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="#lowerinvoke">-lowerinvoke</a></td><td>Lower invoke and unwind, for unwindless code generators</td></tr>
171 <tr><td><a href="#lowersetjmp">-lowersetjmp</a></td><td>Lower Set Jump</td></tr>
172 <tr><td><a href="#lowerswitch">-lowerswitch</a></td><td>Lower SwitchInst's to branches</td></tr>
173 <tr><td><a href="#mem2reg">-mem2reg</a></td><td>Promote Memory to Register</td></tr>
174 <tr><td><a href="#memcpyopt">-memcpyopt</a></td><td>Optimize use of memcpy and friends</td></tr>
175 <tr><td><a href="#mergefunc">-mergefunc</a></td><td>Merge Functions</td></tr>
176 <tr><td><a href="#mergereturn">-mergereturn</a></td><td>Unify function exit nodes</td></tr>
177 <tr><td><a href="#partial-inliner">-partial-inliner</a></td><td>Partial Inliner</td></tr>
178 <tr><td><a href="#partialspecialization">-partialspecialization</a></td><td>Partial Specialization</td></tr>
179 <tr><td><a href="#prune-eh">-prune-eh</a></td><td>Remove unused exception handling info</td></tr>
180 <tr><td><a href="#reassociate">-reassociate</a></td><td>Reassociate expressions</td></tr>
181 <tr><td><a href="#reg2mem">-reg2mem</a></td><td>Demote all values to stack slots</td></tr>
182 <tr><td><a href="#scalarrepl">-scalarrepl</a></td><td>Scalar Replacement of Aggregates</td></tr>
183 <tr><td><a href="#sccp">-sccp</a></td><td>Sparse Conditional Constant Propagation</td></tr>
184 <tr><td><a href="#sink">-sink</a></td><td>Code Sinking</td></tr>
185 <tr><td><a href="#simplify-libcalls">-simplify-libcalls</a></td><td>Simplify well-known library calls</td></tr>
186 <tr><td><a href="#simplify-libcalls-halfpowr">-simplify-libcalls-halfpowr</a></td><td>Simplify half_powr library calls</td></tr>
187 <tr><td><a href="#simplifycfg">-simplifycfg</a></td><td>Simplify the CFG</td></tr>
188 <tr><td><a href="#split-geps">-split-geps</a></td><td>Split complex GEPs into simple GEPs</td></tr>
189 <tr><td><a href="#ssi">-ssi</a></td><td>Static Single Information Construction</td></tr>
190 <tr><td><a href="#ssi-everything">-ssi-everything</a></td><td>Static Single Information Construction (everything, intended for debugging)</td></tr>
191 <tr><td><a href="#strip">-strip</a></td><td>Strip all symbols from a module</td></tr>
192 <tr><td><a href="#strip-dead-debug-info">-strip-dead-debug-info</a></td><td>Strip debug info for unused symbols</td></tr>
193 <tr><td><a href="#strip-dead-prototypes">-strip-dead-prototypes</a></td><td>Remove unused function declarations</td></tr>
194 <tr><td><a href="#strip-debug-declare">-strip-debug-declare</a></td><td>Strip all llvm.dbg.declare intrinsics</td></tr>
195 <tr><td><a href="#strip-nondebug">-strip-nondebug</a></td><td>Strip all symbols, except dbg symbols, from a module</td></tr>
196 <tr><td><a href="#sretpromotion">-sretpromotion</a></td><td>Promote sret arguments</td></tr>
197 <tr><td><a href="#tailcallelim">-tailcallelim</a></td><td>Tail Call Elimination</td></tr>
198 <tr><td><a href="#tailduplicate">-tailduplicate</a></td><td>Tail Duplication</td></tr>
201 <tr><th colspan="2"><b>UTILITY PASSES</b></th></tr>
202 <tr><th>Option</th><th>Name</th></tr>
203 <tr><td><a href="#deadarghaX0r">-deadarghaX0r</a></td><td>Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</td></tr>
204 <tr><td><a href="#extract-blocks">-extract-blocks</a></td><td>Extract Basic Blocks From Module (for bugpoint use)</td></tr>
205 <tr><td><a href="#instnamer">-instnamer</a></td><td>Assign names to anonymous instructions</td></tr>
206 <tr><td><a href="#preverify">-preverify</a></td><td>Preliminary module verification</td></tr>
207 <tr><td><a href="#verify">-verify</a></td><td>Module Verifier</td></tr>
208 <tr><td><a href="#view-cfg">-view-cfg</a></td><td>View CFG of function</td></tr>
209 <tr><td><a href="#view-cfg-only">-view-cfg-only</a></td><td>View CFG of function (with no function bodies)</td></tr>
210 <tr><td><a href="#view-dom">-view-dom</a></td><td>View dominator tree of function</td></tr>
211 <tr><td><a href="#view-dom-only">-view-dom-only</a></td><td>View dominator tree of function (with no function bodies)</td></tr>
212 <tr><td><a href="#view-postdom">-view-postdom</a></td><td>View post dominator tree of function</td></tr>
213 <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>
217 <!-- ======================================================================= -->
218 <div class="doc_section"> <a name="example">Analysis Passes</a></div>
219 <div class="doc_text">
220 <p>This section describes the LLVM Analysis Passes.</p>
223 <!-------------------------------------------------------------------------- -->
224 <div class="doc_subsection">
225 <a name="aa-eval">-aa-eval: Exhaustive Alias Analysis Precision Evaluator</a>
227 <div class="doc_text">
228 <p>This is a simple N^2 alias analysis accuracy evaluator.
229 Basically, for each function in the program, it simply queries to see how the
230 alias analysis implementation answers alias queries between each pair of
231 pointers in the function.</p>
233 <p>This is inspired and adapted from code by: Naveen Neelakantam, Francesco
234 Spadini, and Wojciech Stryjewski.</p>
237 <!-------------------------------------------------------------------------- -->
238 <div class="doc_subsection">
239 <a name="basicaa">-basicaa: Basic Alias Analysis (default AA impl)</a>
241 <div class="doc_text">
243 This is the default implementation of the Alias Analysis interface
244 that simply implements a few identities (two different globals cannot alias,
245 etc), but otherwise does no analysis.
249 <!-------------------------------------------------------------------------- -->
250 <div class="doc_subsection">
251 <a name="basiccg">-basiccg: Basic CallGraph Construction</a>
253 <div class="doc_text">
254 <p>Yet to be written.</p>
257 <!-------------------------------------------------------------------------- -->
258 <div class="doc_subsection">
259 <a name="codegenprepare">-codegenprepare: Optimize for code generation</a>
261 <div class="doc_text">
263 This pass munges the code in the input function to better prepare it for
264 SelectionDAG-based code generation. This works around limitations in it's
265 basic-block-at-a-time approach. It should eventually be removed.
269 <!-------------------------------------------------------------------------- -->
270 <div class="doc_subsection">
271 <a name="count-aa">-count-aa: Count Alias Analysis Query Responses</a>
273 <div class="doc_text">
275 A pass which can be used to count how many alias queries
276 are being made and how the alias analysis implementation being used responds.
280 <!-------------------------------------------------------------------------- -->
281 <div class="doc_subsection">
282 <a name="debug-aa">-debug-aa: AA use debugger</a>
284 <div class="doc_text">
286 This simple pass checks alias analysis users to ensure that if they
287 create a new value, they do not query AA without informing it of the value.
288 It acts as a shim over any other AA pass you want.
292 Yes keeping track of every value in the program is expensive, but this is
297 <!-------------------------------------------------------------------------- -->
298 <div class="doc_subsection">
299 <a name="domfrontier">-domfrontier: Dominance Frontier 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="domtree">-domtree: Dominator Tree Construction</a>
312 <div class="doc_text">
314 This pass is a simple dominator construction algorithm for finding forward
319 <!-------------------------------------------------------------------------- -->
320 <div class="doc_subsection">
321 <a name="dot-callgraph">-dot-callgraph: Print Call Graph to 'dot' file</a>
323 <div class="doc_text">
325 This pass, only available in <code>opt</code>, prints the call graph into a
326 <code>.dot</code> graph. This graph can then be processed with the "dot" tool
327 to convert it to postscript or some other suitable format.
331 <!-------------------------------------------------------------------------- -->
332 <div class="doc_subsection">
333 <a name="dot-cfg">-dot-cfg: Print CFG of function to 'dot' file</a>
335 <div class="doc_text">
337 This pass, only available in <code>opt</code>, prints the control flow graph
338 into a <code>.dot</code> graph. This graph can then be processed with the
339 "dot" tool to convert it to postscript or some other suitable format.
