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="_static/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 <h1>LLVM's Analysis and Transform Passes</h1>
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 <h2><a name="intro">Introduction</a></h2>
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>
74 <tr><th colspan="2"><b>ANALYSIS PASSES</b></th></tr>
75 <tr><th>Option</th><th>Name</th></tr>
76 <tr><td><a href="#aa-eval">-aa-eval</a></td><td>Exhaustive Alias Analysis Precision Evaluator</td></tr>
77 <tr><td><a href="#basicaa">-basicaa</a></td><td>Basic Alias Analysis (stateless AA impl)</td></tr>
78 <tr><td><a href="#basiccg">-basiccg</a></td><td>Basic CallGraph Construction</td></tr>
79 <tr><td><a href="#count-aa">-count-aa</a></td><td>Count Alias Analysis Query Responses</td></tr>
80 <tr><td><a href="#da">-da</a></td><td>Dependence Analysis</td></tr>
81 <tr><td><a href="#debug-aa">-debug-aa</a></td><td>AA use debugger</td></tr>
82 <tr><td><a href="#domfrontier">-domfrontier</a></td><td>Dominance Frontier Construction</td></tr>
83 <tr><td><a href="#domtree">-domtree</a></td><td>Dominator Tree Construction</td></tr>
84 <tr><td><a href="#dot-callgraph">-dot-callgraph</a></td><td>Print Call Graph to 'dot' file</td></tr>
85 <tr><td><a href="#dot-cfg">-dot-cfg</a></td><td>Print CFG of function to 'dot' file</td></tr>
86 <tr><td><a href="#dot-cfg-only">-dot-cfg-only</a></td><td>Print CFG of function to 'dot' file (with no function bodies)</td></tr>
87 <tr><td><a href="#dot-dom">-dot-dom</a></td><td>Print dominance tree of function to 'dot' file</td></tr>
88 <tr><td><a href="#dot-dom-only">-dot-dom-only</a></td><td>Print dominance tree of function to 'dot' file (with no function bodies)</td></tr>
89 <tr><td><a href="#dot-postdom">-dot-postdom</a></td><td>Print postdominance tree of function to 'dot' file</td></tr>
90 <tr><td><a href="#dot-postdom-only">-dot-postdom-only</a></td><td>Print postdominance tree of function to 'dot' file (with no function bodies)</td></tr>
91 <tr><td><a href="#globalsmodref-aa">-globalsmodref-aa</a></td><td>Simple mod/ref analysis for globals</td></tr>
92 <tr><td><a href="#instcount">-instcount</a></td><td>Counts the various types of Instructions</td></tr>
93 <tr><td><a href="#intervals">-intervals</a></td><td>Interval Partition Construction</td></tr>
94 <tr><td><a href="#iv-users">-iv-users</a></td><td>Induction Variable Users</td></tr>
95 <tr><td><a href="#lazy-value-info">-lazy-value-info</a></td><td>Lazy Value Information Analysis</td></tr>
96 <tr><td><a href="#libcall-aa">-libcall-aa</a></td><td>LibCall Alias Analysis</td></tr>
97 <tr><td><a href="#lint">-lint</a></td><td>Statically lint-checks LLVM IR</td></tr>
98 <tr><td><a href="#loops">-loops</a></td><td>Natural Loop Information</td></tr>
99 <tr><td><a href="#memdep">-memdep</a></td><td>Memory Dependence Analysis</td></tr>
100 <tr><td><a href="#module-debuginfo">-module-debuginfo</a></td><td>Decodes module-level debug info</td></tr>
101 <tr><td><a href="#no-aa">-no-aa</a></td><td>No Alias Analysis (always returns 'may' alias)</td></tr>
102 <tr><td><a href="#no-profile">-no-profile</a></td><td>No Profile Information</td></tr>
103 <tr><td><a href="#postdomtree">-postdomtree</a></td><td>Post-Dominator Tree Construction</td></tr>
104 <tr><td><a href="#print-alias-sets">-print-alias-sets</a></td><td>Alias Set Printer</td></tr>
105 <tr><td><a href="#print-callgraph">-print-callgraph</a></td><td>Print a call graph</td></tr>
106 <tr><td><a href="#print-callgraph-sccs">-print-callgraph-sccs</a></td><td>Print SCCs of the Call Graph</td></tr>
107 <tr><td><a href="#print-cfg-sccs">-print-cfg-sccs</a></td><td>Print SCCs of each function CFG</td></tr>
108 <tr><td><a href="#print-dbginfo">-print-dbginfo</a></td><td>Print debug info in human readable form</td></tr>
109 <tr><td><a href="#print-dom-info">-print-dom-info</a></td><td>Dominator Info Printer</td></tr>
110 <tr><td><a href="#print-externalfnconstants">-print-externalfnconstants</a></td><td>Print external fn callsites passed constants</td></tr>
111 <tr><td><a href="#print-function">-print-function</a></td><td>Print function to stderr</td></tr>
112 <tr><td><a href="#print-module">-print-module</a></td><td>Print module to stderr</td></tr>
113 <tr><td><a href="#print-used-types">-print-used-types</a></td><td>Find Used Types</td></tr>
114 <tr><td><a href="#profile-estimator">-profile-estimator</a></td><td>Estimate profiling information</td></tr>
115 <tr><td><a href="#profile-loader">-profile-loader</a></td><td>Load profile information from llvmprof.out</td></tr>
116 <tr><td><a href="#profile-verifier">-profile-verifier</a></td><td>Verify profiling information</td></tr>
117 <tr><td><a href="#regions">-regions</a></td><td>Detect single entry single exit regions</td></tr>
118 <tr><td><a href="#scalar-evolution">-scalar-evolution</a></td><td>Scalar Evolution Analysis</td></tr>
119 <tr><td><a href="#scev-aa">-scev-aa</a></td><td>ScalarEvolution-based Alias Analysis</td></tr>
120 <tr><td><a href="#targetdata">-targetdata</a></td><td>Target Data Layout</td></tr>
123 <tr><th colspan="2"><b>TRANSFORM PASSES</b></th></tr>
124 <tr><th>Option</th><th>Name</th></tr>
125 <tr><td><a href="#adce">-adce</a></td><td>Aggressive Dead Code Elimination</td></tr>
126 <tr><td><a href="#always-inline">-always-inline</a></td><td>Inliner for always_inline functions</td></tr>
127 <tr><td><a href="#argpromotion">-argpromotion</a></td><td>Promote 'by reference' arguments to scalars</td></tr>
128 <tr><td><a href="#bb-vectorize">-bb-vectorize</a></td><td>Combine instructions to form vector instructions within basic blocks</td></tr>
129 <tr><td><a href="#block-placement">-block-placement</a></td><td>Profile Guided Basic Block Placement</td></tr>
130 <tr><td><a href="#break-crit-edges">-break-crit-edges</a></td><td>Break critical edges in CFG</td></tr>
131 <tr><td><a href="#codegenprepare">-codegenprepare</a></td><td>Optimize for code generation</td></tr>
132 <tr><td><a href="#constmerge">-constmerge</a></td><td>Merge Duplicate Global Constants</td></tr>
133 <tr><td><a href="#constprop">-constprop</a></td><td>Simple constant propagation</td></tr>
134 <tr><td><a href="#dce">-dce</a></td><td>Dead Code Elimination</td></tr>
135 <tr><td><a href="#deadargelim">-deadargelim</a></td><td>Dead Argument Elimination</td></tr>
136 <tr><td><a href="#deadtypeelim">-deadtypeelim</a></td><td>Dead Type Elimination</td></tr>
137 <tr><td><a href="#die">-die</a></td><td>Dead Instruction Elimination</td></tr>
138 <tr><td><a href="#dse">-dse</a></td><td>Dead Store Elimination</td></tr>
139 <tr><td><a href="#functionattrs">-functionattrs</a></td><td>Deduce function attributes</td></tr>
140 <tr><td><a href="#globaldce">-globaldce</a></td><td>Dead Global Elimination</td></tr>
141 <tr><td><a href="#globalopt">-globalopt</a></td><td>Global Variable Optimizer</td></tr>
142 <tr><td><a href="#gvn">-gvn</a></td><td>Global Value Numbering</td></tr>
