2 If Passes.html is up to date, the following "one-liner" should print
5 egrep -e '^<tr><td><a href="#.*">-.*</a></td><td>.*</td></tr>$' \
6 -e '^ <a name=".*">.*</a>$' < Passes.html >html; \
7 perl >help <<'EOT' && diff -u help html; rm -f help html
8 open HTML, "<Passes.html" or die "open: Passes.html: $!\n";
10 m:^<tr><td><a href="#(.*)">-.*</a></td><td>.*</td></tr>$: or next;
11 $order{$1} = sprintf("%03d", 1 + int %order);
13 open HELP, "../Release/bin/opt -help|" or die "open: opt -help: $!\n";
15 m:^ -([^ ]+) +- (.*)$: or next;
17 $o = "000" unless defined $o;
18 push @x, "$o<tr><td><a href=\"#$1\">-$1</a></td><td>$2</td></tr>\n";
19 push @y, "$o <a name=\"$1\">-$1: $2</a>\n";
21 @x = map { s/^\d\d\d//; $_ } sort @x;
22 @y = map { s/^\d\d\d//; $_ } sort @y;
26 This (real) one-liner can also be helpful when converting comments to HTML:
28 perl -e '$/ = 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'
30 ====================================
31 LLVM's Analysis and Transform Passes
32 ====================================
40 This document serves as a high level summary of the optimization features that
41 LLVM provides. Optimizations are implemented as Passes that traverse some
42 portion of a program to either collect information or transform the program.
43 The table below divides the passes that LLVM provides into three categories.
44 Analysis passes compute information that other passes can use or for debugging
45 or program visualization purposes. Transform passes can use (or invalidate)
46 the analysis passes. Transform passes all mutate the program in some way.
47 Utility passes provides some utility but don't otherwise fit categorization.
48 For example passes to extract functions to bitcode or write a module to bitcode
49 are neither analysis nor transform passes. The table of contents above
50 provides a quick summary of each pass and links to the more complete pass
51 description later in the document.
56 This section describes the LLVM Analysis Passes.
58 ``-aa-eval``: Exhaustive Alias Analysis Precision Evaluator
59 -----------------------------------------------------------
61 This is a simple N^2 alias analysis accuracy evaluator. Basically, for each
62 function in the program, it simply queries to see how the alias analysis
63 implementation answers alias queries between each pair of pointers in the
66 This is inspired and adapted from code by: Naveen Neelakantam, Francesco
67 Spadini, and Wojciech Stryjewski.
69 ``-basicaa``: Basic Alias Analysis (stateless AA impl)
70 ------------------------------------------------------
72 A basic alias analysis pass that implements identities (two different globals
73 cannot alias, etc), but does no stateful analysis.
75 ``-basiccg``: Basic CallGraph Construction
76 ------------------------------------------
80 ``-count-aa``: Count Alias Analysis Query Responses
81 ---------------------------------------------------
83 A pass which can be used to count how many alias queries are being made and how
84 the alias analysis implementation being used responds.
86 ``-da``: Dependence Analysis
87 ----------------------------
89 Dependence analysis framework, which is used to detect dependences in memory
92 ``-debug-aa``: AA use debugger
93 ------------------------------
95 This simple pass checks alias analysis users to ensure that if they create a
96 new value, they do not query AA without informing it of the value. It acts as
97 a shim over any other AA pass you want.
99 Yes keeping track of every value in the program is expensive, but this is a
102 ``-domfrontier``: Dominance Frontier Construction
103 -------------------------------------------------
105 This pass is a simple dominator construction algorithm for finding forward
108 ``-domtree``: Dominator Tree Construction
109 -----------------------------------------
111 This pass is a simple dominator construction algorithm for finding forward
115 ``-dot-callgraph``: Print Call Graph to "dot" file
116 --------------------------------------------------
118 This pass, only available in ``opt``, prints the call graph into a ``.dot``
119 graph. This graph can then be processed with the "dot" tool to convert it to
120 postscript or some other suitable format.
122 ``-dot-cfg``: Print CFG of function to "dot" file
123 -------------------------------------------------
125 This pass, only available in ``opt``, prints the control flow graph into a
126 ``.dot`` graph. This graph can then be processed with the :program:`dot` tool
127 to convert it to postscript or some other suitable format.
129 ``-dot-cfg-only``: Print CFG of function to "dot" file (with no function bodies)
130 --------------------------------------------------------------------------------
132 This pass, only available in ``opt``, prints the control flow graph into a
133 ``.dot`` graph, omitting the function bodies. This graph can then be processed
134 with the :program:`dot` tool to convert it to postscript or some other suitable
137 ``-dot-dom``: Print dominance tree of function to "dot" file
138 ------------------------------------------------------------
140 This pass, only available in ``opt``, prints the dominator tree into a ``.dot``
141 graph. This graph can then be processed with the :program:`dot` tool to
142 convert it to postscript or some other suitable format.
144 ``-dot-dom-only``: Print dominance tree of function to "dot" file (with no function bodies)
145 -------------------------------------------------------------------------------------------
147 This pass, only available in ``opt``, prints the dominator tree into a ``.dot``
148 graph, omitting the function bodies. This graph can then be processed with the
149 :program:`dot` tool to convert it to postscript or some other suitable format.
151 ``-dot-postdom``: Print postdominance tree of function to "dot" file
152 --------------------------------------------------------------------
154 This pass, only available in ``opt``, prints the post dominator tree into a
155 ``.dot`` graph. This graph can then be processed with the :program:`dot` tool
156 to convert it to postscript or some other suitable format.
