RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2025-02-10 20:42:32 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity
llvm/test/Transforms/InstCombine
History
Evan Cheng de62192843 Throttle back "fold select into operand" transformation. InstCombine should not generate selects of two constants unless they are selects of 0 and 1. 
e.g.
define i32 @t1(i32 %c, i32 %x) nounwind {
       %t1 = icmp eq i32 %c, 0
       %t2 = lshr i32 %x, 18
       %t3 = select i1 %t1, i32 %t2, i32 %x
       ret i32 %t3
}

was turned into

define i32 @t2(i32 %c, i32 %x) nounwind {
       %t1 = icmp eq i32 %c, 0
       %t2 = select i1 %t1, i32 18, i32 0
       %t3 = lshr i32 %x, %t2
       ret i32 %t3
}

For most targets, that means materializing two constants and then a select. e.g. On x86-64

movl    %esi, %eax
shrl    $18, %eax
testl   %edi, %edi
cmovne  %esi, %eax
ret

=>

xorl    %eax, %eax
testl   %edi, %edi
movl    $18, %ecx
cmovne  %eax, %ecx
movl    %esi, %eax
shrl    %cl, %eax
ret

Also, the optimizer and codegen can reason about shl / and / add, etc. by a constant. This optimization will hinder optimizations using ComputeMaskedBits.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@68142 91177308-0d34-0410-b5e6-96231b3b80d8
2009-03-31 20:42:45 +00:00
..
2002-03-11-InstCombineHang.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2002-05-14-SubFailure.ll
don't use 'not' when we can use a positive test
2007-03-28 01:43:43 +00:00
2002-05-14-TouchDeletedInst.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2002-08-02-CastTest.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2002-09-17-GetElementPtrCrash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2002-12-05-MissedConstProp.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-05-26-CastMiscompile.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-05-27-ConstExprCrash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-06-05-BranchInvertInfLoop.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-06-22-ConstantExprCrash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-07-21-ExternalConstant.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2003-08-12-AllocaNonNull.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-09-09-VolatileLoadElim.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-10-23-InstcombineNullFail.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-10-29-CallSiteResolve.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-11-03-VarargsCallBug.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2003-11-13-ConstExprCastCall.ll
Situations can arise when you have a function called that returns a 'void', but
2008-05-14 22:45:20 +00:00
2004-01-13-InstCombineInvokePHI.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2004-02-23-ShiftShiftOverflow.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2004-03-13-InstCombineInfLoop.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2004-04-04-InstCombineReplaceAllUsesWith.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2004-05-07-UnsizedCastLoad.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2004-07-27-ConstantExprMul.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2004-08-09-RemInfLoop.ll
rename *.llx -> *.ll, last batch.
2008-04-19 22:32:52 +00:00
2004-08-10-BoolSetCC.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2004-09-20-BadLoadCombine2.ll
Update old-style syntax in some "not grep" tests.
2008-05-01 23:50:07 +00:00
2004-09-20-BadLoadCombine.ll
rename *.llx -> *.ll, last batch.
2008-04-19 22:32:52 +00:00
2004-09-28-BadShiftAndSetCC.ll
rename *.llx -> *.ll, last batch.
2008-04-19 22:32:52 +00:00
2004-11-22-Missed-and-fold.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2004-11-27-SetCCForCastLargerAndConstant.ll
Eliminate questionable syntax for stdin redirection. This probably also speeds things up a bit.
2008-05-20 22:07:21 +00:00
2004-12-08-InstCombineCrash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2004-12-08-RemInfiniteLoop.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2005-03-04-ShiftOverflow.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2005-04-07-UDivSelectCrash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2005-06-15-DivSelectCrash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2005-06-15-ShiftSetCCCrash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2005-06-16-RangeCrash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2005-06-16-SetCCOrSetCCMiscompile.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2005-07-07-DeadPHILoop.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-02-07-SextZextCrash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-02-13-DemandedMiscompile.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-02-28-Crash.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-03-30-ExtractElement.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-04-01-InfLoop.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2006-04-28-ShiftShiftLongLong.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-05-04-DemandedBitCrash.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2006-05-06-Infloop.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2006-06-28-infloop.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-09-11-EmptyStructCrash.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2006-09-15-CastToBool.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-10-19-SignedToUnsignedCastAndConst-2.ll
Eliminate questionable syntax for stdin redirection. This probably also speeds things up a bit.
2008-05-20 22:07:21 +00:00
2006-10-19-SignedToUnsignedCastAndConst.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-10-20-mask.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-10-26-VectorReassoc.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-11-03-Memmove64.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-11-10-ashr-miscompile.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-11-27-XorBug.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-12-01-BadFPVectorXform.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-12-05-fp-to-int-ext.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-12-08-ICmp-Combining.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2006-12-08-Phi-ICmp-Op-Fold.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2006-12-08-Select-ICmp.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2006-12-10-ICmp-GEP-GEP.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2006-12-15-Range-Test.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2006-12-23-Select-Cmp-Cmp.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
2007-01-13-ExtCompareMiscompile.ll
this test is now compiled into the right thing.