143 <tr><td><a href="#indvars">-indvars</a></td><td>Canonicalize Induction Variables</td></tr>
144 <tr><td><a href="#inline">-inline</a></td><td>Function Integration/Inlining</td></tr>
145 <tr><td><a href="#insert-edge-profiling">-insert-edge-profiling</a></td><td>Insert instrumentation for edge profiling</td></tr>
146 <tr><td><a href="#insert-optimal-edge-profiling">-insert-optimal-edge-profiling</a></td><td>Insert optimal instrumentation for edge profiling</td></tr>
147 <tr><td><a href="#instcombine">-instcombine</a></td><td>Combine redundant instructions</td></tr>
148 <tr><td><a href="#internalize">-internalize</a></td><td>Internalize Global Symbols</td></tr>
149 <tr><td><a href="#ipconstprop">-ipconstprop</a></td><td>Interprocedural constant propagation</td></tr>
150 <tr><td><a href="#ipsccp">-ipsccp</a></td><td>Interprocedural Sparse Conditional Constant Propagation</td></tr>
151 <tr><td><a href="#jump-threading">-jump-threading</a></td><td>Jump Threading</td></tr>
152 <tr><td><a href="#lcssa">-lcssa</a></td><td>Loop-Closed SSA Form Pass</td></tr>
153 <tr><td><a href="#licm">-licm</a></td><td>Loop Invariant Code Motion</td></tr>
154 <tr><td><a href="#loop-deletion">-loop-deletion</a></td><td>Delete dead loops</td></tr>
155 <tr><td><a href="#loop-extract">-loop-extract</a></td><td>Extract loops into new functions</td></tr>
156 <tr><td><a href="#loop-extract-single">-loop-extract-single</a></td><td>Extract at most one loop into a new function</td></tr>
157 <tr><td><a href="#loop-reduce">-loop-reduce</a></td><td>Loop Strength Reduction</td></tr>
158 <tr><td><a href="#loop-rotate">-loop-rotate</a></td><td>Rotate Loops</td></tr>
159 <tr><td><a href="#loop-simplify">-loop-simplify</a></td><td>Canonicalize natural loops</td></tr>
160 <tr><td><a href="#loop-unroll">-loop-unroll</a></td><td>Unroll loops</td></tr>
161 <tr><td><a href="#loop-unswitch">-loop-unswitch</a></td><td>Unswitch loops</td></tr>
162 <tr><td><a href="#loweratomic">-loweratomic</a></td><td>Lower atomic intrinsics to non-atomic form</td></tr>
163 <tr><td><a href="#lowerinvoke">-lowerinvoke</a></td><td>Lower invoke and unwind, for unwindless code generators</td></tr>
164 <tr><td><a href="#lowerswitch">-lowerswitch</a></td><td>Lower SwitchInst's to branches</td></tr>
165 <tr><td><a href="#mem2reg">-mem2reg</a></td><td>Promote Memory to Register</td></tr>
166 <tr><td><a href="#memcpyopt">-memcpyopt</a></td><td>MemCpy Optimization</td></tr>
167 <tr><td><a href="#mergefunc">-mergefunc</a></td><td>Merge Functions</td></tr>
168 <tr><td><a href="#mergereturn">-mergereturn</a></td><td>Unify function exit nodes</td></tr>
169 <tr><td><a href="#partial-inliner">-partial-inliner</a></td><td>Partial Inliner</td></tr>
170 <tr><td><a href="#prune-eh">-prune-eh</a></td><td>Remove unused exception handling info</td></tr>
171 <tr><td><a href="#reassociate">-reassociate</a></td><td>Reassociate expressions</td></tr>
172 <tr><td><a href="#reg2mem">-reg2mem</a></td><td>Demote all values to stack slots</td></tr>
173 <tr><td><a href="#scalarrepl">-scalarrepl</a></td><td>Scalar Replacement of Aggregates (DT)</td></tr>
174 <tr><td><a href="#sccp">-sccp</a></td><td>Sparse Conditional Constant Propagation</td></tr>
175 <tr><td><a href="#simplify-libcalls">-simplify-libcalls</a></td><td>Simplify well-known library calls</td></tr>
176 <tr><td><a href="#simplifycfg">-simplifycfg</a></td><td>Simplify the CFG</td></tr>
177 <tr><td><a href="#sink">-sink</a></td><td>Code sinking</td></tr>
178 <tr><td><a href="#strip">-strip</a></td><td>Strip all symbols from a module</td></tr>
179 <tr><td><a href="#strip-dead-debug-info">-strip-dead-debug-info</a></td><td>Strip debug info for unused symbols</td></tr>
180 <tr><td><a href="#strip-dead-prototypes">-strip-dead-prototypes</a></td><td>Strip Unused Function Prototypes</td></tr>
181 <tr><td><a href="#strip-debug-declare">-strip-debug-declare</a></td><td>Strip all llvm.dbg.declare intrinsics</td></tr>
182 <tr><td><a href="#strip-nondebug">-strip-nondebug</a></td><td>Strip all symbols, except dbg symbols, from a module</td></tr>
183 <tr><td><a href="#tailcallelim">-tailcallelim</a></td><td>Tail Call Elimination</td></tr>
186 <tr><th colspan="2"><b>UTILITY PASSES</b></th></tr>
187 <tr><th>Option</th><th>Name</th></tr>
188 <tr><td><a href="#deadarghaX0r">-deadarghaX0r</a></td><td>Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</td></tr>
189 <tr><td><a href="#extract-blocks">-extract-blocks</a></td><td>Extract Basic Blocks From Module (for bugpoint use)</td></tr>
190 <tr><td><a href="#instnamer">-instnamer</a></td><td>Assign names to anonymous instructions</td></tr>
191 <tr><td><a href="#preverify">-preverify</a></td><td>Preliminary module verification</td></tr>
192 <tr><td><a href="#verify">-verify</a></td><td>Module Verifier</td></tr>
193 <tr><td><a href="#view-cfg">-view-cfg</a></td><td>View CFG of function</td></tr>
194 <tr><td><a href="#view-cfg-only">-view-cfg-only</a></td><td>View CFG of function (with no function bodies)</td></tr>
195 <tr><td><a href="#view-dom">-view-dom</a></td><td>View dominance tree of function</td></tr>
196 <tr><td><a href="#view-dom-only">-view-dom-only</a></td><td>View dominance tree of function (with no function bodies)</td></tr>
197 <tr><td><a href="#view-postdom">-view-postdom</a></td><td>View postdominance tree of function</td></tr>
198 <tr><td><a href="#view-postdom-only">-view-postdom-only</a></td><td>View postdominance tree of function (with no function bodies)</td></tr>
203 <!-- ======================================================================= -->
204 <h2><a name="analyses">Analysis Passes</a></h2>
206 <p>This section describes the LLVM Analysis Passes.</p>
208 <!-------------------------------------------------------------------------- -->
210 <a name="aa-eval">-aa-eval: Exhaustive Alias Analysis Precision Evaluator</a>
213 <p>This is a simple N^2 alias analysis accuracy evaluator.
214 Basically, for each function in the program, it simply queries to see how the
215 alias analysis implementation answers alias queries between each pair of
216 pointers in the function.</p>
218 <p>This is inspired and adapted from code by: Naveen Neelakantam, Francesco
219 Spadini, and Wojciech Stryjewski.</p>
222 <!-------------------------------------------------------------------------- -->
224 <a name="basicaa">-basicaa: Basic Alias Analysis (stateless AA impl)</a>
227 <p>A basic alias analysis pass that implements identities (two different
228 globals cannot alias, etc), but does no stateful analysis.</p>
231 <!-------------------------------------------------------------------------- -->
233 <a name="basiccg">-basiccg: Basic CallGraph Construction</a>
236 <p>Yet to be written.</p>
239 <!-------------------------------------------------------------------------- -->
241 <a name="count-aa">-count-aa: Count Alias Analysis Query Responses</a>
245 A pass which can be used to count how many alias queries
246 are being made and how the alias analysis implementation being used responds.
