointer to the struct
1999 as the first argument of the function, marked with the '<tt>sret</tt>' attribute) and
2000 replaces them with a new function that simply returns each of the elements of
2001 that struct (using multiple return values).
2005 This pass works under a number of conditions:
2009 <li>The returned struct must not contain other structs</li>
2010 <li>The returned struct must only be used to load values from</li>
2011 <li>The placeholder struct passed in is the result of an <tt>alloca</tt></li>
2015 <!-------------------------------------------------------------------------- -->
2016 <div class="doc_subsection">
2017 <a name="tailcallelim">-tailcallelim: Tail Call Elimination</a>
2019 <div class="doc_text">
2021 This file transforms calls of the current function (self recursion) followed
2022 by a return instruction with a branch to the entry of the function, creating
2023 a loop. This pass also implements the following extensions to the basic
2028 <li>Trivial instructions between the call and return do not prevent the
2029 transformation from taking place, though currently the analysis cannot
2030 support moving any really useful instructions (only dead ones).
2031 <li>This pass transforms functions that are prevented from being tail
2032 recursive by an associative expression to use an accumulator variable,
2033 thus compiling the typical naive factorial or <tt>fib</tt> implementation
2034 into efficient code.
2035 <li>TRE is performed if the function returns void, if the return
2036 returns the result returned by the call, or if the function returns a
2037 run-time constant on all exits from the function. It is possible, though
2038 unlikely, that the return returns something else (like constant 0), and
2039 can still be TRE'd. It can be TRE'd if <em>all other</em> return
2040 instructions in the function return the exact same value.
2041 <li>If it can prove that callees do not access theier caller stack frame,
2042 they are marked as eligible for tail call elimination (by the code
2047 <!-------------------------------------------------------------------------- -->
2048 <div class="doc_subsection">
2049 <a name="tailduplicate">-tailduplicate: Tail Duplication</a>
2051 <div class="doc_text">
2053 This pass performs a limited form of tail duplication, intended to simplify
2054 CFGs by removing some unconditional branches. This pass is necessary to
2055 straighten out loops created by the C front-end, but also is capable of
2056 making other code nicer. After this pass is run, the CFG simplify pass
2057 should be run to clean up the mess.
2061 <!-- ======================================================================= -->
2062 <div class="doc_section"> <a name="transform">Utility Passes</a></div>
2063 <div class="doc_text">
2064 <p>This section describes the LLVM Utility Passes.</p>
2067 <!-------------------------------------------------------------------------- -->
2068 <div class="doc_subsection">
2069 <a name="deadarghaX0r">-deadarghaX0r: Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</a>
2071 <div class="doc_text">
2073 Same as dead argument elimination, but deletes arguments to functions which
2074 are external. This is only for use by <a
2075 href="Bugpoint.html">bugpoint</a>.</p>
2078 <!-------------------------------------------------------------------------- -->
2079 <div class="doc_subsection">
2080 <a name="extract-blocks">-extract-blocks: Extract Basic Blocks From Module (for bugpoint use)</a>
2082 <div class="doc_text">
2084 This pass is used by bugpoint to extract all blocks from the module into their
2088 <!-------------------------------------------------------------------------- -->
2089 <div class="doc_subsection">
2090 <a name="instnamer">-instnamer: Assign names to anonymous instructions</a>
2092 <div class="doc_text">
2093 <p>This is a little utility pass that gives instructions names, this is mostly
2094 useful when diffing the effect of an optimization because deleting an
2095 unnamed instruction can change all other instruction numbering, making the
2100 <!-------------------------------------------------------------------------- -->
2101 <div class="doc_subsection">
2102 <a name="preverify">-preverify: Preliminary module verification</a>
2104 <div class="doc_text">
2106 Ensures that the module is in the form required by the <a
2107 href="#verifier">Module Verifier</a> pass.