2008-01-28 17:38:46 +00:00
2007-01-14-FcmpSelf.ll
For PR1319:
2007-04-14 20:13:02 +00:00
2007-01-18-VectorInfLoop.ll
For PR761:
2007-01-26 08:25:06 +00:00
2007-01-27-AndICmp.ll
For PR1319:
2007-04-14 20:13:02 +00:00
2007-02-01-LoadSinkAlloca.ll
For PR1319:
2007-04-14 20:13:02 +00:00
2007-02-07-PointerCast.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
2007-02-23-PhiFoldInfLoop.ll
Remove use of implementation keyword.
2007-03-28 02:38:26 +00:00
2007-03-13-CompareMerge.ll
For PR1319:
2007-04-14 20:13:02 +00:00
2007-03-19-BadTruncChangePR1261.ll
this xform is correct, not an xfail
2007-04-08 08:02:39 +00:00
2007-03-21-SignedRangeTest.ll
For PR1319:
2007-04-14 20:13:02 +00:00
2007-03-25-BadShiftMask.ll
For PR1319:
2007-04-14 20:13:02 +00:00
2007-03-25-DoubleShift.ll
Add a test case for PR1271 (necessary, but not sufficient).
2007-03-25 21:30:41 +00:00
2007-03-26-BadShiftMask.ll
For PR1319:
2007-04-15 07:38:21 +00:00
2007-03-27-PR1280.ll
For PR1319:
2007-04-14 20:13:02 +00:00
2007-03-31-InfiniteLoop.ll
For PR1319:
2007-04-14 20:13:02 +00:00
2007-04-04-BadFoldBitcastIntoMalloc.ll
For PR1319:
2007-04-16 17:36:08 +00:00
2007-04-08-SingleEltVectorCrash.ll
new testcase for PR1304
2007-04-09 01:37:35 +00:00
2007-05-04-Crash.ll
new testacse for PR1384
2007-05-05 01:59:05 +00:00
2007-05-10-icmp-or.ll
new testcase that crashes instcombine
2007-05-11 05:55:38 +00:00
2007-05-14-Crash.ll
New testcase that crashes instcombine
2007-05-15 00:15:49 +00:00
2007-05-18-CastFoldBug.ll
new testcase
2007-05-19 06:50:37 +00:00
2007-06-06-AshrSignBit.ll
Add a test for PR1499.
2007-06-06 17:10:02 +00:00
2007-06-21-DivCompareMiscomp.ll
new testcase miscompiled by instcombine, reduced from perl
2007-06-21 18:09:25 +00:00
2007-08-02-InfiniteLoop.ll
Reduced testcase for PR1594
2007-08-02 17:11:24 +00:00
2007-09-05-EqualGEP.ll
When the two operands of an icmp are equal, there are five possible predicates
2007-09-06 01:10:22 +00:00
2007-09-10-AliasConstFold.ll
Fix a buggy constant folding transformation when handling aliases.
2007-09-10 23:42:42 +00:00
2007-09-11-Trampoline.ll
Test that a call to a trampoline is turned into a call to
2007-09-11 15:07:50 +00:00
2007-09-17-AliasConstFold2.ll
Testcase for PR1678.
2007-09-20 18:56:24 +00:00
2007-10-10-EliminateMemCpy.ll
This test now passes.
2007-10-19 17:11:01 +00:00
2007-10-12-Crash.ll
Fix a bug in my patch last night that broke InstCombine/2007-10-12-Crash.ll
2007-10-12 18:05:47 +00:00
2007-10-28-EmptyField.ll
Fix PR1749 and InstCombine/2007-10-28-EmptyField.ll by handling
2007-10-29 02:40:02 +00:00
2007-10-28-stacksave.ll
Model stacksave and stackrestore as both writing memory, since we
2007-10-29 05:47:52 +00:00
2007-10-31-RangeCrash.ll
Fix InstCombine/2007-10-31-RangeCrash.ll
2007-11-01 02:18:41 +00:00
2007-10-31-StringCrash.ll
Fix InstCombine/2007-10-31-StringCrash.ll by removing an obvious
2007-11-01 02:30:35 +00:00
2007-11-07-OpaqueAlignCrash.ll
Better check
2007-11-08 18:45:15 +00:00
2007-11-15-CompareMiscomp.ll
Fix PR1800 by correcting mistaken logic.
2007-11-16 06:04:17 +00:00
2007-11-22-IcmpCrash.ll
Fix PR1817.
2007-11-22 23:47:13 +00:00
2007-11-25-CompatibleAttributes.ll
Fix PR1816. If a bitcast of a function only exists because of a
2007-11-25 14:10:56 +00:00
2007-12-10-ConstFoldCompare.ll
Fix PR1850 by removing an unsafe transformation from VMCore/ConstantFold.cpp.
2007-12-10 22:53:04 +00:00
2007-12-12-GEPScale.ll
1. "Upgrage" comments.
2007-12-12 15:21:32 +00:00
2007-12-16-AsmNoUnwind.ll
Make instcombine promote inline asm calls to 'nounwind'
2007-12-16 15:51:49 +00:00
2007-12-18-AddSelCmpSub.ll
Remove an orthogonal transformation of the selection condition from my most recent submission.
