Commit Graph

6595 Commits

Author SHA1 Message Date
Gabor Greif df323a51f5 performance: get rid of repeated dereferencing of use_iterator by caching its result
llvm-svn: 100550
2010-04-06 19:32:30 +00:00
Gabor Greif 679728790b make more two predicates constant
llvm-svn: 100549
2010-04-06 19:24:18 +00:00
Gabor Greif 08355d6cda performance: get rid of repeated dereferencing of use_iterator by caching its result
llvm-svn: 100547
2010-04-06 19:14:05 +00:00
Gabor Greif a21bc0fbd5 const-ize predicate ValueIsOnlyUsedLocallyOrStoredToOneGlobal
llvm-svn: 100546
2010-04-06 18:58:22 +00:00
Gabor Greif 0439789023 use CallSite to access calls vs. invokes uniformly
and remove assumptions about operand order

llvm-svn: 100544
2010-04-06 18:45:08 +00:00
Chris Lattner adca608281 fix a really nasty bug that Evan was tracking in SCCP. When resolving
undefs in branches/switches, we have two cases: a branch on a literal
undef or a branch on a symbolic value which is undef.  If we have a
literal undef, the code was correct: forcing it to a constant is the
right thing to do.

If we have a branch on a symbolic value that is undef, we should force
the symbolic value to a constant, which then makes the successor block
live.  Forcing the condition of the branch to being a constant isn't 
safe if later paths become live and the value becomes overdefined.  This
is the case that 'forcedconstant' is designed to handle, so just use it.

This fixes rdar://7765019 but there is no good testcase for this, the
one I have is too insane to be useful in the future.

llvm-svn: 100478
2010-04-05 22:14:48 +00:00
Chris Lattner c832c1bf69 some code cleanups, use SwitchInst::findCaseValue, reduce indentation
llvm-svn: 100468
2010-04-05 21:18:32 +00:00
Evan Cheng ba930449a9 Code clean up.
llvm-svn: 100467
2010-04-05 21:16:25 +00:00
Mon P Wang c576ee9040 Reapply address space patch after fixing an issue in MemCopyOptimizer.
Added support for address spaces and added a isVolatile field to memcpy, memmove, and memset,
e.g., llvm.memcpy.i32(i8*, i8*, i32, i32) -> llvm.memcpy.p0i8.p0i8.i32(i8*, i8*, i32, i32, i1)

llvm-svn: 100304
2010-04-04 03:10:48 +00:00
Chris Lattner ecb536313f require that the branch being controlled by the IV
exits the loop.  With this information we can guarantee 
the iteration count of the loop is bounded by the 
compare.  I think this xforms is finally safe now.

llvm-svn: 100285
2010-04-03 07:21:39 +00:00
Chris Lattner 40060d33f6 add integer overflow check for the fp induction variable
checker.  Amusingly, we already had tests that we should
have rejects because they would be miscompiled in the
testsuite.

The remaining issue with this is that we don't check that
the branch causes us to exit the loop if it fails, so we
don't actually know if we remain in bounds.

llvm-svn: 100284
2010-04-03 07:18:48 +00:00
Chris Lattner 69913466cb add a comment and fix some consistency issues, converting
to a signed vs unsigned value depending on the sign of the
constant fp means that we can't distinguish between a 
truly negative number and a positive number so large the
32nd bit is set.  So, do don't this!

llvm-svn: 100283
2010-04-03 06:41:49 +00:00
Chris Lattner 40ea690f39 fix PR6761, a miscompilation due to the fp->int IV conversion
stuff.  More bugs remain though.

llvm-svn: 100282
2010-04-03 06:30:03 +00:00
Chris Lattner 42202868c3 just eliminate the uitofp checks. This code isn't doing
the required validity checks in the first place, and supporting
a condition large enough to require the 32'nd bit isn't worth it.

llvm-svn: 100280
2010-04-03 06:25:21 +00:00
Chris Lattner ca25b60f4e rename PH -> PN to be consistent with WeakPN and the rest
of llvm.

llvm-svn: 100276
2010-04-03 06:17:08 +00:00
Chris Lattner 774858fc38 improve comment and drop a dead check. If PH had
no uses, it would have been deleted by 
RecursivelyDeleteTriviallyDeadInstructions

llvm-svn: 100275
2010-04-03 06:16:22 +00:00
Chris Lattner 915322bc4a strength reduce a ridiculous use of APInt.
llvm-svn: 100274
2010-04-03 06:13:12 +00:00
Chris Lattner 0b941347f9 rename stuff improve comment grammar.
llvm-svn: 100273
2010-04-03 06:11:07 +00:00
Chris Lattner d77bde5f94 simplify some code and resolve a fixme.
llvm-svn: 100272
2010-04-03 06:06:59 +00:00
Chris Lattner 2ff33f91d5 There is no guarantee that the increment and the branch
are in the same block.  Insert the new increment in the
correct location.