250 <!-------------------------------------------------------------------------- -->
252 <a name="da">-da: Dependence Analysis</a>
255 <p>Dependence analysis framework, which is used to detect dependences in
259 <!-------------------------------------------------------------------------- -->
261 <a name="debug-aa">-debug-aa: AA use debugger</a>
265 This simple pass checks alias analysis users to ensure that if they
266 create a new value, they do not query AA without informing it of the value.
267 It acts as a shim over any other AA pass you want.
271 Yes keeping track of every value in the program is expensive, but this is
276 <!-------------------------------------------------------------------------- -->
278 <a name="domfrontier">-domfrontier: Dominance Frontier Construction</a>
282 This pass is a simple dominator construction algorithm for finding forward
287 <!-------------------------------------------------------------------------- -->
289 <a name="domtree">-domtree: Dominator Tree Construction</a>
293 This pass is a simple dominator construction algorithm for finding forward
298 <!-------------------------------------------------------------------------- -->
300 <a name="dot-callgraph">-dot-callgraph: Print Call Graph to 'dot' file</a>
304 This pass, only available in <code>opt</code>, prints the call graph into a
305 <code>.dot</code> graph. This graph can then be processed with the "dot" tool
306 to convert it to postscript or some other suitable format.
310 <!-------------------------------------------------------------------------- -->
312 <a name="dot-cfg">-dot-cfg: Print CFG of function to 'dot' file</a>
316 This pass, only available in <code>opt</code>, prints the control flow graph
317 into a <code>.dot</code> graph. This graph can then be processed with the
318 "dot" tool to convert it to postscript or some other suitable format.
322 <!-------------------------------------------------------------------------- -->
324 <a name="dot-cfg-only">-dot-cfg-only: Print CFG of function to 'dot' file (with no function bodies)</a>
328 This pass, only available in <code>opt</code>, prints the control flow graph
329 into a <code>.dot</code> graph, omitting the function bodies. This graph can
330 then be processed with the "dot" tool to convert it to postscript or some
331 other suitable format.
335 <!-------------------------------------------------------------------------- -->
337 <a name="dot-dom">-dot-dom: Print dominance tree of function to 'dot' file</a>
341 This pass, only available in <code>opt</code>, prints the dominator tree
342 into a <code>.dot</code> graph. This graph can then be processed with the
343 "dot" tool to convert it to postscript or some other suitable format.
347 <!-------------------------------------------------------------------------- -->
349 <a name="dot-dom-only">-dot-dom-only: Print dominance tree of function to 'dot' file (with no function bodies)</a>
353 This pass, only available in <code>opt</code>, prints the dominator tree
354 into a <code>.dot</code> graph, omitting the function bodies. This graph can
355 then be processed with the "dot" tool to convert it to postscript or some
356 other suitable format.
360 <!-------------------------------------------------------------------------- -->
362 <a name="dot-postdom">-dot-postdom: Print postdominance tree of function to 'dot' file</a>
366 This pass, only available in <code>opt</code>, prints the post dominator tree
367 into a <code>.dot</code> graph. This graph can then be processed with the
368 "dot" tool to convert it to postscript or some other suitable format.
372 <!-------------------------------------------------------------------------- -->
374 <a name="dot-postdom-only">-dot-postdom-only: Print postdominance tree of function to 'dot' file (with no function bodies)</a>
378 This pass, only available in <code>opt</code>, prints the post dominator tree
379 into a <code>.dot</code> graph, omitting the function bodies. This graph can
380 then be processed with the "dot" tool to convert it to postscript or some
381 other suitable format.
385 <!-------------------------------------------------------------------------- -->
387 <a name="globalsmodref-aa">-globalsmodref-aa: Simple mod/ref analysis for globals</a>
391 This simple pass provides alias and mod/ref information for global values
392 that do not have their address taken, and keeps track of whether functions
393 read or write memory (are "pure"). For this simple (but very common) case,
394 we can provide pretty accurate and useful information.
398 <!-------------------------------------------------------------------------- -->
400 <a name="instcount">-instcount: Counts the various types of Instructions</a>
404 This pass collects the count of all instructions and reports them
408 <!-------------------------------------------------------------------------- -->
410 <a name="intervals">-intervals: Interval Partition Construction</a>
414 This analysis calculates and represents the interval partition of a function,
415 or a preexisting interval partition.
419 In this way, the interval partition may be used to reduce a flow graph down
420 to its degenerate single node interval partition (unless it is irreducible).
424 <!-------------------------------------------------------------------------- -->
426 <a name="iv-users">-iv-users: Induction Variable Users</a>
429 <p>Bookkeeping for "interesting" users of expressions computed from
430 induction variables.</p>
433 <!-------------------------------------------------------------------------- -->
435 <a name="lazy-value-info">-lazy-value-info: Lazy Value Information Analysis</a>
438 <p>Interface for lazy computation of value constraint information.</p>
441 <!-------------------------------------------------------------------------- -->
443 <a name="libcall-aa">-libcall-aa: LibCall Alias Analysis</a>
446 <p>LibCall Alias Analysis.</p>
449 <!-------------------------------------------------------------------------- -->
451 <a name="lint">-lint: Statically lint-checks LLVM IR</a>
454 <p>This pass statically checks for common and easily-identified constructs
455 which produce undefined or likely unintended behavior in LLVM IR.</p>
457 <p>It is not a guarantee of correctness, in two ways. First, it isn't
458 comprehensive. There are checks which could be done statically which are
459 not yet implemented. Some of these are indicated by TODO comments, but
460 those aren't comprehensive either. Second, many conditions cannot be
461 checked statically. This pass does no dynamic instrumentation, so it
462 can't check for all possible problems.</p>
464 <p>Another limitation is that it assumes all code will be executed. A store
465 through a null pointer in a basic block which is never reached is harmless,
466 but this pass will warn about it anyway.</p>
468 <p>Optimization passes may make conditions that this pass checks for more or
469 less obvious. If an optimization pass appears to be introducing a warning,
470 it may be that the optimization pass is merely exposing an existing
471 condition in the code.</p>
473 <p>This code may be run before instcombine. In many cases, instcombine checks
474 for the same kinds of things and turns instructions with undefined behavior
475 into unreachable (or equivalent). Because of this, this pass makes some
476 effort to look through bitcasts and so on.
480 <!-------------------------------------------------------------------------- -->
482 <a name="loops">-loops: Natural Loop Information</a>
486 This analysis is used to identify natural loops and determine the loop depth
487 of various nodes of the CFG. Note that the loops identified may actually be
488 several natural loops that share the same header node... not just a single
493 <!-------------------------------------------------------------------------- -->
495 <a name="memdep">-memdep: Memory Dependence Analysis</a>
499 An analysis that determines, for a given memory operation, what preceding
500 memory operations it depends on. It builds on alias analysis information, and
501 tries to provide a lazy, caching interface to a common kind of alias
506 <!-------------------------------------------------------------------------- -->
508 <a name="module-debuginfo">-module-debuginfo: Decodes module-level debug info</a>
511 <p>This pass decodes the debug info metadata in a module and prints in a
512 (sufficiently-prepared-) human-readable form.
514 For example, run this pass from opt along with the -analyze option, and
515 it'll print to standard output.
519 <!-------------------------------------------------------------------------- -->
521 <a name="no-aa">-no-aa: No Alias Analysis (always returns 'may' alias)</a>
525 This is the default implementation of the Alias Analysis interface. It always
526 returns "I don't know" for alias queries. NoAA is unlike other alias analysis
527 implementations, in that it does not chain to a previous analysis. As such it
528 doesn't follow many of the rules that other alias analyses must.
532 <!-------------------------------------------------------------------------- -->
534 <a name="no-profile">-no-profile: No Profile Information</a>
538 The default "no profile" implementation of the abstract
539 <code>ProfileInfo</code> interface.
543 <!-------------------------------------------------------------------------- -->
545 <a name="postdomfrontier">-postdomfrontier: Post-Dominance Frontier Construction</a>
549 This pass is a simple post-dominator construction algorithm for finding
550 post-dominator frontiers.