158 ``-dot-postdom-only``: Print postdominance tree of function to "dot" file (with no function bodies)
159 ---------------------------------------------------------------------------------------------------
161 This pass, only available in ``opt``, prints the post dominator tree into a
162 ``.dot`` graph, omitting the function bodies. This graph can then be processed
163 with the :program:`dot` tool to convert it to postscript or some other suitable
166 ``-globalsmodref-aa``: Simple mod/ref analysis for globals
167 ----------------------------------------------------------
169 This simple pass provides alias and mod/ref information for global values that
170 do not have their address taken, and keeps track of whether functions read or
171 write memory (are "pure"). For this simple (but very common) case, we can
172 provide pretty accurate and useful information.
174 ``-instcount``: Counts the various types of ``Instruction``\ s
175 --------------------------------------------------------------
177 This pass collects the count of all instructions and reports them.
179 ``-intervals``: Interval Partition Construction
180 -----------------------------------------------
182 This analysis calculates and represents the interval partition of a function,
183 or a preexisting interval partition.
185 In this way, the interval partition may be used to reduce a flow graph down to
186 its degenerate single node interval partition (unless it is irreducible).
188 ``-iv-users``: Induction Variable Users
189 ---------------------------------------
191 Bookkeeping for "interesting" users of expressions computed from induction
194 ``-lazy-value-info``: Lazy Value Information Analysis
195 -----------------------------------------------------
197 Interface for lazy computation of value constraint information.
199 ``-libcall-aa``: LibCall Alias Analysis
200 ---------------------------------------
202 LibCall Alias Analysis.
204 ``-lint``: Statically lint-checks LLVM IR
205 -----------------------------------------
207 This pass statically checks for common and easily-identified constructs which
208 produce undefined or likely unintended behavior in LLVM IR.
210 It is not a guarantee of correctness, in two ways. First, it isn't
211 comprehensive. There are checks which could be done statically which are not
212 yet implemented. Some of these are indicated by TODO comments, but those
213 aren't comprehensive either. Second, many conditions cannot be checked
214 statically. This pass does no dynamic instrumentation, so it can't check for
215 all possible problems.
217 Another limitation is that it assumes all code will be executed. A store
218 through a null pointer in a basic block which is never reached is harmless, but
219 this pass will warn about it anyway.
221 Optimization passes may make conditions that this pass checks for more or less
222 obvious. If an optimization pass appears to be introducing a warning, it may
223 be that the optimization pass is merely exposing an existing condition in the
226 This code may be run before :ref:`instcombine <passes-instcombine>`. In many
227 cases, instcombine checks for the same kinds of things and turns instructions
228 with undefined behavior into unreachable (or equivalent). Because of this,
229 this pass makes some effort to look through bitcasts and so on.
231 ``-loops``: Natural Loop Information
232 ------------------------------------
234 This analysis is used to identify natural loops and determine the loop depth of
235 various nodes of the CFG. Note that the loops identified may actually be
236 several natural loops that share the same header node... not just a single
239 ``-memdep``: Memory Dependence Analysis
240 ---------------------------------------
242 An analysis that determines, for a given memory operation, what preceding
243 memory operations it depends on. It builds on alias analysis information, and
244 tries to provide a lazy, caching interface to a common kind of alias
247 ``-module-debuginfo``: Decodes module-level debug info
248 ------------------------------------------------------
250 This pass decodes the debug info metadata in a module and prints in a
251 (sufficiently-prepared-) human-readable form.
253 For example, run this pass from ``opt`` along with the ``-analyze`` option, and
254 it'll print to standard output.
256 ``-no-aa``: No Alias Analysis (always returns 'may' alias)
257 ----------------------------------------------------------
259 This is the default implementation of the Alias Analysis interface. It always
260 returns "I don't know" for alias queries. NoAA is unlike other alias analysis
261 implementations, in that it does not chain to a previous analysis. As such it
262 doesn't follow many of the rules that other alias analyses must.
264 ``-no-profile``: No Profile Information
265 ---------------------------------------
267 The default "no profile" implementation of the abstract ``ProfileInfo``
270 ``-postdomfrontier``: Post-Dominance Frontier Construction
271 ----------------------------------------------------------
273 This pass is a simple post-dominator construction algorithm for finding
274 post-dominator frontiers.
276 ``-postdomtree``: Post-Dominator Tree Construction
277 --------------------------------------------------
279 This pass is a simple post-dominator construction algorithm for finding
282 ``-print-alias-sets``: Alias Set Printer
283 ----------------------------------------
287 ``-print-callgraph``: Print a call graph
288 ----------------------------------------
290 This pass, only available in ``opt``, prints the call graph to standard error
291 in a human-readable form.
293 ``-print-callgraph-sccs``: Print SCCs of the Call Graph
294 -------------------------------------------------------
296 This pass, only available in ``opt``, prints the SCCs of the call graph to
297 standard error in a human-readable form.
299 ``-print-cfg-sccs``: Print SCCs of each function CFG
300 ----------------------------------------------------
302 This pass, only available in ``opt``, printsthe SCCs of each function CFG to
303 standard error in a human-readable fom.