2111 Running the verifier runs this pass automatically, so there should be no need
2116 <!-------------------------------------------------------------------------- -->
2117 <div class="doc_subsection">
2118 <a name="verify">-verify: Module Verifier</a>
2120 <div class="doc_text">
2122 Verifies an LLVM IR code. This is useful to run after an optimization which is
2123 undergoing testing. Note that <tt>llvm-as</tt> verifies its input before
2124 emitting bitcode, and also that malformed bitcode is likely to make LLVM
2125 crash. All language front-ends are therefore encouraged to verify their output
2126 before performing optimizing transformations.
2130 <li>Both of a binary operator's parameters are of the same type.</li>
2131 <li>Verify that the indices of mem access instructions match other
2133 <li>Verify that arithmetic and other things are only performed on
2134 first-class types. Verify that shifts and logicals only happen on
2136 <li>All of the constants in a switch statement are of the correct type.</li>
2137 <li>The code is in valid SSA form.</li>
2138 <li>It is illegal to put a label into any other type (like a structure) or
2140 <li>Only phi nodes can be self referential: <tt>%x = add i32 %x, %x</tt> is
2142 <li>PHI nodes must have an entry for each predecessor, with no extras.</li>
2143 <li>PHI nodes must be the first thing in a basic block, all grouped
2145 <li>PHI nodes must have at least one entry.</li>
2146 <li>All basic blocks should only end with terminator insts, not contain
2148 <li>The entry node to a function must not have predecessors.</li>
2149 <li>All Instructions must be embedded into a basic block.</li>
2150 <li>Functions cannot take a void-typed parameter.</li>
2151 <li>Verify that a function's argument list agrees with its declared
2153 <li>It is illegal to specify a name for a void value.</li>
2154 <li>It is illegal to have a internal global value with no initializer.</li>
2155 <li>It is illegal to have a ret instruction that returns a value that does
2156 not agree with the function return value type.</li>
2157 <li>Function call argument types match the function prototype.</li>
2158 <li>All other things that are tested by asserts spread about the code.</li>
2162 Note that this does not provide full security verification (like Java), but
2163 instead just tries to ensure that code is well-formed.
2167 <!-------------------------------------------------------------------------- -->
2168 <div class="doc_subsection">
2169 <a name="view-cfg">-view-cfg: View CFG of function</a>
2171 <div class="doc_text">
2173 Displays the control flow graph using the GraphViz tool.
2177 <!-------------------------------------------------------------------------- -->
2178 <div class="doc_subsection">
2179 <a name="view-cfg-only">-view-cfg-only: View CFG of function (with no function bodies)</a>
2181 <div class="doc_text">
2183 Displays the control flow graph using the GraphViz tool, but omitting function
2188 <!-------------------------------------------------------------------------- -->
2189 <div class="doc_subsection">
2190 <a name="view-dom">-view-dom: View dominator tree of function</a>
2192 <div class="doc_text">
2194 Displays the dominator tree using the GraphViz tool.
2198 <!-------------------------------------------------------------------------- -->
2199 <div class="doc_subsection">
2200 <a name="view-dom-only">-view-dom-only: View dominator tree of function (with no function
2204 <div class="doc_text">
2206 Displays the dominator tree using the GraphViz tool, but omitting function
2211 <!-------------------------------------------------------------------------- -->
2212 <div class="doc_subsection">
2213 <a name="view-postdom">-view-postdom: View post dominator tree of function</a>
2215 <div class="doc_text">
2217 Displays the post dominator tree using the GraphViz tool.
2221 <!-------------------------------------------------------------------------- -->
2222 <div class="doc_subsection">
2223 <a name="view-postdom-only">-view-postdom-only: View post dominator tree of function (with no
2227 <div class="doc_text">
2229 Displays the post dominator tree using the GraphViz tool, but omitting
2234 <!-- *********************************************************************** -->
2238 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2239 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
2240 <a href="http://validator.w3.org/check/referer"><img
2241 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
2243 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br>
2244 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
2245 Last modified: $Date$