Commit Graph

653 Commits

Author SHA1 Message Date
Chris Lattner
40d31ecb29 simplify code
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@30656 91177308-0d34-0410-b5e6-96231b3b80d8
2006-09-28 22:58:25 +00:00
Chris Lattner
120fce5540 Be far more careful when splitting a loop header, either to form a preheader
or when splitting loops with a common header into multiple loops.  In particular
the old code would always insert the preheader before the old loop header.  This
is disasterous in cases where the loop hasn't been rotated.  For example, it can
produce code like:

        .. outside the loop...
        jmp LBB1_2      #bb13.outer
LBB1_1: #bb1
        movsd 8(%esp,%esi,8), %xmm1
        mulsd (%edi), %xmm1
        addsd %xmm0, %xmm1
        addl $24, %edi
        incl %esi
        jmp LBB1_3      #bb13
LBB1_2: #bb13.outer
        leal (%edx,%eax,8), %edi
        pxor %xmm1, %xmm1
        xorl %esi, %esi
LBB1_3: #bb13
        movapd %xmm1, %xmm0
        cmpl $4, %esi
        jl LBB1_1       #bb1

Note that the loop body is actually LBB1_1 + LBB1_3, which means that the
loop now contains an uncond branch WITHIN it to jump around the inserted
loop header (LBB1_2).  Doh.

This patch changes the preheader insertion code to insert it in the right
spot, producing this code:

        ... outside the loop, fall into the header ...
LBB1_1: #bb13.outer
        leal (%edx,%eax,8), %esi
        pxor %xmm0, %xmm0
        xorl %edi, %edi
        jmp LBB1_3      #bb13
LBB1_2: #bb1
        movsd 8(%esp,%edi,8), %xmm0
        mulsd (%esi), %xmm0
        addsd %xmm1, %xmm0
        addl $24, %esi
        incl %edi
LBB1_3: #bb13
        movapd %xmm0, %xmm1
        cmpl $4, %edi
        jl LBB1_2       #bb1

Totally crazy, no branch in the loop! :)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@30587 91177308-0d34-0410-b5e6-96231b3b80d8
2006-09-23 08:19:21 +00:00
Chris Lattner
c3984578be Teach UpdateDomInfoForRevectoredPreds to handle revectored preds that are not
reachable, making it general purpose enough for use by InsertPreheaderForLoop.
Eliminate custom dominfo updating code in InsertPreheaderForLoop, using
UpdateDomInfoForRevectoredPreds instead.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@30586 91177308-0d34-0410-b5e6-96231b3b80d8
2006-09-23 07:40:52 +00:00
Chris Lattner
a4646b61e4 Second half of the fix for Transforms/Inline/inline_cleanup.ll
This folds unconditional branches that are often produced by code
specialization.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@30307 91177308-0d34-0410-b5e6-96231b3b80d8
2006-09-13 21:27:00 +00:00
Chris Lattner
33bb3c8be3 Implement the first half of Transforms/Inline/inline_cleanup.ll
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@30303 91177308-0d34-0410-b5e6-96231b3b80d8
2006-09-13 19:23:57 +00:00
Chris Lattner
97d2dbd100 Fix Duraid's changes to work when TLI is null. This fixes the failing
lowerinvoke regtests.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@30115 91177308-0d34-0410-b5e6-96231b3b80d8
2006-09-05 17:48:07 +00:00
Duraid Madina
2a0013f59f add setJumpBufSize() and setJumpBufAlignment() to target-lowering.
Call these from your backend to enjoy setjmp/longjmp goodness, see
lib/Target/IA64/IA64ISelLowering.cpp for an example


