[llvm] [VPlan] Add commutative binary OR matcher, use in transform. (PR #92539)
Florian Hahn via llvm-commits
llvm-commits at lists.llvm.org
Sun May 19 10:37:06 PDT 2024
https://github.com/fhahn updated https://github.com/llvm/llvm-project/pull/92539
>From 68a7751ce350e7bb580096d22fca556bbe84718e Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Fri, 17 May 2024 13:58:28 +0100
Subject: [PATCH 1/2] [VPlan] Simplify (X && Y) || (X && !Y) -> X. (PR #89386)
---
.../Transforms/Vectorize/VPlanPatternMatch.h | 8 +++++++-
.../Transforms/Vectorize/VPlanTransforms.cpp | 18 ++++++++++++++++-
.../LoopVectorize/AArch64/masked-call.ll | 9 +++------
.../AArch64/scalable-strict-fadd.ll | 3 +--
.../LoopVectorize/RISCV/uniform-load-store.ll | 20 ++++---------------
.../Transforms/LoopVectorize/uniform-blend.ll | 5 +----
.../unused-blend-mask-for-first-operand.ll | 12 ++---------
.../vplan-sink-scalars-and-merge.ll | 19 +++++-------------
8 files changed, 40 insertions(+), 54 deletions(-)
diff --git a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
index 50b08bbb7ebf7..56cbaa4201297 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
@@ -270,9 +270,15 @@ m_Mul(const Op0_t &Op0, const Op1_t &Op1) {
template <typename Op0_t, typename Op1_t>
inline AllBinaryRecipe_match<Op0_t, Op1_t, Instruction::Or>
-m_Or(const Op0_t &Op0, const Op1_t &Op1) {
+m_BinaryOr(const Op0_t &Op0, const Op1_t &Op1) {
return m_Binary<Instruction::Or, Op0_t, Op1_t>(Op0, Op1);
}
+
+template <typename Op0_t, typename Op1_t>
+inline BinaryVPInstruction_match<Op0_t, Op1_t, VPInstruction::LogicalAnd>
+m_LogicalAnd(const Op0_t &Op0, const Op1_t &Op1) {
+ return m_VPInstruction<VPInstruction::LogicalAnd, Op0_t, Op1_t>(Op0, Op1);
+}
} // namespace VPlanPatternMatch
} // namespace llvm
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index c0eb6d710ad34..5927345546e23 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -935,6 +935,22 @@ static void simplifyRecipe(VPRecipeBase &R, VPTypeAnalysis &TypeInfo) {
#endif
}
+ VPValue *X;
+ VPValue *Y;
+ VPValue *X1;
+ VPValue *Y1;
+ // Simplify (X && Y) || (X && !Y) -> X.
+ // TODO: Split up into simpler, modular combines: (X && Y) || (X && Z) into X
+ // && (Y || Z) and (X || !X) into true. This requires queuing newly created
+ // recipes to be visited during simplification.
+ if (match(&R,
+ m_BinaryOr(m_LogicalAnd(m_VPValue(X), m_VPValue(Y)),
+ m_LogicalAnd(m_VPValue(X1), m_Not(m_VPValue(Y1))))) &&
+ X == X1 && Y == Y1) {
+ R.getVPSingleValue()->replaceAllUsesWith(X);
+ return;
+ }
+
if (match(&R, m_CombineOr(m_Mul(m_VPValue(A), m_SpecificInt(1)),
m_Mul(m_SpecificInt(1), m_VPValue(A)))))
return R.getVPSingleValue()->replaceAllUsesWith(A);
@@ -1402,7 +1418,7 @@ void VPlanTransforms::dropPoisonGeneratingRecipes(
// for dependence analysis). Instead, replace it with an equivalent Add.
// This is possible as all users of the disjoint OR only access lanes
// where the operands are disjoint or poison otherwise.