305 ``-print-dbginfo``: Print debug info in human readable form
306 -----------------------------------------------------------
308 Pass that prints instructions, and associated debug info:
310 #. source/line/col information
311 #. original variable name
312 #. original type name
314 ``-print-dom-info``: Dominator Info Printer
315 -------------------------------------------
317 Dominator Info Printer.
319 ``-print-externalfnconstants``: Print external fn callsites passed constants
320 ----------------------------------------------------------------------------
322 This pass, only available in ``opt``, prints out call sites to external
323 functions that are called with constant arguments. This can be useful when
324 looking for standard library functions we should constant fold or handle in
327 ``-print-function``: Print function to stderr
328 ---------------------------------------------
330 The ``PrintFunctionPass`` class is designed to be pipelined with other
331 ``FunctionPasses``, and prints out the functions of the module as they are
334 ``-print-module``: Print module to stderr
335 -----------------------------------------
337 This pass simply prints out the entire module when it is executed.
339 .. _passes-print-used-types:
341 ``-print-used-types``: Find Used Types
342 --------------------------------------
344 This pass is used to seek out all of the types in use by the program. Note
345 that this analysis explicitly does not include types only used by the symbol
348 ``-profile-estimator``: Estimate profiling information
349 ------------------------------------------------------
351 Profiling information that estimates the profiling information in a very crude
352 and unimaginative way.
354 ``-profile-loader``: Load profile information from ``llvmprof.out``
355 -------------------------------------------------------------------
357 A concrete implementation of profiling information that loads the information
358 from a profile dump file.
360 ``-profile-verifier``: Verify profiling information
361 ---------------------------------------------------
363 Pass that checks profiling information for plausibility.
365 ``-regions``: Detect single entry single exit regions
366 -----------------------------------------------------
368 The ``RegionInfo`` pass detects single entry single exit regions in a function,
369 where a region is defined as any subgraph that is connected to the remaining
370 graph at only two spots. Furthermore, an hierarchical region tree is built.
372 ``-scalar-evolution``: Scalar Evolution Analysis
373 ------------------------------------------------
375 The ``ScalarEvolution`` analysis can be used to analyze and catagorize scalar
376 expressions in loops. It specializes in recognizing general induction
377 variables, representing them with the abstract and opaque ``SCEV`` class.
378 Given this analysis, trip counts of loops and other important properties can be
381 This analysis is primarily useful for induction variable substitution and
384 ``-scev-aa``: ScalarEvolution-based Alias Analysis
385 --------------------------------------------------
387 Simple alias analysis implemented in terms of ``ScalarEvolution`` queries.
389 This differs from traditional loop dependence analysis in that it tests for
390 dependencies within a single iteration of a loop, rather than dependencies
391 between different iterations.
393 ``ScalarEvolution`` has a more complete understanding of pointer arithmetic
394 than ``BasicAliasAnalysis``' collection of ad-hoc analyses.
396 ``-targetdata``: Target Data Layout
397 -----------------------------------
399 Provides other passes access to information on how the size and alignment
400 required by the target ABI for various data types.
405 This section describes the LLVM Transform Passes.
407 ``-adce``: Aggressive Dead Code Elimination
408 -------------------------------------------
410 ADCE aggressively tries to eliminate code. This pass is similar to :ref:`DCE
411 <passes-dce>` but it assumes that values are dead until proven otherwise. This
412 is similar to :ref:`SCCP <passes-sccp>`, except applied to the liveness of
415 ``-always-inline``: Inliner for ``always_inline`` functions
416 -----------------------------------------------------------
418 A custom inliner that handles only functions that are marked as "always
421 ``-argpromotion``: Promote 'by reference' arguments to scalars
422 --------------------------------------------------------------
424 This pass promotes "by reference" arguments to be "by value" arguments. In
425 practice, this means looking for internal functions that have pointer
426 arguments. If it can prove, through the use of alias analysis, that an
427 argument is *only* loaded, then it can pass the value into the function instead
428 of the address of the value. This can cause recursive simplification of code
429 and lead to the elimination of allocas (especially in C++ template code like
432 This pass also handles aggregate arguments that are passed into a function,
433 scalarizing them if the elements of the aggregate are only loaded. Note that
434 it refuses to scalarize aggregates which would require passing in more than
435 three operands to the function, because passing thousands of operands for a
436 large array or structure is unprofitable!
438 Note that this transformation could also be done for arguments that are only
439 stored to (returning the value instead), but does not currently. This case
440 would be best handled when and if LLVM starts supporting multiple return values
443 ``-bb-vectorize``: Basic-Block Vectorization
444 --------------------------------------------
446 This pass combines instructions inside basic blocks to form vector
447 instructions. It iterates over each basic block, attempting to pair compatible
448 instructions, repeating this process until no additional pairs are selected for
449 vectorization. When the outputs of some pair of compatible instructions are
450 used as inputs by some other pair of compatible instructions, those pairs are
451 part of a potential vectorization chain. Instruction pairs are only fused into
452 vector instructions when they are part of a chain longer than some threshold
453 length. Moreover, the pass attempts to find the best possible chain for each
454 pair of compatible instructions. These heuristics are intended to prevent
455 vectorization in cases where it would not yield a performance increase of the
458 ``-block-placement``: Profile Guided Basic Block Placement
459 ----------------------------------------------------------
461 This pass is a very simple profile guided basic block placement algorithm. The
462 idea is to put frequently executed blocks together at the start of the function
463 and hopefully increase the number of fall-through conditional branches. If
464 there is no profile information for a particular function, this pass basically
465 orders blocks in depth-first order.
467 ``-break-crit-edges``: Break critical edges in CFG
468 --------------------------------------------------
470 Break all of the critical edges in the CFG by inserting a dummy basic block.
471 It may be "required" by passes that cannot deal with critical edges. This
472 transformation obviously invalidates the CFG, but can update forward dominator
473 (set, immediate dominators, tree, and frontier) information.
