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="#profile-verifier">-profile-verifier</a></td><td>Verify profiling information</td></tr>
124 <tr><td><a href="#scalar-evolution">-scalar-evolution</a></td><td>Scalar Evolution Analysis</td></tr>
125 <tr><td><a href="#scev-aa">-scev-aa</a></td><td>ScalarEvolution-based Alias Analysis</td></tr>
126 <tr><td><a href="#targetdata">-targetdata</a></td><td>Target Data Layout</td></tr>
129 <tr><th colspan="2"><b>TRANSFORM PASSES</b></th></tr>
130 <tr><th>Option</th><th>Name</th></tr>
131 <tr><td><a href="#abcd">-abcd</a></td><td>Remove redundant conditional branches</td></tr>
132 <tr><td><a href="#adce">-adce</a></td><td>Aggressive Dead Code Elimination</td></tr>
133 <tr><td><a href="#always-inline">-always-inline</a></td><td>Inliner for always_inline functions</td></tr>
134 <tr><td><a href="#argpromotion">-argpromotion</a></td><td>Promote 'by reference' arguments to scalars</td></tr>
135 <tr><td><a href="#block-placement">-block-placement</a></td><td>Profile Guided Basic Block Placement</td></tr>
136 <tr><td><a href="#break-crit-edges">-break-crit-edges</a></td><td>Break critical edges in CFG</td></tr>
137 <tr><td><a href="#codegenprepare">-codegenprepare</a></td><td>Prepare a function for code generation </td></tr>
138 <tr><td><a href="#constmerge">-constmerge</a></td><td>Merge Duplicate Global Constants</td></tr>
139 <tr><td><a href="#constprop">-constprop</a></td><td>Simple constant propagation</td></tr>
140 <tr><td><a href="#dce">-dce</a></td><td>Dead Code Elimination</td></tr>
141 <tr><td><a href="#deadargelim">-deadargelim</a></td><td>Dead Argument Elimination</td></tr>
142 <tr><td><a href="#deadtypeelim">-deadtypeelim</a></td><td>Dead Type Elimination</td></tr>
143 <tr><td><a href="#die">-die</a></td><td>Dead Instruction Elimination</td></tr>
144 <tr><td><a href="#dse">-dse</a></td><td>Dead Store Elimination</td></tr>
145 <tr><td><a href="#functionattrs">-functionattrs</a></td><td>Deduce function attributes</td></tr>
146 <tr><td><a href="#globaldce">-globaldce</a></td><td>Dead Global Elimination</td></tr>
147 <tr><td><a href="#globalopt">-globalopt</a></td><td>Global Variable Optimizer</td></tr>
148 <tr><td><a href="#gvn">-gvn</a></td><td>Global Value Numbering</td></tr>
149 <tr><td><a href="#indvars">-indvars</a></td><td>Canonicalize Induction Variables</td></tr>
150 <tr><td><a href="#inline">-inline</a></td><td>Function Integration/Inlining</td></tr>
151 <tr><td><a href="#insert-edge-profiling">-insert-edge-profiling</a></td><td>Insert instrumentation for edge profiling</td></tr>
152 <tr><td><a href="#insert-optimal-edge-profiling">-insert-optimal-edge-profiling</a></td><td>Insert optimal instrumentation for edge profiling</td></tr>
153 <tr><td><a href="#instcombine">-instcombine</a></td><td>Combine redundant instructions</td></tr>
154 <tr><td><a href="#internalize">-internalize</a></td><td>Internalize Global Symbols</td></tr>
155 <tr><td><a href="#ipconstprop">-ipconstprop</a></td><td>Interprocedural constant propagation</td></tr>
156 <tr><td><a href="#ipsccp">-ipsccp</a></td><td>Interprocedural Sparse Conditional Constant Propagation</td></tr>
157 <tr><td><a href="#jump-threading">-jump-threading</a></td><td>Thread control through conditional blocks </td></tr>
158 <tr><td><a href="#lcssa">-lcssa</a></td><td>Loop-Closed SSA Form Pass</td></tr>
159 <tr><td><a href="#licm">-licm</a></td><td>Loop Invariant Code Motion</td></tr>
160 <tr><td><a href="#loop-deletion">-loop-deletion</a></td><td>Dead Loop Deletion Pass </td></tr>
161 <tr><td><a href="#loop-extract">-loop-extract</a></td><td>Extract loops into new functions</td></tr>
162 <tr><td><a href="#loop-extract-single">-loop-extract-single</a></td><td>Extract at most one loop into a new function</td></tr>
163 <tr><td><a href="#loop-index-split">-loop-index-split</a></td><td>Index Split Loops</td></tr>
164 <tr><td><a href="#loop-reduce">-loop-reduce</a></td><td>Loop Strength Reduction</td></tr>
165 <tr><td><a href="#loop-rotate">-loop-rotate</a></td><td>Rotate Loops</td></tr>
166 <tr><td><a href="#loop-unroll">-loop-unroll</a></td><td>Unroll loops</td></tr>
167 <tr><td><a href="#loop-unswitch">-loop-unswitch</a></td><td>Unswitch loops</td></tr>
168 <tr><td><a href="#loopsimplify">-loopsimplify</a></td><td>Canonicalize natural loops</td></tr>
169 <tr><td><a href="#lowerinvoke">-lowerinvoke</a></td><td>Lower invoke and unwind, for unwindless code generators</td></tr>
170 <tr><td><a href="#lowersetjmp">-lowersetjmp</a></td><td>Lower Set Jump</td></tr>
171 <tr><td><a href="#lowerswitch">-lowerswitch</a></td><td>Lower SwitchInst's to branches</td></tr>
172 <tr><td><a href="#mem2reg">-mem2reg</a></td><td>Promote Memory to Register</td></tr>
173 <tr><td><a href="#memcpyopt">-memcpyopt</a></td><td>Optimize use of memcpy and friends</td></tr>
174 <tr><td><a href="#mergefunc">-mergefunc</a></td><td>Merge Functions</td></tr>
175 <tr><td><a href="#mergereturn">-mergereturn</a></td><td>Unify function exit nodes</td></tr>
176 <tr><td><a href="#partial-inliner">-partial-inliner</a></td><td>Partial Inliner</td></tr>
177 <tr><td><a href="#partialspecialization">-partialspecialization</a></td><td>Partial Specialization</td></tr>
178 <tr><td><a href="#prune-eh">-prune-eh</a></td><td>Remove unused exception handling info</td></tr>
179 <tr><td><a href="#reassociate">-reassociate</a></td><td>Reassociate expressions</td></tr>
180 <tr><td><a href="#reg2mem">-reg2mem</a></td><td>Demote all values to stack slots</td></tr>
181 <tr><td><a href="#scalarrepl">-scalarrepl</a></td><td>Scalar Replacement of Aggregates</td></tr>
182 <tr><td><a href="#sccp">-sccp</a></td><td>Sparse Conditional Constant Propagation</td></tr>
183 <tr><td><a href="#sink">-sink</a></td><td>Code Sinking</td></tr>
184 <tr><td><a href="#simplify-libcalls">-simplify-libcalls</a></td><td>Simplify well-known library calls</td></tr>
185 <tr><td><a href="#simplify-libcalls-halfpowr">-simplify-libcalls-halfpowr</a></td><td>Simplify half_powr library calls</td></tr>
186 <tr><td><a href="#simplifycfg">-simplifycfg</a></td><td>Simplify the CFG</td></tr>
187 <tr><td><a href="#split-geps">-split-geps</a></td><td>Split complex GEPs into simple GEPs</td></tr>
188 <tr><td><a href="#ssi">-ssi</a></td><td>Static Single Information Construction</td></tr>
189 <tr><td><a href="#ssi-everything">-ssi-everything</a></td><td>Static Single Information Construction (everything, intended for debugging)</td></tr>
190 <tr><td><a href="#strip">-strip</a></td><td>Strip all symbols from a module</td></tr>
191 <tr><td><a href="#strip-dead-debug-info">-strip-dead-debug-info</a></td><td>Strip debug info for unused symbols</td></tr>
192 <tr><td><a href="#strip-dead-prototypes">-strip-dead-prototypes</a></td><td>Remove unused function declarations</td></tr>
193 <tr><td><a href="#strip-debug-declare">-strip-debug-declare</a></td><td>Strip all llvm.dbg.declare intrinsics</td></tr>
194 <tr><td><a href="#strip-nondebug">-strip-nondebug</a></td><td>Strip all symbols, except dbg symbols, from a module</td></tr>
195 <tr><td><a href="#sretpromotion">-sretpromotion</a></td><td>Promote sret arguments</td></tr>
196 <tr><td><a href="#tailcallelim">-tailcallelim</a></td><td>Tail Call Elimination</td></tr>
197 <tr><td><a href="#tailduplicate">-tailduplicate</a></td><td>Tail Duplication</td></tr>
200 <tr><th colspan="2"><b>UTILITY PASSES</b></th></tr>
201 <tr><th>Option</th><th>Name</th></tr>
202 <tr><td><a href="#deadarghaX0r">-deadarghaX0r</a></td><td>Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</td></tr>
203 <tr><td><a href="#extract-blocks">-extract-blocks</a></td><td>Extract Basic Blocks From Module (for bugpoint use)</td></tr>
204 <tr><td><a href="#instnamer">-instnamer</a></td><td>Assign names to anonymous instructions</td></tr>
205 <tr><td><a href="#preverify">-preverify</a></td><td>Preliminary module verification</td></tr>
206 <tr><td><a href="#verify">-verify</a></td><td>Module Verifier</td></tr>
207 <tr><td><a href="#view-cfg">-view-cfg</a></td><td>View CFG of function</td></tr>
208 <tr><td><a href="#view-cfg-only">-view-cfg-only</a></td><td>View CFG of function (with no function bodies)</td></tr>
209 <tr><td><a href="#view-dom">-view-dom</a></td><td>View dominator tree of function</td></tr>
210 <tr><td><a href="#view-dom-only">-view-dom-only</a></td><td>View dominator tree of function (with no function bodies)</td></tr>
211 <tr><td><a href="#view-postdom">-view-postdom</a></td><td>View post dominator tree of function</td></tr>
212 <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>
216 <!-- ======================================================================= -->
217 <div class="doc_section"> <a name="example">Analysis Passes</a></div>
218 <div class="doc_text">
219 <p>This section describes the LLVM Analysis Passes.</p>
222 <!-------------------------------------------------------------------------- -->
223 <div class="doc_subsection">
224 <a name="aa-eval">-aa-eval: Exhaustive Alias Analysis Precision Evaluator</a>
226 <div class="doc_text">
227 <p>This is a simple N^2 alias analysis accuracy evaluator.