2007-12-18 20:30:28 +00:00
2007-12-28-IcmpSub2.ll
Add a testcase for my recent InstCombine fix, written by Nicholas.
2007-12-28 21:08:43 +00:00
2008-01-06-BitCastAttributes.ll
Ignore stderr for some more tests that expect warnings there.
2008-06-10 16:13:38 +00:00
2008-01-06-CastCrash.ll
When transforming a call to a bitcast function into
2008-01-06 10:12:28 +00:00
2008-01-06-VoidCast.ll
When transforming a call to a bitcast function into
2008-01-06 10:12:28 +00:00
2008-01-13-AndCmpCmp.ll
Fix PR1907, a nasty miscompilation because instcombine didn't
2008-01-13 20:59:02 +00:00
2008-01-13-NoBitCastAttributes.ll
When turning a call to a bitcast function into a direct call,
2008-01-13 08:02:44 +00:00
2008-01-14-DoubleNest.ll
I noticed that the trampoline straightening transformation could
2008-01-14 19:52:09 +00:00
2008-01-14-VarArgTrampoline.ll
I noticed that the trampoline straightening transformation could
2008-01-14 19:52:09 +00:00
2008-01-21-MismatchedCastAndCompare.ll
Fix two more not-grep tests that were missing llvm-dis.
2008-06-09 22:36:45 +00:00
2008-01-21-MulTrunc.ll
Multiply can be evaluated in a different type, so long as the target type has
2008-01-22 05:08:48 +00:00
2008-01-27-FloatSelect.ll
Fix PR1932 by disabling an xform invalid for fdiv.
2008-01-28 00:58:18 +00:00
2008-01-29-AddICmp.ll
Tag this test with the PR reference.
2008-02-03 16:35:19 +00:00
2008-02-04-GEPIdxBug.ll
Fix a bug compiling PR1978 (perhaps not the only one though) which
2008-02-05 04:45:32 +00:00
2008-02-13-MulURem.ll
Commit the testcase too.
2008-03-06 06:50:03 +00:00
2008-02-16-SDivOverflow2.ll
Correctly fold divide-by-constant, even when faced with overflow.
2008-02-18 22:48:05 +00:00
2008-02-16-SDivOverflow.ll
Correctly fold divide-by-constant, even when faced with overflow.
2008-02-18 22:48:05 +00:00
2008-02-23-MulSub.ll
fix this test so that the fn name doesn't match the regex
2008-02-26 18:13:51 +00:00
2008-02-28-OrFCmpCrash.ll
Folding or(fcmp,fcmp) only works if the operands of the fcmps are the same fp type.
2008-02-29 06:09:11 +00:00
2008-03-13-IntToPtr.ll
The inst combining of inttoptr into GEP with one index was using the bit size of
2008-03-14 05:12:19 +00:00
2008-04-22-ByValBitcast.ll
distill down the essense of this test.
2008-04-23 03:03:42 +00:00
2008-04-28-VolatileStore.ll
don't delete the last store to an alloca if the store is volatile.
2008-04-29 04:58:38 +00:00
2008-04-29-VolatileLoadDontMerge.ll
don't eliminate load from volatile value on paths where the load is dead.
2008-04-29 17:28:22 +00:00
2008-04-29-VolatileLoadMerge.ll
Fix a broken test. Neither load is eliminable without changing the CFG.
2008-07-09 05:01:02 +00:00
2008-05-08-LiveStoreDelete.ll
Make instcombine's DSE respect loads as well as stores. It is not safe to
2008-05-08 17:20:30 +00:00
2008-05-08-StrLenSink.ll
More than just loads can read from memory: readonly calls like strlen
2008-05-08 17:37:37 +00:00
2008-05-09-SinkOfInvoke.ll
Testcase for PR2303.
2008-05-10 16:43:10 +00:00
2008-05-17-InfLoop.ll
Fix PR2339
2008-05-18 04:11:26 +00:00
2008-05-18-FoldIntToPtr.ll
"ret (constexpr)" can't be folded into a Constant. Add a method to
2008-05-25 20:56:15 +00:00
2008-05-22-IDivVector.ll
Fix a recently added optimization to not crash on vectors.
2008-05-23 03:26:47 +00:00
2008-05-22-NegValVector.ll
Constant integer vectors may also be negated.
2008-05-23 04:54:45 +00:00
2008-05-23-CompareFold.ll
Fix a serious brain-o. Obviously no-one reviewed my patch :(
2008-05-24 04:06:28 +00:00
2008-05-31-AddBool.ll
Adding i1 is always Xor.
2008-05-31 17:10:28 +00:00
2008-05-31-Bools.ll
Add more i1 optimizations. add, sub, mul, s/udiv on i1 are now simplified away.
2008-05-31 17:59:52 +00:00
2008-06-05-ashr-crash.ll
Add a test case for opt -instcombine bug fix in revision 52003.
2008-06-05 14:25:11 +00:00
2008-06-08-ICmpPHI.ll
Remove unnecessary target lines.