475 ``-codegenprepare``: Optimize for code generation
476 -------------------------------------------------
478 This pass munges the code in the input function to better prepare it for
479 SelectionDAG-based code generation. This works around limitations in its
480 basic-block-at-a-time approach. It should eventually be removed.
482 ``-constmerge``: Merge Duplicate Global Constants
483 -------------------------------------------------
485 Merges duplicate global constants together into a single constant that is
486 shared. This is useful because some passes (i.e., TraceValues) insert a lot of
487 string constants into the program, regardless of whether or not an existing
490 ``-constprop``: Simple constant propagation
491 -------------------------------------------
493 This pass implements constant propagation and merging. It looks for
494 instructions involving only constant operands and replaces them with a constant
495 value instead of an instruction. For example:
507 NOTE: this pass has a habit of making definitions be dead. It is a good idea
508 to run a :ref:`Dead Instruction Elimination <passes-die>` pass sometime after
513 ``-dce``: Dead Code Elimination
514 -------------------------------
516 Dead code elimination is similar to :ref:`dead instruction elimination
517 <passes-die>`, but it rechecks instructions that were used by removed
518 instructions to see if they are newly dead.
520 ``-deadargelim``: Dead Argument Elimination
521 -------------------------------------------
523 This pass deletes dead arguments from internal functions. Dead argument
524 elimination removes arguments which are directly dead, as well as arguments
525 only passed into function calls as dead arguments of other functions. This
526 pass also deletes dead arguments in a similar way.
528 This pass is often useful as a cleanup pass to run after aggressive
529 interprocedural passes, which add possibly-dead arguments.
531 ``-deadtypeelim``: Dead Type Elimination
532 ----------------------------------------
534 This pass is used to cleanup the output of GCC. It eliminate names for types
535 that are unused in the entire translation unit, using the :ref:`find used types
536 <passes-print-used-types>` pass.
540 ``-die``: Dead Instruction Elimination
541 --------------------------------------
543 Dead instruction elimination performs a single pass over the function, removing
544 instructions that are obviously dead.
546 ``-dse``: Dead Store Elimination
547 --------------------------------
549 A trivial dead store elimination that only considers basic-block local
552 ``-functionattrs``: Deduce function attributes
553 ----------------------------------------------
555 A simple interprocedural pass which walks the call-graph, looking for functions
556 which do not access or only read non-local memory, and marking them
557 ``readnone``/``readonly``. In addition, it marks function arguments (of
558 pointer type) "``nocapture``" if a call to the function does not create any
559 copies of the pointer value that outlive the call. This more or less means
560 that the pointer is only dereferenced, and not returned from the function or
561 stored in a global. This pass is implemented as a bottom-up traversal of the
564 ``-globaldce``: Dead Global Elimination
565 ---------------------------------------
567 This transform is designed to eliminate unreachable internal globals from the
568 program. It uses an aggressive algorithm, searching out globals that are known
569 to be alive. After it finds all of the globals which are needed, it deletes
570 whatever is left over. This allows it to delete recursive chunks of the
571 program which are unreachable.
573 ``-globalopt``: Global Variable Optimizer
574 -----------------------------------------
576 This pass transforms simple global variables that never have their address
577 taken. If obviously true, it marks read/write globals as constant, deletes
578 variables only stored to, etc.
580 ``-gvn``: Global Value Numbering
581 --------------------------------
583 This pass performs global value numbering to eliminate fully and partially
584 redundant instructions. It also performs redundant load elimination.
588 ``-indvars``: Canonicalize Induction Variables
589 ----------------------------------------------
591 This transformation analyzes and transforms the induction variables (and
592 computations derived from them) into simpler forms suitable for subsequent
593 analysis and transformation.
595 This transformation makes the following changes to each loop with an
596 identifiable induction variable:
598 * All loops are transformed to have a *single* canonical induction variable
599 which starts at zero and steps by one.
600 * The canonical induction variable is guaranteed to be the first PHI node in
601 the loop header block.
602 * Any pointer arithmetic recurrences are raised to use array subscripts.
604 If the trip count of a loop is computable, this pass also makes the following
607 * The exit condition for the loop is canonicalized to compare the induction
608 value against the exit value. This turns loops like:
612 for (i = 7; i*i < 1000; ++i)
618 for (i = 0; i != 25; ++i)
620 * Any use outside of the loop of an expression derived from the indvar is
621 changed to compute the derived value outside of the loop, eliminating the
622 dependence on the exit value of the induction variable. If the only purpose
623 of the loop is to compute the exit value of some derived expression, this
624 transformation will make the loop dead.
626 This transformation should be followed by strength reduction after all of the
627 desired loop transformations have been performed. Additionally, on targets
628 where it is profitable, the loop could be transformed to count down to zero
629 (the "do loop" optimization).
631 ``-inline``: Function Integration/Inlining
632 ------------------------------------------
634 Bottom-up inlining of functions into callees.
636 ``-insert-edge-profiling``: Insert instrumentation for edge profiling
637 ---------------------------------------------------------------------
639 This pass instruments the specified program with counters for edge profiling.
640 Edge profiling can give a reasonable approximation of the hot paths through a
641 program, and is used for a wide variety of program transformations.
643 Note that this implementation is very naïve. It inserts a counter for *every*
644 edge in the program, instead of using control flow information to prune the
645 number of counters inserted.
