maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[InstCombine] insert a new shuffle in a safe place (PR25999) 

Limit this transform to a basic block and guard against PHIs.
Hopefully, this fixes the remaining failures in PR25999:
https://llvm.org/bugs/show_bug.cgi?id=25999

llvm-svn: 257133
This commit is contained in:
Sanjay Patel 2016-01-08 01:39:16 +00:00
parent 506ecac085
commit d72a458d28
2 changed files with 57 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
llvm/lib/Transforms/InstCombine/InstCombineVectorOps.cpp
View File

@ -384,23 +384,20 @@ static void replaceExtractElements(InsertElementInst *InsElt,
ConstantVector::get(ExtendMask));
// Insert the new shuffle after the vector operand of the extract is defined
// or at the start of the basic block, so any subsequent extracts can use it.
bool ReplaceAllExtUsers;
if (auto *ExtVecOpInst = dyn_cast<Instruction>(ExtVecOp)) {
// (as long as it's not a PHI) or at the start of the basic block of the
// extract, so any subsequent extracts in the same basic block can use it.
// TODO: Insert before the earliest ExtractElementInst that is replaced.
auto *ExtVecOpInst = dyn_cast<Instruction>(ExtVecOp);
if (ExtVecOpInst && !isa<PHINode>(ExtVecOpInst))
WideVec->insertAfter(ExtVecOpInst);
ReplaceAllExtUsers = true;
} else {
// TODO: Insert at start of function, so it's always safe to replace all?
else
IC.InsertNewInstWith(WideVec, *ExtElt->getParent()->getFirstInsertionPt());
ReplaceAllExtUsers = false;
}
// Replace extracts from the original narrow vector with extracts from the new
// wide vector.
for (User *U : ExtVecOp->users()) {
ExtractElementInst *OldExt = dyn_cast<ExtractElementInst>(U);
if (!OldExt ||
(!ReplaceAllExtUsers && OldExt->getParent() != WideVec->getParent()))
if (!OldExt || OldExt->getParent() != WideVec->getParent())
continue;
auto *NewExt = ExtractElementInst::Create(WideVec, OldExt->getOperand(1));
NewExt->insertAfter(WideVec);

50
llvm/test/Transforms/InstCombine/insert-extract-shuffle.ll
View File

@ -125,3 +125,53 @@ end:
ret <8 x i16> %t6
}
; The widening shuffle must be inserted at a valid point (after the PHIs).
define <4 x double> @pr25999_phis1(i1 %c, <2 x double> %a, <4 x double> %b) {
; CHECK-LABEL: @pr25999_phis1(
; CHECK: %tmp1 = phi <2 x double> [ %a, %bb1 ], [ %r, %bb2 ]
; CHECK-NEXT: %tmp2 = phi <4 x double> [ %b, %bb1 ], [ zeroinitializer, %bb2 ]
; CHECK-NEXT: %[[WIDEVEC:.*]] = shufflevector <2 x double> %tmp1, <2 x double> undef, <4 x i32> <i32 0, i32 undef, i32 undef, i32 undef>
; CHECK-NEXT: %tmp4 = shufflevector <4 x double> %tmp2, <4 x double> %[[WIDEVEC]], <4 x i32> <i32 0, i32 1, i32 4, i32 3>
; CHECK-NEXT: ret <4 x double> %tmp4
bb1:
br i1 %c, label %bb2, label %bb3
bb2:
%r = call <2 x double> @dummy(<2 x double> %a)
br label %bb3
bb3:
%tmp1 = phi <2 x double> [ %a, %bb1 ], [ %r, %bb2 ]
%tmp2 = phi <4 x double> [ %b, %bb1 ], [ zeroinitializer, %bb2 ]
%tmp3 = extractelement <2 x double> %tmp1, i32 0
%tmp4 = insertelement <4 x double> %tmp2, double %tmp3, i32 2
ret <4 x double> %tmp4
}
declare <2 x double> @dummy(<2 x double>)
define <4 x double> @pr25999_phis2(i1 %c, <2 x double> %a, <4 x double> %b) {
; CHECK-LABEL: @pr25999_phis2(
; CHECK: %tmp1 = phi <2 x double> [ %a, %bb1 ], [ %r, %bb2 ]
; CHECK-NEXT: %tmp2 = phi <4 x double> [ %b, %bb1 ], [ zeroinitializer, %bb2 ]
; CHECK-NEXT: %d = fadd <2 x double> %tmp1, %tmp1
; CHECK-NEXT: %[[WIDEVEC:.*]] = shufflevector <2 x double> %d, <2 x double> undef, <4 x i32> <i32 0, i32 undef, i32 undef, i32 undef>
; CHECK-NEXT: %tmp4 = shufflevector <4 x double> %tmp2, <4 x double> %[[WIDEVEC]], <4 x i32> <i32 0, i32 1, i32 4, i32 3>
; CHECK-NEXT: ret <4 x double> %tmp4
bb1:
br i1 %c, label %bb2, label %bb3
bb2:
%r = call <2 x double> @dummy(<2 x double> %a)
br label %bb3
bb3:
%tmp1 = phi <2 x double> [ %a, %bb1 ], [ %r, %bb2 ]
%tmp2 = phi <4 x double> [ %b, %bb1 ], [ zeroinitializer, %bb2 ]
%d = fadd <2 x double> %tmp1, %tmp1
%tmp3 = extractelement <2 x double> %d, i32 0
%tmp4 = insertelement <4 x double> %tmp2, double %tmp3, i32 2
ret <4 x double> %tmp4
}