343 <!-------------------------------------------------------------------------- -->
344 <div class="doc_subsection">
345 <a name="dot-cfg-only">-dot-cfg-only: Print CFG of function to 'dot' file (with no function bodies)</a>
347 <div class="doc_text">
349 This pass, only available in <code>opt</code>, prints the control flow graph
350 into a <code>.dot</code> graph, omitting the function bodies. This graph can
351 then be processed with the "dot" tool to convert it to postscript or some
352 other suitable format.
356 <!-------------------------------------------------------------------------- -->
357 <div class="doc_subsection">
358 <a name="dot-dom">-dot-dom: Print dominator tree of function to 'dot' file</a>
360 <div class="doc_text">
362 This pass, only available in <code>opt</code>, prints the dominator tree
363 into a <code>.dot</code> graph. This graph can then be processed with the
364 "dot" tool to convert it to postscript or some other suitable format.
368 <!-------------------------------------------------------------------------- -->
369 <div class="doc_subsection">
370 <a name="dot-dom-only">-dot-dom-only: Print dominator tree of function to 'dot' file (with no
373 <div class="doc_text">
375 This pass, only available in <code>opt</code>, prints the dominator tree
376 into a <code>.dot</code> graph, omitting the function bodies. This graph can
377 then be processed with the "dot" tool to convert it to postscript or some
378 other suitable format.
382 <!-------------------------------------------------------------------------- -->
383 <div class="doc_subsection">
384 <a name="dot-postdom">dot-postdom: Print post dominator tree of function to 'dot' file</a>
386 <div class="doc_text">
388 This pass, only available in <code>opt</code>, prints the post dominator tree
389 into a <code>.dot</code> graph. This graph can then be processed with the
390 "dot" tool to convert it to postscript or some other suitable format.
394 <!-------------------------------------------------------------------------- -->
395 <div class="doc_subsection">
396 <a name="dot-postdom-only">dot-postdom-only: Print post dominator tree of function to 'dot' file
397 (with no function bodies)</a>
399 <div class="doc_text">
401 This pass, only available in <code>opt</code>, prints the post dominator tree
402 into a <code>.dot</code> graph, omitting the function bodies. This graph can
403 then be processed with the "dot" tool to convert it to postscript or some
404 other suitable format.
408 <!-------------------------------------------------------------------------- -->
409 <div class="doc_subsection">
410 <a name="globalsmodref-aa">-globalsmodref-aa: Simple mod/ref analysis for globals</a>
412 <div class="doc_text">
414 This simple pass provides alias and mod/ref information for global values
415 that do not have their address taken, and keeps track of whether functions
416 read or write memory (are "pure"). For this simple (but very common) case,
417 we can provide pretty accurate and useful information.
421 <!-------------------------------------------------------------------------- -->
422 <div class="doc_subsection">
423 <a name="instcount">-instcount: Counts the various types of Instructions</a>
425 <div class="doc_text">
427 This pass collects the count of all instructions and reports them
431 <!-------------------------------------------------------------------------- -->
432 <div class="doc_subsection">
433 <a name="interprocedural-aa-eval">-interprocedural-aa-eval: Exhaustive Interprocedural Alias Analysis Precision Evaluator</a>
435 <div class="doc_text">
436 <p>This pass implements a simple N^2 alias analysis accuracy evaluator.
437 Basically, for each function in the program, it simply queries to see how the
438 alias analysis implementation answers alias queries between each pair of
439 pointers in the function.
443 <!-------------------------------------------------------------------------- -->
444 <div class="doc_subsection">
445 <a name="interprocedural-basic-aa">-interprocedural-basic-aa: Interprocedural Basic Alias Analysis</a>
447 <div class="doc_text">
448 <p>This pass defines the default implementation of the Alias Analysis interface
449 that simply implements a few identities (two different globals cannot alias,
450 etc), but otherwise does no analysis.
454 <!-------------------------------------------------------------------------- -->
455 <div class="doc_subsection">
456 <a name="intervals">-intervals: Interval Partition Construction</a>
458 <div class="doc_text">
460 This analysis calculates and represents the interval partition of a function,
461 or a preexisting interval partition.
465 In this way, the interval partition may be used to reduce a flow graph down
466 to its degenerate single node interval partition (unless it is irreducible).
470 <!-------------------------------------------------------------------------- -->
471 <div class="doc_subsection">
472 <a name="iv-users">-iv-users: Induction Variable Users</a>
474 <div class="doc_text">
475 <p>Bookkeeping for "interesting" users of expressions computed from
476 induction variables.</p>
479 <!-------------------------------------------------------------------------- -->
480 <div class="doc_subsection">
481 <a name="lazy-value-info">-lazy-value-info: Lazy Value Information Analysis</a>
483 <div class="doc_text">
484 <p>Interface for lazy computation of value constraint information.</p>
487 <!-------------------------------------------------------------------------- -->
488 <div class="doc_subsection">
489 <a name="lda">-lda: Loop Dependence Analysis</a>
491 <div class="doc_text">
492 <p>Loop dependence analysis framework, which is used to detect dependences in
493 memory accesses in loops.</p>
496 <!-------------------------------------------------------------------------- -->
497 <div class="doc_subsection">
498 <a name="libcall-aa">-libcall-aa: LibCall Alias Analysis</a>
500 <div class="doc_text">
501 <p>LibCall Alias Analysis.</p>
504 <!-------------------------------------------------------------------------- -->
505 <div class="doc_subsection">
506 <a name="lint">-lint: Check for common errors in LLVM IR</a>
508 <div class="doc_text">
509 <p>This pass statically checks for common and easily-identified constructs
510 which produce undefined or likely unintended behavior in LLVM IR.</p>
512 <p>It is not a guarantee of correctness, in two ways. First, it isn't
513 comprehensive. There are checks which could be done statically which are
514 not yet implemented. Some of these are indicated by TODO comments, but
515 those aren't comprehensive either. Second, many conditions cannot be
516 checked statically. This pass does no dynamic instrumentation, so it
517 can't check for all possible problems.</p>
519 <p>Another limitation is that it assumes all code will be executed. A store
520 through a null pointer in a basic block which is never reached is harmless,
521 but this pass will warn about it anyway.</p>
523 <p>Optimization passes may make conditions that this pass checks for more or
524 less obvious. If an optimization pass appears to be introducing a warning,
525 it may be that the optimization pass is merely exposing an existing
526 condition in the code.</p>
528 <p>This code may be run before instcombine. In many cases, instcombine checks
529 for the same kinds of things and turns instructions with undefined behavior
530 into unreachable (or equivalent). Because of this, this pass makes some
531 effort to look through bitcasts and so on.
535 <!-------------------------------------------------------------------------- -->
536 <div class="doc_subsection">
537 <a name="live-values">-live-values: Values Liveness Analysis</a>
539 <div class="doc_text">
540 <p>LLVM IR Value liveness analysis pass.</p>
543 <!-------------------------------------------------------------------------- -->
544 <div class="doc_subsection">
545 <a name="loops">-loops: Natural Loop Construction</a>
547 <div class="doc_text">
549 This analysis is used to identify natural loops and determine the loop depth
550 of various nodes of the CFG. Note that the loops identified may actually be
551 several natural loops that share the same header node... not just a single
556 <!-------------------------------------------------------------------------- -->
557 <div class="doc_subsection">
558 <a name="memdep">-memdep: Memory Dependence Analysis</a>
560 <div class="doc_text">
562 An analysis that determines, for a given memory operation, what preceding
563 memory operations it depends on. It builds on alias analysis information, and
564 tries to provide a lazy, caching interface to a common kind of alias
569 <!-------------------------------------------------------------------------- -->
570 <div class="doc_subsection">
571 <a name="module-debuginfo">-module-debuginfo: Prints module debug info metadata</a>
573 <div class="doc_text">
574 <p>This pass decodes the debug info metadata in a module and prints in a
575 (sufficiently-prepared-) human-readable form.
577 For example, run this pass from opt along with the -analyze option, and
578 it'll print to standard output.
582 <!-------------------------------------------------------------------------- -->
583 <div class="doc_subsection">
584 <a name="no-aa">-no-aa: No Alias Analysis (always returns 'may' alias)</a>
586 <div class="doc_text">
588 Always returns "I don't know" for alias queries. NoAA is unlike other alias
589 analysis implementations, in that it does not chain to a previous analysis. As
590 such it doesn't follow many of the rules that other alias analyses must.
594 <!-------------------------------------------------------------------------- -->
595 <div class="doc_subsection">
596 <a name="no-profile">-no-profile: No Profile Information</a>
598 <div class="doc_text">
600 The default "no profile" implementation of the abstract
601 <code>ProfileInfo</code> interface.