647 ``-insert-optimal-edge-profiling``: Insert optimal instrumentation for edge profiling
648 -------------------------------------------------------------------------------------
650 This pass instruments the specified program with counters for edge profiling.
651 Edge profiling can give a reasonable approximation of the hot paths through a
652 program, and is used for a wide variety of program transformations.
654 .. _passes-instcombine:
656 ``-instcombine``: Combine redundant instructions
657 ------------------------------------------------
659 Combine instructions to form fewer, simple instructions. This pass does not
660 modify the CFG This pass is where algebraic simplification happens.
662 This pass combines things like:
675 This is a simple worklist driven algorithm.
677 This pass guarantees that the following canonicalizations are performed on the
680 #. If a binary operator has a constant operand, it is moved to the right-hand
682 #. Bitwise operators with constant operands are always grouped so that shifts
683 are performed first, then ``or``\ s, then ``and``\ s, then ``xor``\ s.
684 #. Compare instructions are converted from ``<``, ``>``, ``≤``, or ``≥`` to
685 ``=`` or ``≠`` if possible.
686 #. All ``cmp`` instructions on boolean values are replaced with logical
688 #. ``add X, X`` is represented as ``mul X, 2`` ⇒ ``shl X, 1``
689 #. Multiplies with a constant power-of-two argument are transformed into
693 ``-internalize``: Internalize Global Symbols
694 --------------------------------------------
696 This pass loops over all of the functions in the input module, looking for a
697 main function. If a main function is found, all other functions and all global
698 variables with initializers are marked as internal.
700 ``-ipconstprop``: Interprocedural constant propagation
701 ------------------------------------------------------
703 This pass implements an *extremely* simple interprocedural constant propagation
704 pass. It could certainly be improved in many different ways, like using a
705 worklist. This pass makes arguments dead, but does not remove them. The
706 existing dead argument elimination pass should be run after this to clean up
709 ``-ipsccp``: Interprocedural Sparse Conditional Constant Propagation
710 --------------------------------------------------------------------
712 An interprocedural variant of :ref:`Sparse Conditional Constant Propagation
715 ``-jump-threading``: Jump Threading
716 -----------------------------------
718 Jump threading tries to find distinct threads of control flow running through a
719 basic block. This pass looks at blocks that have multiple predecessors and
720 multiple successors. If one or more of the predecessors of the block can be
721 proven to always cause a jump to one of the successors, we forward the edge
722 from the predecessor to the successor by duplicating the contents of this
725 An example of when this can occur is code like this:
734 In this case, the unconditional branch at the end of the first if can be
735 revectored to the false side of the second if.
737 ``-lcssa``: Loop-Closed SSA Form Pass
738 -------------------------------------
740 This pass transforms loops by placing phi nodes at the end of the loops for all
741 values that are live across the loop boundary. For example, it turns the left
751 X3 = phi(X1, X2) X3 = phi(X1, X2)
752 ... = X3 + 4 X4 = phi(X3)
755 This is still valid LLVM; the extra phi nodes are purely redundant, and will be
756 trivially eliminated by ``InstCombine``. The major benefit of this
757 transformation is that it makes many other loop optimizations, such as
758 ``LoopUnswitch``\ ing, simpler.
762 ``-licm``: Loop Invariant Code Motion
763 -------------------------------------
765 This pass performs loop invariant code motion, attempting to remove as much
766 code from the body of a loop as possible. It does this by either hoisting code
767 into the preheader block, or by sinking code to the exit blocks if it is safe.
768 This pass also promotes must-aliased memory locations in the loop to live in
769 registers, thus hoisting and sinking "invariant" loads and stores.
771 This pass uses alias analysis for two purposes:
773 #. Moving loop invariant loads and calls out of loops. If we can determine
774 that a load or call inside of a loop never aliases anything stored to, we
775 can hoist it or sink it like any other instruction.
777 #. Scalar Promotion of Memory. If there is a store instruction inside of the
778 loop, we try to move the store to happen AFTER the loop instead of inside of
779 the loop. This can only happen if a few conditions are true:
781 #. The pointer stored through is loop invariant.
782 #. There are no stores or loads in the loop which *may* alias the pointer.
783 There are no calls in the loop which mod/ref the pointer.
785 If these conditions are true, we can promote the loads and stores in the
786 loop of the pointer to use a temporary alloca'd variable. We then use the
787 :ref:`mem2reg <passes-mem2reg>` functionality to construct the appropriate
788 SSA form for the variable.
790 ``-loop-deletion``: Delete dead loops
791 -------------------------------------
793 This file implements the Dead Loop Deletion Pass. This pass is responsible for
794 eliminating loops with non-infinite computable trip counts that have no side
795 effects or volatile instructions, and do not contribute to the computation of
796 the function's return value.
798 .. _passes-loop-extract:
800 ``-loop-extract``: Extract loops into new functions
801 ---------------------------------------------------
803 A pass wrapper around the ``ExtractLoop()`` scalar transformation to extract
804 each top-level loop into its own new function. If the loop is the *only* loop
805 in a given function, it is not touched. This is a pass most useful for
806 debugging via bugpoint.
808 ``-loop-extract-single``: Extract at most one loop into a new function
809 ----------------------------------------------------------------------
811 Similar to :ref:`Extract loops into new functions <passes-loop-extract>`, this
812 pass extracts one natural loop from the program into a function if it can.
813 This is used by :program:`bugpoint`.
815 ``-loop-reduce``: Loop Strength Reduction
816 -----------------------------------------
818 This pass performs a strength reduction on array references inside loops that
819 have as one or more of their components the loop induction variable. This is
820 accomplished by creating a new value to hold the initial value of the array
821 access for the first iteration, and then creating a new GEP instruction in the
822 loop to increment the value by the appropriate amount.