554 <!-------------------------------------------------------------------------- -->
556 <a name="postdomtree">-postdomtree: Post-Dominator Tree Construction</a>
560 This pass is a simple post-dominator construction algorithm for finding
565 <!-------------------------------------------------------------------------- -->
567 <a name="print-alias-sets">-print-alias-sets: Alias Set Printer</a>
570 <p>Yet to be written.</p>
573 <!-------------------------------------------------------------------------- -->
575 <a name="print-callgraph">-print-callgraph: Print a call graph</a>
579 This pass, only available in <code>opt</code>, prints the call graph to
580 standard error in a human-readable form.
584 <!-------------------------------------------------------------------------- -->
586 <a name="print-callgraph-sccs">-print-callgraph-sccs: Print SCCs of the Call Graph</a>
590 This pass, only available in <code>opt</code>, prints the SCCs of the call
591 graph to standard error in a human-readable form.
595 <!-------------------------------------------------------------------------- -->
597 <a name="print-cfg-sccs">-print-cfg-sccs: Print SCCs of each function CFG</a>
601 This pass, only available in <code>opt</code>, prints the SCCs of each
602 function CFG to standard error in a human-readable form.
606 <!-------------------------------------------------------------------------- -->
608 <a name="print-dbginfo">-print-dbginfo: Print debug info in human readable form</a>
611 <p>Pass that prints instructions, and associated debug info:</p>
614 <li>source/line/col information</li>
615 <li>original variable name</li>
616 <li>original type name</li>
620 <!-------------------------------------------------------------------------- -->
622 <a name="print-dom-info">-print-dom-info: Dominator Info Printer</a>
625 <p>Dominator Info Printer.</p>
628 <!-------------------------------------------------------------------------- -->
630 <a name="print-externalfnconstants">-print-externalfnconstants: Print external fn callsites passed constants</a>
634 This pass, only available in <code>opt</code>, prints out call sites to
635 external functions that are called with constant arguments. This can be
636 useful when looking for standard library functions we should constant fold
637 or handle in alias analyses.
641 <!-------------------------------------------------------------------------- -->
643 <a name="print-function">-print-function: Print function to stderr</a>
647 The <code>PrintFunctionPass</code> class is designed to be pipelined with
648 other <code>FunctionPass</code>es, and prints out the functions of the module
649 as they are processed.
653 <!-------------------------------------------------------------------------- -->
655 <a name="print-module">-print-module: Print module to stderr</a>
659 This pass simply prints out the entire module when it is executed.
663 <!-------------------------------------------------------------------------- -->
665 <a name="print-used-types">-print-used-types: Find Used Types</a>
669 This pass is used to seek out all of the types in use by the program. Note
670 that this analysis explicitly does not include types only used by the symbol
674 <!-------------------------------------------------------------------------- -->
676 <a name="profile-estimator">-profile-estimator: Estimate profiling information</a>
679 <p>Profiling information that estimates the profiling information
680 in a very crude and unimaginative way.
684 <!-------------------------------------------------------------------------- -->
686 <a name="profile-loader">-profile-loader: Load profile information from llvmprof.out</a>
690 A concrete implementation of profiling information that loads the information
691 from a profile dump file.
695 <!-------------------------------------------------------------------------- -->
697 <a name="profile-verifier">-profile-verifier: Verify profiling information</a>
700 <p>Pass that checks profiling information for plausibility.</p>
703 <a name="regions">-regions: Detect single entry single exit regions</a>
707 The <code>RegionInfo</code> pass detects single entry single exit regions in a
708 function, where a region is defined as any subgraph that is connected to the
709 remaining graph at only two spots. Furthermore, an hierarchical region tree is
714 <!-------------------------------------------------------------------------- -->
716 <a name="scalar-evolution">-scalar-evolution: Scalar Evolution Analysis</a>
720 The <code>ScalarEvolution</code> analysis can be used to analyze and
721 catagorize scalar expressions in loops. It specializes in recognizing general
722 induction variables, representing them with the abstract and opaque
723 <code>SCEV</code> class. Given this analysis, trip counts of loops and other
724 important properties can be obtained.
728 This analysis is primarily useful for induction variable substitution and
733 <!-------------------------------------------------------------------------- -->
735 <a name="scev-aa">-scev-aa: ScalarEvolution-based Alias Analysis</a>
738 <p>Simple alias analysis implemented in terms of ScalarEvolution queries.
740 This differs from traditional loop dependence analysis in that it tests
741 for dependencies within a single iteration of a loop, rather than
742 dependencies between different iterations.
744 ScalarEvolution has a more complete understanding of pointer arithmetic
745 than BasicAliasAnalysis' collection of ad-hoc analyses.
749 <!-------------------------------------------------------------------------- -->
751 <a name="targetdata">-targetdata: Target Data Layout</a>
754 <p>Provides other passes access to information on how the size and alignment
755 required by the target ABI for various data types.</p>
760 <!-- ======================================================================= -->
761 <h2><a name="transforms">Transform Passes</a></h2>
763 <p>This section describes the LLVM Transform Passes.</p>
765 <!-------------------------------------------------------------------------- -->
767 <a name="adce">-adce: Aggressive Dead Code Elimination</a>
770 <p>ADCE aggressively tries to eliminate code. This pass is similar to
771 <a href="#dce">DCE</a> but it assumes that values are dead until proven
772 otherwise. This is similar to <a href="#sccp">SCCP</a>, except applied to
773 the liveness of values.</p>
776 <!-------------------------------------------------------------------------- -->
778 <a name="always-inline">-always-inline: Inliner for always_inline functions</a>
781 <p>A custom inliner that handles only functions that are marked as
785 <!-------------------------------------------------------------------------- -->
787 <a name="argpromotion">-argpromotion: Promote 'by reference' arguments to scalars</a>
791 This pass promotes "by reference" arguments to be "by value" arguments. In
792 practice, this means looking for internal functions that have pointer
793 arguments. If it can prove, through the use of alias analysis, that an
794 argument is *only* loaded, then it can pass the value into the function
795 instead of the address of the value. This can cause recursive simplification
796 of code and lead to the elimination of allocas (especially in C++ template
801 This pass also handles aggregate arguments that are passed into a function,
802 scalarizing them if the elements of the aggregate are only loaded. Note that
803 it refuses to scalarize aggregates which would require passing in more than
804 three operands to the function, because passing thousands of operands for a
805 large array or structure is unprofitable!
809 Note that this transformation could also be done for arguments that are only
810 stored to (returning the value instead), but does not currently. This case
811 would be best handled when and if LLVM starts supporting multiple return
812 values from functions.
816 <!-------------------------------------------------------------------------- -->
818 <a name="bb-vectorize">-bb-vectorize: Basic-Block Vectorization</a>
821 <p>This pass combines instructions inside basic blocks to form vector
822 instructions. It iterates over each basic block, attempting to pair
823 compatible instructions, repeating this process until no additional
824 pairs are selected for vectorization. When the outputs of some pair
825 of compatible instructions are used as inputs by some other pair of
826 compatible instructions, those pairs are part of a potential
827 vectorization chain. Instruction pairs are only fused into vector
828 instructions when they are part of a chain longer than some
829 threshold length. Moreover, the pass attempts to find the best
830 possible chain for each pair of compatible instructions. These
831 heuristics are intended to prevent vectorization in cases where
832 it would not yield a performance increase of the resulting code.
836 <!-------------------------------------------------------------------------- -->
838 <a name="block-placement">-block-placement: Profile Guided Basic Block Placement</a>
841 <p>This pass is a very simple profile guided basic block placement algorithm.
842 The idea is to put frequently executed blocks together at the start of the
843 function and hopefully increase the number of fall-through conditional
844 branches. If there is no profile information for a particular function, this
845 pass basically orders blocks in depth-first order.</p>
848 <!-------------------------------------------------------------------------- -->
850 <a name="break-crit-edges">-break-crit-edges: Break critical edges in CFG</a>
854 Break all of the critical edges in the CFG by inserting a dummy basic block.
855 It may be "required" by passes that cannot deal with critical edges. This
856 transformation obviously invalidates the CFG, but can update forward dominator
857 (set, immediate dominators, tree, and frontier) information.
