6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-09-10 22:55:57 +00:00
Code Issues Projects Releases Wiki Activity
3dd08734c1
llvm-6502/test/Transforms/InstCombine
History
Chris Lattner 3dd08734c1 optimize bitcasts from large integers to vector into vector 
element insertion from the pieces that feed into the vector.
This handles a pattern that occurs frequently due to code
generated for the x86-64 abi.  We now compile something like
this:

struct S { float A, B, C, D; };
struct S g;
struct S bar() { 
  struct S A = g;
  ++A.A;
  ++A.C;
  return A;
}

into all nice vector operations:

_bar:                                   ## @bar
## BB#0:                                ## %entry
	movq	_g@GOTPCREL(%rip), %rax
	movss	LCPI1_0(%rip), %xmm1
	movss	(%rax), %xmm0
	addss	%xmm1, %xmm0
	pshufd	$16, %xmm0, %xmm0
	movss	4(%rax), %xmm2
	movss	12(%rax), %xmm3
	pshufd	$16, %xmm2, %xmm2
	unpcklps	%xmm2, %xmm0
	addss	8(%rax), %xmm1
	pshufd	$16, %xmm1, %xmm1
	pshufd	$16, %xmm3, %xmm2
	unpcklps	%xmm2, %xmm1
	ret

instead of icky integer operations:

_bar:                                   ## @bar
	movq	_g@GOTPCREL(%rip), %rax
	movss	LCPI1_0(%rip), %xmm1
	movss	(%rax), %xmm0
	addss	%xmm1, %xmm0
	movd	%xmm0, %ecx
	movl	4(%rax), %edx
	movl	12(%rax), %esi
	shlq	$32, %rdx
	addq	%rcx, %rdx
	movd	%rdx, %xmm0
	addss	8(%rax), %xmm1
	movd	%xmm1, %eax
	shlq	$32, %rsi
	addq	%rax, %rsi
	movd	%rsi, %xmm1
	ret

This resolves rdar://8360454



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@112343 91177308-0d34-0410-b5e6-96231b3b80d8
2010-08-28 01:20:38 +00:00
..
2002-03-11-InstCombineHang.ll
2002-05-14-SubFailure.ll
2002-08-02-CastTest.ll
2002-12-05-MissedConstProp.ll
2003-05-26-CastMiscompile.ll
2003-05-27-ConstExprCrash.ll
2003-06-05-BranchInvertInfLoop.ll
2003-07-21-ExternalConstant.ll
2003-08-12-AllocaNonNull.ll
2003-09-09-VolatileLoadElim.ll
2003-10-29-CallSiteResolve.ll
2003-11-03-VarargsCallBug.ll
2003-11-13-ConstExprCastCall.ll Make opt default to not adding a target data string and update tests that depend on target data to supply it within the test 2009-11-03 15:29:06 +00:00
2004-01-13-InstCombineInvokePHI.ll
2004-02-23-ShiftShiftOverflow.ll
2004-03-13-InstCombineInfLoop.ll
2004-04-04-InstCombineReplaceAllUsesWith.ll
2004-05-07-UnsizedCastLoad.ll
2004-07-27-ConstantExprMul.ll
2004-08-09-RemInfLoop.ll
2004-08-10-BoolSetCC.ll
2004-09-20-BadLoadCombine2.ll
2004-09-20-BadLoadCombine.ll
2004-09-28-BadShiftAndSetCC.ll
2004-11-22-Missed-and-fold.ll
2004-11-27-SetCCForCastLargerAndConstant.ll Allow instcombine to combine "sext(a) >u const" to "a >u trunc(const)". 2009-12-17 22:42:29 +00:00
2004-12-08-RemInfiniteLoop.ll
2005-03-04-ShiftOverflow.ll
2005-04-07-UDivSelectCrash.ll
2005-06-15-DivSelectCrash.ll
2005-06-15-ShiftSetCCCrash.ll
2005-06-16-RangeCrash.ll
2005-06-16-SetCCOrSetCCMiscompile.ll
2005-07-07-DeadPHILoop.ll
2006-02-13-DemandedMiscompile.ll
2006-02-28-Crash.ll
2006-03-30-ExtractElement.ll
2006-04-28-ShiftShiftLongLong.ll
2006-05-04-DemandedBitCrash.ll
2006-09-15-CastToBool.ll
2006-10-19-SignedToUnsignedCastAndConst-2.ll
2006-10-20-mask.ll
2006-10-26-VectorReassoc.ll
2006-11-03-Memmove64.ll
2006-11-10-ashr-miscompile.ll
2006-12-01-BadFPVectorXform.ll
2006-12-05-fp-to-int-ext.ll
