mirror of
https://github.com/c64scene-ar/llvm-6502.git
synced 2025-06-17 04:24:00 +00:00
Add a convenience method to copy wrapping, exact, and fast-math flags (NFC).
The loop vectorizer preserves wrapping, exact, and fast-math properties of scalar instructions. This patch adds a convenience method to make that operation easier because we need to do this in the loop vectorizer, SLP vectorizer, and possibly other places. Although this is a 'no functional change' patch, I've added a testcase to verify that the exact flag is preserved by the loop vectorizer. The wrapping and fast-math flags are already checked in existing testcases. Differential Revision: http://reviews.llvm.org/D5138 git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@216886 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
@ -358,6 +358,10 @@ public:
|
|||||||
/// isExact - Determine whether the exact flag is set.
|
/// isExact - Determine whether the exact flag is set.
|
||||||
bool isExact() const;
|
bool isExact() const;
|
||||||
|
|
||||||
|
/// Convenience method to copy wrapping, exact, and fast-math flag values
|
||||||
|
/// from V to this instruction.
|
||||||
|
void copyFlags(const Value *V);
|
||||||
|
|
||||||
// Methods for support type inquiry through isa, cast, and dyn_cast:
|
// Methods for support type inquiry through isa, cast, and dyn_cast:
|
||||||
static inline bool classof(const Instruction *I) {
|
static inline bool classof(const Instruction *I) {
|
||||||
return I->isBinaryOp();
|
return I->isBinaryOp();
|
||||||
|
@ -2030,6 +2030,22 @@ bool BinaryOperator::isExact() const {
|
|||||||
return cast<PossiblyExactOperator>(this)->isExact();
|
return cast<PossiblyExactOperator>(this)->isExact();
|
||||||
}
|
}
|
||||||
|
|
||||||
|
void BinaryOperator::copyFlags(const Value *V) {
|
||||||
|
// Copy the wrapping flags.
|
||||||
|
if (auto *OB = dyn_cast<OverflowingBinaryOperator>(V)) {
|
||||||
|
setHasNoSignedWrap(OB->hasNoSignedWrap());
|
||||||
|
setHasNoUnsignedWrap(OB->hasNoUnsignedWrap());
|
||||||
|
}
|
||||||
|
|
||||||
|
// Copy the exact flag.
|
||||||
|
if (auto *PE = dyn_cast<PossiblyExactOperator>(V))
|
||||||
|
setIsExact(PE->isExact());
|
||||||
|
|
||||||
|
// Copy the fast-math flags.
|
||||||
|
if (auto *FP = dyn_cast<FPMathOperator>(V))
|
||||||
|
setFastMathFlags(FP->getFastMathFlags());
|
||||||
|
}
|
||||||
|
|
||||||
//===----------------------------------------------------------------------===//
|
//===----------------------------------------------------------------------===//
|
||||||
// FPMathOperator Class
|
// FPMathOperator Class
|
||||||
//===----------------------------------------------------------------------===//
|
//===----------------------------------------------------------------------===//
|
||||||
|
@ -3248,18 +3248,8 @@ void InnerLoopVectorizer::vectorizeBlockInLoop(BasicBlock *BB, PhiVector *PV) {
|
|||||||
for (unsigned Part = 0; Part < UF; ++Part) {
|
for (unsigned Part = 0; Part < UF; ++Part) {
|
||||||
Value *V = Builder.CreateBinOp(BinOp->getOpcode(), A[Part], B[Part]);
|
Value *V = Builder.CreateBinOp(BinOp->getOpcode(), A[Part], B[Part]);
|
||||||
|
|
||||||
// Update the NSW, NUW and Exact flags. Notice: V can be an Undef.
|
if (BinaryOperator *VecOp = dyn_cast<BinaryOperator>(V))
|
||||||
BinaryOperator *VecOp = dyn_cast<BinaryOperator>(V);
|
VecOp->copyFlags(BinOp);
|
||||||
if (VecOp && isa<OverflowingBinaryOperator>(BinOp)) {
|
|
||||||
VecOp->setHasNoSignedWrap(BinOp->hasNoSignedWrap());
|
|
||||||
VecOp->setHasNoUnsignedWrap(BinOp->hasNoUnsignedWrap());
|
|
||||||
}
|
|
||||||
if (VecOp && isa<PossiblyExactOperator>(VecOp))
|
|
||||||
VecOp->setIsExact(BinOp->isExact());
|
|
||||||
|
|
||||||
// Copy the fast-math flags.
|
|
||||||
if (VecOp && isa<FPMathOperator>(V))
|
|
||||||
VecOp->setFastMathFlags(it->getFastMathFlags());
|
|
||||||
|
|
||||||
Entry[Part] = V;
|
Entry[Part] = V;
|
||||||
}
|
}
|
||||||
|
24
test/Transforms/LoopVectorize/exact.ll
Normal file
24
test/Transforms/LoopVectorize/exact.ll
Normal file
@ -0,0 +1,24 @@
|
|||||||
|
; RUN: opt < %s -loop-vectorize -force-vector-width=4 -S | FileCheck %s
|
||||||
|
|
||||||
|
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
|
||||||
|
target triple = "x86_64-apple-macosx10.9.0"
|
||||||
|
|
||||||
|
; CHECK-LABEL: @lshr_exact(
|
||||||
|
; CHECK: lshr exact <4 x i32>
|
||||||
|
define void @lshr_exact(i32* %x) {
|
||||||
|
entry:
|
||||||
|
br label %for.body
|
||||||
|
|
||||||
|
for.body:
|
||||||
|
%iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
|
||||||
|
%arrayidx = getelementptr inbounds i32* %x, i64 %iv
|
||||||
|
%0 = load i32* %arrayidx, align 4
|
||||||
|
%conv1 = lshr exact i32 %0, 1
|
||||||
|
store i32 %conv1, i32* %arrayidx, align 4
|
||||||
|
%iv.next = add nuw nsw i64 %iv, 1
|
||||||
|
%exitcond = icmp eq i64 %iv.next, 256
|
||||||
|
br i1 %exitcond, label %for.end, label %for.body
|
||||||
|
|
||||||
|
for.end:
|
||||||
|
ret void
|
||||||
|
}
|
Reference in New Issue
Block a user