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@30095 91177308-0d34-0410-b5e6-96231b3b80d8
2006-09-04 06:21:35 +00:00
Chris Lattner
7f8897f22e eliminate RegisterOpt. It does the same thing as RegisterPass.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29925 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-27 22:42:52 +00:00
Chris Lattner
a4f0b3a084 s|llvm/Support/Visibility.h|llvm/Support/Compiler.h|
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29911 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-27 12:54:02 +00:00
Chris Lattner
3bb4657488 Don't attempt to split subloops out of a loop with a huge number of backedges.
Not only will this take huge amounts of compile time, the resultant loop nests
won't be useful for optimization.  This reduces loopsimplify time on
Transforms/LoopSimplify/2006-08-11-LoopSimplifyLongTime.ll from ~32s to ~0.4s
with a debug build of llvm on a 2.7Ghz G5.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29647 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-12 05:25:00 +00:00
Chris Lattner
fa78946482 Reimplement the loopsimplify code which deletes edges from unreachable
blocks that target loop blocks.

Before, the code was run once per loop, and depended on the number of
predecessors each block in the loop had.  Unfortunately, scanning preds can
be really slow when huge numbers of phis exist or when phis with huge numbers
of inputs exist.

Now, the code is run once per function and scans successors instead of preds,
which is far faster.  In addition, the new code is simpler and is goto free,
woo.

This change speeds up a nasty testcase Duraid provided me from taking hours to
taking ~72s with a debug build.  The functionality this implements is already
tested in the testsuite as Transforms/CodeExtractor/2004-03-13-LoopExtractorCrash.ll.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29644 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-12 04:51:20 +00:00
Chris Lattner
8cfe6335e4 Fix PR867 (and maybe 868) and testcsae:
Transforms/SimplifyCFG/2006-08-03-Crash.ll


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29515 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-03 21:40:24 +00:00
Chris Lattner
f9ba401bf5 Add special check to avoid isLoop call. Simple, but doesn't seem to speed
up lcssa much in practice.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29465 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-02 00:16:47 +00:00
Chris Lattner
d41ae8bc0c Replace the SSA update code in LCSSA with a bottom-up approach instead of a top
down approach, inspired by discussions with Tanya.

This approach is significantly faster, because it does not need dominator
frontiers and it does not insert extraneous unused PHI nodes.  For example, on
252.eon, in a release-asserts build, this speeds up LCSSA (which is the slowest
pass in gccas) from 9.14s to 0.74s on my G5.  This code is also slightly smaller
and significantly simpler than the old code.

Amusingly, in a normal Release build (which includes the
"assert(L->isLCSSAForm());" assertion), asserting that the result of LCSSA
is in LCSSA form is actually slower than the LCSSA transformation pass
itself on 252.eon.  I will see if Loop::isLCSSAForm can be sped up next.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29463 91177308-0d34-0410-b5e6-96231b3b80d8
2006-08-02 00:06:09 +00:00
Chris Lattner
21ba23d0d8 silence warnings in a release build
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29189 91177308-0d34-0410-b5e6-96231b3b80d8
2006-07-18 21:48:57 +00:00
Chris Lattner
27d672136a eliminate some ugly code, using ConstantExpr::getWithOperands instead.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29149 91177308-0d34-0410-b5e6-96231b3b80d8
2006-07-14 22:21:31 +00:00
Chris Lattner
981418bf15 Handle instructions in the map, but that map to a null pointer.
This unbreaks smg2000.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29127 91177308-0d34-0410-b5e6-96231b3b80d8
2006-07-12 21:37:11 +00:00
Chris Lattner
1bb3a40257 In addition to deleting calls, the inliner can constant fold them as well.
Handle this case, which doesn't require a new callgraph edge.  This fixes
a crash compiling MallocBench/gs.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29121 91177308-0d34-0410-b5e6-96231b3b80d8
2006-07-12 18:37:18 +00:00
Chris Lattner
d85340f4ec Change the callgraph representation to store the callsite along with the
target CG node.  This allows the inliner to properly update the callgraph
when using the pruning inliner.  The pruning inliner may not copy over all
call sites from a callee to a caller, so the edges corresponding to those
call sites should not be copied over either.