605 <!-------------------------------------------------------------------------- -->
606 <div class="doc_subsection">
607 <a name="pointertracking">-pointertracking: Track pointer bounds.</a>
609 <div class="doc_text">
610 <p>Tracking of pointer bounds.
614 <!-------------------------------------------------------------------------- -->
615 <div class="doc_subsection">
616 <a name="postdomfrontier">-postdomfrontier: Post-Dominance Frontier Construction</a>
618 <div class="doc_text">
620 This pass is a simple post-dominator construction algorithm for finding
621 post-dominator frontiers.
625 <!-------------------------------------------------------------------------- -->
626 <div class="doc_subsection">
627 <a name="postdomtree">-postdomtree: Post-Dominator Tree Construction</a>
629 <div class="doc_text">
631 This pass is a simple post-dominator construction algorithm for finding
636 <!-------------------------------------------------------------------------- -->
637 <div class="doc_subsection">
638 <a name="print-alias-sets">-print-alias-sets: Alias Set Printer</a>
640 <div class="doc_text">
641 <p>Yet to be written.</p>
644 <!-------------------------------------------------------------------------- -->
645 <div class="doc_subsection">
646 <a name="print-callgraph">-print-callgraph: Print a call graph</a>
648 <div class="doc_text">
650 This pass, only available in <code>opt</code>, prints the call graph to
651 standard output in a human-readable form.
655 <!-------------------------------------------------------------------------- -->
656 <div class="doc_subsection">
657 <a name="print-callgraph-sccs">-print-callgraph-sccs: Print SCCs of the Call Graph</a>
659 <div class="doc_text">
661 This pass, only available in <code>opt</code>, prints the SCCs of the call
662 graph to standard output in a human-readable form.
666 <!-------------------------------------------------------------------------- -->
667 <div class="doc_subsection">
668 <a name="print-cfg-sccs">-print-cfg-sccs: Print SCCs of each function CFG</a>
670 <div class="doc_text">
672 This pass, only available in <code>opt</code>, prints the SCCs of each
673 function CFG to standard output in a human-readable form.
677 <!-------------------------------------------------------------------------- -->
678 <div class="doc_subsection">
679 <a name="print-dbginfo">-print-dbginfo: Print debug info in human readable form</a>
681 <div class="doc_text">
682 <p>Pass that prints instructions, and associated debug info:
685 <li>source/line/col information</li>
686 <li>original variable name</li>
687 <li>original type name</li>
693 <!-------------------------------------------------------------------------- -->
694 <div class="doc_subsection">
695 <a name="print-dom-info">-print-dom-info: Dominator Info Printer</a>
697 <div class="doc_text">
698 <p>Dominator Info Printer.</p>
701 <!-------------------------------------------------------------------------- -->
702 <div class="doc_subsection">
703 <a name="print-externalfnconstants">-print-externalfnconstants: Print external fn callsites passed constants</a>
705 <div class="doc_text">
707 This pass, only available in <code>opt</code>, prints out call sites to
708 external functions that are called with constant arguments. This can be
709 useful when looking for standard library functions we should constant fold
710 or handle in alias analyses.
714 <!-------------------------------------------------------------------------- -->
715 <div class="doc_subsection">
716 <a name="print-function">-print-function: Print function to stderr</a>
718 <div class="doc_text">
720 The <code>PrintFunctionPass</code> class is designed to be pipelined with
721 other <code>FunctionPass</code>es, and prints out the functions of the module
722 as they are processed.
726 <!-------------------------------------------------------------------------- -->
727 <div class="doc_subsection">
728 <a name="print-module">-print-module: Print module to stderr</a>
730 <div class="doc_text">
732 This pass simply prints out the entire module when it is executed.
736 <!-------------------------------------------------------------------------- -->
737 <div class="doc_subsection">
738 <a name="print-used-types">-print-used-types: Find Used Types</a>
740 <div class="doc_text">
742 This pass is used to seek out all of the types in use by the program. Note
743 that this analysis explicitly does not include types only used by the symbol
747 <!-------------------------------------------------------------------------- -->
748 <div class="doc_subsection">
749 <a name="profile-estimator">-profile-estimator: Estimate profiling information</a>
751 <div class="doc_text">
752 <p>Profiling information that estimates the profiling information
753 in a very crude and unimaginative way.
757 <!-------------------------------------------------------------------------- -->
758 <div class="doc_subsection">
759 <a name="profile-loader">-profile-loader: Load profile information from llvmprof.out</a>
761 <div class="doc_text">
763 A concrete implementation of profiling information that loads the information
764 from a profile dump file.
768 <!-------------------------------------------------------------------------- -->
769 <div class="doc_subsection">
770 <a name="profile-verifier">-profile-verifier: Verify profiling information</a>
772 <div class="doc_text">
773 <p>Pass that checks profiling information for plausibility.</p>
775 <div class="doc_subsection">
776 <a name="regions">-regions: Detect single entry single exit regions in a function</a>
778 <div class="doc_text">
780 The <code>RegionInfo</code> pass detects single entry single exit regions in a
781 function, where a region is defined as any subgraph that is connected to the
782 remaining graph at only two spots. Furthermore, an hierarchical region tree is
787 <!-------------------------------------------------------------------------- -->
788 <div class="doc_subsection">
789 <a name="scalar-evolution">-scalar-evolution: Scalar Evolution Analysis</a>
791 <div class="doc_text">
793 The <code>ScalarEvolution</code> analysis can be used to analyze and
794 catagorize scalar expressions in loops. It specializes in recognizing general
795 induction variables, representing them with the abstract and opaque
796 <code>SCEV</code> class. Given this analysis, trip counts of loops and other
797 important properties can be obtained.
801 This analysis is primarily useful for induction variable substitution and
806 <!-------------------------------------------------------------------------- -->
807 <div class="doc_subsection">
808 <a name="scev-aa">-scev-aa: </a>
810 <div class="doc_text">
811 <p>Simple alias analysis implemented in terms of ScalarEvolution queries.
813 This differs from traditional loop dependence analysis in that it tests
814 for dependencies within a single iteration of a loop, rather than
815 dependencies between different iterations.
817 ScalarEvolution has a more complete understanding of pointer arithmetic
818 than BasicAliasAnalysis' collection of ad-hoc analyses.
822 <!-------------------------------------------------------------------------- -->
823 <div class="doc_subsection">
824 <a name="strip-dead-debug-info">-strip-dead-debug-info: Strip debug info for unused symbols</a>
826 <div class="doc_text">
828 performs code stripping. this transformation can delete:
832 <li>names for virtual registers</li>
833 <li>symbols for internal globals and functions</li>
834 <li>debug information</li>
838 note that this transformation makes code much less readable, so it should
839 only be used in situations where the <tt>strip</tt> utility would be used,
840 such as reducing code size or making it harder to reverse engineer code.
844 <!-------------------------------------------------------------------------- -->
845 <div class="doc_subsection">
846 <a name="targetdata">-targetdata: Target Data Layout</a>
848 <div class="doc_text">
849 <p>Provides other passes access to information on how the size and alignment
850 required by the the target ABI for various data types.</p>
853 <!-- ======================================================================= -->
854 <div class="doc_section"> <a name="transform">Transform Passes</a></div>
855 <div class="doc_text">
856 <p>This section describes the LLVM Transform Passes.</p>
859 <!-------------------------------------------------------------------------- -->
860 <div class="doc_subsection">
861 <a name="abcd">-abcd: Remove redundant conditional branches</a>
863 <div class="doc_text">
864 <p>ABCD removes conditional branch instructions that can be proved redundant.
865 With the SSI representation, each variable has a constraint. By analyzing these
866 constraints we can prove that a branch is redundant. When a branch is proved
867 redundant it means that one direction will always be taken; thus, we can change
868 this branch into an unconditional jump.</p>
869 <p>It is advisable to run <a href="#simplifycfg">SimplifyCFG</a> and
870 <a href="#adce">Aggressive Dead Code Elimination</a> after ABCD
871 to clean up the code.</p>
874 <!-------------------------------------------------------------------------- -->
875 <div class="doc_subsection">
876 <a name="adce">-adce: Aggressive Dead Code Elimination</a>
878 <div class="doc_text">
879 <p>ADCE aggressively tries to eliminate code. This pass is similar to
880 <a href="#dce">DCE</a> but it assumes that values are dead until proven
881 otherwise. This is similar to <a href="#sccp">SCCP</a>, except applied to
882 the liveness of values.</p>
885 <!-------------------------------------------------------------------------- -->
886 <div class="doc_subsection">
887 <a name="always-inline">-always-inline: Inliner for always_inline functions</a>
889 <div class="doc_text">
890 <p>A custom inliner that handles only functions that are marked as
894 <!-------------------------------------------------------------------------- -->
895 <div class="doc_subsection">
896 <a name="argpromotion">-argpromotion: Promote 'by reference' arguments to scalars</a>
898 <div class="doc_text">
900 This pass promotes "by reference" arguments to be "by value" arguments. In
901 practice, this means looking for internal functions that have pointer
902 arguments. If it can prove, through the use of alias analysis, that an
903 argument is *only* loaded, then it can pass the value into the function
904 instead of the address of the value. This can cause recursive simplification
905 of code and lead to the elimination of allocas (especially in C++ template
910 This pass also handles aggregate arguments that are passed into a function,
911 scalarizing them if the elements of the aggregate are only loaded. Note that
912 it refuses to scalarize aggregates which would require passing in more than
913 three operands to the function, because passing thousands of operands for a
914 large array or structure is unprofitable!