228 Basically, for each function in the program, it simply queries to see how the
229 alias analysis implementation answers alias queries between each pair of
230 pointers in the function.</p>
232 <p>This is inspired and adapted from code by: Naveen Neelakantam, Francesco
233 Spadini, and Wojciech Stryjewski.</p>
236 <!-------------------------------------------------------------------------- -->
237 <div class="doc_subsection">
238 <a name="basicaa">-basicaa: Basic Alias Analysis (default AA impl)</a>
240 <div class="doc_text">
242 This is the default implementation of the Alias Analysis interface
243 that simply implements a few identities (two different globals cannot alias,
244 etc), but otherwise does no analysis.
248 <!-------------------------------------------------------------------------- -->
249 <div class="doc_subsection">
250 <a name="basiccg">-basiccg: Basic CallGraph Construction</a>
252 <div class="doc_text">
253 <p>Yet to be written.</p>
256 <!-------------------------------------------------------------------------- -->
257 <div class="doc_subsection">
258 <a name="codegenprepare">-codegenprepare: Optimize for code generation</a>
260 <div class="doc_text">
262 This pass munges the code in the input function to better prepare it for
263 SelectionDAG-based code generation. This works around limitations in it's
264 basic-block-at-a-time approach. It should eventually be removed.
268 <!-------------------------------------------------------------------------- -->
269 <div class="doc_subsection">
270 <a name="count-aa">-count-aa: Count Alias Analysis Query Responses</a>
272 <div class="doc_text">
274 A pass which can be used to count how many alias queries
275 are being made and how the alias analysis implementation being used responds.
279 <!-------------------------------------------------------------------------- -->
280 <div class="doc_subsection">
281 <a name="debug-aa">-debug-aa: AA use debugger</a>
283 <div class="doc_text">
285 This simple pass checks alias analysis users to ensure that if they
286 create a new value, they do not query AA without informing it of the value.
287 It acts as a shim over any other AA pass you want.
291 Yes keeping track of every value in the program is expensive, but this is
296 <!-------------------------------------------------------------------------- -->
297 <div class="doc_subsection">
298 <a name="domfrontier">-domfrontier: Dominance Frontier Construction</a>
300 <div class="doc_text">
302 This pass is a simple dominator construction algorithm for finding forward
307 <!-------------------------------------------------------------------------- -->
308 <div class="doc_subsection">
309 <a name="domtree">-domtree: Dominator Tree Construction</a>
311 <div class="doc_text">
313 This pass is a simple dominator construction algorithm for finding forward
318 <!-------------------------------------------------------------------------- -->
319 <div class="doc_subsection">
320 <a name="dot-callgraph">-dot-callgraph: Print Call Graph to 'dot' file</a>
322 <div class="doc_text">
324 This pass, only available in <code>opt</code>, prints the call graph into a
325 <code>.dot</code> graph. This graph can then be processed with the "dot" tool
326 to convert it to postscript or some other suitable format.
330 <!-------------------------------------------------------------------------- -->
331 <div class="doc_subsection">
332 <a name="dot-cfg">-dot-cfg: Print CFG of function to 'dot' file</a>
334 <div class="doc_text">
336 This pass, only available in <code>opt</code>, prints the control flow graph
337 into a <code>.dot</code> graph. This graph can then be processed with the
338 "dot" tool to convert it to postscript or some other suitable format.
342 <!-------------------------------------------------------------------------- -->
343 <div class="doc_subsection">
344 <a name="dot-cfg-only">-dot-cfg-only: Print CFG of function to 'dot' file (with no function bodies)</a>
346 <div class="doc_text">
348 This pass, only available in <code>opt</code>, prints the control flow graph
349 into a <code>.dot</code> graph, omitting the function bodies. This graph can
350 then be processed with the "dot" tool to convert it to postscript or some
351 other suitable format.
355 <!-------------------------------------------------------------------------- -->
356 <div class="doc_subsection">
357 <a name="dot-dom">-dot-dom: Print dominator tree of function to 'dot' file</a>
359 <div class="doc_text">
361 This pass, only available in <code>opt</code>, prints the dominator tree
362 into a <code>.dot</code> graph. This graph can then be processed with the
363 "dot" tool to convert it to postscript or some other suitable format.
367 <!-------------------------------------------------------------------------- -->
368 <div class="doc_subsection">
369 <a name="dot-dom-only">-dot-dom-only: Print dominator tree of function to 'dot' file (with no
372 <div class="doc_text">
374 This pass, only available in <code>opt</code>, prints the dominator tree
375 into a <code>.dot</code> graph, omitting the function bodies. This graph can
376 then be processed with the "dot" tool to convert it to postscript or some
377 other suitable format.
381 <!-------------------------------------------------------------------------- -->
382 <div class="doc_subsection">
383 <a name="dot-postdom">dot-postdom: Print post dominator tree of function to 'dot' file</a>
385 <div class="doc_text">
387 This pass, only available in <code>opt</code>, prints the post dominator tree
388 into a <code>.dot</code> graph. This graph can then be processed with the
389 "dot" tool to convert it to postscript or some other suitable format.
393 <!-------------------------------------------------------------------------- -->
394 <div class="doc_subsection">
395 <a name="dot-postdom-only">dot-postdom-only: Print post dominator tree of function to 'dot' file
396 (with no function bodies)</a>
398 <div class="doc_text">
400 This pass, only available in <code>opt</code>, prints the post dominator tree
401 into a <code>.dot</code> graph, omitting the function bodies. This graph can
402 then be processed with the "dot" tool to convert it to postscript or some
403 other suitable format.
407 <!-------------------------------------------------------------------------- -->
408 <div class="doc_subsection">
409 <a name="globalsmodref-aa">-globalsmodref-aa: Simple mod/ref analysis for globals</a>
411 <div class="doc_text">
413 This simple pass provides alias and mod/ref information for global values
414 that do not have their address taken, and keeps track of whether functions
415 read or write memory (are "pure"). For this simple (but very common) case,
416 we can provide pretty accurate and useful information.
420 <!-------------------------------------------------------------------------- -->
421 <div class="doc_subsection">
422 <a name="instcount">-instcount: Counts the various types of Instructions</a>
424 <div class="doc_text">
426 This pass collects the count of all instructions and reports them
430 <!-------------------------------------------------------------------------- -->
431 <div class="doc_subsection">
432 <a name="interprocedural-aa-eval">-interprocedural-aa-eval: Exhaustive Interprocedural Alias Analysis Precision Evaluator</a>
434 <div class="doc_text">
435 <p>This pass implements a simple N^2 alias analysis accuracy evaluator.
436 Basically, for each function in the program, it simply queries to see how the
437 alias analysis implementation answers alias queries between each pair of
438 pointers in the function.
442 <!-------------------------------------------------------------------------- -->
443 <div class="doc_subsection">
444 <a name="interprocedural-basic-aa">-interprocedural-basic-aa: Interprocedural Basic Alias Analysis</a>
446 <div class="doc_text">
447 <p>This pass defines the default implementation of the Alias Analysis interface
448 that simply implements a few identities (two different globals cannot alias,
449 etc), but otherwise does no analysis.
453 <!-------------------------------------------------------------------------- -->
454 <div class="doc_subsection">
455 <a name="intervals">-intervals: Interval Partition Construction</a>
457 <div class="doc_text">
459 This analysis calculates and represents the interval partition of a function,
460 or a preexisting interval partition.
464 In this way, the interval partition may be used to reduce a flow graph down
465 to its degenerate single node interval partition (unless it is irreducible).
