Commit Graph

6683 Commits

Author SHA1 Message Date
Dan Gohman
deff621abd Use getConstant instead of getIntegerSCEV. The two are basically the
same, now that getConstant has overloads consistent with ConstantInt::get.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102965 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-03 22:09:21 +00:00
Devang Patel
86b6f80a4c Check for side effects before splitting loop.
Patch by Jakub Staszak!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102928 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-03 18:06:58 +00:00
Chris Lattner
83f66fe614 revert r102831. We already delete dead readonly calls in
other places, killing a valid transformation is not the right
answer.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102850 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-01 17:19:38 +00:00
Owen Anderson
1b4a38646f Disable the call-deletion transformation introduced in r86975. Without
halting analysis, it is illegal to delete a call to a read-only function.
The correct solution is almost certainly to add a "must halt" attribute and
only allow deletions in its presence.

XFAIL the relevant testcase for now.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102831 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-01 08:34:28 +00:00
Chris Lattner
0ca2f28458 rename InlineInfo.DevirtualizedCalls -> InlinedCalls to
reflect that it includes all inlined calls now, not just
devirtualized ones.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102824 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-01 01:26:13 +00:00
Chris Lattner
6da12e6767 Implement rdar://6295824 and PR6724 with two tiny changes
that can have a big effect :).  The first is to enable the
iterative SCC passmanager juice that kicks in when the
scc passmgr detects that a function pass has devirtualized
a call.  In this case, it will rerun all the passes it 
manages on the SCC, up to the iteration count limit (4). This
is useful because a function pass may devirualize a call, and
we want the inliner to inline it, or pruneeh to infer stuff
about it, etc.

The second patch is to add *all* call sites to the 
DevirtualizedCalls list the inliner uses.  This list is
about to get renamed, but the jist of this is that the 
inliner now reconsiders *all* inlined call sites as candidates
for further inlining.  The intuition is this that in cases 
like this:

f() { g(1); }     g(int x) { h(x); }

We analyze this bottom up, and may decide that it isn't 
profitable to inline H into G.  Next step, we decide that it is
profitable to inline G into F, and do so, which means that F 
now calls H.  Even though the call from G -> H may not have been
profitable to inline, the call from F -> H may be (in this case
because a constant allows folding etc).

In my spot checks, this doesn't have a big impact on code.  For
example, the LLC output for 252.eon grew from 0.02% (from
317252 to 317308) and 176.gcc actually shrunk by .3% (from 1525612
to 1520964 bytes).  252.eon never iterated in the SCC Passmgr,
176.gcc iterated at most 1 time.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102823 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-01 01:15:56 +00:00
Chris Lattner
159528702a The inliner has traditionally not considered call sites
that appear due to inlining a callee as candidates for
futher inlining, but a recent patch made it do this if
those call sites were indirect and became direct.

Unfortunately, in bizarre cases (see testcase) doing this
can cause us to infinitely inline mutually recursive
functions into callers not in the cycle.  Fix this by
keeping track of the inline history from which callsite
inline candidates got inlined from.

This shouldn't affect any "real world" code, but is required
for a follow on patch that is coming up next.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102822 91177308-0d34-0410-b5e6-96231b3b80d8
2010-05-01 01:05:10 +00:00
Devang Patel
0aa885d3ab Preserve debug info attached with call instruction while eliminating dead argument.
Radar 7927803


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102760 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-30 20:23:54 +00:00
Chris Lattner
2d4b8ee1d9 further clarify alignment of globals, fix instcombine
to not increase the alignment of globals with an assigned
alignment and section.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102476 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-28 00:31:12 +00:00
Chris Lattner
c6f0aadc3b Fix a problem that lower invoke has with allocas (PR6694), and
add a version of createLowerInvokePass that allows the client
to specify whether it wants "expensive" or "cheap" lowering.

Patch by Alex Mac!