2006-12-08-Phi-ICmp-Op-Fold.ll
2006-12-08-Select-ICmp.ll
2006-12-15-Range-Test.ll
2006-12-23-Select-Cmp-Cmp.ll
2007-01-13-ExtCompareMiscompile.ll
2007-01-14-FcmpSelf.ll
2007-01-18-VectorInfLoop.ll
2007-01-27-AndICmp.ll
2007-02-01-LoadSinkAlloca.ll
2007-02-07-PointerCast.ll
2007-02-23-PhiFoldInfLoop.ll
2007-03-13-CompareMerge.ll
2007-03-19-BadTruncChangePR1261.ll
2007-03-21-SignedRangeTest.ll
2007-03-25-BadShiftMask.ll
2007-03-25-DoubleShift.ll
2007-03-26-BadShiftMask.ll
2007-04-04-BadFoldBitcastIntoMalloc.ll
2007-04-08-SingleEltVectorCrash.ll
2007-05-04-Crash.ll
2007-05-10-icmp-or.ll
2007-05-14-Crash.ll
2007-05-18-CastFoldBug.ll
2007-06-06-AshrSignBit.ll
2007-06-21-DivCompareMiscomp.ll
2007-08-02-InfiniteLoop.ll
2007-09-10-AliasConstFold.ll
2007-09-11-Trampoline.ll
2007-09-17-AliasConstFold2.ll
2007-10-10-EliminateMemCpy.ll Make opt default to not adding a target data string and update tests that depend on target data to supply it within the test 2009-11-03 15:29:06 +00:00
2007-10-12-Crash.ll
2007-10-28-stacksave.ll
2007-10-31-RangeCrash.ll
2007-10-31-StringCrash.ll
2007-11-07-OpaqueAlignCrash.ll
2007-11-15-CompareMiscomp.ll
2007-11-22-IcmpCrash.ll
2007-11-25-CompatibleAttributes.ll
2007-12-10-ConstFoldCompare.ll
2007-12-12-GEPScale.ll
2007-12-16-AsmNoUnwind.ll
2007-12-18-AddSelCmpSub.ll
2007-12-28-IcmpSub2.ll add newlines at the end of files. 2010-04-07 22:53:17 +00:00
2008-01-06-BitCastAttributes.ll
2008-01-06-CastCrash.ll
2008-01-06-VoidCast.ll
2008-01-13-AndCmpCmp.ll
2008-01-13-NoBitCastAttributes.ll
2008-01-14-DoubleNest.ll
2008-01-14-VarArgTrampoline.ll
2008-01-21-MismatchedCastAndCompare.ll
2008-01-21-MulTrunc.ll make instcombine only rewrite a chain of computation 2009-11-07 19:11:46 +00:00
2008-01-27-FloatSelect.ll
2008-01-29-AddICmp.ll
2008-02-13-MulURem.ll
2008-02-16-SDivOverflow2.ll
2008-02-16-SDivOverflow.ll
2008-02-23-MulSub.ll
2008-02-28-OrFCmpCrash.ll
2008-03-13-IntToPtr.ll
2008-04-22-ByValBitcast.ll
2008-04-28-VolatileStore.ll
2008-04-29-VolatileLoadDontMerge.ll
2008-04-29-VolatileLoadMerge.ll
2008-05-08-LiveStoreDelete.ll
2008-05-08-StrLenSink.ll
2008-05-09-SinkOfInvoke.ll
2008-05-17-InfLoop.ll
2008-05-18-FoldIntToPtr.ll
2008-05-22-IDivVector.ll
2008-05-22-NegValVector.ll
2008-05-23-CompareFold.ll
2008-05-31-AddBool.ll
2008-05-31-Bools.ll
2008-06-05-ashr-crash.ll
2008-06-08-ICmpPHI.ll
2008-06-13-InfiniteLoopStore.ll
2008-06-13-ReadOnlyCallStore.ll
2008-06-19-UncondLoad.ll
2008-06-21-CompareMiscomp.ll
2008-06-24-StackRestore.ll
2008-07-08-AndICmp.ll
2008-07-08-ShiftOneAndOne.ll
2008-07-08-SubAnd.ll
2008-07-08-VolatileLoadMerge.ll
2008-07-09-SubAndError.ll
2008-07-10-CastSextBool.ll
2008-07-10-ICmpBinOp.ll
2008-07-11-RemAnd.ll
2008-07-13-DivZero.ll
2008-07-16-fsub.ll
2008-07-16-sse2_storel_dq.ll
2008-08-05-And.ll
2008-08-17-ICmpXorSignbit.ll
2008-09-02-VectorCrash.ll
2008-09-29-FoldingOr.ll
2008-10-11-DivCompareFold.ll
2008-10-23-ConstFoldWithoutMask.ll
2008-11-01-SRemDemandedBits.ll
2008-11-08-FCmp.ll
2008-11-20-DivMulRem.ll
2008-11-27-IDivVector.ll
2008-11-27-MultiplyIntVec.ll
2008-11-27-UDivNegative.ll
2008-12-17-SRemNegConstVec.ll
2009-01-05-i128-crash.ll
2009-01-08-AlignAlloca.ll
2009-01-16-PointerAddrSpace.ll