469 <!-------------------------------------------------------------------------- -->
470 <div class="doc_subsection">
471 <a name="iv-users">-iv-users: Induction Variable Users</a>
473 <div class="doc_text">
474 <p>Bookkeeping for "interesting" users of expressions computed from
475 induction variables.</p>
478 <!-------------------------------------------------------------------------- -->
479 <div class="doc_subsection">
480 <a name="lazy-value-info">-lazy-value-info: Lazy Value Information Analysis</a>
482 <div class="doc_text">
483 <p>Interface for lazy computation of value constraint information.</p>
486 <!-------------------------------------------------------------------------- -->
487 <div class="doc_subsection">
488 <a name="lda">-lda: Loop Dependence Analysis</a>
490 <div class="doc_text">
491 <p>Loop dependence analysis framework, which is used to detect dependences in
492 memory accesses in loops.</p>
495 <!-------------------------------------------------------------------------- -->
496 <div class="doc_subsection">
497 <a name="libcall-aa">-libcall-aa: LibCall Alias Analysis</a>
499 <div class="doc_text">
500 <p>LibCall Alias Analysis.</p>
503 <!-------------------------------------------------------------------------- -->
504 <div class="doc_subsection">
505 <a name="lint">-lint: Check for common errors in LLVM IR</a>
507 <div class="doc_text">
508 <p>This pass statically checks for common and easily-identified constructs
509 which produce undefined or likely unintended behavior in LLVM IR.</p>
511 <p>It is not a guarantee of correctness, in two ways. First, it isn't
512 comprehensive. There are checks which could be done statically which are
513 not yet implemented. Some of these are indicated by TODO comments, but
514 those aren't comprehensive either. Second, many conditions cannot be
515 checked statically. This pass does no dynamic instrumentation, so it
516 can't check for all possible problems.</p>
518 <p>Another limitation is that it assumes all code will be executed. A store
519 through a null pointer in a basic block which is never reached is harmless,
520 but this pass will warn about it anyway.</p>
522 <p>Optimization passes may make conditions that this pass checks for more or
523 less obvious. If an optimization pass appears to be introducing a warning,
524 it may be that the optimization pass is merely exposing an existing
525 condition in the code.</p>
527 <p>This code may be run before instcombine. In many cases, instcombine checks
528 for the same kinds of things and turns instructions with undefined behavior
529 into unreachable (or equivalent). Because of this, this pass makes some
530 effort to look through bitcasts and so on.
534 <!-------------------------------------------------------------------------- -->
535 <div class="doc_subsection">
536 <a name="live-values">-live-values: Values Liveness Analysis</a>
538 <div class="doc_text">
539 <p>LLVM IR Value liveness analysis pass.</p>
542 <!-------------------------------------------------------------------------- -->
543 <div class="doc_subsection">
544 <a name="loops">-loops: Natural Loop Construction</a>
546 <div class="doc_text">
548 This analysis is used to identify natural loops and determine the loop depth
549 of various nodes of the CFG. Note that the loops identified may actually be
550 several natural loops that share the same header node... not just a single
555 <!-------------------------------------------------------------------------- -->
556 <div class="doc_subsection">
557 <a name="memdep">-memdep: Memory Dependence Analysis</a>
559 <div class="doc_text">
561 An analysis that determines, for a given memory operation, what preceding
562 memory operations it depends on. It builds on alias analysis information, and
563 tries to provide a lazy, caching interface to a common kind of alias
568 <!-------------------------------------------------------------------------- -->
569 <div class="doc_subsection">
570 <a name="module-debuginfo">-module-debuginfo: Prints module debug info metadata</a>
572 <div class="doc_text">
573 <p>This pass decodes the debug info metadata in a module and prints in a
574 (sufficiently-prepared-) human-readable form.
576 For example, run this pass from opt along with the -analyze option, and
577 it'll print to standard output.
581 <!-------------------------------------------------------------------------- -->
582 <div class="doc_subsection">
583 <a name="no-aa">-no-aa: No Alias Analysis (always returns 'may' alias)</a>
585 <div class="doc_text">
587 Always returns "I don't know" for alias queries. NoAA is unlike other alias
588 analysis implementations, in that it does not chain to a previous analysis. As
589 such it doesn't follow many of the rules that other alias analyses must.
593 <!-------------------------------------------------------------------------- -->
594 <div class="doc_subsection">
595 <a name="no-profile">-no-profile: No Profile Information</a>
597 <div class="doc_text">
599 The default "no profile" implementation of the abstract
600 <code>ProfileInfo</code> interface.
604 <!-------------------------------------------------------------------------- -->
605 <div class="doc_subsection">
606 <a name="pointertracking">-pointertracking: Track pointer bounds.</a>
608 <div class="doc_text">
609 <p>Tracking of pointer bounds.
613 <!-------------------------------------------------------------------------- -->
614 <div class="doc_subsection">
615 <a name="postdomfrontier">-postdomfrontier: Post-Dominance Frontier Construction</a>
617 <div class="doc_text">
619 This pass is a simple post-dominator construction algorithm for finding
620 post-dominator frontiers.
624 <!-------------------------------------------------------------------------- -->
625 <div class="doc_subsection">
626 <a name="postdomtree">-postdomtree: Post-Dominator Tree Construction</a>
628 <div class="doc_text">
630 This pass is a simple post-dominator construction algorithm for finding
635 <!-------------------------------------------------------------------------- -->
636 <div class="doc_subsection">
637 <a name="print-alias-sets">-print-alias-sets: Alias Set Printer</a>
639 <div class="doc_text">
640 <p>Yet to be written.</p>
643 <!-------------------------------------------------------------------------- -->
644 <div class="doc_subsection">
645 <a name="print-callgraph">-print-callgraph: Print a call graph</a>
647 <div class="doc_text">
649 This pass, only available in <code>opt</code>, prints the call graph to
650 standard output in a human-readable form.
654 <!-------------------------------------------------------------------------- -->
655 <div class="doc_subsection">
656 <a name="print-callgraph-sccs">-print-callgraph-sccs: Print SCCs of the Call Graph</a>
658 <div class="doc_text">
660 This pass, only available in <code>opt</code>, prints the SCCs of the call
661 graph to standard output in a human-readable form.
665 <!-------------------------------------------------------------------------- -->
666 <div class="doc_subsection">
667 <a name="print-cfg-sccs">-print-cfg-sccs: Print SCCs of each function CFG</a>
669 <div class="doc_text">
671 This pass, only available in <code>opt</code>, prints the SCCs of each
672 function CFG to standard output in a human-readable form.
676 <!-------------------------------------------------------------------------- -->
677 <div class="doc_subsection">
678 <a name="print-dbginfo">-print-dbginfo: Print debug info in human readable form</a>
680 <div class="doc_text">
681 <p>Pass that prints instructions, and associated debug info:
684 <li>source/line/col information</li>
685 <li>original variable name</li>
686 <li>original type name</li>
692 <!-------------------------------------------------------------------------- -->
693 <div class="doc_subsection">
694 <a name="print-dom-info">-print-dom-info: Dominator Info Printer</a>
696 <div class="doc_text">
697 <p>Dominator Info Printer.</p>
700 <!-------------------------------------------------------------------------- -->
701 <div class="doc_subsection">
702 <a name="print-externalfnconstants">-print-externalfnconstants: Print external fn callsites passed constants</a>
704 <div class="doc_text">
706 This pass, only available in <code>opt</code>, prints out call sites to
707 external functions that are called with constant arguments. This can be
708 useful when looking for standard library functions we should constant fold
709 or handle in alias analyses.
713 <!-------------------------------------------------------------------------- -->
714 <div class="doc_subsection">
715 <a name="print-function">-print-function: Print function to stderr</a>
717 <div class="doc_text">
719 The <code>PrintFunctionPass</code> class is designed to be pipelined with
720 other <code>FunctionPass</code>es, and prints out the functions of the module
721 as they are processed.
725 <!-------------------------------------------------------------------------- -->
726 <div class="doc_subsection">
727 <a name="print-module">-print-module: Print module to stderr</a>
729 <div class="doc_text">
731 This pass simply prints out the entire module when it is executed.
735 <!-------------------------------------------------------------------------- -->
736 <div class="doc_subsection">
737 <a name="print-used-types">-print-used-types: Find Used Types</a>
739 <div class="doc_text">
741 This pass is used to seek out all of the types in use by the program. Note
742 that this analysis explicitly does not include types only used by the symbol
746 <!-------------------------------------------------------------------------- -->
747 <div class="doc_subsection">
748 <a name="profile-estimator">-profile-estimator: Estimate profiling information</a>
750 <div class="doc_text">
751 <p>Profiling information that estimates the profiling information
752 in a very crude and unimaginative way.
756 <!-------------------------------------------------------------------------- -->
757 <div class="doc_subsection">
758 <a name="profile-loader">-profile-loader: Load profile information from llvmprof.out</a>
760 <div class="doc_text">
762 A concrete implementation of profiling information that loads the information
763 from a profile dump file.
767 <!-------------------------------------------------------------------------- -->
768 <div class="doc_subsection">
769 <a name="profile-verifier">-profile-verifier: Verify profiling information</a>
771 <div class="doc_text">
772 <p>Pass that checks profiling information for plausibility.</p>
775 <!-------------------------------------------------------------------------- -->
776 <div class="doc_subsection">
777 <a name="scalar-evolution">-scalar-evolution: Scalar Evolution Analysis</a>
779 <div class="doc_text">
781 The <code>ScalarEvolution</code> analysis can be used to analyze and
782 catagorize scalar expressions in loops. It specializes in recognizing general
783 induction variables, representing them with the abstract and opaque
784 <code>SCEV</code> class. Given this analysis, trip counts of loops and other
785 important properties can be obtained.