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102402 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-26 23:49:32 +00:00
Chris Lattner
8581c260d3 fix PR6940: sitofp(undef) folds to 0.0, not undef.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102358 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-26 18:21:23 +00:00
Chris Lattner
d89673c028 remove #if 1's.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102296 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-25 04:43:02 +00:00
Dan Gohman
1d367988e2 Generalize LSR's OptimizeMax to handle the new kinds of max expressions
that indvars may use, now that indvars is recognizing le and ge loops.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102235 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-24 03:13:44 +00:00
Chris Lattner
5d7da43215 enable my inliner change: add newly devirtualized call sites to
the worklist, making them inline candidates.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102213 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-23 21:16:07 +00:00
Chris Lattner
076863225c switch InlineInfo.DevirtualizedCalls's list to be of WeakVH.
This fixes a bug where calls inlined into an invoke would get
changed into an invoke but the array would keep pointing to
the (now dead) call.  The improved inliner behavior is still
disabled for now.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102196 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-23 18:37:01 +00:00
Dan Gohman
fafb890ee2 Fix LSR to tolerate cases where ScalarEvolution initially
misses an opportunity to fold add operands, but folds them
after LSR has separated them out. This fixes rdar://7886751.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102157 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-23 01:55:05 +00:00
Chris Lattner
62cc838b90 disable my previous inliner patch, it appears to be busting self-host.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102153 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-23 00:41:03 +00:00
Chris Lattner
fe9af3b1f7 The inliner was choosing to not consider call sites
that appear in the SCC as a result of inlining as candidates
for inlining.  Change this so that it *does* consider call 
sites that change from being indirect to being direct as a
result of inlining.  This allows it to completely 
"devirtualize" the testcase.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102146 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-22 23:37:35 +00:00
Chris Lattner
60915146f4 refactor the interface to InlineFunction so that most of the in/out
arguments are handled with a new InlineFunctionInfo class.  This 
makes it easier to extend InlineFunction to return more info in the
future.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102137 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-22 23:07:58 +00:00
Chris Lattner
b957a5e41e when inlining something like this:
define void @f3(void (i8*)* %__f) ssp {
entry:
  call void %__f(i8* undef)
  unreachable
}

define void @f4(i8* %this) ssp align 2 {
entry:
  call void @f3(void (i8*)* @f2) ssp
  ret void
}

The inliner is turning the indirect call to %__f into a direct
call to F2.  Make the call graph more precise when this happens.

The inliner doesn't revisit call sites introduced by inlining,
so there isn't an easy way to test for this, but a more precise
callgraph is a good thing.




git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102131 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-22 21:31:00 +00:00
Chris Lattner
f8b99cb03b eliminate dead #include.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102119 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-22 20:41:10 +00:00
Bob Wilson
b913863a95 Fix a performance problem with the new SSAUpdater. This showed up in the
GCCAS time for MultiSource/Benchmarks/ASCI_Purple/SMG2000.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@102009 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-21 18:39:03 +00:00
Devang Patel
3943084136 Rename ValueMapTy as ValueToValueMapTy to clearly indicate that this has no replationship with ADT/ValueMap.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101950 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-20 22:24:18 +00:00
Devang Patel
cd9b492451 There is no need to install ValueMapper.h header.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101949 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-20 22:18:31 +00:00
Gabor Greif
a9b2313c13 use abstract accessors to CallInst
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101899 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-20 13:13:04 +00:00
Chris Lattner
4f1f6f6342 remove a bunch of ad-hoc code to simplify instructions from
loop unswitch, and use inst simplify instead.  It is more
powerful and less duplication.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101874 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-20 05:33:18 +00:00
Chris Lattner
047542669a move some select simplifications out out instcombine into
inst simplify.  No functionality change.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101873 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-20 05:32:14 +00:00
Chris Lattner
a78130c320 RewriteLoopBodyWithConditionConstant can end up rewriting the
condition we're unswitching on.  In this case, don't try to
simplify the second copy of the loop which may be dead or not,
but is probably a constant now.  This fixes PR6879