824 ``-loop-rotate``: Rotate Loops
825 ------------------------------
827 A simple loop rotation transformation.
829 ``-loop-simplify``: Canonicalize natural loops
830 ----------------------------------------------
832 This pass performs several transformations to transform natural loops into a
833 simpler form, which makes subsequent analyses and transformations simpler and
836 Loop pre-header insertion guarantees that there is a single, non-critical entry
837 edge from outside of the loop to the loop header. This simplifies a number of
838 analyses and transformations, such as :ref:`LICM <passes-licm>`.
840 Loop exit-block insertion guarantees that all exit blocks from the loop (blocks
841 which are outside of the loop that have predecessors inside of the loop) only
842 have predecessors from inside of the loop (and are thus dominated by the loop
843 header). This simplifies transformations such as store-sinking that are built
846 This pass also guarantees that loops will have exactly one backedge.
848 Note that the :ref:`simplifycfg <passes-simplifycfg>` pass will clean up blocks
849 which are split out but end up being unnecessary, so usage of this pass should
850 not pessimize generated code.
852 This pass obviously modifies the CFG, but updates loop information and
853 dominator information.
855 ``-loop-unroll``: Unroll loops
856 ------------------------------
858 This pass implements a simple loop unroller. It works best when loops have
859 been canonicalized by the :ref:`indvars <passes-indvars>` pass, allowing it to
860 determine the trip counts of loops easily.
862 ``-loop-unswitch``: Unswitch loops
863 ----------------------------------
865 This pass transforms loops that contain branches on loop-invariant conditions
866 to have multiple loops. For example, it turns the left into the right code:
877 This can increase the size of the code exponentially (doubling it every time a
878 loop is unswitched) so we only unswitch if the resultant code will be smaller
881 This pass expects :ref:`LICM <passes-licm>` to be run before it to hoist
882 invariant conditions out of the loop, to make the unswitching opportunity
885 ``-loweratomic``: Lower atomic intrinsics to non-atomic form
886 ------------------------------------------------------------
888 This pass lowers atomic intrinsics to non-atomic form for use in a known
889 non-preemptible environment.
891 The pass does not verify that the environment is non-preemptible (in general
892 this would require knowledge of the entire call graph of the program including
893 any libraries which may not be available in bitcode form); it simply lowers
894 every atomic intrinsic.
896 ``-lowerinvoke``: Lower invoke and unwind, for unwindless code generators
897 -------------------------------------------------------------------------
899 This transformation is designed for use by code generators which do not yet
900 support stack unwinding. This pass supports two models of exception handling
901 lowering, the "cheap" support and the "expensive" support.
903 "Cheap" exception handling support gives the program the ability to execute any
904 program which does not "throw an exception", by turning "``invoke``"
905 instructions into calls and by turning "``unwind``" instructions into calls to
906 ``abort()``. If the program does dynamically use the "``unwind``" instruction,
907 the program will print a message then abort.
909 "Expensive" exception handling support gives the full exception handling
910 support to the program at the cost of making the "``invoke``" instruction
911 really expensive. It basically inserts ``setjmp``/``longjmp`` calls to emulate
912 the exception handling as necessary.
914 Because the "expensive" support slows down programs a lot, and EH is only used
915 for a subset of the programs, it must be specifically enabled by the
916 ``-enable-correct-eh-support`` option.
918 Note that after this pass runs the CFG is not entirely accurate (exceptional
919 control flow edges are not correct anymore) so only very simple things should
920 be done after the ``lowerinvoke`` pass has run (like generation of native
921 code). This should not be used as a general purpose "my LLVM-to-LLVM pass
922 doesn't support the ``invoke`` instruction yet" lowering pass.
924 ``-lowerswitch``: Lower ``SwitchInst``\ s to branches
925 -----------------------------------------------------
927 Rewrites switch instructions with a sequence of branches, which allows targets
928 to get away with not implementing the switch instruction until it is
933 ``-mem2reg``: Promote Memory to Register
934 ----------------------------------------
936 This file promotes memory references to be register references. It promotes
937 alloca instructions which only have loads and stores as uses. An ``alloca`` is
938 transformed by using dominator frontiers to place phi nodes, then traversing
939 the function in depth-first order to rewrite loads and stores as appropriate.
940 This is just the standard SSA construction algorithm to construct "pruned" SSA
943 ``-memcpyopt``: MemCpy Optimization
944 -----------------------------------
946 This pass performs various transformations related to eliminating ``memcpy``
947 calls, or transforming sets of stores into ``memset``\ s.
949 ``-mergefunc``: Merge Functions
950 -------------------------------
952 This pass looks for equivalent functions that are mergable and folds them.
954 A hash is computed from the function, based on its type and number of basic
957 Once all hashes are computed, we perform an expensive equality comparison on
958 each function pair. This takes n^2/2 comparisons per bucket, so it's important
959 that the hash function be high quality. The equality comparison iterates
960 through each instruction in each basic block.
962 When a match is found the functions are folded. If both functions are
963 overridable, we move the functionality into a new internal function and leave
964 two overridable thunks to it.
966 ``-mergereturn``: Unify function exit nodes
967 -------------------------------------------
969 Ensure that functions have at most one ``ret`` instruction in them.
970 Additionally, it keeps track of which node is the new exit node of the CFG.
972 ``-partial-inliner``: Partial Inliner
973 -------------------------------------
975 This pass performs partial inlining, typically by inlining an ``if`` statement
976 that surrounds the body of the function.