861 <!-------------------------------------------------------------------------- -->
863 <a name="codegenprepare">-codegenprepare: Optimize for code generation</a>
866 This pass munges the code in the input function to better prepare it for
867 SelectionDAG-based code generation. This works around limitations in it's
868 basic-block-at-a-time approach. It should eventually be removed.
871 <!-------------------------------------------------------------------------- -->
873 <a name="constmerge">-constmerge: Merge Duplicate Global Constants</a>
877 Merges duplicate global constants together into a single constant that is
878 shared. This is useful because some passes (ie TraceValues) insert a lot of
879 string constants into the program, regardless of whether or not an existing
884 <!-------------------------------------------------------------------------- -->
886 <a name="constprop">-constprop: Simple constant propagation</a>
889 <p>This file implements constant propagation and merging. It looks for
890 instructions involving only constant operands and replaces them with a
891 constant value instead of an instruction. For example:</p>
892 <blockquote><pre>add i32 1, 2</pre></blockquote>
894 <blockquote><pre>i32 3</pre></blockquote>
895 <p>NOTE: this pass has a habit of making definitions be dead. It is a good
896 idea to to run a <a href="#die">DIE</a> (Dead Instruction Elimination) pass
897 sometime after running this pass.</p>
900 <!-------------------------------------------------------------------------- -->
902 <a name="dce">-dce: Dead Code Elimination</a>
906 Dead code elimination is similar to <a href="#die">dead instruction
907 elimination</a>, but it rechecks instructions that were used by removed
908 instructions to see if they are newly dead.
912 <!-------------------------------------------------------------------------- -->
914 <a name="deadargelim">-deadargelim: Dead Argument Elimination</a>
918 This pass deletes dead arguments from internal functions. Dead argument
919 elimination removes arguments which are directly dead, as well as arguments
920 only passed into function calls as dead arguments of other functions. This
921 pass also deletes dead arguments in a similar way.
925 This pass is often useful as a cleanup pass to run after aggressive
926 interprocedural passes, which add possibly-dead arguments.
930 <!-------------------------------------------------------------------------- -->
932 <a name="deadtypeelim">-deadtypeelim: Dead Type Elimination</a>
936 This pass is used to cleanup the output of GCC. It eliminate names for types
937 that are unused in the entire translation unit, using the <a
938 href="#findusedtypes">find used types</a> pass.
942 <!-------------------------------------------------------------------------- -->
944 <a name="die">-die: Dead Instruction Elimination</a>
948 Dead instruction elimination performs a single pass over the function,
949 removing instructions that are obviously dead.
953 <!-------------------------------------------------------------------------- -->
955 <a name="dse">-dse: Dead Store Elimination</a>
959 A trivial dead store elimination that only considers basic-block local
964 <!-------------------------------------------------------------------------- -->
966 <a name="functionattrs">-functionattrs: Deduce function attributes</a>
969 <p>A simple interprocedural pass which walks the call-graph, looking for
970 functions which do not access or only read non-local memory, and marking them
971 readnone/readonly. In addition, it marks function arguments (of pointer type)
972 'nocapture' if a call to the function does not create any copies of the pointer
973 value that outlive the call. This more or less means that the pointer is only
974 dereferenced, and not returned from the function or stored in a global.
975 This pass is implemented as a bottom-up traversal of the call-graph.
979 <!-------------------------------------------------------------------------- -->
981 <a name="globaldce">-globaldce: Dead Global Elimination</a>
985 This transform is designed to eliminate unreachable internal globals from the
986 program. It uses an aggressive algorithm, searching out globals that are
987 known to be alive. After it finds all of the globals which are needed, it
988 deletes whatever is left over. This allows it to delete recursive chunks of
989 the program which are unreachable.
993 <!-------------------------------------------------------------------------- -->
995 <a name="globalopt">-globalopt: Global Variable Optimizer</a>
999 This pass transforms simple global variables that never have their address
1000 taken. If obviously true, it marks read/write globals as constant, deletes
1001 variables only stored to, etc.
1005 <!-------------------------------------------------------------------------- -->
1007 <a name="gvn">-gvn: Global Value Numbering</a>
1011 This pass performs global value numbering to eliminate fully and partially
1012 redundant instructions. It also performs redundant load elimination.
1016 <!-------------------------------------------------------------------------- -->
1018 <a name="indvars">-indvars: Canonicalize Induction Variables</a>
1022 This transformation analyzes and transforms the induction variables (and
1023 computations derived from them) into simpler forms suitable for subsequent
1024 analysis and transformation.
1028 This transformation makes the following changes to each loop with an
1029 identifiable induction variable:
1033 <li>All loops are transformed to have a <em>single</em> canonical
1034 induction variable which starts at zero and steps by one.</li>
1035 <li>The canonical induction variable is guaranteed to be the first PHI node
1036 in the loop header block.</li>
1037 <li>Any pointer arithmetic recurrences are raised to use array
1042 If the trip count of a loop is computable, this pass also makes the following
1047 <li>The exit condition for the loop is canonicalized to compare the
1048 induction value against the exit value. This turns loops like:
1049 <blockquote><pre>for (i = 7; i*i < 1000; ++i)</pre></blockquote>
1051 <blockquote><pre>for (i = 0; i != 25; ++i)</pre></blockquote></li>
1052 <li>Any use outside of the loop of an expression derived from the indvar
1053 is changed to compute the derived value outside of the loop, eliminating
1054 the dependence on the exit value of the induction variable. If the only
1055 purpose of the loop is to compute the exit value of some derived
1056 expression, this transformation will make the loop dead.</li>
1060 This transformation should be followed by strength reduction after all of the
1061 desired loop transformations have been performed. Additionally, on targets
1062 where it is profitable, the loop could be transformed to count down to zero
1063 (the "do loop" optimization).
1067 <!-------------------------------------------------------------------------- -->
1069 <a name="inline">-inline: Function Integration/Inlining</a>
1073 Bottom-up inlining of functions into callees.
1077 <!-------------------------------------------------------------------------- -->
1079 <a name="insert-edge-profiling">-insert-edge-profiling: Insert instrumentation for edge profiling</a>
1083 This pass instruments the specified program with counters for edge profiling.
1084 Edge profiling can give a reasonable approximation of the hot paths through a
1085 program, and is used for a wide variety of program transformations.
1089 Note that this implementation is very naïve. It inserts a counter for
1090 <em>every</em> edge in the program, instead of using control flow information
1091 to prune the number of counters inserted.
1095 <!-------------------------------------------------------------------------- -->
1097 <a name="insert-optimal-edge-profiling">-insert-optimal-edge-profiling: Insert optimal instrumentation for edge profiling</a>
1100 <p>This pass instruments the specified program with counters for edge profiling.
1101 Edge profiling can give a reasonable approximation of the hot paths through a
1102 program, and is used for a wide variety of program transformations.
1106 <!-------------------------------------------------------------------------- -->
1108 <a name="instcombine">-instcombine: Combine redundant instructions</a>
1112 Combine instructions to form fewer, simple
1113 instructions. This pass does not modify the CFG This pass is where algebraic
1114 simplification happens.
1118 This pass combines things like:
1123 %Z = add i32 %Y, 1</pre></blockquote>
1130 >%Z = add i32 %X, 2</pre></blockquote>
1133 This is a simple worklist driven algorithm.
1137 This pass guarantees that the following canonicalizations are performed on
1142 <li>If a binary operator has a constant operand, it is moved to the right-
1144 <li>Bitwise operators with constant operands are always grouped so that
1145 shifts are performed first, then <code>or</code>s, then
1146 <code>and</code>s, then <code>xor</code>s.</li>
1147 <li>Compare instructions are converted from <code><</code>,
1148 <code>></code>, <code>≤</code>, or <code>≥</code> to
1149 <code>=</code> or <code>≠</code> if possible.</li>
1150 <li>All <code>cmp</code> instructions on boolean values are replaced with
1151 logical operations.</li>
1152 <li><code>add <var>X</var>, <var>X</var></code> is represented as
1153 <code>mul <var>X</var>, 2</code> ⇒ <code>shl <var>X</var>, 1</code></li>
1154 <li>Multiplies with a constant power-of-two argument are transformed into
1160 <!-------------------------------------------------------------------------- -->
1162 <a name="internalize">-internalize: Internalize Global Symbols</a>
1166 This pass loops over all of the functions in the input module, looking for a
1167 main function. If a main function is found, all other functions and all
1168 global variables with initializers are marked as internal.