918 Note that this transformation could also be done for arguments that are only
919 stored to (returning the value instead), but does not currently. This case
920 would be best handled when and if LLVM starts supporting multiple return
921 values from functions.
925 <!-------------------------------------------------------------------------- -->
926 <div class="doc_subsection">
927 <a name="block-placement">-block-placement: Profile Guided Basic Block Placement</a>
929 <div class="doc_text">
930 <p>This pass is a very simple profile guided basic block placement algorithm.
931 The idea is to put frequently executed blocks together at the start of the
932 function and hopefully increase the number of fall-through conditional
933 branches. If there is no profile information for a particular function, this
934 pass basically orders blocks in depth-first order.</p>
937 <!-------------------------------------------------------------------------- -->
938 <div class="doc_subsection">
939 <a name="break-crit-edges">-break-crit-edges: Break critical edges in CFG</a>
941 <div class="doc_text">
943 Break all of the critical edges in the CFG by inserting a dummy basic block.
944 It may be "required" by passes that cannot deal with critical edges. This
945 transformation obviously invalidates the CFG, but can update forward dominator
946 (set, immediate dominators, tree, and frontier) information.
950 <!-------------------------------------------------------------------------- -->
951 <div class="doc_subsection">
952 <a name="codegenprepare">-codegenprepare: Prepare a function for code generation</a>
954 <div class="doc_text">
955 This pass munges the code in the input function to better prepare it for
956 SelectionDAG-based code generation. This works around limitations in it's
957 basic-block-at-a-time approach. It should eventually be removed.
960 <!-------------------------------------------------------------------------- -->
961 <div class="doc_subsection">
962 <a name="constmerge">-constmerge: Merge Duplicate Global Constants</a>
964 <div class="doc_text">
966 Merges duplicate global constants together into a single constant that is
967 shared. This is useful because some passes (ie TraceValues) insert a lot of
968 string constants into the program, regardless of whether or not an existing
973 <!-------------------------------------------------------------------------- -->
974 <div class="doc_subsection">
975 <a name="constprop">-constprop: Simple constant propagation</a>
977 <div class="doc_text">
978 <p>This file implements constant propagation and merging. It looks for
979 instructions involving only constant operands and replaces them with a
980 constant value instead of an instruction. For example:</p>
981 <blockquote><pre>add i32 1, 2</pre></blockquote>
983 <blockquote><pre>i32 3</pre></blockquote>
984 <p>NOTE: this pass has a habit of making definitions be dead. It is a good
985 idea to to run a <a href="#die">DIE</a> (Dead Instruction Elimination) pass
986 sometime after running this pass.</p>
989 <!-------------------------------------------------------------------------- -->
990 <div class="doc_subsection">
991 <a name="dce">-dce: Dead Code Elimination</a>
993 <div class="doc_text">
995 Dead code elimination is similar to <a href="#die">dead instruction
996 elimination</a>, but it rechecks instructions that were used by removed
997 instructions to see if they are newly dead.
1001 <!-------------------------------------------------------------------------- -->
1002 <div class="doc_subsection">
1003 <a name="deadargelim">-deadargelim: Dead Argument Elimination</a>
1005 <div class="doc_text">
1007 This pass deletes dead arguments from internal functions. Dead argument
1008 elimination removes arguments which are directly dead, as well as arguments
1009 only passed into function calls as dead arguments of other functions. This
1010 pass also deletes dead arguments in a similar way.
1014 This pass is often useful as a cleanup pass to run after aggressive
1015 interprocedural passes, which add possibly-dead arguments.
1019 <!-------------------------------------------------------------------------- -->
1020 <div class="doc_subsection">
1021 <a name="deadtypeelim">-deadtypeelim: Dead Type Elimination</a>
1023 <div class="doc_text">
1025 This pass is used to cleanup the output of GCC. It eliminate names for types
1026 that are unused in the entire translation unit, using the <a
1027 href="#findusedtypes">find used types</a> pass.
1031 <!-------------------------------------------------------------------------- -->
1032 <div class="doc_subsection">
1033 <a name="die">-die: Dead Instruction Elimination</a>
1035 <div class="doc_text">
1037 Dead instruction elimination performs a single pass over the function,
1038 removing instructions that are obviously dead.
1042 <!-------------------------------------------------------------------------- -->
1043 <div class="doc_subsection">
1044 <a name="dse">-dse: Dead Store Elimination</a>
1046 <div class="doc_text">
1048 A trivial dead store elimination that only considers basic-block local
1053 <!-------------------------------------------------------------------------- -->
1054 <div class="doc_subsection">
1055 <a name="functionattrs">-functionattrs: Deduce function attributes</a>
1057 <div class="doc_text">
1058 <p>A simple interprocedural pass which walks the call-graph, looking for
1059 functions which do not access or only read non-local memory, and marking them
1060 readnone/readonly. In addition, it marks function arguments (of pointer type)
1061 'nocapture' if a call to the function does not create any copies of the pointer
1062 value that outlive the call. This more or less means that the pointer is only
1063 dereferenced, and not returned from the function or stored in a global.
1064 This pass is implemented as a bottom-up traversal of the call-graph.
1068 <!-------------------------------------------------------------------------- -->
1069 <div class="doc_subsection">
1070 <a name="globaldce">-globaldce: Dead Global Elimination</a>
1072 <div class="doc_text">
1074 This transform is designed to eliminate unreachable internal globals from the
1075 program. It uses an aggressive algorithm, searching out globals that are
1076 known to be alive. After it finds all of the globals which are needed, it
1077 deletes whatever is left over. This allows it to delete recursive chunks of
1078 the program which are unreachable.
1082 <!-------------------------------------------------------------------------- -->
1083 <div class="doc_subsection">
1084 <a name="globalopt">-globalopt: Global Variable Optimizer</a>
1086 <div class="doc_text">
1088 This pass transforms simple global variables that never have their address
1089 taken. If obviously true, it marks read/write globals as constant, deletes
1090 variables only stored to, etc.
1094 <!-------------------------------------------------------------------------- -->
1095 <div class="doc_subsection">
1096 <a name="gvn">-gvn: Global Value Numbering</a>
1098 <div class="doc_text">
1100 This pass performs global value numbering to eliminate fully and partially
1101 redundant instructions. It also performs redundant load elimination.
1105 <!-------------------------------------------------------------------------- -->
1106 <div class="doc_subsection">
1107 <a name="indvars">-indvars: Canonicalize Induction Variables</a>
1109 <div class="doc_text">
1111 This transformation analyzes and transforms the induction variables (and
1112 computations derived from them) into simpler forms suitable for subsequent
1113 analysis and transformation.
1117 This transformation makes the following changes to each loop with an
1118 identifiable induction variable:
1122 <li>All loops are transformed to have a <em>single</em> canonical
1123 induction variable which starts at zero and steps by one.</li>
1124 <li>The canonical induction variable is guaranteed to be the first PHI node
1125 in the loop header block.</li>
1126 <li>Any pointer arithmetic recurrences are raised to use array
1131 If the trip count of a loop is computable, this pass also makes the following
1136 <li>The exit condition for the loop is canonicalized to compare the
1137 induction value against the exit value. This turns loops like:
1138 <blockquote><pre>for (i = 7; i*i < 1000; ++i)</pre></blockquote>
1140 <blockquote><pre>for (i = 0; i != 25; ++i)</pre></blockquote></li>
1141 <li>Any use outside of the loop of an expression derived from the indvar
1142 is changed to compute the derived value outside of the loop, eliminating
1143 the dependence on the exit value of the induction variable. If the only
1144 purpose of the loop is to compute the exit value of some derived
1145 expression, this transformation will make the loop dead.</li>
1149 This transformation should be followed by strength reduction after all of the
1150 desired loop transformations have been performed. Additionally, on targets
1151 where it is profitable, the loop could be transformed to count down to zero
1152 (the "do loop" optimization).