789 This analysis is primarily useful for induction variable substitution and
794 <!-------------------------------------------------------------------------- -->
795 <div class="doc_subsection">
796 <a name="scev-aa">-scev-aa: </a>
798 <div class="doc_text">
799 <p>Simple alias analysis implemented in terms of ScalarEvolution queries.
801 This differs from traditional loop dependence analysis in that it tests
802 for dependencies within a single iteration of a loop, rather than
803 dependencies between different iterations.
805 ScalarEvolution has a more complete understanding of pointer arithmetic
806 than BasicAliasAnalysis' collection of ad-hoc analyses.
810 <!-------------------------------------------------------------------------- -->
811 <div class="doc_subsection">
812 <a name="strip-dead-debug-info">-strip-dead-debug-info: Strip debug info for unused symbols</a>
814 <div class="doc_text">
816 performs code stripping. this transformation can delete:
820 <li>names for virtual registers</li>
821 <li>symbols for internal globals and functions</li>
822 <li>debug information</li>
826 note that this transformation makes code much less readable, so it should
827 only be used in situations where the <tt>strip</tt> utility would be used,
828 such as reducing code size or making it harder to reverse engineer code.
832 <!-------------------------------------------------------------------------- -->
833 <div class="doc_subsection">
834 <a name="targetdata">-targetdata: Target Data Layout</a>
836 <div class="doc_text">
837 <p>Provides other passes access to information on how the size and alignment
838 required by the the target ABI for various data types.</p>
841 <!-- ======================================================================= -->
842 <div class="doc_section"> <a name="transform">Transform Passes</a></div>
843 <div class="doc_text">
844 <p>This section describes the LLVM Transform Passes.</p>
847 <!-------------------------------------------------------------------------- -->
848 <div class="doc_subsection">
849 <a name="abcd">-abcd: Remove redundant conditional branches</a>
851 <div class="doc_text">
852 <p>ABCD removes conditional branch instructions that can be proved redundant.
853 With the SSI representation, each variable has a constraint. By analyzing these
854 constraints we can prove that a branch is redundant. When a branch is proved
855 redundant it means that one direction will always be taken; thus, we can change
856 this branch into an unconditional jump.</p>
857 <p>It is advisable to run <a href="#simplifycfg">SimplifyCFG</a> and
858 <a href="#adce">Aggressive Dead Code Elimination</a> after ABCD
859 to clean up the code.</p>
862 <!-------------------------------------------------------------------------- -->
863 <div class="doc_subsection">
864 <a name="adce">-adce: Aggressive Dead Code Elimination</a>
866 <div class="doc_text">
867 <p>ADCE aggressively tries to eliminate code. This pass is similar to
868 <a href="#dce">DCE</a> but it assumes that values are dead until proven
869 otherwise. This is similar to <a href="#sccp">SCCP</a>, except applied to
870 the liveness of values.</p>
873 <!-------------------------------------------------------------------------- -->
874 <div class="doc_subsection">
875 <a name="always-inline">-always-inline: Inliner for always_inline functions</a>
877 <div class="doc_text">
878 <p>A custom inliner that handles only functions that are marked as
882 <!-------------------------------------------------------------------------- -->
883 <div class="doc_subsection">
884 <a name="argpromotion">-argpromotion: Promote 'by reference' arguments to scalars</a>
886 <div class="doc_text">
888 This pass promotes "by reference" arguments to be "by value" arguments. In
889 practice, this means looking for internal functions that have pointer
890 arguments. If it can prove, through the use of alias analysis, that an
891 argument is *only* loaded, then it can pass the value into the function
892 instead of the address of the value. This can cause recursive simplification
893 of code and lead to the elimination of allocas (especially in C++ template
898 This pass also handles aggregate arguments that are passed into a function,
899 scalarizing them if the elements of the aggregate are only loaded. Note that
900 it refuses to scalarize aggregates which would require passing in more than
901 three operands to the function, because passing thousands of operands for a
902 large array or structure is unprofitable!
906 Note that this transformation could also be done for arguments that are only
907 stored to (returning the value instead), but does not currently. This case
908 would be best handled when and if LLVM starts supporting multiple return
909 values from functions.
913 <!-------------------------------------------------------------------------- -->
914 <div class="doc_subsection">
915 <a name="block-placement">-block-placement: Profile Guided Basic Block Placement</a>
917 <div class="doc_text">
918 <p>This pass is a very simple profile guided basic block placement algorithm.
919 The idea is to put frequently executed blocks together at the start of the
920 function and hopefully increase the number of fall-through conditional
921 branches. If there is no profile information for a particular function, this
922 pass basically orders blocks in depth-first order.</p>
925 <!-------------------------------------------------------------------------- -->
926 <div class="doc_subsection">
927 <a name="break-crit-edges">-break-crit-edges: Break critical edges in CFG</a>
929 <div class="doc_text">
931 Break all of the critical edges in the CFG by inserting a dummy basic block.
932 It may be "required" by passes that cannot deal with critical edges. This
933 transformation obviously invalidates the CFG, but can update forward dominator
934 (set, immediate dominators, tree, and frontier) information.
938 <!-------------------------------------------------------------------------- -->
939 <div class="doc_subsection">
940 <a name="codegenprepare">-codegenprepare: Prepare a function for code generation</a>
942 <div class="doc_text">
943 This pass munges the code in the input function to better prepare it for
944 SelectionDAG-based code generation. This works around limitations in it's
945 basic-block-at-a-time approach. It should eventually be removed.
948 <!-------------------------------------------------------------------------- -->
949 <div class="doc_subsection">
950 <a name="constmerge">-constmerge: Merge Duplicate Global Constants</a>
952 <div class="doc_text">
954 Merges duplicate global constants together into a single constant that is
955 shared. This is useful because some passes (ie TraceValues) insert a lot of
956 string constants into the program, regardless of whether or not an existing
961 <!-------------------------------------------------------------------------- -->
962 <div class="doc_subsection">
963 <a name="constprop">-constprop: Simple constant propagation</a>
965 <div class="doc_text">
966 <p>This file implements constant propagation and merging. It looks for
967 instructions involving only constant operands and replaces them with a
968 constant value instead of an instruction. For example:</p>
969 <blockquote><pre>add i32 1, 2</pre></blockquote>
971 <blockquote><pre>i32 3</pre></blockquote>
972 <p>NOTE: this pass has a habit of making definitions be dead. It is a good
973 idea to to run a <a href="#die">DIE</a> (Dead Instruction Elimination) pass
974 sometime after running this pass.</p>
977 <!-------------------------------------------------------------------------- -->
978 <div class="doc_subsection">
979 <a name="dce">-dce: Dead Code Elimination</a>
981 <div class="doc_text">
983 Dead code elimination is similar to <a href="#die">dead instruction
984 elimination</a>, but it rechecks instructions that were used by removed
985 instructions to see if they are newly dead.
989 <!-------------------------------------------------------------------------- -->
990 <div class="doc_subsection">
991 <a name="deadargelim">-deadargelim: Dead Argument Elimination</a>
993 <div class="doc_text">
995 This pass deletes dead arguments from internal functions. Dead argument
996 elimination removes arguments which are directly dead, as well as arguments
997 only passed into function calls as dead arguments of other functions. This
998 pass also deletes dead arguments in a similar way.
1002 This pass is often useful as a cleanup pass to run after aggressive
1003 interprocedural passes, which add possibly-dead arguments.
1007 <!-------------------------------------------------------------------------- -->
1008 <div class="doc_subsection">
1009 <a name="deadtypeelim">-deadtypeelim: Dead Type Elimination</a>
1011 <div class="doc_text">
1013 This pass is used to cleanup the output of GCC. It eliminate names for types
1014 that are unused in the entire translation unit, using the <a
1015 href="#findusedtypes">find used types</a> pass.
1019 <!-------------------------------------------------------------------------- -->
1020 <div class="doc_subsection">
1021 <a name="die">-die: Dead Instruction Elimination</a>
1023 <div class="doc_text">
1025 Dead instruction elimination performs a single pass over the function,
1026 removing instructions that are obviously dead.
1030 <!-------------------------------------------------------------------------- -->
1031 <div class="doc_subsection">
1032 <a name="dse">-dse: Dead Store Elimination</a>
1034 <div class="doc_text">
1036 A trivial dead store elimination that only considers basic-block local
1041 <!-------------------------------------------------------------------------- -->
1042 <div class="doc_subsection">
1043 <a name="functionattrs">-functionattrs: Deduce function attributes</a>
1045 <div class="doc_text">
1046 <p>A simple interprocedural pass which walks the call-graph, looking for
1047 functions which do not access or only read non-local memory, and marking them
1048 readnone/readonly. In addition, it marks function arguments (of pointer type)
1049 'nocapture' if a call to the function does not create any copies of the pointer
1050 value that outlive the call. This more or less means that the pointer is only
1051 dereferenced, and not returned from the function or stored in a global.