Also, more cleanups.

llvm-svn: 100271
2010-04-03 06:05:10 +00:00
Chris Lattner c558b49f14 first half of a pass through IndVarSimplify::HandleFloatingPointIV,
this cleans up a bunch of code and also fixes several crashes and
miscompiles.  More to come unfortunately, this optimization
is quite broken.

llvm-svn: 100270
2010-04-03 05:54:59 +00:00
Chris Lattner 2e23e5284c don't internalize available_externally functions, they are
really just declarations.  This is related to PR6524

llvm-svn: 100269
2010-04-03 05:24:50 +00:00
Bob Wilson f1aa4743d9 Revert all my SSAUpdater patches. The PHI placement algorithm is not correct
(what was I thinking?) and there's also a problem with LCSSA.  I'll try again
later with fixes.

--- Reverse-merging r100263 into '.':
U    lib/Transforms/Utils/SSAUpdater.cpp
--- Reverse-merging r100177 into '.':
G    lib/Transforms/Utils/SSAUpdater.cpp
--- Reverse-merging r100148 into '.':
G    lib/Transforms/Utils/SSAUpdater.cpp
--- Reverse-merging r100147 into '.':
U    include/llvm/Transforms/Utils/SSAUpdater.h
G    lib/Transforms/Utils/SSAUpdater.cpp
--- Reverse-merging r100131 into '.':
G    include/llvm/Transforms/Utils/SSAUpdater.h
G    lib/Transforms/Utils/SSAUpdater.cpp
--- Reverse-merging r100130 into '.':
G    lib/Transforms/Utils/SSAUpdater.cpp
--- Reverse-merging r100126 into '.':
G    include/llvm/Transforms/Utils/SSAUpdater.h
G    lib/Transforms/Utils/SSAUpdater.cpp
--- Reverse-merging r100050 into '.':
D    test/Transforms/GVN/2010-03-31-RedundantPHIs.ll
--- Reverse-merging r100047 into '.':
G    include/llvm/Transforms/Utils/SSAUpdater.h
G    lib/Transforms/Utils/SSAUpdater.cpp

llvm-svn: 100264
2010-04-03 03:50:38 +00:00
Bob Wilson 25f1aefd5b Add a DEBUG_TYPE for the SSAUpdater.
llvm-svn: 100263
2010-04-03 03:28:44 +00:00
Evan Cheng ed66db3f9b Code refactoring.
llvm-svn: 100262
2010-04-03 02:23:43 +00:00
Mon P Wang 999c1b927b Revert r100191 since it breaks objc in clang
llvm-svn: 100199
2010-04-02 18:43:02 +00:00
Mon P Wang a972ab8564 Reapply address space patch after fixing an issue in MemCopyOptimizer.
Added support for address spaces and added a isVolatile field to memcpy, memmove, and memset,
e.g., llvm.memcpy.i32(i8*, i8*, i32, i32) -> llvm.memcpy.p0i8.p0i8.i32(i8*, i8*, i32, i32, i1)

llvm-svn: 100191
2010-04-02 18:04:15 +00:00
Dan Gohman f7239102fe Manually notify ScalarEvolution before making an operand replacement, since
it can't currently observe such changes automatically.

llvm-svn: 100186
2010-04-02 14:48:31 +00:00
Bob Wilson 3c54edf9b3 Recommit 100158 now that the buildbots are happy again.
llvm-svn: 100177
2010-04-02 05:09:46 +00:00
Dan Gohman 4bd755419f Revert the recent alignment changes. They're broken for -Os because,
in particular, they end up aligning strings at 16-byte boundaries, and
there's no way for GlobalOpt to check OptForSize.

llvm-svn: 100172
2010-04-02 03:04:37 +00:00
Bob Wilson 0389adcd73 Revert 100158 in case it is causing some of the buildbot problems.
llvm-svn: 100164
2010-04-02 01:22:49 +00:00
Dan Gohman c671347fcb Make globalopt refine global variable alignment.
llvm-svn: 100160
2010-04-02 00:14:16 +00:00
Bob Wilson 9af4e118c6 Check for terminating conditions before adding PHIs to the worklists.
This is more efficient than adding them to the worklist and then ignoring them.