- if (match(RecWithFlags, m_Or(m_VPValue(A), m_VPValue(B))) &&
+ if (match(RecWithFlags, m_BinaryOr(m_VPValue(A), m_VPValue(B))) &&
RecWithFlags->isDisjoint()) {
VPBuilder Builder(RecWithFlags);
VPInstruction *New = Builder.createOverflowingOp(
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll b/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll
index b91579106261a..d335ac4b69709 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll
@@ -223,10 +223,9 @@ define void @test_if_then(ptr noalias %a, ptr readnone %b) #4 {
; TFCOMMON-NEXT: [[TMP10:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[TMP9]])
; TFCOMMON-NEXT: [[TMP11:%.*]] = xor <vscale x 2 x i1> [[TMP8]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
; TFCOMMON-NEXT: [[TMP12:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP11]], <vscale x 2 x i1> zeroinitializer
-; TFCOMMON-NEXT: [[TMP13:%.*]] = or <vscale x 2 x i1> [[TMP9]], [[TMP12]]
; TFCOMMON-NEXT: [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP12]], <vscale x 2 x i64> zeroinitializer, <vscale x 2 x i64> [[TMP10]]
; TFCOMMON-NEXT: [[TMP14:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
-; TFCOMMON-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP14]], i32 8, <vscale x 2 x i1> [[TMP13]])
+; TFCOMMON-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP14]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
; TFCOMMON-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
; TFCOMMON-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
; TFCOMMON-NEXT: [[TMP15:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
@@ -272,16 +271,14 @@ define void @test_if_then(ptr noalias %a, ptr readnone %b) #4 {
; TFA_INTERLEAVE-NEXT: [[TMP20:%.*]] = xor <vscale x 2 x i1> [[TMP14]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
; TFA_INTERLEAVE-NEXT: [[TMP21:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP19]], <vscale x 2 x i1> zeroinitializer
; TFA_INTERLEAVE-NEXT: [[TMP22:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i1> [[TMP20]], <vscale x 2 x i1> zeroinitializer
-; TFA_INTERLEAVE-NEXT: [[TMP23:%.*]] = or <vscale x 2 x i1> [[TMP15]], [[TMP21]]
-; TFA_INTERLEAVE-NEXT: [[TMP24:%.*]] = or <vscale x 2 x i1> [[TMP16]], [[TMP22]]
; TFA_INTERLEAVE-NEXT: [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP21]], <vscale x 2 x i64> zeroinitializer, <vscale x 2 x i64> [[TMP17]]
; TFA_INTERLEAVE-NEXT: [[PREDPHI4:%.*]] = select <vscale x 2 x i1> [[TMP22]], <vscale x 2 x i64> zeroinitializer, <vscale x 2 x i64> [[TMP18]]
; TFA_INTERLEAVE-NEXT: [[TMP25:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
; TFA_INTERLEAVE-NEXT: [[TMP26:%.*]] = call i64 @llvm.vscale.i64()
; TFA_INTERLEAVE-NEXT: [[TMP27:%.*]] = mul i64 [[TMP26]], 2
; TFA_INTERLEAVE-NEXT: [[TMP28:%.*]] = getelementptr inbounds i64, ptr [[TMP25]], i64 [[TMP27]]
-; TFA_INTERLEAVE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP25]], i32 8, <vscale x 2 x i1> [[TMP23]])
-; TFA_INTERLEAVE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI4]], ptr [[TMP28]], i32 8, <vscale x 2 x i1> [[TMP24]])
+; TFA_INTERLEAVE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP25]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI4]], ptr [[TMP28]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
; TFA_INTERLEAVE-NEXT: [[INDEX_NEXT:%.*]] = add i64 [[INDEX]], [[TMP6]]
; TFA_INTERLEAVE-NEXT: [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
; TFA_INTERLEAVE-NEXT: [[TMP30:%.*]] = mul i64 [[TMP29]], 2
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll
index ddc004657ed5b..bcf8096f1b738 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll
@@ -1241,9 +1241,8 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
; CHECK-ORDERED-TF-NEXT: [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP16]], i32 4, <vscale x 4 x i1> [[TMP15]], <vscale x 4 x float> poison)
; CHECK-ORDERED-TF-NEXT: [[TMP17:%.*]] = xor <vscale x 4 x i1> [[TMP13]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
; CHECK-ORDERED-TF-NEXT: [[TMP18:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP17]], <vscale x 4 x i1> zeroinitializer
-; CHECK-ORDERED-TF-NEXT: [[TMP19:%.*]] = or <vscale x 4 x i1> [[TMP15]], [[TMP18]]
; CHECK-ORDERED-TF-NEXT: [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP18]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float 3.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer), <vscale x 4 x float> [[WIDE_MASKED_LOAD1]]
-; CHECK-ORDERED-TF-NEXT: [[TMP20:%.*]] = select <vscale x 4 x i1> [[TMP19]], <vscale x 4 x float> [[PREDPHI]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
+; CHECK-ORDERED-TF-NEXT: [[TMP20:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[PREDPHI]], <vscale x 4 x float> shufflevector (<vscale x 4 x float> insertelement (<vscale x 4 x float> poison, float -0.000000e+00, i64 0), <vscale x 4 x float> poison, <vscale x 4 x i32> zeroinitializer)