1156 <!-------------------------------------------------------------------------- -->
1157 <div class="doc_subsection">
1158 <a name="inline">-inline: Function Integration/Inlining</a>
1160 <div class="doc_text">
1162 Bottom-up inlining of functions into callees.
1166 <!-------------------------------------------------------------------------- -->
1167 <div class="doc_subsection">
1168 <a name="insert-edge-profiling">-insert-edge-profiling: Insert instrumentation for edge profiling</a>
1170 <div class="doc_text">
1172 This pass instruments the specified program with counters for edge profiling.
1173 Edge profiling can give a reasonable approximation of the hot paths through a
1174 program, and is used for a wide variety of program transformations.
1178 Note that this implementation is very naïve. It inserts a counter for
1179 <em>every</em> edge in the program, instead of using control flow information
1180 to prune the number of counters inserted.
1184 <!-------------------------------------------------------------------------- -->
1185 <div class="doc_subsection">
1186 <a name="insert-optimal-edge-profiling">-insert-optimal-edge-profiling: Insert optimal instrumentation for edge profiling</a>
1188 <div class="doc_text">
1189 <p>This pass instruments the specified program with counters for edge profiling.
1190 Edge profiling can give a reasonable approximation of the hot paths through a
1191 program, and is used for a wide variety of program transformations.
1195 <!-------------------------------------------------------------------------- -->
1196 <div class="doc_subsection">
1197 <a name="instcombine">-instcombine: Combine redundant instructions</a>
1199 <div class="doc_text">
1201 Combine instructions to form fewer, simple
1202 instructions. This pass does not modify the CFG This pass is where algebraic
1203 simplification happens.
1207 This pass combines things like:
1212 %Z = add i32 %Y, 1</pre></blockquote>
1219 >%Z = add i32 %X, 2</pre></blockquote>
1222 This is a simple worklist driven algorithm.
1226 This pass guarantees that the following canonicalizations are performed on
1231 <li>If a binary operator has a constant operand, it is moved to the right-
1233 <li>Bitwise operators with constant operands are always grouped so that
1234 shifts are performed first, then <code>or</code>s, then
1235 <code>and</code>s, then <code>xor</code>s.</li>
1236 <li>Compare instructions are converted from <code><</code>,
1237 <code>></code>, <code>≤</code>, or <code>≥</code> to
1238 <code>=</code> or <code>≠</code> if possible.</li>
1239 <li>All <code>cmp</code> instructions on boolean values are replaced with
1240 logical operations.</li>
1241 <li><code>add <var>X</var>, <var>X</var></code> is represented as
1242 <code>mul <var>X</var>, 2</code> ⇒ <code>shl <var>X</var>, 1</code></li>
1243 <li>Multiplies with a constant power-of-two argument are transformed into
1249 <!-------------------------------------------------------------------------- -->
1250 <div class="doc_subsection">
1251 <a name="internalize">-internalize: Internalize Global Symbols</a>
1253 <div class="doc_text">
1255 This pass loops over all of the functions in the input module, looking for a
1256 main function. If a main function is found, all other functions and all
1257 global variables with initializers are marked as internal.
1261 <!-------------------------------------------------------------------------- -->
1262 <div class="doc_subsection">
1263 <a name="ipconstprop">-ipconstprop: Interprocedural constant propagation</a>
1265 <div class="doc_text">
1267 This pass implements an <em>extremely</em> simple interprocedural constant
1268 propagation pass. It could certainly be improved in many different ways,
1269 like using a worklist. This pass makes arguments dead, but does not remove
1270 them. The existing dead argument elimination pass should be run after this
1271 to clean up the mess.
1275 <!-------------------------------------------------------------------------- -->
1276 <div class="doc_subsection">
1277 <a name="ipsccp">-ipsccp: Interprocedural Sparse Conditional Constant Propagation</a>
1279 <div class="doc_text">
1281 An interprocedural variant of <a href="#sccp">Sparse Conditional Constant
1286 <!-------------------------------------------------------------------------- -->
1287 <div class="doc_subsection">
1288 <a name="jump-threading">-jump-threading: Thread control through conditional blocks</a>
1290 <div class="doc_text">
1292 Jump threading tries to find distinct threads of control flow running through
1293 a basic block. This pass looks at blocks that have multiple predecessors and
1294 multiple successors. If one or more of the predecessors of the block can be
1295 proven to always cause a jump to one of the successors, we forward the edge
1296 from the predecessor to the successor by duplicating the contents of this
1300 An example of when this can occur is code like this:
1307 if (X < 3) {</pre>
1310 In this case, the unconditional branch at the end of the first if can be
1311 revectored to the false side of the second if.
1315 <!-------------------------------------------------------------------------- -->
1316 <div class="doc_subsection">
1317 <a name="lcssa">-lcssa: Loop-Closed SSA Form Pass</a>
1319 <div class="doc_text">
1321 This pass transforms loops by placing phi nodes at the end of the loops for
1322 all values that are live across the loop boundary. For example, it turns
1323 the left into the right code:
1327 >for (...) for (...)
1332 X3 = phi(X1, X2) X3 = phi(X1, X2)
1333 ... = X3 + 4 X4 = phi(X3)
1337 This is still valid LLVM; the extra phi nodes are purely redundant, and will
1338 be trivially eliminated by <code>InstCombine</code>. The major benefit of
1339 this transformation is that it makes many other loop optimizations, such as
1340 LoopUnswitching, simpler.
1344 <!-------------------------------------------------------------------------- -->
1345 <div class="doc_subsection">
1346 <a name="licm">-licm: Loop Invariant Code Motion</a>
1348 <div class="doc_text">
1350 This pass performs loop invariant code motion, attempting to remove as much
1351 code from the body of a loop as possible. It does this by either hoisting
1352 code into the preheader block, or by sinking code to the exit blocks if it is
1353 safe. This pass also promotes must-aliased memory locations in the loop to
1354 live in registers, thus hoisting and sinking "invariant" loads and stores.
1358 This pass uses alias analysis for two purposes:
1362 <li>Moving loop invariant loads and calls out of loops. If we can determine
1363 that a load or call inside of a loop never aliases anything stored to,
1364 we can hoist it or sink it like any other instruction.</li>
1365 <li>Scalar Promotion of Memory - If there is a store instruction inside of
1366 the loop, we try to move the store to happen AFTER the loop instead of
1367 inside of the loop. This can only happen if a few conditions are true:
1369 <li>The pointer stored through is loop invariant.</li>
1370 <li>There are no stores or loads in the loop which <em>may</em> alias
1371 the pointer. There are no calls in the loop which mod/ref the
1374 If these conditions are true, we can promote the loads and stores in the
1375 loop of the pointer to use a temporary alloca'd variable. We then use
1376 the mem2reg functionality to construct the appropriate SSA form for the
1380 <!-------------------------------------------------------------------------- -->
1381 <div class="doc_subsection">
1382 <a name="loop-deletion">-loop-deletion: Dead Loop Deletion Pass</a>
1384 <div class="doc_text">
1386 This file implements the Dead Loop Deletion Pass. This pass is responsible
1387 for eliminating loops with non-infinite computable trip counts that have no
1388 side effects or volatile instructions, and do not contribute to the
1389 computation of the function's return value.
1393 <!-------------------------------------------------------------------------- -->
1394 <div class="doc_subsection">
1395 <a name="loop-extract">-loop-extract: Extract loops into new functions</a>
1397 <div class="doc_text">
1399 A pass wrapper around the <code>ExtractLoop()</code> scalar transformation to
1400 extract each top-level loop into its own new function. If the loop is the
1401 <em>only</em> loop in a given function, it is not touched. This is a pass most
1402 useful for debugging via bugpoint.
1406 <!-------------------------------------------------------------------------- -->
1407 <div class="doc_subsection">
1408 <a name="loop-extract-single">-loop-extract-single: Extract at most one loop into a new function</a>
1410 <div class="doc_text">
1412 Similar to <a href="#loop-extract">Extract loops into new functions</a>,
1413 this pass extracts one natural loop from the program into a function if it
1414 can. This is used by bugpoint.
1418 <!-------------------------------------------------------------------------- -->
1419 <div class="doc_subsection">
1420 <a name="loop-index-split">-loop-index-split: Index Split Loops</a>
1422 <div class="doc_text">
1424 This pass divides loop's iteration range by spliting loop such that each
1425 individual loop is executed efficiently.