2008-06-13 22:10:32 +00:00
2008-06-13-InfiniteLoopStore.ll
Remove unnecessary target lines.
2008-06-13 22:12:16 +00:00
2008-06-13-ReadOnlyCallStore.ll
Don't skip over instructions other than loads that might read memory
2008-06-13 22:02:12 +00:00
2008-06-19-UncondLoad.ll
Fix PR2471, which is a bug involving an invalid promotion from a conditional load.
2008-06-20 05:12:56 +00:00
2008-06-21-CompareMiscomp.ll
Fix for PR2479: correctly optimize expressions like (a > 13) & (a ==
2008-06-21 23:36:13 +00:00
2008-06-24-StackRestore.ll
Fix PR2488, a case where we deleted stack restores too aggressively.
2008-06-25 05:59:28 +00:00
2008-07-08-AndICmp.ll
Reinstate this optimization, but without the miscompile. Thanks to Bill for
2008-08-06 04:54:03 +00:00
2008-07-08-ShiftOneAndOne.ll
Fold ((1 << a) & 1) to (a == 0).
2008-07-09 05:20:13 +00:00
2008-07-08-SubAnd.ll
Reduce x - y to -y when we know the 'x' part will get masked off anyways.
2008-07-09 04:32:37 +00:00
2008-07-08-VolatileLoadMerge.ll
new testcase for PR2496
2008-07-08 17:18:05 +00:00
2008-07-09-SubAndError.ll
Fix overzealous optimization. Thanks to Duncan Sands for pointing out my error!
2008-07-10 05:51:40 +00:00
2008-07-10-CastSextBool.ll
Fix folding of icmp's of i1 where the comparison is signed. The code
2008-07-11 04:20:58 +00:00
2008-07-10-ICmpBinOp.ll
Add another optimization from PR2330. Also catch some missing cases that are
2008-07-11 07:20:53 +00:00
2008-07-11-RemAnd.ll
Enhance analysis of srem.
2008-07-12 05:04:38 +00:00
2008-07-13-DivZero.ll
Fix PR2506 by being a bit more careful about reverse fact propagation when
2008-07-14 00:15:52 +00:00
2008-07-16-fsub.ll
Fix PR2553
2008-07-17 06:07:20 +00:00
2008-07-16-sse2_storel_dq.ll
Fix PR2296. Do not transform x86_sse2_storel_dq into a full-width store.
2008-07-16 07:28:14 +00:00
2008-08-05-And.ll
Just grep for through the LL code instead of the ASM code
2008-08-06 00:10:32 +00:00
2008-08-17-ICmpXorSignbit.ll
Reinstate this optimization to fold icmp of xor when possible. Don't try to
2009-01-31 21:30:05 +00:00
2008-09-02-VectorCrash.ll
Don't crash when trying to constant fold a vector with some elements that can't
2008-09-03 05:54:33 +00:00
2008-09-29-FoldingOr.ll
Fix misoptimization of: xor i1 (icmp eq (X, C1), icmp s[lg]t (X, C2))
2008-09-30 06:08:34 +00:00
2008-10-11-DivCompareFold.ll
Fix PR2697 by rewriting the '(X / pos) op neg' logic. This also changes
2008-10-11 22:55:00 +00:00
2008-10-23-ConstFoldWithoutMask.ll
Don't try to create a mask when we don't need one. Fixes a crash.
2008-10-24 06:14:27 +00:00
2008-11-01-SRemDemandedBits.ll
Fix demanded bits analysis with srem by negative number. Based on a patch
2008-11-02 02:41:50 +00:00
2008-11-08-FCmp.ll
If the LHS of the FCMP is coming from a UIToFP instruction, then we don't want
2008-11-09 04:26:50 +00:00
2008-11-20-DivMulRem.ll
Optimize (x/y)*y into x-(x%y) in general. Div and rem are about the same, and
2008-11-21 07:33:58 +00:00
2008-11-27-IDivVector.ll
Add a couple of missed optimizations on integer vectors. Multiply and divide
2008-11-27 20:21:08 +00:00
2008-11-27-MultiplyIntVec.ll
Add a couple of missed optimizations on integer vectors. Multiply and divide
2008-11-27 20:21:08 +00:00
2008-11-27-UDivNegative.ll
Chris prefers icmp/select over udiv!
2008-11-27 22:41:10 +00:00
2008-12-17-SRemNegConstVec.ll
Make all the vector elements positive in an srem of constant vector.
2008-12-18 06:31:11 +00:00
2009-01-05-i128-crash.ll
make m_ConstantInt(int64_t) safely match ConstantInt's that are larger than i64.
2009-01-05 23:45:50 +00:00
2009-01-08-AlignAlloca.ll
Make instcombine ensure that all allocas are explicitly aligned at at
2009-01-13 20:18:38 +00:00
2009-01-16-PointerAddrSpace.ll
Fix PR3335 by not turning a store to one address space into a store to another.
2009-01-16 20:12:52 +00:00
2009-01-19-fmod-constant-float-specials.ll
Do not use host floating point types when emitting
2009-01-21 20:32:55 +00:00
2009-01-19-fmod-constant-float.ll
Calls to fmod, it turns out, are constant-folded by
2009-01-20 21:58:13 +00:00
2009-01-24-EmptyStruct.ll
testcase for PR3381.