This fixes PR827 and Transforms/Inline/2006-07-12-InlinePruneCGUpdate.ll


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29120 91177308-0d34-0410-b5e6-96231b3b80d8
2006-07-12 18:29:36 +00:00
Owen Anderson
f8dad7a926 Fix typo in the comment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29078 91177308-0d34-0410-b5e6-96231b3b80d8
2006-07-09 21:35:40 +00:00
Owen Anderson
e4e1ecd37c Add a fix for an issue where LCSSA would fail to insert undef's in some corner
cases.  Ideally, this issue will go away in the future as LCSSA gets smarter
about which Phi nodes it inserts.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@29076 91177308-0d34-0410-b5e6-96231b3b80d8
2006-07-09 08:14:06 +00:00
Chris Lattner
9525528a7d Use hidden visibility to make symbols in an anonymous namespace get
dropped.  This shrinks libllvmgcc.dylib another 67K


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28975 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-28 23:17:24 +00:00
Chris Lattner
f4b546110c Shrink libllvmgcc.dylib by another 23K
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28972 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-28 22:08:15 +00:00
Chris Lattner
13a68e4257 Use the PotDoms map to memoize 'dominating value' lookup. With this patch,
LCSSA is still the slowest pass when gccas'ing 252.eon, but now it only takes
39s instead of 289s. :)


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28776 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-14 01:13:57 +00:00
Owen Anderson
8b92d0cae1 Fix another instance where PHI nodes need special treatment.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28774 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-13 20:50:09 +00:00
Owen Anderson
7be3f1e078 Fix a bug that was causing major slowdowns in povray. This was due to LCSSA
not handling PHI nodes correctly when determining if a value was live-out.