1172 <!-------------------------------------------------------------------------- -->
1174 <a name="ipconstprop">-ipconstprop: Interprocedural constant propagation</a>
1178 This pass implements an <em>extremely</em> simple interprocedural constant
1179 propagation pass. It could certainly be improved in many different ways,
1180 like using a worklist. This pass makes arguments dead, but does not remove
1181 them. The existing dead argument elimination pass should be run after this
1182 to clean up the mess.
1186 <!-------------------------------------------------------------------------- -->
1188 <a name="ipsccp">-ipsccp: Interprocedural Sparse Conditional Constant Propagation</a>
1192 An interprocedural variant of <a href="#sccp">Sparse Conditional Constant
1197 <!-------------------------------------------------------------------------- -->
1199 <a name="jump-threading">-jump-threading: Jump Threading</a>
1203 Jump threading tries to find distinct threads of control flow running through
1204 a basic block. This pass looks at blocks that have multiple predecessors and
1205 multiple successors. If one or more of the predecessors of the block can be
1206 proven to always cause a jump to one of the successors, we forward the edge
1207 from the predecessor to the successor by duplicating the contents of this
1211 An example of when this can occur is code like this:
1218 if (X < 3) {</pre>
1221 In this case, the unconditional branch at the end of the first if can be
1222 revectored to the false side of the second if.
1226 <!-------------------------------------------------------------------------- -->
1228 <a name="lcssa">-lcssa: Loop-Closed SSA Form Pass</a>
1232 This pass transforms loops by placing phi nodes at the end of the loops for
1233 all values that are live across the loop boundary. For example, it turns
1234 the left into the right code:
1238 >for (...) for (...)
1243 X3 = phi(X1, X2) X3 = phi(X1, X2)
1244 ... = X3 + 4 X4 = phi(X3)
1248 This is still valid LLVM; the extra phi nodes are purely redundant, and will
1249 be trivially eliminated by <code>InstCombine</code>. The major benefit of
1250 this transformation is that it makes many other loop optimizations, such as
1251 LoopUnswitching, simpler.
1255 <!-------------------------------------------------------------------------- -->
1257 <a name="licm">-licm: Loop Invariant Code Motion</a>
1261 This pass performs loop invariant code motion, attempting to remove as much
1262 code from the body of a loop as possible. It does this by either hoisting
1263 code into the preheader block, or by sinking code to the exit blocks if it is
1264 safe. This pass also promotes must-aliased memory locations in the loop to
1265 live in registers, thus hoisting and sinking "invariant" loads and stores.
1269 This pass uses alias analysis for two purposes:
1273 <li>Moving loop invariant loads and calls out of loops. If we can determine
1274 that a load or call inside of a loop never aliases anything stored to,
1275 we can hoist it or sink it like any other instruction.</li>
1276 <li>Scalar Promotion of Memory - If there is a store instruction inside of
1277 the loop, we try to move the store to happen AFTER the loop instead of
1278 inside of the loop. This can only happen if a few conditions are true:
1280 <li>The pointer stored through is loop invariant.</li>
1281 <li>There are no stores or loads in the loop which <em>may</em> alias
1282 the pointer. There are no calls in the loop which mod/ref the
1285 If these conditions are true, we can promote the loads and stores in the
1286 loop of the pointer to use a temporary alloca'd variable. We then use
1287 the mem2reg functionality to construct the appropriate SSA form for the
1292 <!-------------------------------------------------------------------------- -->
1294 <a name="loop-deletion">-loop-deletion: Delete dead loops</a>
1298 This file implements the Dead Loop Deletion Pass. This pass is responsible
1299 for eliminating loops with non-infinite computable trip counts that have no
1300 side effects or volatile instructions, and do not contribute to the
1301 computation of the function's return value.
1305 <!-------------------------------------------------------------------------- -->
1307 <a name="loop-extract">-loop-extract: Extract loops into new functions</a>
1311 A pass wrapper around the <code>ExtractLoop()</code> scalar transformation to
1312 extract each top-level loop into its own new function. If the loop is the
1313 <em>only</em> loop in a given function, it is not touched. This is a pass most
1314 useful for debugging via bugpoint.
1318 <!-------------------------------------------------------------------------- -->
1320 <a name="loop-extract-single">-loop-extract-single: Extract at most one loop into a new function</a>
1324 Similar to <a href="#loop-extract">Extract loops into new functions</a>,
1325 this pass extracts one natural loop from the program into a function if it
1326 can. This is used by bugpoint.
1330 <!-------------------------------------------------------------------------- -->
1332 <a name="loop-reduce">-loop-reduce: Loop Strength Reduction</a>
1336 This pass performs a strength reduction on array references inside loops that
1337 have as one or more of their components the loop induction variable. This is
1338 accomplished by creating a new value to hold the initial value of the array
1339 access for the first iteration, and then creating a new GEP instruction in
1340 the loop to increment the value by the appropriate amount.
1344 <!-------------------------------------------------------------------------- -->
1346 <a name="loop-rotate">-loop-rotate: Rotate Loops</a>
1349 <p>A simple loop rotation transformation.</p>
1352 <!-------------------------------------------------------------------------- -->
1354 <a name="loop-simplify">-loop-simplify: Canonicalize natural loops</a>
1358 This pass performs several transformations to transform natural loops into a
1359 simpler form, which makes subsequent analyses and transformations simpler and
1364 Loop pre-header insertion guarantees that there is a single, non-critical
1365 entry edge from outside of the loop to the loop header. This simplifies a
1366 number of analyses and transformations, such as LICM.
1370 Loop exit-block insertion guarantees that all exit blocks from the loop
1371 (blocks which are outside of the loop that have predecessors inside of the
1372 loop) only have predecessors from inside of the loop (and are thus dominated
1373 by the loop header). This simplifies transformations such as store-sinking
1374 that are built into LICM.
1378 This pass also guarantees that loops will have exactly one backedge.
1382 Note that the simplifycfg pass will clean up blocks which are split out but
1383 end up being unnecessary, so usage of this pass should not pessimize
1388 This pass obviously modifies the CFG, but updates loop information and
1389 dominator information.
1393 <!-------------------------------------------------------------------------- -->
1395 <a name="loop-unroll">-loop-unroll: Unroll loops</a>
1399 This pass implements a simple loop unroller. It works best when loops have
1400 been canonicalized by the <a href="#indvars"><tt>-indvars</tt></a> pass,
1401 allowing it to determine the trip counts of loops easily.
1405 <!-------------------------------------------------------------------------- -->
1407 <a name="loop-unswitch">-loop-unswitch: Unswitch loops</a>
1411 This pass transforms loops that contain branches on loop-invariant conditions
1412 to have multiple loops. For example, it turns the left into the right code:
1424 This can increase the size of the code exponentially (doubling it every time
1425 a loop is unswitched) so we only unswitch if the resultant code will be
1426 smaller than a threshold.
1430 This pass expects LICM to be run before it to hoist invariant conditions out
1431 of the loop, to make the unswitching opportunity obvious.
1435 <!-------------------------------------------------------------------------- -->
1437 <a name="loweratomic">-loweratomic: Lower atomic intrinsics to non-atomic form</a>
1441 This pass lowers atomic intrinsics to non-atomic form for use in a known
1442 non-preemptible environment.
1446 The pass does not verify that the environment is non-preemptible (in
1447 general this would require knowledge of the entire call graph of the
1448 program including any libraries which may not be available in bitcode form);
1449 it simply lowers every atomic intrinsic.
1453 <!-------------------------------------------------------------------------- -->
1455 <a name="lowerinvoke">-lowerinvoke: Lower invoke and unwind, for unwindless code generators</a>
1459 This transformation is designed for use by code generators which do not yet
1460 support stack unwinding. This pass supports two models of exception handling
1461 lowering, the 'cheap' support and the 'expensive' support.