2009-01-24 17:16:04 +00:00
2009-01-31-InfIterate.ll
reduce testcase.
2009-02-02 06:55:45 +00:00
2009-01-31-Pressure.ll
Reinstate this optimization to fold icmp of xor when possible. Don't try to
2009-01-31 21:30:05 +00:00
2009-02-04-FPBitcast.ll
Allow the inverse transform x86_fp80 -> i80 (also
2009-02-04 11:17:06 +00:00
2009-02-20-InstCombine-SROA.ll
rename a function to indicate that it checks for profitability as well
2009-02-21 00:46:50 +00:00
2009-02-21-LoadCST.ll
Don't sign extend the char when expanding char -> int during
2009-02-21 20:50:42 +00:00
2009-02-25-CrashZeroSizeArray.ll
Fix PR3667
2009-02-25 18:20:01 +00:00
2009-03-02-VarLengthArrayGEP.ll
Fix PR3694: add an instcombine micro-optimization that helps
2009-03-02 09:18:21 +00:00
2009-03-18-vector-ashr-crash.ll
Fix PR3826 - InstComb assert with vector shift, by not calling ComputeNumSignBits on a vector.
2009-03-18 16:32:19 +00:00
2009-03-20-AShrOverShift.ll
Fix instcombine to not introduce undefined shifts when merging two
2009-03-20 22:41:15 +00:00
2009-03-24-InfLoop.ll
Fix PR3874 by restoring a condition I removed, but making it more
2009-03-25 00:28:58 +00:00
add2.ll
be more aggressive about transforming add -> or when the operands have no
2008-05-19 20:01:56 +00:00
add-shrink.ll
Teach instcombine 4 new xforms:
2008-05-20 05:46:13 +00:00
add-sitofp.ll
Generalize the new code in instcombine's ComputeNumSignBits for handling
2008-05-23 02:28:01 +00:00
add.ll
Teach masked value is zero about add and sub, and use MVIZ to
2008-03-21 05:19:58 +00:00
addnegneg.ll
Fold (-x + -y) -> -(x+y) which promotes better association, fixing
2008-02-17 21:03:36 +00:00
adjust-for-sminmax.ll
Improve instcombine's handling of integer min and max in two ways:
2008-09-16 18:46:06 +00:00
align-2d-gep.ll
Teach InstCombine's ComputeMaskedBits to handle pointer expressions
2008-04-10 18:43:06 +00:00
align-addr.ll
Teach InstCombine's ComputeMaskedBits to handle pointer expressions
2008-04-10 18:43:06 +00:00
align-inc.ll
Change these tests to use regular loads instead of llvm.x86.sse2.loadu.dq.
2009-02-16 00:44:23 +00:00
alloca.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
and2.ll
Implement a couple of foldings for ordered and unordered comparisons,
2007-10-24 05:38:08 +00:00
and-compare.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
and-fcmp.ll
- Somehow I forgot about one / une.
2008-10-14 18:13:38 +00:00
and-not-or.ll
simplify these patterns using m_Specific. No need to grep for
2008-12-01 05:16:26 +00:00
and-or-and.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
and-or-not.ll
Convert tests using "| wc -l | grep ..." to use the count script.
2007-08-15 13:36:28 +00:00
and-or.ll
Implement ((A|B)&1)|(B&-2) -> (A&1) | B transformation. This also takes care of
2008-12-01 01:07:11 +00:00
and-xor-merge.ll
Convert tests using "| wc -l | grep ..." to use the count script.
2007-08-15 13:36:28 +00:00
and.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
apint-add1.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-add2.ll
For PR1319:
2007-04-15 07:38:21 +00:00
apint-and1.ll
For PR1319:
2007-04-15 07:38:21 +00:00
apint-and2.ll
For PR1319:
2007-04-15 07:38:21 +00:00
apint-and-compare.ll
Convert tests using "| wc -l | grep ..." to use the count script.
2007-08-15 13:36:28 +00:00
apint-and-or-and.ll
For PR1319:
2007-04-15 07:38:21 +00:00
apint-and-xor-merge.ll
Convert tests using "| wc -l | grep ..." to use the count script.
2007-08-15 13:36:28 +00:00
apint-call-cast-target.ll
When simplifying a call to a bitcast function, tighten up
2008-06-01 07:38:42 +00:00
apint-cast-and-cast.ll
Add test case for testing InstCombine with arbitrary precision integer
2007-03-23 20:48:34 +00:00
apint-cast-cast-to-and.ll
Add test case for testing InstCombine with arbitrary precision integer
2007-03-23 20:48:34 +00:00
apint-cast.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-div1.ll
bool -> i1 (found by llvm.exp)
2007-04-14 18:30:06 +00:00
apint-div2.ll
bool -> i1 (found by llvm.exp)
2007-04-14 18:30:06 +00:00
apint-elim-logicalops.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-mul1.ll
Remove use of implementation keyword.
2007-03-28 02:38:26 +00:00
apint-mul2.ll
Remove use of implementation keyword.