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101870 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-20 05:09:16 +00:00
Chris Lattner
4471136e4d make the inliner do less work for leaf functions.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101846 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-20 00:47:08 +00:00
Chris Lattner
eae220259f Fix rdar://7879828 - crash in CallGraph, a self host issue.
Arg promotion was deleting call graph nodes that still had references
from the 'indirect' CGN.  Like the inliner, it should only delete the
function if all references are gone.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101845 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-20 00:46:50 +00:00
Dan Gohman
c056454ecf Remove the Expr member from IVUsers. Instead of remembering the expression,
just ask ScalarEvolution for it on demand. This helps IVUsers be more robust
in the case of expressions changing underneath it. This fixes PR6862.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101819 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-19 21:48:58 +00:00
Bob Wilson
84bd6b0c31 Re-commit my previous SSAUpdater changes. The previous version naively tried
to determine where to place PHIs by iteratively comparing reaching definitions
at each block.  That was just plain wrong.  This version now computes the
dominator tree within the subset of the CFG where PHIs may need to be placed,
and then places the PHIs in the iterated dominance frontier of each definition.
The rest of the patch is mostly the same, with a few more performance
improvements added in.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101612 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-17 03:08:24 +00:00
Eric Christopher
551754c495 Revert 101465, it broke internal OpenGL testing.
Probably the best way to know that all getOperand() calls have been handled
is to replace that API instead of updating.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101579 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 23:37:20 +00:00
Chris Lattner
2decb22222 introduce a new CallGraphSCC class, and pass it around
to CallGraphSCCPass's instead of passing around a
std::vector<CallGraphNode*>.  No functionality change,
but now we have a much tidier interface.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101558 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 22:42:17 +00:00
Dan Gohman
ea25b48af3 Refine the detection of seemingly infinitely recursive calls where the
callee is expected to be expanded to something else by codegen, so that
normal infinitely recursive calls are still transformed.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101468 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 15:57:50 +00:00
Gabor Greif
4ec2258ffb reapply r101434
with a fix for self-hosting

rotate CallInst operands, i.e. move callee to the back
of the operand array

the motivation for this patch are laid out in my mail to llvm-commits:
more efficient access to operands and callee, faster callgraph-construction,
smaller compiler binary


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101465 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 15:33:14 +00:00
Chris Lattner
a0bada729f fix comment noticed by Bob
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101437 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 02:32:17 +00:00
Gabor Greif
607a7ab3da back out r101423 and r101397, they break llvm-gcc self-host on darwin10
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101434 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 01:16:20 +00:00
Chris Lattner
d55c1c1659 fix PR6832: we were using the alignment of a pointer when we
wanted the alignment of the pointee.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101432 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 01:05:38 +00:00
Chris Lattner
a001b66498 improve comments.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101429 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 00:38:19 +00:00
Chris Lattner
4cc576bc5e pull all the ConvertToScalarInfo code together into one
place.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101427 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 00:24:57 +00:00
Chris Lattner
593375d04a more refactoring: suck some stuff out of SRoA into
ConvertToScalarInfo.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101425 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 00:20:00 +00:00
Gabor Greif
8db3f2c2ab shift intrinsic operand
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101423 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-16 00:06:45 +00:00
Chris Lattner
c447207264 introduce a new ConvertToScalarInfo struct to simplify
CanConvertToScalar/MergeInType.  Eliminate a pointless
LLVMContext argument to MergeInType.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101422 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-15 23:50:26 +00:00
Chris Lattner
31d80103d5 tidy interface to isOnlyCopiedFromConstantGlobal
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101405 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-15 21:59:20 +00:00
Gabor Greif
2ff961f668 reapply r101364, which has been backed out in r101368
with a fix

rotate CallInst operands, i.e. move callee to the back
of the operand array

the motivation for this patch are laid out in my mail to llvm-commits:
more efficient access to operands and callee, faster callgraph-construction,
smaller compiler binary


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101397 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-15 20:51:13 +00:00
Anton Korobeynikov
764f62ba60 Revert r100896 and around - this breaks the only mingw32 buildbot we have.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101387 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-15 19:51:42 +00:00
Dan Gohman
b35798347e Fix a bunch of namespace polution.
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101376 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-15 17:08:50 +00:00
Gabor Greif
9ee1720811 back out r101364, as it trips the linux nightlybot on some clang C++ tests
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101368 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-15 12:46:56 +00:00
Gabor Greif
165dac08d1 rotate CallInst operands, i.e. move callee to the back
of the operand array

the motivation for this patch are laid out in my mail to llvm-commits:
more efficient access to operands and callee, faster callgraph-construction,
smaller compiler binary


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@101364 91177308-0d34-0410-b5e6-96231b3b80d8
2010-04-15 10:49:53 +00:00