1052 This pass is implemented as a bottom-up traversal of the call-graph.
1056 <!-------------------------------------------------------------------------- -->
1057 <div class="doc_subsection">
1058 <a name="globaldce">-globaldce: Dead Global Elimination</a>
1060 <div class="doc_text">
1062 This transform is designed to eliminate unreachable internal globals from the
1063 program. It uses an aggressive algorithm, searching out globals that are
1064 known to be alive. After it finds all of the globals which are needed, it
1065 deletes whatever is left over. This allows it to delete recursive chunks of
1066 the program which are unreachable.
1070 <!-------------------------------------------------------------------------- -->
1071 <div class="doc_subsection">
1072 <a name="globalopt">-globalopt: Global Variable Optimizer</a>
1074 <div class="doc_text">
1076 This pass transforms simple global variables that never have their address
1077 taken. If obviously true, it marks read/write globals as constant, deletes
1078 variables only stored to, etc.
1082 <!-------------------------------------------------------------------------- -->
1083 <div class="doc_subsection">
1084 <a name="gvn">-gvn: Global Value Numbering</a>
1086 <div class="doc_text">
1088 This pass performs global value numbering to eliminate fully and partially
1089 redundant instructions. It also performs redundant load elimination.
1093 <!-------------------------------------------------------------------------- -->
1094 <div class="doc_subsection">
1095 <a name="indvars">-indvars: Canonicalize Induction Variables</a>
1097 <div class="doc_text">
1099 This transformation analyzes and transforms the induction variables (and
1100 computations derived from them) into simpler forms suitable for subsequent
1101 analysis and transformation.
1105 This transformation makes the following changes to each loop with an
1106 identifiable induction variable:
1110 <li>All loops are transformed to have a <em>single</em> canonical
1111 induction variable which starts at zero and steps by one.</li>
1112 <li>The canonical induction variable is guaranteed to be the first PHI node
1113 in the loop header block.</li>
1114 <li>Any pointer arithmetic recurrences are raised to use array
1119 If the trip count of a loop is computable, this pass also makes the following
1124 <li>The exit condition for the loop is canonicalized to compare the
1125 induction value against the exit value. This turns loops like:
1126 <blockquote><pre>for (i = 7; i*i < 1000; ++i)</pre></blockquote>
1128 <blockquote><pre>for (i = 0; i != 25; ++i)</pre></blockquote></li>
1129 <li>Any use outside of the loop of an expression derived from the indvar
1130 is changed to compute the derived value outside of the loop, eliminating
1131 the dependence on the exit value of the induction variable. If the only
1132 purpose of the loop is to compute the exit value of some derived
1133 expression, this transformation will make the loop dead.</li>
1137 This transformation should be followed by strength reduction after all of the
1138 desired loop transformations have been performed. Additionally, on targets
1139 where it is profitable, the loop could be transformed to count down to zero
1140 (the "do loop" optimization).
1144 <!-------------------------------------------------------------------------- -->
1145 <div class="doc_subsection">
1146 <a name="inline">-inline: Function Integration/Inlining</a>
1148 <div class="doc_text">
1150 Bottom-up inlining of functions into callees.
1154 <!-------------------------------------------------------------------------- -->
1155 <div class="doc_subsection">
1156 <a name="insert-edge-profiling">-insert-edge-profiling: Insert instrumentation for edge profiling</a>
1158 <div class="doc_text">
1160 This pass instruments the specified program with counters for edge profiling.
1161 Edge profiling can give a reasonable approximation of the hot paths through a
1162 program, and is used for a wide variety of program transformations.
1166 Note that this implementation is very naïve. It inserts a counter for
1167 <em>every</em> edge in the program, instead of using control flow information
1168 to prune the number of counters inserted.
1172 <!-------------------------------------------------------------------------- -->
1173 <div class="doc_subsection">
1174 <a name="insert-optimal-edge-profiling">-insert-optimal-edge-profiling: Insert optimal instrumentation for edge profiling</a>
1176 <div class="doc_text">
1177 <p>This pass instruments the specified program with counters for edge profiling.
1178 Edge profiling can give a reasonable approximation of the hot paths through a
1179 program, and is used for a wide variety of program transformations.
1183 <!-------------------------------------------------------------------------- -->
1184 <div class="doc_subsection">
1185 <a name="instcombine">-instcombine: Combine redundant instructions</a>
1187 <div class="doc_text">
1189 Combine instructions to form fewer, simple
1190 instructions. This pass does not modify the CFG This pass is where algebraic
1191 simplification happens.
1195 This pass combines things like:
1200 %Z = add i32 %Y, 1</pre></blockquote>
1207 >%Z = add i32 %X, 2</pre></blockquote>
1210 This is a simple worklist driven algorithm.
1214 This pass guarantees that the following canonicalizations are performed on
1219 <li>If a binary operator has a constant operand, it is moved to the right-
1221 <li>Bitwise operators with constant operands are always grouped so that
1222 shifts are performed first, then <code>or</code>s, then
1223 <code>and</code>s, then <code>xor</code>s.</li>
1224 <li>Compare instructions are converted from <code><</code>,
1225 <code>></code>, <code>≤</code>, or <code>≥</code> to
1226 <code>=</code> or <code>≠</code> if possible.</li>
1227 <li>All <code>cmp</code> instructions on boolean values are replaced with
1228 logical operations.</li>
1229 <li><code>add <var>X</var>, <var>X</var></code> is represented as
1230 <code>mul <var>X</var>, 2</code> ⇒ <code>shl <var>X</var>, 1</code></li>
1231 <li>Multiplies with a constant power-of-two argument are transformed into
1237 <!-------------------------------------------------------------------------- -->
1238 <div class="doc_subsection">
1239 <a name="internalize">-internalize: Internalize Global Symbols</a>
1241 <div class="doc_text">
1243 This pass loops over all of the functions in the input module, looking for a
1244 main function. If a main function is found, all other functions and all
1245 global variables with initializers are marked as internal.
1249 <!-------------------------------------------------------------------------- -->
1250 <div class="doc_subsection">
1251 <a name="ipconstprop">-ipconstprop: Interprocedural constant propagation</a>
1253 <div class="doc_text">
1255 This pass implements an <em>extremely</em> simple interprocedural constant
1256 propagation pass. It could certainly be improved in many different ways,
1257 like using a worklist. This pass makes arguments dead, but does not remove
1258 them. The existing dead argument elimination pass should be run after this
1259 to clean up the mess.
1263 <!-------------------------------------------------------------------------- -->
1264 <div class="doc_subsection">
1265 <a name="ipsccp">-ipsccp: Interprocedural Sparse Conditional Constant Propagation</a>
1267 <div class="doc_text">
1269 An interprocedural variant of <a href="#sccp">Sparse Conditional Constant
1274 <!-------------------------------------------------------------------------- -->
1275 <div class="doc_subsection">
1276 <a name="jump-threading">-jump-threading: Thread control through conditional blocks</a>
1278 <div class="doc_text">
1280 Jump threading tries to find distinct threads of control flow running through
1281 a basic block. This pass looks at blocks that have multiple predecessors and
1282 multiple successors. If one or more of the predecessors of the block can be
1283 proven to always cause a jump to one of the successors, we forward the edge
1284 from the predecessor to the successor by duplicating the contents of this
1288 An example of when this can occur is code like this:
1295 if (X < 3) {</pre>
1298 In this case, the unconditional branch at the end of the first if can be
1299 revectored to the false side of the second if.
1303 <!-------------------------------------------------------------------------- -->
1304 <div class="doc_subsection">
1305 <a name="lcssa">-lcssa: Loop-Closed SSA Form Pass</a>
1307 <div class="doc_text">
1309 This pass transforms loops by placing phi nodes at the end of the loops for
1310 all values that are live across the loop boundary. For example, it turns
1311 the left into the right code:
1315 >for (...) for (...)
1320 X3 = phi(X1, X2) X3 = phi(X1, X2)
1321 ... = X3 + 4 X4 = phi(X3)
1325 This is still valid LLVM; the extra phi nodes are purely redundant, and will
1326 be trivially eliminated by <code>InstCombine</code>. The major benefit of
1327 this transformation is that it makes many other loop optimizations, such as
1328 LoopUnswitching, simpler.
1332 <!-------------------------------------------------------------------------- -->
1333 <div class="doc_subsection">
1334 <a name="licm">-licm: Loop Invariant Code Motion</a>
1336 <div class="doc_text">
1338 This pass performs loop invariant code motion, attempting to remove as much
1339 code from the body of a loop as possible. It does this by either hoisting
1340 code into the preheader block, or by sinking code to the exit blocks if it is
1341 safe. This pass also promotes must-aliased memory locations in the loop to
1342 live in registers, thus hoisting and sinking "invariant" loads and stores.
1346 This pass uses alias analysis for two purposes:
1350 <li>Moving loop invariant loads and calls out of loops. If we can determine
1351 that a load or call inside of a loop never aliases anything stored to,
1352 we can hoist it or sink it like any other instruction.</li>
1353 <li>Scalar Promotion of Memory - If there is a store instruction inside of
1354 the loop, we try to move the store to happen AFTER the loop instead of
1355 inside of the loop. This can only happen if a few conditions are true:
1357 <li>The pointer stored through is loop invariant.</li>
1358 <li>There are no stores or loads in the loop which <em>may</em> alias
1359 the pointer. There are no calls in the loop which mod/ref the
1362 If these conditions are true, we can promote the loads and stores in the
1363 loop of the pointer to use a temporary alloca'd variable. We then use
1364 the mem2reg functionality to construct the appropriate SSA form for the
1368 <!-------------------------------------------------------------------------- -->
1369 <div class="doc_subsection">
1370 <a name="loop-deletion">-loop-deletion: Dead Loop Deletion Pass</a>
1372 <div class="doc_text">
1374 This file implements the Dead Loop Deletion Pass. This pass is responsible
1375 for eliminating loops with non-infinite computable trip counts that have no
1376 side effects or volatile instructions, and do not contribute to the
1377 computation of the function's return value.