2007-03-28 02:38:26 +00:00
apint-not.ll
Fix testsuite hang.
2007-03-26 10:59:13 +00:00
apint-or1.ll
Remove use of implementation keyword.
2007-03-28 02:38:26 +00:00
apint-or2.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-rem1.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-rem2.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-select.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-shift-simplify.ll
Convert tests using "| wc -l | grep ..." to use the count script.
2007-08-15 13:36:28 +00:00
apint-shift.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-shl-trunc.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-sub.ll
Instcombine was illegally transforming -X/C into X/-C when either X or C
2008-11-30 03:42:12 +00:00
apint-xor1.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-xor2.ll
For PR1319:
2007-04-15 07:38:21 +00:00
apint-zext1.ll
For PR1319:
2007-04-14 20:13:02 +00:00
apint-zext2.ll
For PR1319:
2007-04-14 20:13:02 +00:00
ashr-nop.ll
Fix a ValueTracking rule: RHS means operand 1, not 0. Add a simple
2009-02-24 02:00:40 +00:00
binop-cast.ll
Fix a test test llvm.exp found.
2007-04-14 18:33:31 +00:00
bit-tracking.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
bitcast-gep.ll
Fix part 3/2 of PR3290, making instcombine zap (gep(bitcast)) when possible.
2009-01-09 05:44:56 +00:00
bitcast-vector-fold.ll
Implement constant folding if vector<->vector bitcasts where the number
2007-12-11 07:29:44 +00:00
bitcount.ll
This is the patch to provide clean intrinsic function overloading support in LLVM. It cleans up the intrinsic definitions and generally smooths the process for more complicated intrinsic writing. It will be used by the upcoming atomic intrinsics as well as vector and float intrinsics in the future.
2007-08-04 01:51:18 +00:00
bittest.ll
Update old-style syntax in some "not grep" tests.
2008-05-01 23:50:07 +00:00
bswap-fold.ll
implement some simple bswap optimizations, rdar://5992453
2008-06-18 04:33:20 +00:00
bswap.ll
rewrite bswap matching to be more general, allowing arbitrary
2008-10-05 02:13:19 +00:00
call2.ll
Test for transformConstExprCastCall fix.
2007-04-04 19:18:16 +00:00
call-cast-target.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
call-intrinsics.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
call.ll
Ignore stderr for some more tests that expect warnings there.
2008-06-10 16:13:38 +00:00
canonicalize_branch.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
cast2.ll
For PR1319:
2007-04-14 20:13:02 +00:00
cast_ld_addr_space.ll
Instrcombine should not change load(cast p) to cast(load p) if the cast
2009-02-07 22:19:29 +00:00
cast_ptr.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
cast-and-cast.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
cast-cast-to-and.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
cast-load-gep.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
cast-malloc.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
cast-mul-select.ll
Fix missed optimization opportunity when analyzing cast of mul and select.
2008-07-05 21:19:34 +00:00
cast-propagate.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
cast-set.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
cast-sext-zext.ll
Fix PR3452 (an infinite loop bootstrapping) by disabling the recent
2009-01-31 19:05:27 +00:00
cast-store-gep.ll
Make InstCombineStoreToCast handle aggregates more aggressively,
2009-01-24 01:00:13 +00:00
cast.ll
Clean up previous cast optimization a bit. Also make zext elimination a bit more aggressive: if it's not necessary to emit an AND (i.e. high bits are already zero), it's profitable to evaluate the operand at a different type.
2009-01-16 02:11:43 +00:00
CPP_min_max.ll
rename *.llx -> *.ll, last batch.
2008-04-19 22:32:52 +00:00
dce-iterate.ll
make sure to set Changed=true when instcombine hacks on the code,
2009-01-31 07:04:22 +00:00
deadcode.ll
dead calls to llvm.stacksave can be deleted, even though they
2007-12-29 00:59:12 +00:00
dg.exp
sabre brings to my attention that the 'tr' suffix is also obsolete
2008-05-20 21:00:03 +00:00
div-cmp-overflow.ll
Fix an icmp+sdiv optimization to check for and handle an overflow
2008-09-10 23:30:57 +00:00
div.ll
implement PR2328.
2008-05-16 02:59:42 +00:00
enforce-known-alignment.ll
Change the argument type in this test to something less convoluted,
2009-02-18 04:25:04 +00:00
extractvalue.ll
Add a few cases to instcombine's extractvalue testcase.
2008-07-16 12:57:25 +00:00
fold-vector-zero.ll
Add support for folding binary operators with vector zero operands.
2007-10-30 19:00:49 +00:00
fp-ret-bitcast.ll
Situations can arise when you have a function called that returns a 'void', but
2008-05-14 22:45:20 +00:00
fpcast.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
fpextend.ll
Fold fptrunc(add (fpextend x), (fpextend y)) -> add(x,y), as GCC does.
2008-01-27 05:29:54 +00:00
GEPIdxCanon.ll
Remove GCSE and LoadVN from the testsuite.
2008-08-16 00:00:54 +00:00
getelementptr_cast.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
getelementptr_const.ll
Remove GCSE and LoadVN from the testsuite.