2009-01-19-fmod-constant-float-specials.ll Teach APFloat how to create both QNaNs and SNaNs and with arbitrary-width 2010-02-28 02:51:25 +00:00
2009-01-19-fmod-constant-float.ll
2009-01-24-EmptyStruct.ll
2009-01-31-InfIterate.ll
2009-01-31-Pressure.ll
2009-02-04-FPBitcast.ll
2009-02-20-InstCombine-SROA.ll
2009-02-21-LoadCST.ll
2009-02-25-CrashZeroSizeArray.ll
2009-03-18-vector-ashr-crash.ll
2009-03-20-AShrOverShift.ll
2009-03-24-InfLoop.ll
2009-04-07-MulPromoteToI96.ll
2009-05-23-FCmpToICmp.ll
2009-06-11-StoreAddrSpace.ll
2009-06-16-SRemDemandedBits.ll
2009-07-02-MaskedIntVector.ll
2009-12-17-CmpSelectNull.ll Optimize icmp of null and select of two constants even if the select has 2009-12-18 08:22:35 +00:00
2010-01-28-NegativeSRem.ll Fix PR6165. The bug was that LHSKnownZero was being and'd with DemandedMask 2010-01-28 17:22:42 +00:00
2010-03-03-ExtElim.ll Make the 'icmp pred trunc(ext(X)), CST --> icmp pred X, ext(trunc(CST))' 2010-03-04 06:54:10 +00:00
2010-07-19-sqrt.ll Testcase for r108687. 2010-07-19 08:14:26 +00:00
2010-08-19-StoreNarrowing.ll Re-apply r111568 with a fix for the clang self-host. 2010-08-20 18:24:43 +00:00
add2.ll
add3.ll
add-shrink.ll
add-sitofp.ll
add.ll
addnegneg.ll
adjust-for-sminmax.ll
align-2d-gep.ll Make opt default to not adding a target data string and update tests that depend on target data to supply it within the test 2009-11-03 15:29:06 +00:00
align-addr.ll Make instcombine set explicit alignments on load or store 2010-08-03 18:20:32 +00:00
align-external.ll
alloca.ll Teach instcombine to promote alloca array sizes. 2010-05-28 15:09:00 +00:00
and2.ll convert to filecheck. 2010-02-11 06:24:37 +00:00
and-compare.ll
and-fcmp.ll
and-not-or.ll
and-or-and.ll
and-or-not.ll
and-or.ll
and-xor-merge.ll
and.ll
apint-add1.ll
apint-add2.ll
apint-and1.ll
apint-and2.ll
apint-and-compare.ll
apint-and-or-and.ll
apint-and-xor-merge.ll
apint-call-cast-target.ll
apint-cast-and-cast.ll
apint-cast-cast-to-and.ll
apint-cast.ll make instcombine only rewrite a chain of computation 2009-11-07 19:11:46 +00:00
apint-div1.ll
apint-div2.ll
apint-mul1.ll
apint-mul2.ll
apint-not.ll
apint-or1.ll
apint-or2.ll
apint-rem1.ll
apint-rem2.ll
apint-select.ll
apint-shift-simplify.ll
apint-shift.ll change the preferred canonical form for a sign extension to be 2010-01-10 07:08:30 +00:00
apint-shl-trunc.ll
apint-sub.ll
apint-xor1.ll
apint-xor2.ll
apint-zext1.ll
apint-zext2.ll
ashr-nop.ll
badmalloc.ll Teach instCombine to remove malloc+free if malloc's only uses are comparisons 2010-05-27 19:09:06 +00:00
binop-cast.ll
bit-checks.ll Re-apply the infamous r108614, with a fix pointed out by Dirk Steinke. 2010-08-02 09:32:13 +00:00
bit-tracking.ll
bitcast-sext-vector.ll InstCombine should not fold sext/zext of a vector and a bitcast to a scalar to a sext/zext 2010-01-23 04:35:57 +00:00
bitcast-vec-canon.ll
bitcast-vector-fold.ll
bitcast.ll optimize bitcasts from large integers to vector into vector 2010-08-28 01:20:38 +00:00
bitcount.ll
bittest.ll
bswap-fold.ll Fold bswap(undef) to undef. 2010-02-17 00:54:58 +00:00
bswap.ll remove two trunc xforms that are subsumed by EvaluateInDifferentType. 2010-01-05 22:01:41 +00:00
call2.ll
call-cast-target.ll
call-intrinsics.ll