1465 'Cheap' exception handling support gives the program the ability to execute
1466 any program which does not "throw an exception", by turning 'invoke'
1467 instructions into calls and by turning 'unwind' instructions into calls to
1468 abort(). If the program does dynamically use the unwind instruction, the
1469 program will print a message then abort.
1473 'Expensive' exception handling support gives the full exception handling
1474 support to the program at the cost of making the 'invoke' instruction
1475 really expensive. It basically inserts setjmp/longjmp calls to emulate the
1476 exception handling as necessary.
1480 Because the 'expensive' support slows down programs a lot, and EH is only
1481 used for a subset of the programs, it must be specifically enabled by the
1482 <tt>-enable-correct-eh-support</tt> option.
1486 Note that after this pass runs the CFG is not entirely accurate (exceptional
1487 control flow edges are not correct anymore) so only very simple things should
1488 be done after the lowerinvoke pass has run (like generation of native code).
1489 This should not be used as a general purpose "my LLVM-to-LLVM pass doesn't
1490 support the invoke instruction yet" lowering pass.
1494 <!-------------------------------------------------------------------------- -->
1496 <a name="lowerswitch">-lowerswitch: Lower SwitchInst's to branches</a>
1500 Rewrites <tt>switch</tt> instructions with a sequence of branches, which
1501 allows targets to get away with not implementing the switch instruction until
1506 <!-------------------------------------------------------------------------- -->
1508 <a name="mem2reg">-mem2reg: Promote Memory to Register</a>
1512 This file promotes memory references to be register references. It promotes
1513 <tt>alloca</tt> instructions which only have <tt>load</tt>s and
1514 <tt>store</tt>s as uses. An <tt>alloca</tt> is transformed by using dominator
1515 frontiers to place <tt>phi</tt> nodes, then traversing the function in
1516 depth-first order to rewrite <tt>load</tt>s and <tt>store</tt>s as
1517 appropriate. This is just the standard SSA construction algorithm to construct
1522 <!-------------------------------------------------------------------------- -->
1524 <a name="memcpyopt">-memcpyopt: MemCpy Optimization</a>
1528 This pass performs various transformations related to eliminating memcpy
1529 calls, or transforming sets of stores into memset's.
1533 <!-------------------------------------------------------------------------- -->
1535 <a name="mergefunc">-mergefunc: Merge Functions</a>
1538 <p>This pass looks for equivalent functions that are mergable and folds them.
1540 A hash is computed from the function, based on its type and number of
1543 Once all hashes are computed, we perform an expensive equality comparison
1544 on each function pair. This takes n^2/2 comparisons per bucket, so it's
1545 important that the hash function be high quality. The equality comparison
1546 iterates through each instruction in each basic block.
1548 When a match is found the functions are folded. If both functions are
1549 overridable, we move the functionality into a new internal function and
1550 leave two overridable thunks to it.
1554 <!-------------------------------------------------------------------------- -->
1556 <a name="mergereturn">-mergereturn: Unify function exit nodes</a>
1560 Ensure that functions have at most one <tt>ret</tt> instruction in them.
1561 Additionally, it keeps track of which node is the new exit node of the CFG.
1565 <!-------------------------------------------------------------------------- -->
1567 <a name="partial-inliner">-partial-inliner: Partial Inliner</a>
1570 <p>This pass performs partial inlining, typically by inlining an if
1571 statement that surrounds the body of the function.
1575 <!-------------------------------------------------------------------------- -->
1577 <a name="prune-eh">-prune-eh: Remove unused exception handling info</a>
1581 This file implements a simple interprocedural pass which walks the call-graph,
1582 turning <tt>invoke</tt> instructions into <tt>call</tt> instructions if and
1583 only if the callee cannot throw an exception. It implements this as a
1584 bottom-up traversal of the call-graph.
1588 <!-------------------------------------------------------------------------- -->
1590 <a name="reassociate">-reassociate: Reassociate expressions</a>
1594 This pass reassociates commutative expressions in an order that is designed
1595 to promote better constant propagation, GCSE, LICM, PRE, etc.
1599 For example: 4 + (<var>x</var> + 5) ⇒ <var>x</var> + (4 + 5)
1603 In the implementation of this algorithm, constants are assigned rank = 0,
1604 function arguments are rank = 1, and other values are assigned ranks
1605 corresponding to the reverse post order traversal of current function
1606 (starting at 2), which effectively gives values in deep loops higher rank
1607 than values not in loops.
1611 <!-------------------------------------------------------------------------- -->
1613 <a name="reg2mem">-reg2mem: Demote all values to stack slots</a>
1617 This file demotes all registers to memory references. It is intended to be
1618 the inverse of <a href="#mem2reg"><tt>-mem2reg</tt></a>. By converting to
1619 <tt>load</tt> instructions, the only values live across basic blocks are
1620 <tt>alloca</tt> instructions and <tt>load</tt> instructions before
1621 <tt>phi</tt> nodes. It is intended that this should make CFG hacking much
1622 easier. To make later hacking easier, the entry block is split into two, such
1623 that all introduced <tt>alloca</tt> instructions (and nothing else) are in the
1628 <!-------------------------------------------------------------------------- -->
1630 <a name="scalarrepl">-scalarrepl: Scalar Replacement of Aggregates (DT)</a>
1634 The well-known scalar replacement of aggregates transformation. This
1635 transform breaks up <tt>alloca</tt> instructions of aggregate type (structure
1636 or array) into individual <tt>alloca</tt> instructions for each member if
1637 possible. Then, if possible, it transforms the individual <tt>alloca</tt>
1638 instructions into nice clean scalar SSA form.
1642 This combines a simple scalar replacement of aggregates algorithm with the <a
1643 href="#mem2reg"><tt>mem2reg</tt></a> algorithm because often interact,
1644 especially for C++ programs. As such, iterating between <tt>scalarrepl</tt>,
1645 then <a href="#mem2reg"><tt>mem2reg</tt></a> until we run out of things to
1650 <!-------------------------------------------------------------------------- -->
1652 <a name="sccp">-sccp: Sparse Conditional Constant Propagation</a>
1656 Sparse conditional constant propagation and merging, which can be summarized
1661 <li>Assumes values are constant unless proven otherwise</li>
1662 <li>Assumes BasicBlocks are dead unless proven otherwise</li>
1663 <li>Proves values to be constant, and replaces them with constants</li>
1664 <li>Proves conditional branches to be unconditional</li>
1668 Note that this pass has a habit of making definitions be dead. It is a good
1669 idea to to run a DCE pass sometime after running this pass.
1673 <!-------------------------------------------------------------------------- -->
1675 <a name="simplify-libcalls">-simplify-libcalls: Simplify well-known library calls</a>
1679 Applies a variety of small optimizations for calls to specific well-known
1680 function calls (e.g. runtime library functions). For example, a call
1681 <tt>exit(3)</tt> that occurs within the <tt>main()</tt> function can be
1682 transformed into simply <tt>return 3</tt>.
1686 <!-------------------------------------------------------------------------- -->
1688 <a name="simplifycfg">-simplifycfg: Simplify the CFG</a>
1692 Performs dead code elimination and basic block merging. Specifically:
1696 <li>Removes basic blocks with no predecessors.</li>
1697 <li>Merges a basic block into its predecessor if there is only one and the
1698 predecessor only has one successor.</li>
1699 <li>Eliminates PHI nodes for basic blocks with a single predecessor.</li>
1700 <li>Eliminates a basic block that only contains an unconditional
1705 <!-------------------------------------------------------------------------- -->
1707 <a name="sink">-sink: Code sinking</a>
1710 <p>This pass moves instructions into successor blocks, when possible, so that
1711 they aren't executed on paths where their results aren't needed.
1715 <!-------------------------------------------------------------------------- -->
1717 <a name="strip">-strip: Strip all symbols from a module</a>
1721 performs code stripping. this transformation can delete:
1725 <li>names for virtual registers</li>
1726 <li>symbols for internal globals and functions</li>
1727 <li>debug information</li>
1731 note that this transformation makes code much less readable, so it should
1732 only be used in situations where the <tt>strip</tt> utility would be used,
1733 such as reducing code size or making it harder to reverse engineer code.
