[llvm] [VectorCombine] Fold binop(shuffle(V1, Mask), shuffle(V2, Mask)) -> shuffle(binop(V1, V2), Mask) (PR #211530)
Yingwei Zheng via llvm-commits
llvm-commits at lists.llvm.org
Thu Jul 23 07:32:29 PDT 2026
- Previous message: [llvm] [VectorCombine] Fold binop(shuffle(V1, Mask), shuffle(V2, Mask)) -> shuffle(binop(V1, V2), Mask) (PR #211530)
- Next message: [llvm] [VectorCombine] Fold binop(shuffle(V1, Mask), shuffle(V2, Mask)) -> shuffle(binop(V1, V2), Mask) (PR #211530)
- Messages sorted by:
[ date ]
[ thread ]
[ subject ]
[ author ]
================
@@ -2443,6 +2444,70 @@ bool VectorCombine::foldConcatOfBoolMasks(Instruction &I) {
return true;
}
+// Fold binop(shuffle(V0, Mask), shuffle(V1, Mask))
+// --> shuffle(binop(V0, V1), Mask)
+bool VectorCombine::foldBinopOfShuffles(Instruction &I) {
+ // It is not safe to transform things like div, urem, etc. because we may
+ // create undefined behavior when executing those on unknown vector elements.
+ if (!isSafeToSpeculativelyExecute(&I))
+ return false;
+
+ // If both arguments of the binary operation are shuffles that use the same
+ // mask and shuffle within a single vector, move the shuffle after the binop.
+ ArrayRef<int> Mask;
+ Value *LHS, *RHS, *V0, *V1;
+ if (!match(&I, m_BinOp(m_Value(LHS), m_Value(RHS))) ||
+ !match(LHS, m_Shuffle(m_Value(V0), m_Poison(), m_Mask(Mask))) ||
+ !match(RHS, m_Shuffle(m_Value(V1), m_Poison(), m_SpecificMask(Mask))))
+ return false;
+
+ if (V0->getType() != V1->getType())
+ return false;
+
+ auto *BO = cast<BinaryOperator>(&I);
+ auto *OpTy = cast<VectorType>(V0->getType());
+ auto *DstTy = cast<VectorType>(I.getType());
+ Instruction::BinaryOps Opcode = BO->getOpcode();
+
+ InstructionCost OldCost;
+ OldCost += TTI.getInstructionCost(&I, CostKind);
+ OldCost += TTI.getInstructionCost(cast<Instruction>(LHS), CostKind);
+ if (LHS != RHS)
+ OldCost += TTI.getInstructionCost(cast<Instruction>(RHS), CostKind);
----------------
dtcxzyw wrote:
Is it covered by any test cases?
https://github.com/llvm/llvm-project/pull/211530
- Previous message: [llvm] [VectorCombine] Fold binop(shuffle(V1, Mask), shuffle(V2, Mask)) -> shuffle(binop(V1, V2), Mask) (PR #211530)
- Next message: [llvm] [VectorCombine] Fold binop(shuffle(V1, Mask), shuffle(V2, Mask)) -> shuffle(binop(V1, V2), Mask) (PR #211530)
- Messages sorted by:
[ date ]
[ thread ]
[ subject ]
[ author ]
More information about the llvm-commits
mailing list