978 ``-prune-eh``: Remove unused exception handling info
979 ----------------------------------------------------
981 This file implements a simple interprocedural pass which walks the call-graph,
982 turning invoke instructions into call instructions if and only if the callee
983 cannot throw an exception. It implements this as a bottom-up traversal of the
986 ``-reassociate``: Reassociate expressions
987 -----------------------------------------
989 This pass reassociates commutative expressions in an order that is designed to
990 promote better constant propagation, GCSE, :ref:`LICM <passes-licm>`, PRE, etc.
992 For example: 4 + (x + 5) ⇒ x + (4 + 5)
994 In the implementation of this algorithm, constants are assigned rank = 0,
995 function arguments are rank = 1, and other values are assigned ranks
996 corresponding to the reverse post order traversal of current function (starting
997 at 2), which effectively gives values in deep loops higher rank than values not
1000 ``-reg2mem``: Demote all values to stack slots
1001 ----------------------------------------------
1003 This file demotes all registers to memory references. It is intended to be the
1004 inverse of :ref:`mem2reg <passes-mem2reg>`. By converting to ``load``
1005 instructions, the only values live across basic blocks are ``alloca``
1006 instructions and ``load`` instructions before ``phi`` nodes. It is intended
1007 that this should make CFG hacking much easier. To make later hacking easier,
1008 the entry block is split into two, such that all introduced ``alloca``
1009 instructions (and nothing else) are in the entry block.
1011 ``-scalarrepl``: Scalar Replacement of Aggregates (DT)
1012 ------------------------------------------------------
1014 The well-known scalar replacement of aggregates transformation. This transform
1015 breaks up ``alloca`` instructions of aggregate type (structure or array) into
1016 individual ``alloca`` instructions for each member if possible. Then, if
1017 possible, it transforms the individual ``alloca`` instructions into nice clean
1020 This combines a simple scalar replacement of aggregates algorithm with the
1021 :ref:`mem2reg <passes-mem2reg>` algorithm because they often interact,
1022 especially for C++ programs. As such, iterating between ``scalarrepl``, then
1023 :ref:`mem2reg <passes-mem2reg>` until we run out of things to promote works
1028 ``-sccp``: Sparse Conditional Constant Propagation
1029 --------------------------------------------------
1031 Sparse conditional constant propagation and merging, which can be summarized
1034 * Assumes values are constant unless proven otherwise
1035 * Assumes BasicBlocks are dead unless proven otherwise
1036 * Proves values to be constant, and replaces them with constants
1037 * Proves conditional branches to be unconditional
1039 Note that this pass has a habit of making definitions be dead. It is a good
1040 idea to run a :ref:`DCE <passes-dce>` pass sometime after running this pass.
1042 ``-simplify-libcalls``: Simplify well-known library calls
1043 ---------------------------------------------------------
1045 Applies a variety of small optimizations for calls to specific well-known
1046 function calls (e.g. runtime library functions). For example, a call
1047 ``exit(3)`` that occurs within the ``main()`` function can be transformed into
1048 simply ``return 3``.
1050 .. _passes-simplifycfg:
1052 ``-simplifycfg``: Simplify the CFG
1053 ----------------------------------
1055 Performs dead code elimination and basic block merging. Specifically:
1057 * Removes basic blocks with no predecessors.
1058 * Merges a basic block into its predecessor if there is only one and the
1059 predecessor only has one successor.
1060 * Eliminates PHI nodes for basic blocks with a single predecessor.
1061 * Eliminates a basic block that only contains an unconditional branch.
1063 ``-sink``: Code sinking
1064 -----------------------
1066 This pass moves instructions into successor blocks, when possible, so that they
1067 aren't executed on paths where their results aren't needed.
1069 ``-strip``: Strip all symbols from a module
1070 -------------------------------------------
1072 Performs code stripping. This transformation can delete:
1074 * names for virtual registers
1075 * symbols for internal globals and functions
1078 Note that this transformation makes code much less readable, so it should only
1079 be used in situations where the strip utility would be used, such as reducing
1080 code size or making it harder to reverse engineer code.
1082 ``-strip-dead-debug-info``: Strip debug info for unused symbols
1083 ---------------------------------------------------------------
1085 .. FIXME: this description is the same as for -strip
1087 performs code stripping. this transformation can delete:
1089 * names for virtual registers
1090 * symbols for internal globals and functions
1093 note that this transformation makes code much less readable, so it should only
1094 be used in situations where the strip utility would be used, such as reducing
1095 code size or making it harder to reverse engineer code.
1097 ``-strip-dead-prototypes``: Strip Unused Function Prototypes
1098 ------------------------------------------------------------
1100 This pass loops over all of the functions in the input module, looking for dead
1101 declarations and removes them. Dead declarations are declarations of functions
1102 for which no implementation is available (i.e., declarations for unused library
1105 ``-strip-debug-declare``: Strip all ``llvm.dbg.declare`` intrinsics
1106 -------------------------------------------------------------------
1108 .. FIXME: this description is the same as for -strip
1110 This pass implements code stripping. Specifically, it can delete:
1112 #. names for virtual registers
1113 #. symbols for internal globals and functions
1114 #. debug information
1116 Note that this transformation makes code much less readable, so it should only
1117 be used in situations where the 'strip' utility would be used, such as reducing
1118 code size or making it harder to reverse engineer code.