1429 <!-------------------------------------------------------------------------- -->
1430 <div class="doc_subsection">
1431 <a name="loop-reduce">-loop-reduce: Loop Strength Reduction</a>
1433 <div class="doc_text">
1435 This pass performs a strength reduction on array references inside loops that
1436 have as one or more of their components the loop induction variable. This is
1437 accomplished by creating a new value to hold the initial value of the array
1438 access for the first iteration, and then creating a new GEP instruction in
1439 the loop to increment the value by the appropriate amount.
1443 <!-------------------------------------------------------------------------- -->
1444 <div class="doc_subsection">
1445 <a name="loop-rotate">-loop-rotate: Rotate Loops</a>
1447 <div class="doc_text">
1448 <p>A simple loop rotation transformation.</p>
1451 <!-------------------------------------------------------------------------- -->
1452 <div class="doc_subsection">
1453 <a name="loop-unroll">-loop-unroll: Unroll loops</a>
1455 <div class="doc_text">
1457 This pass implements a simple loop unroller. It works best when loops have
1458 been canonicalized by the <a href="#indvars"><tt>-indvars</tt></a> pass,
1459 allowing it to determine the trip counts of loops easily.
1463 <!-------------------------------------------------------------------------- -->
1464 <div class="doc_subsection">
1465 <a name="loop-unswitch">-loop-unswitch: Unswitch loops</a>
1467 <div class="doc_text">
1469 This pass transforms loops that contain branches on loop-invariant conditions
1470 to have multiple loops. For example, it turns the left into the right code:
1482 This can increase the size of the code exponentially (doubling it every time
1483 a loop is unswitched) so we only unswitch if the resultant code will be
1484 smaller than a threshold.
1488 This pass expects LICM to be run before it to hoist invariant conditions out
1489 of the loop, to make the unswitching opportunity obvious.
1493 <!-------------------------------------------------------------------------- -->
1494 <div class="doc_subsection">
1495 <a name="loopsimplify">-loopsimplify: Canonicalize natural loops</a>
1497 <div class="doc_text">
1499 This pass performs several transformations to transform natural loops into a
1500 simpler form, which makes subsequent analyses and transformations simpler and
1505 Loop pre-header insertion guarantees that there is a single, non-critical
1506 entry edge from outside of the loop to the loop header. This simplifies a
1507 number of analyses and transformations, such as LICM.
1511 Loop exit-block insertion guarantees that all exit blocks from the loop
1512 (blocks which are outside of the loop that have predecessors inside of the
1513 loop) only have predecessors from inside of the loop (and are thus dominated
1514 by the loop header). This simplifies transformations such as store-sinking
1515 that are built into LICM.
1519 This pass also guarantees that loops will have exactly one backedge.
1523 Note that the simplifycfg pass will clean up blocks which are split out but
1524 end up being unnecessary, so usage of this pass should not pessimize
1529 This pass obviously modifies the CFG, but updates loop information and
1530 dominator information.
1534 <!-------------------------------------------------------------------------- -->
1535 <div class="doc_subsection">
1536 <a name="lowerallocs">-lowerallocs: Lower allocations from instructions to calls</a>
1538 <div class="doc_text">
1540 Turn <tt>malloc</tt> and <tt>free</tt> instructions into <tt>@malloc</tt> and
1541 <tt>@free</tt> calls.
1545 This is a target-dependent tranformation because it depends on the size of
1546 data types and alignment constraints.
1550 <!-------------------------------------------------------------------------- -->
1551 <div class="doc_subsection">
1552 <a name="lowerinvoke">-lowerinvoke: Lower invoke and unwind, for unwindless code generators</a>
1554 <div class="doc_text">
1556 This transformation is designed for use by code generators which do not yet
1557 support stack unwinding. This pass supports two models of exception handling
1558 lowering, the 'cheap' support and the 'expensive' support.
1562 'Cheap' exception handling support gives the program the ability to execute
1563 any program which does not "throw an exception", by turning 'invoke'
1564 instructions into calls and by turning 'unwind' instructions into calls to
1565 abort(). If the program does dynamically use the unwind instruction, the
1566 program will print a message then abort.
1570 'Expensive' exception handling support gives the full exception handling
1571 support to the program at the cost of making the 'invoke' instruction
1572 really expensive. It basically inserts setjmp/longjmp calls to emulate the
1573 exception handling as necessary.
1577 Because the 'expensive' support slows down programs a lot, and EH is only
1578 used for a subset of the programs, it must be specifically enabled by the
1579 <tt>-enable-correct-eh-support</tt> option.
1583 Note that after this pass runs the CFG is not entirely accurate (exceptional
1584 control flow edges are not correct anymore) so only very simple things should
1585 be done after the lowerinvoke pass has run (like generation of native code).
1586 This should not be used as a general purpose "my LLVM-to-LLVM pass doesn't
1587 support the invoke instruction yet" lowering pass.
1591 <!-------------------------------------------------------------------------- -->
1592 <div class="doc_subsection">
1593 <a name="lowersetjmp">-lowersetjmp: Lower Set Jump</a>
1595 <div class="doc_text">
1597 Lowers <tt>setjmp</tt> and <tt>longjmp</tt> to use the LLVM invoke and unwind
1598 instructions as necessary.
1602 Lowering of <tt>longjmp</tt> is fairly trivial. We replace the call with a
1603 call to the LLVM library function <tt>__llvm_sjljeh_throw_longjmp()</tt>.
1604 This unwinds the stack for us calling all of the destructors for
1605 objects allocated on the stack.
1609 At a <tt>setjmp</tt> call, the basic block is split and the <tt>setjmp</tt>
1610 removed. The calls in a function that have a <tt>setjmp</tt> are converted to
1611 invoke where the except part checks to see if it's a <tt>longjmp</tt>
1612 exception and, if so, if it's handled in the function. If it is, then it gets
1613 the value returned by the <tt>longjmp</tt> and goes to where the basic block
1614 was split. <tt>invoke</tt> instructions are handled in a similar fashion with
1615 the original except block being executed if it isn't a <tt>longjmp</tt>
1616 except that is handled by that function.
1620 <!-------------------------------------------------------------------------- -->
1621 <div class="doc_subsection">
1622 <a name="lowerswitch">-lowerswitch: Lower SwitchInst's to branches</a>
1624 <div class="doc_text">
1626 Rewrites <tt>switch</tt> instructions with a sequence of branches, which
1627 allows targets to get away with not implementing the switch instruction until
1632 <!-------------------------------------------------------------------------- -->
1633 <div class="doc_subsection">
1634 <a name="mem2reg">-mem2reg: Promote Memory to Register</a>
1636 <div class="doc_text">
1638 This file promotes memory references to be register references. It promotes
1639 <tt>alloca</tt> instructions which only have <tt>load</tt>s and
1640 <tt>store</tt>s as uses. An <tt>alloca</tt> is transformed by using dominator
1641 frontiers to place <tt>phi</tt> nodes, then traversing the function in
1642 depth-first order to rewrite <tt>load</tt>s and <tt>store</tt>s as
1643 appropriate. This is just the standard SSA construction algorithm to construct
1648 <!-------------------------------------------------------------------------- -->
1649 <div class="doc_subsection">
1650 <a name="memcpyopt">-memcpyopt: Optimize use of memcpy and friend</a>
1652 <div class="doc_text">
1654 This pass performs various transformations related to eliminating memcpy
1655 calls, or transforming sets of stores into memset's.
1659 <!-------------------------------------------------------------------------- -->
1660 <div class="doc_subsection">
1661 <a name="mergefunc">-mergefunc: Merge Functions</a>
1663 <div class="doc_text">
1664 <p>This pass looks for equivalent functions that are mergable and folds them.
1666 A hash is computed from the function, based on its type and number of
1669 Once all hashes are computed, we perform an expensive equality comparison
1670 on each function pair. This takes n^2/2 comparisons per bucket, so it's
1671 important that the hash function be high quality. The equality comparison
1672 iterates through each instruction in each basic block.
1674 When a match is found the functions are folded. If both functions are
1675 overridable, we move the functionality into a new internal function and
1676 leave two overridable thunks to it.
1680 <!-------------------------------------------------------------------------- -->
1681 <div class="doc_subsection">
1682 <a name="mergereturn">-mergereturn: Unify function exit nodes</a>
1684 <div class="doc_text">
1686 Ensure that functions have at most one <tt>ret</tt> instruction in them.
1687 Additionally, it keeps track of which node is the new exit node of the CFG.
1691 <!-------------------------------------------------------------------------- -->
1692 <div class="doc_subsection">
1693 <a name="partial-inliner">-partial-inliner: Partial Inliner</a>
1695 <div class="doc_text">
1696 <p>This pass performs partial inlining, typically by inlining an if
1697 statement that surrounds the body of the function.