2008-08-16 00:00:54 +00:00
getelementptr_index.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
getelementptr_promote.ll
On 64-bit targets, change 32-bit getelementptr indices to be 64-bit
2008-09-11 23:06:38 +00:00
getelementptr-setcc.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
getelementptr-seteq.ll
optimize "p != gep p, ..." better. This allows us to compile
2008-04-22 02:53:33 +00:00
getelementptr.ll
Byebye llvm-upgrade!
2008-03-25 04:26:08 +00:00
hoist_instr.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
icmp.ll
Implement rdar://6480391, extending of equality icmp's to avoid a truncation.
2009-01-09 07:47:06 +00:00
IntPtrCast.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
JavaCompare.ll
make this condition more precise.
2008-07-11 03:54:57 +00:00
known_align.ll
Fix a bug in GetKnownAlignment of packed structs.
2007-07-31 20:13:21 +00:00
load2.ll
Transform a load from an undef/zero global into an undef/global even if we
2007-08-11 18:48:48 +00:00
load3.ll
Teach instcombine's visitLoad to scan back several instructions
2008-10-15 23:19:35 +00:00
load.ll
Teach instcombine's visitLoad to scan back several instructions
2008-10-15 23:19:35 +00:00
loadstore-alignment.ll
Make instcombine ensure that all allocas are explicitly aligned at at
2009-01-13 20:18:38 +00:00
logical-select.ll
Handle the case where there is no "not". It is possible it got
2008-11-16 04:25:26 +00:00
malloc2.ll
For PR1319:
2007-04-14 20:13:02 +00:00
malloc3.ll
testcase for PR1728
2007-10-12 05:29:53 +00:00
malloc-free-delete.ll
For PR1319:
2007-04-14 20:13:02 +00:00
malloc.ll
new testcase, update old one.
2007-04-06 18:56:54 +00:00
memcpy-to-load.ll
Turn a memcpy from a double* into a load/store of double instead of
2008-01-14 00:28:35 +00:00
memmove.ll
Implement PR2370: memmove(x,x,size) -> noop.
2008-05-28 05:30:41 +00:00
memset.ll
move lowering of llvm.memset -> store from simplify libcalls
2008-04-30 06:39:11 +00:00
mul-masked-bits.ll
Fix a bug in the ComputeMaskedBits logic for multiply.
2008-05-07 00:35:55 +00:00
mul.ll
Implement support for simplifying vector comparisons by 0.0 and 1.0 like we
2008-08-11 22:06:05 +00:00
multi-use-or.ll
now that all the pieces are in place, teach instcombine's
2009-01-31 08:40:03 +00:00
narrow.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
not-fcmp.ll
It's safe to fold not of fcmp.
2007-08-06 20:04:16 +00:00
not.ll
Implement PR2298. This transforms:
2008-05-09 05:19:28 +00:00
or2.ll
Implement a couple of foldings for ordered and unordered comparisons,
2007-10-24 05:38:08 +00:00
or-fcmp.ll
Combine (fcmp cc0 x, y) | (fcmp cc1 x, y) into a single fcmp when possible.
2008-10-14 18:44:08 +00:00
or-to-xor.ll
Use m_Specific() instead of double matching.
2008-12-01 08:09:47 +00:00
or.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
phi-merge.ll
Implement PR1777 by detecting dependent phis that
2007-11-06 21:52:06 +00:00
phi.ll
Teach inst combine to merge GEPs through PHIs. This is really
2008-12-01 02:34:36 +00:00
pr2645-0.ll
Fix a vectorshuffle instcombine bug introduced by r55995.
2008-09-11 22:47:57 +00:00
pr2645-1.ll
Fix a vectorshuffle instcombine bug introduced by r55995.
2008-09-11 22:47:57 +00:00
pr2996.ll
Fix this recently moved code to use the correct type. CI is now a
2008-11-02 00:17:33 +00:00
preserve-sminmax.ll
Improve instcombine's handling of integer min and max in two ways:
2008-09-16 18:46:06 +00:00
ptr-int-cast.ll
canonicalize inttoptr and ptrtoint instructions which cast pointers
2009-03-24 18:35:40 +00:00
README.txt
*** empty log message ***
2002-08-03 20:24:45 +00:00
rem.ll
Don't try to simplify urem and srem using arithmetic rules that don't work
2008-03-06 06:48:30 +00:00
sdiv-1.ll
Strengthen check for div inst-combining.
2008-11-30 04:33:53 +00:00
sdiv-2.ll
getSExtValue() doesn't work for ConstantInts with bitwidth > 64 bits. Use all
2008-11-30 12:38:24 +00:00
select-2.ll
Throttle back "fold select into operand" transformation. InstCombine should not generate selects of two constants unless they are selects of 0 and 1.
2009-03-31 20:42:45 +00:00
select.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
set.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
setcc-cast-cast.ll
Update old-style syntax in some "not grep" tests.
2008-05-01 23:50:07 +00:00
setcc-strength-reduce.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
sext-misc.ll
optimize a common idiom generated by clang for bitfield access, PR2638.