call.ll Remove arm_apcscc from the test files. It is the default and doing this 2010-06-17 15:18:27 +00:00
canonicalize_branch.ll filecheckize this. 2010-01-18 22:00:46 +00:00
cast_ptr.ll merge some tests. 2010-01-05 21:54:09 +00:00
cast-mul-select.ll unify the code that determines whether it is a good idea to change the type 2009-11-10 07:23:37 +00:00
cast-set.ll make instcombine only rewrite a chain of computation 2009-11-07 19:11:46 +00:00
cast.ll fix PR7311 by avoiding breaking casts when a bitcast from scalar->vector 2010-07-12 01:19:22 +00:00
compare-signs.ll Fix some CHECK lines which were ignored by accident. 2009-12-12 09:25:50 +00:00
constant-fold-compare.ll Aggressively flip compare constant expressions where appropriate; constant 2009-12-17 06:07:04 +00:00
constant-fold-gep.ll Make opt default to not adding a target data string and update tests that depend on target data to supply it within the test 2009-11-03 15:29:06 +00:00
CPP_min_max.ll
crash.ll Remove arm_apcscc from the test files. It is the default and doing this 2010-06-17 15:18:27 +00:00
dce-iterate.ll
deadcode.ll
dg.exp
div-cmp-overflow.ll
div.ll
enforce-known-alignment.ll
exact-sdiv.ll
extractvalue.ll
fcmp-select.ll Don't do (X != Y) ? X : Y -> X for floating-point values; it doesn't 2010-02-23 17:17:57 +00:00
fcmp-special.ll Constant-fold certain comparisons with infinity and negative infinity. 2010-02-22 04:06:03 +00:00
fold-bin-operand.ll Make opt default to not adding a target data string and update tests that depend on target data to supply it within the test 2009-11-03 15:29:06 +00:00
fold-vector-zero.ll
fp-ret-bitcast.ll Make opt default to not adding a target data string and update tests that depend on target data to supply it within the test 2009-11-03 15:29:06 +00:00
fpcast.ll
fpextend.ll
fsub.ll Reduce fsub-fadd.ll and merge it into fsub-fsub.ll. Rename fsub-fsub.ll to 2010-01-17 00:21:21 +00:00
gepgep.ll When constant folding GEP of GEP, do not crash if an index of 2010-03-12 17:55:20 +00:00
getelementptr.ll Add a testcase for getelementptr index promotion. 2010-05-28 15:07:59 +00:00
hoist_instr.ll
icmp.ll Constant fold x == undef to undef. 2010-06-28 21:30:07 +00:00
idioms.ll optimize ~(~X >>s Y) --> (X >>s Y), patch by Edmund Grimley 2010-01-19 18:16:19 +00:00
IntPtrCast.ll
intrinsics.ll PR7853: fix a silly mistake introduced in r101899, and add a test to make sure 2010-08-09 20:49:43 +00:00
invariant.ll Print empty structs as {} rather than { }. 2010-04-08 18:03:05 +00:00
JavaCompare.ll make these less sensitive to temporary naming. 2010-03-05 08:43:33 +00:00
known_align.ll
load3.ll fix PR7429, a crash turning a load from a string into a float. 2010-07-12 00:22:51 +00:00
load-cmp.ll make these less sensitive to temporary naming. 2010-03-05 08:43:33 +00:00
load-select.ll Remove ARM-specific calling convention from this test. Target data is 2010-01-30 00:40:23 +00:00
load.ll merge two tests. 2010-07-12 00:19:47 +00:00
loadstore-alignment.ll Delete useless trailing semicolons. 2010-01-05 17:55:26 +00:00
logical-select.ll fix logical-select to invoke filecheck right, and fix hte instcombine 2010-02-05 19:53:02 +00:00
lshr-phi.ll
malloc2.ll
malloc3.ll
malloc-free-delete.ll Teach instCombine to remove malloc+free if malloc's only uses are comparisons 2010-05-27 19:09:06 +00:00