This patch reduces the number of detected live-out variables in the testcase
from 6565 to 485.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28771 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-13 19:37:18 +00:00
Chris Lattner
7f2e1dd5d4 Fix an infinite loop on Transforms/SimplifyCFG/2006-06-12-InfLoop.ll
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28758 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-12 20:18:01 +00:00
Owen Anderson
3d2aa47bd3 Fix for 2006-06-26-MultipleExitsSingleBlock.
If a single exit block has multiple predecessors within the loop, it will
appear in the exit blocks list more than once.  LCSSA needs to take that into
account so that it doesn't double process that exit block.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28750 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-12 07:10:16 +00:00
Owen Anderson
c2cc15cf9d Re-commit the safe parts of my 6/9 patch. Still working on fixing the unsafe parts.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28748 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-11 19:22:28 +00:00
Evan Cheng
b9b2b309d3 Back out Owen's 6/9 changes. They broke MultiSource/Benchmarks/Prolangs-C/bison (and perhaps others).
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28747 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-11 09:32:57 +00:00
Owen Anderson
f25c19c6b5 Make Loop able to verify that it is in LCSSA-form, and have the LCSSA pass assert
on this.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28738 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-09 18:33:30 +00:00
Owen Anderson
a452932171 Update some comments, and expose LCSSAID in preparation for having other passes
require LCSSA.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28734 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-08 20:02:53 +00:00
Owen Anderson
e9d93d5d70 Fix some formatting, and use inLoop() when appropriate.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28694 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-06 04:36:36 +00:00
Owen Anderson
92deacf2f7 Stop a memory leak, and update some comments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28693 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-06 04:28:30 +00:00
Owen Anderson
c14c1d472a Some more clean-up, and squash an IDF-Phi related bug.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28680 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-04 00:55:19 +00:00
Owen Anderson
9e1c1ddd4b Various clean-ups suggested by Chris.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28678 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-04 00:02:23 +00:00
Owen Anderson
30019c88f4 Fix a bug in Phi-noded insertion. Also, update some comments to reflect what's
actually going on.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28677 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-03 23:22:50 +00:00
Chris Lattner
191b0ba97e Force anything that #includes llvm/Transforms/Utils/UnifyFunctionExitNodes.h
to link in the implementation.  Thanks to Anton Korobeynikov for figuring out
what was going on here.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28660 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-02 18:40:06 +00:00
Chris Lattner
ecf626fedf Remove dead #include
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28642 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-01 20:02:28 +00:00
Chris Lattner
35033efd08 Make the "pruning cloner" smarter. As it propagates constants through the
code (while cloning) it often gets the branch/switch instructions.  Since it
knows that edges of the CFG are dead, it need not clone (or even look) at
the obviously dead blocks.  This should speed up the inliner substantially on
code where there are lots of inlinable calls to functions with constant
arguments.  On C++ code in particular, this kicks in.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28641 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-01 19:19:23 +00:00
Owen Anderson
ff99366919 Remove a FIXME that was fixed with my last patch.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28619 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-01 06:07:40 +00:00
Owen Anderson
2824da4738 More cleanups. Also, add a special case for updating PHI nodes, and
reimplement getValueDominatingFunction to walk the DominanceTree rather than
just searching blindly.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28618 91177308-0d34-0410-b5e6-96231b3b80d8
2006-06-01 06:05:47 +00:00
Owen Anderson
408a4061d8 Extract a huge loop into a helper method. Fix a few iterator-invalidation bugs.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28599 91177308-0d34-0410-b5e6-96231b3b80d8
2006-05-31 20:55:06 +00:00
Owen Anderson
00ea74c27e Add Use replacement. Assuming there is nothing horribly wrong with this, LCSSA
is now theoretically feature-complete.  It has not, however, been thoroughly
test, and is still considered experimental.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28529 91177308-0d34-0410-b5e6-96231b3b80d8
2006-05-29 01:00:00 +00:00
Owen Anderson
bd82277cbb Major think-o. Iterate over all live out-of-loop values, and perform the
other calculations on each individually, rather than trying to delay it and do
them all at the end.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28527 91177308-0d34-0410-b5e6-96231b3b80d8
2006-05-28 19:33:28 +00:00
Owen Anderson
2f21e07915 Make LCSSA insert proper Phi nodes throughout the rest of the CFG by computing
the iterated Dominance Frontier of the loop-closure Phi's.  This is the
second phase of the LCSSA pass.  The third phase (coming soon) will be to
update all uses of loop variables to use the loop-closure Phi's instead.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28524 91177308-0d34-0410-b5e6-96231b3b80d8
2006-05-27 18:47:11 +00:00
Chris Lattner
8d8d6530b1 Fix some regression from the inliner patch I committed last night. This fixes
ldecod, lencod, and SPASS.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28523 91177308-0d34-0410-b5e6-96231b3b80d8
2006-05-27 17:28:13 +00:00
Chris Lattner
5b5bc3032f Switch the inliner over to using CloneAndPruneFunctionInto. This effectively
makes it so that it constant folds instructions on the fly.  This is good
for several reasons:

0. Many instructions are constant foldable after inlining, particularly if
   inlining a call with constant arguments.
1. Without this, the inliner has to allocate memory for all of the instructions
   that can be constant folded, then a subsequent pass has to delete them.  This
   gets the job done without this extra work.
2. This makes the inliner *pass* a bit more aggressive: in particular, it
   partially solves a phase order issue where the inliner would inline lots
   of code that folds away to nothing, but think that the resultant function
   is big because of this code that will be gone.  Now the code never exists.

This is the first part of a 2-step process.  The second part will be smart
enough to see when this implicit constant folding propagates a constant into
a branch or switch instruction, making CFG edges dead.

This implements Transforms/Inline/inline_constprop.ll


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28521 91177308-0d34-0410-b5e6-96231b3b80d8
2006-05-27 01:28:04 +00:00
Chris Lattner
83f03bfc3f Implement a new method, CloneAndPruneFunctionInto, as documented.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28519 91177308-0d34-0410-b5e6-96231b3b80d8
2006-05-27 01:22:24 +00:00
Chris Lattner
4b30fcb5f5 Refactor some code to expose an interface to constant fold and instruction given it's opcode, typeand operands.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@28517 91177308-0d34-0410-b5e6-96231b3b80d8
2006-05-27 01:18:04 +00:00