llvm-svn: 100158
2010-04-02 00:10:41 +00:00
Bob Wilson 737195069a Remove trailing whitespace.
llvm-svn: 100148
2010-04-01 23:06:38 +00:00
Bob Wilson 37b73d9d3e Rewrite another SSAUpdater function to avoid recursion.
llvm-svn: 100147
2010-04-01 23:05:58 +00:00
Bob Wilson 8409feadf0 Change another SSAUpdater function to avoid recursion.
llvm-svn: 100131
2010-04-01 20:04:30 +00:00
Bob Wilson 043c0406f7 Simplify the code to check for existing PHIs, now that it is only used in
one place.  This removes the template function added in svn 94690.

llvm-svn: 100130
2010-04-01 19:53:48 +00:00
Bob Wilson 38fc88ee5d The SSAUpdater should avoid recursive traversals of the CFG, since that may
blow out the stack for really big functions.  Start by fixing an easy case.

llvm-svn: 100126
2010-04-01 18:46:59 +00:00
Gabor Greif 5d5db5342b Introduce ImmutableCallSite, useful for contexts where no mutation
is necessary. Inherits from new templated baseclass CallSiteBase<>
which is highly customizable. Base CallSite on it too, in a configuration
that allows full mutation.
Adapt some call sites in analyses to employ ImmutableCallSite.

llvm-svn: 100100
2010-04-01 08:21:08 +00:00
Nick Lewycky bfb50a0d43 Clean up this file a little, no functionality change. This is a subset of my
patch back in r94322.

llvm-svn: 100097
2010-04-01 07:34:00 +00:00
Bob Wilson ac229124f4 Rewrite part of the SSAUpdater to be more careful about inserting redundant
PHIs.  The previous algorithm was unable to reliably detect when existing
PHIs in a cycle can be reused.  I'm still working on reducing a testcase.
Radar 7711900.

llvm-svn: 100047
2010-03-31 20:51:00 +00:00
Dale Johannesen b67a6e6620 Fix a nasty dangling-pointer heisenbug that could
generate wrong code pretty much anywhere AFAICT.
A case that hits the bug reproducibly is impossible,
but the situation was like this:
Addr = ...
Store -> Addr
Addr2 = GEP , 0, 0
Store -> Addr2
Handling the first store, the code changed replaced Addr
with a sunkaddr and deleted Addr, but not its table
entry.  Code in OptimizedBlock replaced Addr2 with a
bitcast; if that happened to reuse the memory of Addr,
the old table entry was erroneously found when handling
the second store.

llvm-svn: 100044
2010-03-31 20:37:15 +00:00
Bob Wilson 6f7fd28824 Revert Mon Ping's change 99928, since it broke all the llvm-gcc buildbots.
llvm-svn: 99948
2010-03-30 22:27:04 +00:00
Mon P Wang 7460571381 Added support for address spaces and added a isVolatile field to memcpy, memmove, and memset,
e.g., llvm.memcpy.i32(i8*, i8*, i32, i32) -> llvm.memcpy.p0i8.p0i8.i32(i8*, i8*, i32, i32, i1)
A update of langref will occur in a subsequent checkin.

llvm-svn: 99928
2010-03-30 20:55:56 +00:00
Dan Gohman 39027c403c Fix a grammaro.
llvm-svn: 99917
2010-03-30 20:04:57 +00:00
Gabor Greif b469818279 fix two cases where the arguments were extracted from the wrong range out of the InvokeInst
spotted by baldrick -- thanks\!

llvm-svn: 99914
2010-03-30 19:20:53 +00:00
Jeffrey Yasskin 12fd516e51 Remove another memory leak from ABCD by using Edges by value instead of
pointer.  There was also a SmallPtrSet whose settiness wasn't being used, so I
changed it to a SmallVector.

llvm-svn: 99713
2010-03-27 09:09:17 +00:00
Jeffrey Yasskin 97e613b6da In ABCD, change the non-null Bound*s to Bound&s.
llvm-svn: 99711
2010-03-27 08:15:46 +00:00
Jeffrey Yasskin 33bc7e4cb5 Fix a memory leak in ABCD by giving ownership of Bound objects to the
MemoizedResultChart.

llvm-svn: 99710
2010-03-27 08:09:24 +00:00
Eric Christopher 81c03447fc When we promote a load of an argument make sure to take the alignment
of the previous load - it's usually important.  For example, we don't want
to blindly turn an unaligned load into an aligned one.

llvm-svn: 99699
2010-03-27 01:54:00 +00:00