malloc.ll
memcpy-to-load.ll
memcpy.ll Slightly generalize transformation of memmove(a,a,n) so that it also applies 2009-12-17 21:07:31 +00:00
memmove.ll
memset_chk.ll Reapply address space patch after fixing an issue in MemCopyOptimizer. 2010-04-04 03:10:48 +00:00
memset.ll
mul-masked-bits.ll
mul.ll
multi-use-or.ll Floating-point add, sub, and mul are now spelled fadd, fsub, and fmul, 2010-03-02 01:11:08 +00:00
narrow.ll
no-negzero.ll
not-fcmp.ll
not.ll
nothrow.ll
nsw.ll
objsize.ll Fix declarations in a few more tests. 2010-04-17 21:29:25 +00:00
odr-linkage.ll add newlines at the end of files. 2010-04-07 22:53:17 +00:00
or-fcmp.ll
or-to-xor.ll
or.ll revert r108320, I see the failures now... 2010-07-14 06:16:35 +00:00
phi-merge-gep.ll
phi.ll fix PR6512, a case where instcombine would incorrectly merge loads 2010-03-05 18:53:28 +00:00
pr2645-0.ll
pr2645-1.ll
pr2996.ll
preserve-sminmax.ll
ptr-int-cast.ll Convert this test to FileCheck and add a testcase for PR3574. 2010-02-23 01:28:09 +00:00
README.txt
rem.ll
sdiv-1.ll
sdiv-2.ll
sdiv-shift.ll
select-2.ll
select-load-call.ll
select.ll Teach instcombine to transform 2010-07-08 11:39:10 +00:00
set.ll
setcc-strength-reduce.ll
sext.ll merge sext-2 into sext.ll 2009-12-02 05:34:35 +00:00
shift-sra.ll Fix a typo in ValueTracking that's causing instcombine to delete needed shift instructions. 2010-03-13 02:20:29 +00:00
shift.ll tidy up test. 2010-08-27 23:15:21 +00:00
shufflemask-undef.ll
shufflevec-constant.ll Make ConstantFoldConstantExpression recursively visit the entire 2009-11-23 16:22:21 +00:00
signed-comparison.ll
signext.ll Simplify/generalize the xor+add->sign-extend instcombine. 2010-01-31 04:29:12 +00:00
simplify-demanded-bits-pointer.ll
sink_instruction.ll
sitofp.ll
srem1.ll
srem-simplify-bug.ll
srem.ll
stack-overalign.ll
stacksaverestore.ll
store.ll Fix PR5471 by removing an instcombine xform. Some pieces of the code 2009-11-26 22:04:42 +00:00
strcpy_chk.ll Verify function prototypes before trying to optimize functions. We also 2010-04-12 04:48:00 +00:00
sub.ll Add a small transform: transform -(X<<Y) to (-X<<Y) when the shift has a single 2010-01-31 02:30:23 +00:00
trunc.ll teach the truncation optimization that an entire chain of 2010-08-27 20:32:06 +00:00
udiv_select_to_select_shift.ll
udiv-simplify-bug-0.ll
udiv-simplify-bug-1.ll
udivrem-change-width.ll make instcombine only rewrite a chain of computation 2009-11-07 19:11:46 +00:00
urem-simplify-bug.ll
urem.ll
vec_demanded_elts-2.ll
vec_demanded_elts-3.ll
vec_demanded_elts.ll
vec_extract_elt.ll
vec_insertelt.ll
vec_narrow.ll Floating-point add, sub, and mul are now spelled fadd, fsub, and fmul, 2010-03-02 01:11:08 +00:00
vec_shuffle.ll Temporarily revert r110987 as it's causing some miscompares in 2010-08-17 22:55:27 +00:00
vector-casts.ll Rename ValueRequiresCast to ShouldOptimizeCast, to better reflect 2010-02-11 06:26:33 +00:00
vector-srem.ll
volatile_store.ll
xor2.ll Revert r108141 again, sigh. 2010-07-12 14:42:04 +00:00
xor-undef.ll
xor.ll
zero-point-zero-add.ll
zeroext-and-reduce.ll
zext-bool-add-sub.ll
zext-fold.ll
zext-or-icmp.ll
zext.ll

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.