1701 <!-------------------------------------------------------------------------- -->
1702 <div class="doc_subsection">
1703 <a name="partialspecialization">-partialspecialization: Partial Specialization</a>
1705 <div class="doc_text">
1706 <p>This pass finds function arguments that are often a common constant and
1707 specializes a version of the called function for that constant.
1709 This pass simply does the cloning for functions it specializes. It depends
1710 on <a href="#ipsccp">IPSCCP</a> and <a href="#deadargelim">DAE</a> to clean up the results.
1712 The initial heuristic favors constant arguments that are used in control
1717 <!-------------------------------------------------------------------------- -->
1718 <div class="doc_subsection">
1719 <a name="prune-eh">-prune-eh: Remove unused exception handling info</a>
1721 <div class="doc_text">
1723 This file implements a simple interprocedural pass which walks the call-graph,
1724 turning <tt>invoke</tt> instructions into <tt>call</tt> instructions if and
1725 only if the callee cannot throw an exception. It implements this as a
1726 bottom-up traversal of the call-graph.
1730 <!-------------------------------------------------------------------------- -->
1731 <div class="doc_subsection">
1732 <a name="reassociate">-reassociate: Reassociate expressions</a>
1734 <div class="doc_text">
1736 This pass reassociates commutative expressions in an order that is designed
1737 to promote better constant propagation, GCSE, LICM, PRE, etc.
1741 For example: 4 + (<var>x</var> + 5) ⇒ <var>x</var> + (4 + 5)
1745 In the implementation of this algorithm, constants are assigned rank = 0,
1746 function arguments are rank = 1, and other values are assigned ranks
1747 corresponding to the reverse post order traversal of current function
1748 (starting at 2), which effectively gives values in deep loops higher rank
1749 than values not in loops.
1753 <!-------------------------------------------------------------------------- -->
1754 <div class="doc_subsection">
1755 <a name="reg2mem">-reg2mem: Demote all values to stack slots</a>
1757 <div class="doc_text">
1759 This file demotes all registers to memory references. It is intented to be
1760 the inverse of <a href="#mem2reg"><tt>-mem2reg</tt></a>. By converting to
1761 <tt>load</tt> instructions, the only values live across basic blocks are
1762 <tt>alloca</tt> instructions and <tt>load</tt> instructions before
1763 <tt>phi</tt> nodes. It is intended that this should make CFG hacking much
1764 easier. To make later hacking easier, the entry block is split into two, such
1765 that all introduced <tt>alloca</tt> instructions (and nothing else) are in the
1770 <!-------------------------------------------------------------------------- -->
1771 <div class="doc_subsection">
1772 <a name="scalarrepl">-scalarrepl: Scalar Replacement of Aggregates</a>
1774 <div class="doc_text">
1776 The well-known scalar replacement of aggregates transformation. This
1777 transform breaks up <tt>alloca</tt> instructions of aggregate type (structure
1778 or array) into individual <tt>alloca</tt> instructions for each member if
1779 possible. Then, if possible, it transforms the individual <tt>alloca</tt>
1780 instructions into nice clean scalar SSA form.
1784 This combines a simple scalar replacement of aggregates algorithm with the <a
1785 href="#mem2reg"><tt>mem2reg</tt></a> algorithm because often interact,
1786 especially for C++ programs. As such, iterating between <tt>scalarrepl</tt>,
1787 then <a href="#mem2reg"><tt>mem2reg</tt></a> until we run out of things to
1792 <!-------------------------------------------------------------------------- -->
1793 <div class="doc_subsection">
1794 <a name="sccp">-sccp: Sparse Conditional Constant Propagation</a>
1796 <div class="doc_text">
1798 Sparse conditional constant propagation and merging, which can be summarized
1803 <li>Assumes values are constant unless proven otherwise</li>
1804 <li>Assumes BasicBlocks are dead unless proven otherwise</li>
1805 <li>Proves values to be constant, and replaces them with constants</li>
1806 <li>Proves conditional branches to be unconditional</li>
1810 Note that this pass has a habit of making definitions be dead. It is a good
1811 idea to to run a DCE pass sometime after running this pass.
1815 <!-------------------------------------------------------------------------- -->
1816 <div class="doc_subsection">
1817 <a name="sink">-sink: Code Sinking</a>
1819 <div class="doc_text">
1820 <p>This pass moves instructions into successor blocks, when possible, so that
1821 they aren't executed on paths where their results aren't needed.
1825 <!-------------------------------------------------------------------------- -->
1826 <div class="doc_subsection">
1827 <a name="simplify-libcalls">-simplify-libcalls: Simplify well-known library calls</a>
1829 <div class="doc_text">
1831 Applies a variety of small optimizations for calls to specific well-known
1832 function calls (e.g. runtime library functions). For example, a call
1833 <tt>exit(3)</tt> that occurs within the <tt>main()</tt> function can be
1834 transformed into simply <tt>return 3</tt>.
1838 <!-------------------------------------------------------------------------- -->
1839 <div class="doc_subsection">
1840 <a name="simplify-libcalls-halfpowr">-simplify-libcalls-halfpowr: Simplify half_powr library calls</a>
1842 <div class="doc_text">
1843 <p>Simple pass that applies an experimental transformation on calls
1844 to specific functions.
1848 <!-------------------------------------------------------------------------- -->
1849 <div class="doc_subsection">
1850 <a name="simplifycfg">-simplifycfg: Simplify the CFG</a>
1852 <div class="doc_text">
1854 Performs dead code elimination and basic block merging. Specifically:
1858 <li>Removes basic blocks with no predecessors.</li>
1859 <li>Merges a basic block into its predecessor if there is only one and the
1860 predecessor only has one successor.</li>
1861 <li>Eliminates PHI nodes for basic blocks with a single predecessor.</li>
1862 <li>Eliminates a basic block that only contains an unconditional
1867 <!-------------------------------------------------------------------------- -->
1868 <div class="doc_subsection">
1869 <a name="split-geps">-split-geps: Split complex GEPs into simple GEPs</a>
1871 <div class="doc_text">
1872 <p>This function breaks GEPs with more than 2 non-zero operands into smaller
1873 GEPs each with no more than 2 non-zero operands. This exposes redundancy
1874 between GEPs with common initial operand sequences.
1878 <!-------------------------------------------------------------------------- -->
1879 <div class="doc_subsection">
1880 <a name="ssi">-ssi: Static Single Information Construction</a>
1882 <div class="doc_text">
1883 <p>This pass converts a list of variables to the Static Single Information
1886 We are building an on-demand representation, that is, we do not convert
1887 every single variable in the target function to SSI form. Rather, we receive
1888 a list of target variables that must be converted. We also do not
1889 completely convert a target variable to the SSI format. Instead, we only
1890 change the variable in the points where new information can be attached
1891 to its live range, that is, at branch points.
1895 <!-------------------------------------------------------------------------- -->
1896 <div class="doc_subsection">
1897 <a name="ssi-everything">-ssi-everything: Static Single Information Construction (everything, intended for debugging)</a>
1899 <div class="doc_text">
1900 <p>A pass that runs <a href="#ssi">SSI</a> on every non-void variable, intended for debugging.
1904 <!-------------------------------------------------------------------------- -->
1905 <div class="doc_subsection">
1906 <a name="strip">-strip: Strip all symbols from a module</a>
1908 <div class="doc_text">
1910 performs code stripping. this transformation can delete:
1914 <li>names for virtual registers</li>
1915 <li>symbols for internal globals and functions</li>
1916 <li>debug information</li>
1920 note that this transformation makes code much less readable, so it should
1921 only be used in situations where the <tt>strip</tt> utility would be used,
1922 such as reducing code size or making it harder to reverse engineer code.
1926 <!-------------------------------------------------------------------------- -->
1927 <div class="doc_subsection">
1928 <a name="strip-dead-prototypes">-strip-dead-prototypes: Remove unused function declarations</a>
1930 <div class="doc_text">
1932 This pass loops over all of the functions in the input module, looking for
1933 dead declarations and removes them. Dead declarations are declarations of
1934 functions for which no implementation is available (i.e., declarations for
1935 unused library functions).
1939 <!-------------------------------------------------------------------------- -->
1940 <div class="doc_subsection">
1941 <a name="strip-debug-declare">-strip-debug-declare: Strip all llvm.dbg.declare intrinsics</a>
1943 <div class="doc_text">
1944 <p>This pass implements code stripping. Specifically, it can delete:
1946 <li>names for virtual registers</li>
1947 <li>symbols for internal globals and functions</li>
1948 <li>debug information</li>
1950 Note that this transformation makes code much less readable, so it should
1951 only be used in situations where the 'strip' utility would be used, such as
1952 reducing code size or making it harder to reverse engineer code.