1381 <!-------------------------------------------------------------------------- -->
1382 <div class="doc_subsection">
1383 <a name="loop-extract">-loop-extract: Extract loops into new functions</a>
1385 <div class="doc_text">
1387 A pass wrapper around the <code>ExtractLoop()</code> scalar transformation to
1388 extract each top-level loop into its own new function. If the loop is the
1389 <em>only</em> loop in a given function, it is not touched. This is a pass most
1390 useful for debugging via bugpoint.
1394 <!-------------------------------------------------------------------------- -->
1395 <div class="doc_subsection">
1396 <a name="loop-extract-single">-loop-extract-single: Extract at most one loop into a new function</a>
1398 <div class="doc_text">
1400 Similar to <a href="#loop-extract">Extract loops into new functions</a>,
1401 this pass extracts one natural loop from the program into a function if it
1402 can. This is used by bugpoint.
1406 <!-------------------------------------------------------------------------- -->
1407 <div class="doc_subsection">
1408 <a name="loop-index-split">-loop-index-split: Index Split Loops</a>
1410 <div class="doc_text">
1412 This pass divides loop's iteration range by spliting loop such that each
1413 individual loop is executed efficiently.
1417 <!-------------------------------------------------------------------------- -->
1418 <div class="doc_subsection">
1419 <a name="loop-reduce">-loop-reduce: Loop Strength Reduction</a>
1421 <div class="doc_text">
1423 This pass performs a strength reduction on array references inside loops that
1424 have as one or more of their components the loop induction variable. This is
1425 accomplished by creating a new value to hold the initial value of the array
1426 access for the first iteration, and then creating a new GEP instruction in
1427 the loop to increment the value by the appropriate amount.
1431 <!-------------------------------------------------------------------------- -->
1432 <div class="doc_subsection">
1433 <a name="loop-rotate">-loop-rotate: Rotate Loops</a>
1435 <div class="doc_text">
1436 <p>A simple loop rotation transformation.</p>
1439 <!-------------------------------------------------------------------------- -->
1440 <div class="doc_subsection">
1441 <a name="loop-unroll">-loop-unroll: Unroll loops</a>
1443 <div class="doc_text">
1445 This pass implements a simple loop unroller. It works best when loops have
1446 been canonicalized by the <a href="#indvars"><tt>-indvars</tt></a> pass,
1447 allowing it to determine the trip counts of loops easily.
1451 <!-------------------------------------------------------------------------- -->
1452 <div class="doc_subsection">
1453 <a name="loop-unswitch">-loop-unswitch: Unswitch loops</a>
1455 <div class="doc_text">
1457 This pass transforms loops that contain branches on loop-invariant conditions
1458 to have multiple loops. For example, it turns the left into the right code:
1470 This can increase the size of the code exponentially (doubling it every time
1471 a loop is unswitched) so we only unswitch if the resultant code will be
1472 smaller than a threshold.
1476 This pass expects LICM to be run before it to hoist invariant conditions out
1477 of the loop, to make the unswitching opportunity obvious.
1481 <!-------------------------------------------------------------------------- -->
1482 <div class="doc_subsection">
1483 <a name="loopsimplify">-loopsimplify: Canonicalize natural loops</a>
1485 <div class="doc_text">
1487 This pass performs several transformations to transform natural loops into a
1488 simpler form, which makes subsequent analyses and transformations simpler and
1493 Loop pre-header insertion guarantees that there is a single, non-critical
1494 entry edge from outside of the loop to the loop header. This simplifies a
1495 number of analyses and transformations, such as LICM.
1499 Loop exit-block insertion guarantees that all exit blocks from the loop
1500 (blocks which are outside of the loop that have predecessors inside of the
1501 loop) only have predecessors from inside of the loop (and are thus dominated
1502 by the loop header). This simplifies transformations such as store-sinking
1503 that are built into LICM.
1507 This pass also guarantees that loops will have exactly one backedge.
1511 Note that the simplifycfg pass will clean up blocks which are split out but
1512 end up being unnecessary, so usage of this pass should not pessimize
1517 This pass obviously modifies the CFG, but updates loop information and
1518 dominator information.
1522 <!-------------------------------------------------------------------------- -->
1523 <div class="doc_subsection">
1524 <a name="lowerallocs">-lowerallocs: Lower allocations from instructions to calls</a>
1526 <div class="doc_text">
1528 Turn <tt>malloc</tt> and <tt>free</tt> instructions into <tt>@malloc</tt> and
1529 <tt>@free</tt> calls.
1533 This is a target-dependent tranformation because it depends on the size of
1534 data types and alignment constraints.
1538 <!-------------------------------------------------------------------------- -->
1539 <div class="doc_subsection">
1540 <a name="lowerinvoke">-lowerinvoke: Lower invoke and unwind, for unwindless code generators</a>
1542 <div class="doc_text">
1544 This transformation is designed for use by code generators which do not yet
1545 support stack unwinding. This pass supports two models of exception handling
1546 lowering, the 'cheap' support and the 'expensive' support.
1550 'Cheap' exception handling support gives the program the ability to execute
1551 any program which does not "throw an exception", by turning 'invoke'
1552 instructions into calls and by turning 'unwind' instructions into calls to
1553 abort(). If the program does dynamically use the unwind instruction, the
1554 program will print a message then abort.
1558 'Expensive' exception handling support gives the full exception handling
1559 support to the program at the cost of making the 'invoke' instruction
1560 really expensive. It basically inserts setjmp/longjmp calls to emulate the
1561 exception handling as necessary.
1565 Because the 'expensive' support slows down programs a lot, and EH is only
1566 used for a subset of the programs, it must be specifically enabled by the
1567 <tt>-enable-correct-eh-support</tt> option.
1571 Note that after this pass runs the CFG is not entirely accurate (exceptional
1572 control flow edges are not correct anymore) so only very simple things should
1573 be done after the lowerinvoke pass has run (like generation of native code).
1574 This should not be used as a general purpose "my LLVM-to-LLVM pass doesn't
1575 support the invoke instruction yet" lowering pass.
1579 <!-------------------------------------------------------------------------- -->
1580 <div class="doc_subsection">
1581 <a name="lowersetjmp">-lowersetjmp: Lower Set Jump</a>
1583 <div class="doc_text">
1585 Lowers <tt>setjmp</tt> and <tt>longjmp</tt> to use the LLVM invoke and unwind
1586 instructions as necessary.
1590 Lowering of <tt>longjmp</tt> is fairly trivial. We replace the call with a
1591 call to the LLVM library function <tt>__llvm_sjljeh_throw_longjmp()</tt>.
1592 This unwinds the stack for us calling all of the destructors for
1593 objects allocated on the stack.
1597 At a <tt>setjmp</tt> call, the basic block is split and the <tt>setjmp</tt>
1598 removed. The calls in a function that have a <tt>setjmp</tt> are converted to
1599 invoke where the except part checks to see if it's a <tt>longjmp</tt>
1600 exception and, if so, if it's handled in the function. If it is, then it gets
1601 the value returned by the <tt>longjmp</tt> and goes to where the basic block
1602 was split. <tt>invoke</tt> instructions are handled in a similar fashion with
1603 the original except block being executed if it isn't a <tt>longjmp</tt>
1604 except that is handled by that function.
1608 <!-------------------------------------------------------------------------- -->
1609 <div class="doc_subsection">
1610 <a name="lowerswitch">-lowerswitch: Lower SwitchInst's to branches</a>
1612 <div class="doc_text">
1614 Rewrites <tt>switch</tt> instructions with a sequence of branches, which
1615 allows targets to get away with not implementing the switch instruction until
1620 <!-------------------------------------------------------------------------- -->
1621 <div class="doc_subsection">
1622 <a name="mem2reg">-mem2reg: Promote Memory to Register</a>
1624 <div class="doc_text">
1626 This file promotes memory references to be register references. It promotes
1627 <tt>alloca</tt> instructions which only have <tt>load</tt>s and
1628 <tt>store</tt>s as uses. An <tt>alloca</tt> is transformed by using dominator
1629 frontiers to place <tt>phi</tt> nodes, then traversing the function in
1630 depth-first order to rewrite <tt>load</tt>s and <tt>store</tt>s as
1631 appropriate. This is just the standard SSA construction algorithm to construct
1636 <!-------------------------------------------------------------------------- -->
1637 <div class="doc_subsection">
1638 <a name="memcpyopt">-memcpyopt: Optimize use of memcpy and friend</a>
1640 <div class="doc_text">
1642 This pass performs various transformations related to eliminating memcpy
1643 calls, or transforming sets of stores into memset's.