1120 ``-strip-nondebug``: Strip all symbols, except dbg symbols, from a module
1121 -------------------------------------------------------------------------
1123 .. FIXME: this description is the same as for -strip
1125 This pass implements code stripping. Specifically, it can delete:
1127 #. names for virtual registers
1128 #. symbols for internal globals and functions
1129 #. debug information
1131 Note that this transformation makes code much less readable, so it should only
1132 be used in situations where the 'strip' utility would be used, such as reducing
1133 code size or making it harder to reverse engineer code.
1135 ``-tailcallelim``: Tail Call Elimination
1136 ----------------------------------------
1138 This file transforms calls of the current function (self recursion) followed by
1139 a return instruction with a branch to the entry of the function, creating a
1140 loop. This pass also implements the following extensions to the basic
1143 #. Trivial instructions between the call and return do not prevent the
1144 transformation from taking place, though currently the analysis cannot
1145 support moving any really useful instructions (only dead ones).
1146 #. This pass transforms functions that are prevented from being tail recursive
1147 by an associative expression to use an accumulator variable, thus compiling
1148 the typical naive factorial or fib implementation into efficient code.
1149 #. TRE is performed if the function returns void, if the return returns the
1150 result returned by the call, or if the function returns a run-time constant
1151 on all exits from the function. It is possible, though unlikely, that the
1152 return returns something else (like constant 0), and can still be TRE'd. It
1153 can be TRE'd if *all other* return instructions in the function return the
1155 #. If it can prove that callees do not access theier caller stack frame, they
1156 are marked as eligible for tail call elimination (by the code generator).
1161 This section describes the LLVM Utility Passes.
1163 ``-deadarghaX0r``: Dead Argument Hacking (BUGPOINT USE ONLY; DO NOT USE)
1164 ------------------------------------------------------------------------
1166 Same as dead argument elimination, but deletes arguments to functions which are
1167 external. This is only for use by :doc:`bugpoint <Bugpoint>`.
1169 ``-extract-blocks``: Extract Basic Blocks From Module (for bugpoint use)
1170 ------------------------------------------------------------------------
1172 This pass is used by bugpoint to extract all blocks from the module into their
1175 ``-instnamer``: Assign names to anonymous instructions
1176 ------------------------------------------------------
1178 This is a little utility pass that gives instructions names, this is mostly
1179 useful when diffing the effect of an optimization because deleting an unnamed
1180 instruction can change all other instruction numbering, making the diff very
1183 ``-preverify``: Preliminary module verification
1184 -----------------------------------------------
1186 Ensures that the module is in the form required by the :ref:`Module Verifier
1187 <passes-verify>` pass. Running the verifier runs this pass automatically, so
1188 there should be no need to use it directly.
1192 ``-verify``: Module Verifier
1193 ----------------------------
1195 Verifies an LLVM IR code. This is useful to run after an optimization which is
1196 undergoing testing. Note that llvm-as verifies its input before emitting
1197 bitcode, and also that malformed bitcode is likely to make LLVM crash. All
1198 language front-ends are therefore encouraged to verify their output before
1199 performing optimizing transformations.
1201 #. Both of a binary operator's parameters are of the same type.
1202 #. Verify that the indices of mem access instructions match other operands.
1203 #. Verify that arithmetic and other things are only performed on first-class
1204 types. Verify that shifts and logicals only happen on integrals f.e.
1205 #. All of the constants in a switch statement are of the correct type.
1206 #. The code is in valid SSA form.
1207 #. It is illegal to put a label into any other type (like a structure) or to
1209 #. Only phi nodes can be self referential: ``%x = add i32 %x``, ``%x`` is
1211 #. PHI nodes must have an entry for each predecessor, with no extras.
1212 #. PHI nodes must be the first thing in a basic block, all grouped together.
1213 #. PHI nodes must have at least one entry.
1214 #. All basic blocks should only end with terminator insts, not contain them.
1215 #. The entry node to a function must not have predecessors.
1216 #. All Instructions must be embedded into a basic block.
1217 #. Functions cannot take a void-typed parameter.
1218 #. Verify that a function's argument list agrees with its declared type.
1219 #. It is illegal to specify a name for a void value.
1220 #. It is illegal to have an internal global value with no initializer.
1221 #. It is illegal to have a ``ret`` instruction that returns a value that does
1222 not agree with the function return value type.
1223 #. Function call argument types match the function prototype.
1224 #. All other things that are tested by asserts spread about the code.
1226 Note that this does not provide full security verification (like Java), but
1227 instead just tries to ensure that code is well-formed.
1229 ``-view-cfg``: View CFG of function
1230 -----------------------------------
1232 Displays the control flow graph using the GraphViz tool.
1234 ``-view-cfg-only``: View CFG of function (with no function bodies)
1235 ------------------------------------------------------------------
1237 Displays the control flow graph using the GraphViz tool, but omitting function
1240 ``-view-dom``: View dominance tree of function
1241 ----------------------------------------------
1243 Displays the dominator tree using the GraphViz tool.
1245 ``-view-dom-only``: View dominance tree of function (with no function bodies)
1246 -----------------------------------------------------------------------------
1248 Displays the dominator tree using the GraphViz tool, but omitting function
1251 ``-view-postdom``: View postdominance tree of function
1252 ------------------------------------------------------
1254 Displays the post dominator tree using the GraphViz tool.
1256 ``-view-postdom-only``: View postdominance tree of function (with no function bodies)
1257 -------------------------------------------------------------------------------------
1259 Displays the post dominator tree using the GraphViz tool, but omitting function