2008-08-06 07:35:52 +00:00
shift-simplify.ll
Convert tests using "| wc -l | grep ..." to use the count script.
2007-08-15 13:36:28 +00:00
shift-sra.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
shift-trunc-shift.ll
implement InstCombine/shift-trunc-shift.ll. This allows
2007-12-22 09:07:47 +00:00
shift.ll
two changes:
2009-03-24 18:15:30 +00:00
shl-icmp.ll
Don't do: "(X & 4) >> 1 == 2 --> (X & 4) == 4" if there are more than one uses of the shift result.
2008-04-23 00:38:06 +00:00
shl-trunc.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
shufflemask-undef.ll
Fix a shufflevector instcombine that was emitting invalid masks indices
2008-08-06 18:17:32 +00:00
shufflevec-constant.ll
Add suppport for ConstantExprs of shufflevectors whose result type is not equal to the
2009-02-12 21:28:33 +00:00
signext.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
sink_instruction.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
sitofp.ll
Zap sitofp/fptoui pairs. In all cases when the sign difference
2008-08-06 05:13:06 +00:00
srem1.ll
Fix a bogus srem rule - a negative value srem'd by a power-of-2
2008-08-13 23:12:35 +00:00
srem-simplify-bug.ll
Correct the value of LowBits in srem and urem handling in
2008-05-06 00:51:48 +00:00
srem.ll
Instcombine x-((x/y)*y) into a remainder operator.
2007-09-17 17:31:57 +00:00
stack-overalign.ll
Add a testcase for the problem fixed in r65289.
2009-02-24 02:17:42 +00:00
stacksaverestore.ll
optimize away stackrestore calls that have no intervening alloca or call.
2008-02-18 06:12:38 +00:00
store-merge.ll
Convert tests using "| wc -l | grep ..." to use the count script.
2007-08-15 13:36:28 +00:00
store.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
sub.ll
Instcombine was illegally transforming -X/C into X/-C when either X or C
2008-11-30 03:42:12 +00:00
udiv_select_to_select_shift.ll
Fix PR2506 by being a bit more careful about reverse fact propagation when
2008-07-14 00:15:52 +00:00
udiv-simplify-bug-0.ll
Fix a mistake in the computation of leading zeros for udiv.
2008-05-02 21:30:02 +00:00
udiv-simplify-bug-1.ll
Chris prefers icmp/select over udiv!
2008-11-27 22:41:10 +00:00
urem-simplify-bug.ll
Fix an overaggressive SimplifyDemandedBits optimization on urem. This
2008-05-01 19:13:24 +00:00
urem.ll
Instcombine x-((x/y)*y) into a remainder operator.
2007-09-17 17:31:57 +00:00
vec_demanded_elts-2.ll
APInt'fy SimplifyDemandedVectorElts so it can analyze vectors with more than 64 elements.
2009-02-03 10:05:09 +00:00
vec_demanded_elts.ll
manually upgrade test. Add a new test2. I have no way to see if this works
2007-04-14 22:27:33 +00:00
vec_extract_elt.ll
new testcase
2007-04-14 23:00:51 +00:00
vec_insert_to_shuffle.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
vec_insertelt.ll
For PR1319:
2007-04-14 20:13:02 +00:00
vec_narrow.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
vec_shuffle2.ll
Fixed optimization of combining two shuffles where the first shuffle inputs
2009-01-26 04:39:00 +00:00
vec_shuffle.ll
For PR1319:
2007-04-15 07:38:21 +00:00
vector-srem.ll
Fix an abort in instcombine when folding creates a vector rem instruction.
2007-11-05 23:16:33 +00:00
volatile_store.ll
Don't zap back to back volatile load/stores
2007-09-07 05:33:03 +00:00
xor2.ll
For PR1319:
2007-04-14 20:13:02 +00:00
xor-demorgans.ll
XFAIL this test. The xform was removed.
2009-01-04 06:32:28 +00:00
xor-undef.ll
Handle a special case xor undef, undef -> 0. Technically this should be transformed to undef. But this is such a common idiom (misuse) we are going to handle it.
2008-03-25 20:07:13 +00:00
xor.ll
Remove llvm-upgrade and update tests.
2008-03-09 08:16:40 +00:00
zero-point-zero-add.ll
eliminate additions of 0.0 when they are obviously dead. This has to be careful to
2008-01-29 06:52:45 +00:00
zeroext-and-reduce.ll
Remove llvm-upgrade and update test cases.
2008-03-01 09:15:35 +00:00
zext-fold.ll
Peer through sext/zext when looking for not(cmp).
2008-05-31 19:01:33 +00:00
zext-or-icmp.ll
Transform (zext (or (icmp), (icmp))) to (or (zext (cimp), (zext icmp))) if at least one of the (zext icmp) can be transformed to eliminate an icmp.
2008-03-24 00:21:34 +00:00
zext.ll
commit a tweaked version of Daniel's patch for PR3599. We now
2009-02-17 20:47:23 +00:00

README.txt 

This directory contains test cases for the instcombine transformation.  The
dated tests are actual bug tests, whereas the named tests are used to test
for features that the this pass should be capable of performing.