1647 <!-------------------------------------------------------------------------- -->
1648 <div class="doc_subsection">
1649 <a name="mergefunc">-mergefunc: Merge Functions</a>
1651 <div class="doc_text">
1652 <p>This pass looks for equivalent functions that are mergable and folds them.
1654 A hash is computed from the function, based on its type and number of
1657 Once all hashes are computed, we perform an expensive equality comparison
1658 on each function pair. This takes n^2/2 comparisons per bucket, so it's
1659 important that the hash function be high quality. The equality comparison
1660 iterates through each instruction in each basic block.
1662 When a match is found the functions are folded. If both functions are
1663 overridable, we move the functionality into a new internal function and
1664 leave two overridable thunks to it.
1668 <!-------------------------------------------------------------------------- -->
1669 <div class="doc_subsection">
1670 <a name="mergereturn">-mergereturn: Unify function exit nodes</a>
1672 <div class="doc_text">
1674 Ensure that functions have at most one <tt>ret</tt> instruction in them.
1675 Additionally, it keeps track of which node is the new exit node of the CFG.
1679 <!-------------------------------------------------------------------------- -->
1680 <div class="doc_subsection">
1681 <a name="partial-inliner">-partial-inliner: Partial Inliner</a>
1683 <div class="doc_text">
1684 <p>This pass performs partial inlining, typically by inlining an if
1685 statement that surrounds the body of the function.
1689 <!-------------------------------------------------------------------------- -->
1690 <div class="doc_subsection">
1691 <a name="partialspecialization">-partialspecialization: Partial Specialization</a>
1693 <div class="doc_text">
1694 <p>This pass finds function arguments that are often a common constant and
1695 specializes a version of the called function for that constant.
1697 This pass simply does the cloning for functions it specializes. It depends
1698 on <a href="#ipsccp">IPSCCP</a> and <a href="#deadargelim">DAE</a> to clean up the results.
1700 The initial heuristic favors constant arguments that are used in control
1705 <!-------------------------------------------------------------------------- -->
1706 <div class="doc_subsection">
1707 <a name="prune-eh">-prune-eh: Remove unused exception handling info</a>
1709 <div class="doc_text">
1711 This file implements a simple interprocedural pass which walks the call-graph,
1712 turning <tt>invoke</tt> instructions into <tt>call</tt> instructions if and
1713 only if the callee cannot throw an exception. It implements this as a
1714 bottom-up traversal of the call-graph.
1718 <!-------------------------------------------------------------------------- -->
1719 <div class="doc_subsection">
1720 <a name="reassociate">-reassociate: Reassociate expressions</a>
1722 <div class="doc_text">
1724 This pass reassociates commutative expressions in an order that is designed
1725 to promote better constant propagation, GCSE, LICM, PRE, etc.
1729 For example: 4 + (<var>x</var> + 5) ⇒ <var>x</var> + (4 + 5)
1733 In the implementation of this algorithm, constants are assigned rank = 0,
1734 function arguments are rank = 1, and other values are assigned ranks
1735 corresponding to the reverse post order traversal of current function
1736 (starting at 2), which effectively gives values in deep loops higher rank
1737 than values not in loops.
1741 <!-------------------------------------------------------------------------- -->
1742 <div class="doc_subsection">
1743 <a name="reg2mem">-reg2mem: Demote all values to stack slots</a>
1745 <div class="doc_text">
1747 This file demotes all registers to memory references. It is intented to be
1748 the inverse of <a href="#mem2reg"><tt>-mem2reg</tt></a>. By converting to
1749 <tt>load</tt> instructions, the only values live across basic blocks are
1750 <tt>alloca</tt> instructions and <tt>load</tt> instructions before
1751 <tt>phi</tt> nodes. It is intended that this should make CFG hacking much
1752 easier. To make later hacking easier, the entry block is split into two, such
1753 that all introduced <tt>alloca</tt> instructions (and nothing else) are in the
1758 <!-------------------------------------------------------------------------- -->
1759 <div class="doc_subsection">
1760 <a name="scalarrepl">-scalarrepl: Scalar Replacement of Aggregates</a>
1762 <div class="doc_text">
1764 The well-known scalar replacement of aggregates transformation. This
1765 transform breaks up <tt>alloca</tt> instructions of aggregate type (structure
1766 or array) into individual <tt>alloca</tt> instructions for each member if
1767 possible. Then, if possible, it transforms the individual <tt>alloca</tt>
1768 instructions into nice clean scalar SSA form.
1772 This combines a simple scalar replacement of aggregates algorithm with the <a
1773 href="#mem2reg"><tt>mem2reg</tt></a> algorithm because often interact,
1774 especially for C++ programs. As such, iterating between <tt>scalarrepl</tt>,
1775 then <a href="#mem2reg"><tt>mem2reg</tt></a> until we run out of things to
1780 <!-------------------------------------------------------------------------- -->
1781 <div class="doc_subsection">
1782 <a name="sccp">-sccp: Sparse Conditional Constant Propagation</a>
1784 <div class="doc_text">
1786 Sparse conditional constant propagation and merging, which can be summarized
1791 <li>Assumes values are constant unless proven otherwise</li>
1792 <li>Assumes BasicBlocks are dead unless proven otherwise</li>
1793 <li>Proves values to be constant, and replaces them with constants</li>
1794 <li>Proves conditional branches to be unconditional</li>
1798 Note that this pass has a habit of making definitions be dead. It is a good
1799 idea to to run a DCE pass sometime after running this pass.
1803 <!-------------------------------------------------------------------------- -->
1804 <div class="doc_subsection">
1805 <a name="sink">-sink: Code Sinking</a>
1807 <div class="doc_text">
1808 <p>This pass moves instructions into successor blocks, when possible, so that
1809 they aren't executed on paths where their results aren't needed.
1813 <!-------------------------------------------------------------------------- -->
1814 <div class="doc_subsection">
1815 <a name="simplify-libcalls">-simplify-libcalls: Simplify well-known library calls</a>
1817 <div class="doc_text">
1819 Applies a variety of small optimizations for calls to specific well-known
1820 function calls (e.g. runtime library functions). For example, a call
1821 <tt>exit(3)</tt> that occurs within the <tt>main()</tt> function can be
1822 transformed into simply <tt>return 3</tt>.
1826 <!-------------------------------------------------------------------------- -->
1827 <div class="doc_subsection">
1828 <a name="simplify-libcalls-halfpowr">-simplify-libcalls-halfpowr: Simplify half_powr library calls</a>
1830 <div class="doc_text">
1831 <p>Simple pass that applies an experimental transformation on calls
1832 to specific functions.
1836 <!-------------------------------------------------------------------------- -->
1837 <div class="doc_subsection">
1838 <a name="simplifycfg">-simplifycfg: Simplify the CFG</a>
1840 <div class="doc_text">
1842 Performs dead code elimination and basic block merging. Specifically:
1846 <li>Removes basic blocks with no predecessors.</li>
1847 <li>Merges a basic block into its predecessor if there is only one and the
1848 predecessor only has one successor.</li>
1849 <li>Eliminates PHI nodes for basic blocks with a single predecessor.</li>
1850 <li>Eliminates a basic block that only contains an unconditional
1855 <!-------------------------------------------------------------------------- -->
1856 <div class="doc_subsection">
1857 <a name="split-geps">-split-geps: Split complex GEPs into simple GEPs</a>
1859 <div class="doc_text">
1860 <p>This function breaks GEPs with more than 2 non-zero operands into smaller
1861 GEPs each with no more than 2 non-zero operands. This exposes redundancy
1862 between GEPs with common initial operand sequences.
1866 <!-------------------------------------------------------------------------- -->
1867 <div class="doc_subsection">
1868 <a name="ssi">-ssi: Static Single Information Construction</a>
1870 <div class="doc_text">
1871 <p>This pass converts a list of variables to the Static Single Information
1874 We are building an on-demand representation, that is, we do not convert
1875 every single variable in the target function to SSI form. Rather, we receive
1876 a list of target variables that must be converted. We also do not
1877 completely convert a target variable to the SSI format. Instead, we only
1878 change the variable in the points where new information can be attached
1879 to its live range, that is, at branch points.
1883 <!-------------------------------------------------------------------------- -->
1884 <div class="doc_subsection">
1885 <a name="ssi-everything">-ssi-everything: Static Single Information Construction (everything, intended for debugging)</a>
1887 <div class="doc_text">
1888 <p>A pass that runs <a href="#ssi">SSI</a> on every non-void variable, intended for debugging.
1892 <!-------------------------------------------------------------------------- -->
1893 <div class="doc_subsection">
1894 <a name="strip">-strip: Strip all symbols from a module</a>
1896 <div class="doc_text">
1898 performs code stripping. this transformation can delete:
1902 <li>names for virtual registers</li>
1903 <li>symbols for internal globals and functions</li>
1904 <li>debug information</li>
1908 note that this transformation makes code much less readable, so it should
1909 only be used in situations where the <tt>strip</tt> utility would be used,
1910 such as reducing code size or making it harder to reverse engineer code.
1914 <!-------------------------------------------------------------------------- -->
1915 <div class="doc_subsection">
1916 <a name="strip-dead-prototypes">-strip-dead-prototypes: Remove unused function declarations</a>
1918 <div class="doc_text">
1920 This pass loops over all of the functions in the input module, looking for
1921 dead declarations and removes them. Dead declarations are declarations of
1922 functions for which no implementation is available (i.e., declarations for
1923 unused library functions).