1737 <!-------------------------------------------------------------------------- -->
1739 <a name="strip-dead-debug-info">-strip-dead-debug-info: Strip debug info for unused symbols</a>
1743 performs code stripping. this transformation can delete:
1747 <li>names for virtual registers</li>
1748 <li>symbols for internal globals and functions</li>
1749 <li>debug information</li>
1753 note that this transformation makes code much less readable, so it should
1754 only be used in situations where the <tt>strip</tt> utility would be used,
1755 such as reducing code size or making it harder to reverse engineer code.
1759 <!-------------------------------------------------------------------------- -->
1761 <a name="strip-dead-prototypes">-strip-dead-prototypes: Strip Unused Function Prototypes</a>
1765 This pass loops over all of the functions in the input module, looking for
1766 dead declarations and removes them. Dead declarations are declarations of
1767 functions for which no implementation is available (i.e., declarations for
1768 unused library functions).
1772 <!-------------------------------------------------------------------------- -->
1774 <a name="strip-debug-declare">-strip-debug-declare: Strip all llvm.dbg.declare intrinsics</a>
1777 <p>This pass implements code stripping. Specifically, it can delete:</p>
1779 <li>names for virtual registers</li>
1780 <li>symbols for internal globals and functions</li>
1781 <li>debug information</li>
1784 Note that this transformation makes code much less readable, so it should
1785 only be used in situations where the 'strip' utility would be used, such as
1786 reducing code size or making it harder to reverse engineer code.
1790 <!-------------------------------------------------------------------------- -->
1792 <a name="strip-nondebug">-strip-nondebug: Strip all symbols, except dbg symbols, from a module</a>
1795 <p>This pass implements code stripping. Specifically, it can delete:</p>
1797 <li>names for virtual registers</li>
1798 <li>symbols for internal globals and functions</li>
1799 <li>debug information</li>
1802 Note that this transformation makes code much less readable, so it should
1803 only be used in situations where the 'strip' utility would be used, such as
1804 reducing code size or making it harder to reverse engineer code.
1808 <!-------------------------------------------------------------------------- -->
1810 <a name="tailcallelim">-tailcallelim: Tail Call Elimination</a>
1814 This file transforms calls of the current function (self recursion) followed
1815 by a return instruction with a branch to the entry of the function, creating
1816 a loop. This pass also implements the following extensions to the basic
1821 <li>Trivial instructions between the call and return do not prevent the
1822 transformation from taking place, though currently the analysis cannot
1823 support moving any really useful instructions (only dead ones).
1824 <li>This pass transforms functions that are prevented from being tail
1825 recursive by an associative expression to use an accumulator variable,
1826 thus compiling the typical naive factorial or <tt>fib</tt> implementation
1827 into efficient code.
1828 <li>TRE is performed if the function returns void, if the return
1829 returns the result returned by the call, or if the function returns a
1830 run-time constant on all exits from the function. It is possible, though
1831 unlikely, that the return returns something else (like constant 0), and
1832 can still be TRE'd. It can be TRE'd if <em>all other</em> return
1833 instructions in the function return the exact same value.
1834 <li>If it can prove that callees do not access theier caller stack frame,
1835 they are marked as eligible for tail call elimination (by the code
1840 <!-- ======================================================================= -->
1841 <h2><a name="utilities">Utility Passes</a></h2>
1843 <p>This section describes the LLVM Utility Passes.</p>
1845 <!-------------------------------------------------------------------------- -->
1847 <a name="deadarghaX0r">-deadarghaX0r: Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)</a>
1851 Same as dead argument elimination, but deletes arguments to functions which
1852 are external. This is only for use by <a
1853 href="Bugpoint.html">bugpoint</a>.</p>
1856 <!-------------------------------------------------------------------------- -->
1858 <a name="extract-blocks">-extract-blocks: Extract Basic Blocks From Module (for bugpoint use)</a>
1862 This pass is used by bugpoint to extract all blocks from the module into their
1866 <!-------------------------------------------------------------------------- -->
1868 <a name="instnamer">-instnamer: Assign names to anonymous instructions</a>
1871 <p>This is a little utility pass that gives instructions names, this is mostly
1872 useful when diffing the effect of an optimization because deleting an
1873 unnamed instruction can change all other instruction numbering, making the
1878 <!-------------------------------------------------------------------------- -->
1880 <a name="preverify">-preverify: Preliminary module verification</a>
1884 Ensures that the module is in the form required by the <a
1885 href="#verifier">Module Verifier</a> pass.
1889 Running the verifier runs this pass automatically, so there should be no need
1894 <!-------------------------------------------------------------------------- -->
1896 <a name="verify">-verify: Module Verifier</a>
1900 Verifies an LLVM IR code. This is useful to run after an optimization which is
1901 undergoing testing. Note that <tt>llvm-as</tt> verifies its input before
1902 emitting bitcode, and also that malformed bitcode is likely to make LLVM
1903 crash. All language front-ends are therefore encouraged to verify their output
1904 before performing optimizing transformations.
1908 <li>Both of a binary operator's parameters are of the same type.</li>
1909 <li>Verify that the indices of mem access instructions match other
1911 <li>Verify that arithmetic and other things are only performed on
1912 first-class types. Verify that shifts and logicals only happen on
1914 <li>All of the constants in a switch statement are of the correct type.</li>
1915 <li>The code is in valid SSA form.</li>
1916 <li>It is illegal to put a label into any other type (like a structure) or
1918 <li>Only phi nodes can be self referential: <tt>%x = add i32 %x, %x</tt> is
1920 <li>PHI nodes must have an entry for each predecessor, with no extras.</li>
1921 <li>PHI nodes must be the first thing in a basic block, all grouped
1923 <li>PHI nodes must have at least one entry.</li>
1924 <li>All basic blocks should only end with terminator insts, not contain
1926 <li>The entry node to a function must not have predecessors.</li>
1927 <li>All Instructions must be embedded into a basic block.</li>
1928 <li>Functions cannot take a void-typed parameter.</li>
1929 <li>Verify that a function's argument list agrees with its declared
1931 <li>It is illegal to specify a name for a void value.</li>
1932 <li>It is illegal to have an internal global value with no initializer.</li>
1933 <li>It is illegal to have a ret instruction that returns a value that does
1934 not agree with the function return value type.</li>
1935 <li>Function call argument types match the function prototype.</li>
1936 <li>All other things that are tested by asserts spread about the code.</li>
1940 Note that this does not provide full security verification (like Java), but
1941 instead just tries to ensure that code is well-formed.
1945 <!-------------------------------------------------------------------------- -->
1947 <a name="view-cfg">-view-cfg: View CFG of function</a>
1951 Displays the control flow graph using the GraphViz tool.
1955 <!-------------------------------------------------------------------------- -->
1957 <a name="view-cfg-only">-view-cfg-only: View CFG of function (with no function bodies)</a>
1961 Displays the control flow graph using the GraphViz tool, but omitting function
1966 <!-------------------------------------------------------------------------- -->
1968 <a name="view-dom">-view-dom: View dominance tree of function</a>
1972 Displays the dominator tree using the GraphViz tool.
1976 <!-------------------------------------------------------------------------- -->
1978 <a name="view-dom-only">-view-dom-only: View dominance tree of function (with no function bodies)</a>
1982 Displays the dominator tree using the GraphViz tool, but omitting function
1987 <!-------------------------------------------------------------------------- -->
1989 <a name="view-postdom">-view-postdom: View postdominance tree of function</a>
1993 Displays the post dominator tree using the GraphViz tool.
1997 <!-------------------------------------------------------------------------- -->
1999 <a name="view-postdom-only">-view-postdom-only: View postdominance tree of function (with no function bodies)</a>
2003 Displays the post dominator tree using the GraphViz tool, but omitting
2010 <!-- *********************************************************************** -->
2014 <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
2015 src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
2016 <a href="http://validator.w3.org/check/referer"><img
2017 src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
2019 <a href="mailto:rspencer@x10sys.com">Reid Spencer</a><br>
2020 <a href="http://llvm.org/">LLVM Compiler Infrastructure</a><br>
2021 Last modified: $Date$