1956 <!-------------------------------------------------------------------------- -->
1957 <div class="doc_subsection">
1958 <a name="strip-nondebug">-strip-nondebug: Strip all symbols, except dbg symbols, from a module</a>
1960 <div class="doc_text">
1961 <p>This pass implements code stripping. Specifically, it can delete:
1963 <li>names for virtual registers</li>
1964 <li>symbols for internal globals and functions</li>
1965 <li>debug information</li>
1967 Note that this transformation makes code much less readable, so it should
1968 only be used in situations where the 'strip' utility would be used, such as
1969 reducing code size or making it harder to reverse engineer code.
1973 <!-------------------------------------------------------------------------- -->
1974 <div class="doc_subsection">
1975 <a name="sretpromotion">-sretpromotion: Promote sret arguments</a>
1977 <div class="doc_text">
1979 This pass finds functions that return a struct (using a pointer to the struct
1980 as the first argument of the function, marked with the '<tt>sret</tt>' attribute) and
1981 replaces them with a new function that simply returns each of the elements of
1982 that struct (using multiple return values).
1986 This pass works under a number of conditions:
1990 <li>The returned struct must not contain other structs</li>
1991 <li>The returned struct must only be used to load values from</li>
1992 <li>The placeholder struct passed in is the result of an <tt>alloca</tt></li>
1996 <!-------------------------------------------------------------------------- -->
1997 <div class="doc_subsection">
1998 <a name="tailcallelim">-tailcallelim: Tail Call Elimination</a>
2000 <div class="doc_text">
2002 This file transforms calls of the current function (self recursion) followed
2003 by a return instruction with a branch to the entry of the function, creating
2004 a loop. This pass also implements the following extensions to the basic
2009 <li>Trivial instructions between the call and return do not prevent the
2010 transformation from taking place, though currently the analysis cannot
2011 support moving any really useful instructions (only dead ones).
2012 <li>This pass transforms functions that are prevented from being tail
2013 recursive by an associative expression to use an accumulator variable,
2014 thus compiling the typical naive factorial or <tt>fib</tt> implementation
2015 into efficient code.
2016 <li>TRE is performed if the function returns void, if the return
2017 returns the result returned by the call, or if the function returns a
2018 run-time constant on all exits from the function. It is possible, though
2019 unlikely, that the return returns something else (like constant 0), and
2020 can still be TRE'd. It can be TRE'd if <em>all other</em> return
2021 instructions in the function return the exact same value.
2022 <li>If it can prove that callees do not access theier caller stack frame,
2023 they are marked as eligible for tail call elimination (by the code
2028 <!-------------------------------------------------------------------------- -->
2029 <div class="doc_subsection">
2030 <a name="tailduplicate">-tailduplicate: Tail Duplication</a>
2032 <div class="doc_text">
2034 This pass performs a limited form of tail duplication, intended to simplify
2035 CFGs by removing some unconditional branches. This pass is necessary to
2036 straighten out loops created by the C front-end, but also is capable of
2037 making other code nicer. After this pass is run, the CFG simplify pass
2038 should be run to clean up the mess.
2042 <!-- ======================================================================= -->
2043 <div class="doc_section"> <a name="transform">Utility Passes</a></div>
2044 <div class="doc_text">
2045 <p>This section describes the LLVM Utility Passes.</p>
2048 <!-------------------------------------------------------------------------- -->
2049 <div class="doc_subsection">
2050 <a name="deadarghaX0r">-deadarghaX0r: Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</a>
2052 <div class="doc_text">
2054 Same as dead argument elimination, but deletes arguments to functions which
2055 are external. This is only for use by <a
2056 href="Bugpoint.html">bugpoint</a>.</p>
2059 <!-------------------------------------------------------------------------- -->
2060 <div class="doc_subsection">
2061 <a name="extract-blocks">-extract-blocks: Extract Basic Blocks From Module (for bugpoint use)</a>
2063 <div class="doc_text">
2065 This pass is used by bugpoint to extract all blocks from the module into their
2069 <!-------------------------------------------------------------------------- -->
2070 <div class="doc_subsection">
2071 <a name="instnamer">-instnamer: Assign names to anonymous instructions</a>
2073 <div class="doc_text">
2074 <p>This is a little utility pass that gives instructions names, this is mostly
2075 useful when diffing the effect of an optimization because deleting an
2076 unnamed instruction can change all other instruction numbering, making the
2081 <!-------------------------------------------------------------------------- -->
2082 <div class="doc_subsection">
2083 <a name="preverify">-preverify: Preliminary module verification</a>
2085 <div class="doc_text">
2087 Ensures that the module is in the form required by the <a
2088 href="#verifier">Module Verifier</a> pass.
2092 Running the verifier runs this pass automatically, so there should be no need
2097 <!-------------------------------------------------------------------------- -->
2098 <div class="doc_subsection">
2099 <a name="verify">-verify: Module Verifier</a>
2101 <div class="doc_text">
2103 Verifies an LLVM IR code. This is useful to run after an optimization which is
2104 undergoing testing. Note that <tt>llvm-as</tt> verifies its input before
2105 emitting bitcode, and also that malformed bitcode is likely to make LLVM
2106 crash. All language front-ends are therefore encouraged to verify their output
2107 before performing optimizing transformations.
2111 <li>Both of a binary operator's parameters are of the same type.</li>
2112 <li>Verify that the indices of mem access instructions match other
2114 <li>Verify that arithmetic and other things are only performed on
2115 first-class types. Verify that shifts and logicals only happen on
2117 <li>All of the constants in a switch statement are of the correct type.</li>
2118 <li>The code is in valid SSA form.</li>
2119 <li>It is illegal to put a label into any other type (like a structure) or
2121 <li>Only phi nodes can be self referential: <tt>%x = add i32 %x, %x</tt> is
2123 <li>PHI nodes must have an entry for each predecessor, with no extras.</li>
2124 <li>PHI nodes must be the first thing in a basic block, all grouped
2126 <li>PHI nodes must have at least one entry.</li>
2127 <li>All basic blocks should only end with terminator insts, not contain
2129 <li>The entry node to a function must not have predecessors.</li>
2130 <li>All Instructions must be embedded into a basic block.</li>
2131 <li>Functions cannot take a void-typed parameter.</li>
2132 <li>Verify that a function's argument list agrees with its declared
2134 <li>It is illegal to specify a name for a void value.</li>
2135 <li>It is illegal to have a internal global value with no initializer.</li>
2136 <li>It is illegal to have a ret instruction that returns a value that does
2137 not agree with the function return value type.</li>
2138 <li>Function call argument types match the function prototype.</li>
2139 <li>All other things that are tested by asserts spread about the code.</li>
2143 Note that this does not provide full security verification (like Java), but
2144 instead just tries to ensure that code is well-formed.
2148 <!-------------------------------------------------------------------------- -->
2149 <div class="doc_subsection">
2150 <a name="view-cfg">-view-cfg: View CFG of function</a>
2152 <div class="doc_text">
2154 Displays the control flow graph using the GraphViz tool.
2158 <!-------------------------------------------------------------------------- -->
2159 <div class="doc_subsection">
2160 <a name="view-cfg-only">-view-cfg-only: View CFG of function (with no function bodies)</a>
2162 <div class="doc_text">
2164 Displays the control flow graph using the GraphViz tool, but omitting function
2169 <!-------------------------------------------------------------------------- -->
2170 <div class="doc_subsection">
2171 <a name="view-dom">-view-dom: View dominator tree of function</a>
2173 <div class="doc_text">
2175 Displays the dominator tree using the GraphViz tool.
2179 <!-------------------------------------------------------------------------- -->
2180 <div class="doc_subsection">
2181 <a name="view-dom-only">-view-dom-only: View dominator tree of function (with no function
2185 <div class="doc_text">
2187 Displays the dominator tree using the GraphViz tool, but omitting function
2192 <!-------------------------------------------------------------------------- -->
2193 <div class="doc_subsection">
2194 <a name="view-postdom">-view-postdom: View post dominator tree of function</a>
2196 <div class="doc_text">
2198 Displays the post dominator tree using the GraphViz tool.
2202 <!-------------------------------------------------------------------------- -->
2203 <div class="doc_subsection">
2204 <a name="view-postdom-only">-view-postdom-only: View post dominator tree of function (with no
2208 <div class="doc_text">
2210 Displays the post dominator tree using the GraphViz tool, but omitting
2215 <!-- *********************************************************************** -->
2219 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2220 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
2221 <a href="http://validator.w3.org/check/referer"><img
2222 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
2224 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br>
2225 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
2226 Last modified: $Date$