1927 <!-------------------------------------------------------------------------- -->
1928 <div class="doc_subsection">
1929 <a name="strip-debug-declare">-strip-debug-declare: Strip all llvm.dbg.declare intrinsics</a>
1931 <div class="doc_text">
1932 <p>This pass implements code stripping. Specifically, it can delete:
1934 <li>names for virtual registers</li>
1935 <li>symbols for internal globals and functions</li>
1936 <li>debug information</li>
1938 Note that this transformation makes code much less readable, so it should
1939 only be used in situations where the 'strip' utility would be used, such as
1940 reducing code size or making it harder to reverse engineer code.
1944 <!-------------------------------------------------------------------------- -->
1945 <div class="doc_subsection">
1946 <a name="strip-nondebug">-strip-nondebug: Strip all symbols, except dbg symbols, from a module</a>
1948 <div class="doc_text">
1949 <p>This pass implements code stripping. Specifically, it can delete:
1951 <li>names for virtual registers</li>
1952 <li>symbols for internal globals and functions</li>
1953 <li>debug information</li>
1955 Note that this transformation makes code much less readable, so it should
1956 only be used in situations where the 'strip' utility would be used, such as
1957 reducing code size or making it harder to reverse engineer code.
1961 <!-------------------------------------------------------------------------- -->
1962 <div class="doc_subsection">
1963 <a name="sretpromotion">-sretpromotion: Promote sret arguments</a>
1965 <div class="doc_text">
1967 This pass finds functions that return a struct (using a pointer to the struct
1968 as the first argument of the function, marked with the '<tt>sret</tt>' attribute) and
1969 replaces them with a new function that simply returns each of the elements of
1970 that struct (using multiple return values).
1974 This pass works under a number of conditions:
1978 <li>The returned struct must not contain other structs</li>
1979 <li>The returned struct must only be used to load values from</li>
1980 <li>The placeholder struct passed in is the result of an <tt>alloca</tt></li>
1984 <!-------------------------------------------------------------------------- -->
1985 <div class="doc_subsection">
1986 <a name="tailcallelim">-tailcallelim: Tail Call Elimination</a>
1988 <div class="doc_text">
1990 This file transforms calls of the current function (self recursion) followed
1991 by a return instruction with a branch to the entry of the function, creating
1992 a loop. This pass also implements the following extensions to the basic
1997 <li>Trivial instructions between the call and return do not prevent the
1998 transformation from taking place, though currently the analysis cannot
1999 support moving any really useful instructions (only dead ones).
2000 <li>This pass transforms functions that are prevented from being tail
2001 recursive by an associative expression to use an accumulator variable,
2002 thus compiling the typical naive factorial or <tt>fib</tt> implementation
2003 into efficient code.
2004 <li>TRE is performed if the function returns void, if the return
2005 returns the result returned by the call, or if the function returns a
2006 run-time constant on all exits from the function. It is possible, though
2007 unlikely, that the return returns something else (like constant 0), and
2008 can still be TRE'd. It can be TRE'd if <em>all other</em> return
2009 instructions in the function return the exact same value.
2010 <li>If it can prove that callees do not access theier caller stack frame,
2011 they are marked as eligible for tail call elimination (by the code
2016 <!-------------------------------------------------------------------------- -->
2017 <div class="doc_subsection">
2018 <a name="tailduplicate">-tailduplicate: Tail Duplication</a>
2020 <div class="doc_text">
2022 This pass performs a limited form of tail duplication, intended to simplify
2023 CFGs by removing some unconditional branches. This pass is necessary to
2024 straighten out loops created by the C front-end, but also is capable of
2025 making other code nicer. After this pass is run, the CFG simplify pass
2026 should be run to clean up the mess.
2030 <!-- ======================================================================= -->
2031 <div class="doc_section"> <a name="transform">Utility Passes</a></div>
2032 <div class="doc_text">
2033 <p>This section describes the LLVM Utility Passes.</p>
2036 <!-------------------------------------------------------------------------- -->
2037 <div class="doc_subsection">
2038 <a name="deadarghaX0r">-deadarghaX0r: Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</a>
2040 <div class="doc_text">
2042 Same as dead argument elimination, but deletes arguments to functions which
2043 are external. This is only for use by <a
2044 href="Bugpoint.html">bugpoint</a>.</p>
2047 <!-------------------------------------------------------------------------- -->
2048 <div class="doc_subsection">
2049 <a name="extract-blocks">-extract-blocks: Extract Basic Blocks From Module (for bugpoint use)</a>
2051 <div class="doc_text">
2053 This pass is used by bugpoint to extract all blocks from the module into their
2057 <!-------------------------------------------------------------------------- -->
2058 <div class="doc_subsection">
2059 <a name="instnamer">-instnamer: Assign names to anonymous instructions</a>
2061 <div class="doc_text">
2062 <p>This is a little utility pass that gives instructions names, this is mostly
2063 useful when diffing the effect of an optimization because deleting an
2064 unnamed instruction can change all other instruction numbering, making the
2069 <!-------------------------------------------------------------------------- -->
2070 <div class="doc_subsection">
2071 <a name="preverify">-preverify: Preliminary module verification</a>
2073 <div class="doc_text">
2075 Ensures that the module is in the form required by the <a
2076 href="#verifier">Module Verifier</a> pass.
2080 Running the verifier runs this pass automatically, so there should be no need
2085 <!-------------------------------------------------------------------------- -->
2086 <div class="doc_subsection">
2087 <a name="verify">-verify: Module Verifier</a>
2089 <div class="doc_text">
2091 Verifies an LLVM IR code. This is useful to run after an optimization which is
2092 undergoing testing. Note that <tt>llvm-as</tt> verifies its input before
2093 emitting bitcode, and also that malformed bitcode is likely to make LLVM
2094 crash. All language front-ends are therefore encouraged to verify their output
2095 before performing optimizing transformations.
2099 <li>Both of a binary operator's parameters are of the same type.</li>
2100 <li>Verify that the indices of mem access instructions match other
2102 <li>Verify that arithmetic and other things are only performed on
2103 first-class types. Verify that shifts and logicals only happen on
2105 <li>All of the constants in a switch statement are of the correct type.</li>
2106 <li>The code is in valid SSA form.</li>
2107 <li>It is illegal to put a label into any other type (like a structure) or
2109 <li>Only phi nodes can be self referential: <tt>%x = add i32 %x, %x</tt> is
2111 <li>PHI nodes must have an entry for each predecessor, with no extras.</li>
2112 <li>PHI nodes must be the first thing in a basic block, all grouped
2114 <li>PHI nodes must have at least one entry.</li>
2115 <li>All basic blocks should only end with terminator insts, not contain
2117 <li>The entry node to a function must not have predecessors.</li>
2118 <li>All Instructions must be embedded into a basic block.</li>
2119 <li>Functions cannot take a void-typed parameter.</li>
2120 <li>Verify that a function's argument list agrees with its declared
2122 <li>It is illegal to specify a name for a void value.</li>
2123 <li>It is illegal to have a internal global value with no initializer.</li>
2124 <li>It is illegal to have a ret instruction that returns a value that does
2125 not agree with the function return value type.</li>
2126 <li>Function call argument types match the function prototype.</li>
2127 <li>All other things that are tested by asserts spread about the code.</li>
2131 Note that this does not provide full security verification (like Java), but
2132 instead just tries to ensure that code is well-formed.
2136 <!-------------------------------------------------------------------------- -->
2137 <div class="doc_subsection">
2138 <a name="view-cfg">-view-cfg: View CFG of function</a>
2140 <div class="doc_text">
2142 Displays the control flow graph using the GraphViz tool.
2146 <!-------------------------------------------------------------------------- -->
2147 <div class="doc_subsection">
2148 <a name="view-cfg-only">-view-cfg-only: View CFG of function (with no function bodies)</a>
2150 <div class="doc_text">
2152 Displays the control flow graph using the GraphViz tool, but omitting function
2157 <!-------------------------------------------------------------------------- -->
2158 <div class="doc_subsection">
2159 <a name="view-dom">-view-dom: View dominator tree of function</a>
2161 <div class="doc_text">
2163 Displays the dominator tree using the GraphViz tool.
2167 <!-------------------------------------------------------------------------- -->
2168 <div class="doc_subsection">
2169 <a name="view-dom-only">-view-dom-only: View dominator tree of function (with no function
2173 <div class="doc_text">
2175 Displays the dominator tree using the GraphViz tool, but omitting function
2180 <!-------------------------------------------------------------------------- -->
2181 <div class="doc_subsection">
2182 <a name="view-postdom">-view-postdom: View post dominator tree of function</a>
2184 <div class="doc_text">
2186 Displays the post dominator tree using the GraphViz tool.
2190 <!-------------------------------------------------------------------------- -->
2191 <div class="doc_subsection">
2192 <a name="view-postdom-only">-view-postdom-only: View post dominator tree of function (with no
2196 <div class="doc_text">
2198 Displays the post dominator tree using the GraphViz tool, but omitting
2203 <!-- *********************************************************************** -->
2207 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2208 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
2209 <a href="http://validator.w3.org/check/referer"><img
2210 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
2212 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br>
2213 <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
2214 Last modified: $Date$