; CHECK-ORDERED-TF-NEXT: [[TMP21]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP20]])
; CHECK-ORDERED-TF-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP23]]
; CHECK-ORDERED-TF-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP9]])
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll b/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll
index 1ce4cb928e808..ee70f4aa35850 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll
@@ -462,13 +462,10 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca
; TF-SCALABLE-NEXT: [[TMP12:%.*]] = icmp ugt <vscale x 2 x i64> [[VEC_IND]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 10, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
; TF-SCALABLE-NEXT: [[TMP13:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP12]], <vscale x 2 x i1> zeroinitializer
; TF-SCALABLE-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 2 x i64> @llvm.masked.gather.nxv2i64.nxv2p0(<vscale x 2 x ptr> [[BROADCAST_SPLAT]], i32 8, <vscale x 2 x i1> [[TMP13]], <vscale x 2 x i64> poison)
-; TF-SCALABLE-NEXT: [[TMP14:%.*]] = xor <vscale x 2 x i1> [[TMP12]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TF-SCALABLE-NEXT: [[TMP15:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP14]], <vscale x 2 x i1> zeroinitializer
-; TF-SCALABLE-NEXT: [[TMP17:%.*]] = or <vscale x 2 x i1> [[TMP13]], [[TMP15]]
; TF-SCALABLE-NEXT: [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP13]], <vscale x 2 x i64> [[WIDE_MASKED_GATHER]], <vscale x 2 x i64> zeroinitializer
; TF-SCALABLE-NEXT: [[TMP16:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[TMP11]]
; TF-SCALABLE-NEXT: [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[TMP16]], i32 0
-; TF-SCALABLE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP18]], i32 8, <vscale x 2 x i1> [[TMP17]])
+; TF-SCALABLE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP18]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
; TF-SCALABLE-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP20]]
; TF-SCALABLE-NEXT: [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[DOTSPLAT]]
; TF-SCALABLE-NEXT: [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -510,13 +507,10 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca
; TF-FIXEDLEN-NEXT: [[TMP1:%.*]] = icmp ugt <4 x i64> [[VEC_IND]], <i64 10, i64 10, i64 10, i64 10>
; TF-FIXEDLEN-NEXT: [[TMP2:%.*]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i1> [[TMP1]], <4 x i1> zeroinitializer
; TF-FIXEDLEN-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <4 x i64> @llvm.masked.gather.v4i64.v4p0(<4 x ptr> [[BROADCAST_SPLAT]], i32 8, <4 x i1> [[TMP2]], <4 x i64> poison)
-; TF-FIXEDLEN-NEXT: [[TMP3:%.*]] = xor <4 x i1> [[TMP1]], <i1 true, i1 true, i1 true, i1 true>
-; TF-FIXEDLEN-NEXT: [[TMP4:%.*]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i1> [[TMP3]], <4 x i1> zeroinitializer
-; TF-FIXEDLEN-NEXT: [[TMP6:%.*]] = or <4 x i1> [[TMP2]], [[TMP4]]
; TF-FIXEDLEN-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP2]], <4 x i64> [[WIDE_MASKED_GATHER]], <4 x i64> zeroinitializer
; TF-FIXEDLEN-NEXT: [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[TMP0]]
; TF-FIXEDLEN-NEXT: [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
-; TF-FIXEDLEN-NEXT: call void @llvm.masked.store.v4i64.p0(<4 x i64> [[PREDPHI]], ptr [[TMP7]], i32 8, <4 x i1> [[TMP6]])
+; TF-FIXEDLEN-NEXT: call void @llvm.masked.store.v4i64.p0(<4 x i64> [[PREDPHI]], ptr [[TMP7]], i32 8, <4 x i1> [[ACTIVE_LANE_MASK]])
; TF-FIXEDLEN-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4
; TF-FIXEDLEN-NEXT: [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], <i64 4, i64 4, i64 4, i64 4>
; TF-FIXEDLEN-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1028
@@ -1296,12 +1290,9 @@ define void @conditional_uniform_store(ptr noalias nocapture %a, ptr noalias noc
; TF-SCALABLE-NEXT: [[TMP12:%.*]] = icmp ugt <vscale x 2 x i64> [[VEC_IND]], shufflevector (<vscale x 2 x i64> insertelement (<vscale x 2 x i64> poison, i64 10, i64 0), <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer)
; TF-SCALABLE-NEXT: [[TMP13:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP12]], <vscale x 2 x i1> zeroinitializer
; TF-SCALABLE-NEXT: call void @llvm.masked.scatter.nxv2i64.nxv2p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], <vscale x 2 x ptr> [[BROADCAST_SPLAT2]], i32 8, <vscale x 2 x i1> [[TMP13]])
-; TF-SCALABLE-NEXT: [[TMP15:%.*]] = xor <vscale x 2 x i1> [[TMP12]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
-; TF-SCALABLE-NEXT: [[TMP16:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP15]], <vscale x 2 x i1> zeroinitializer
-; TF-SCALABLE-NEXT: [[TMP17:%.*]] = or <vscale x 2 x i1> [[TMP13]], [[TMP16]]
; TF-SCALABLE-NEXT: [[TMP14:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[TMP11]]
; TF-SCALABLE-NEXT: [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[TMP14]], i32 0
-; TF-SCALABLE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP18]], i32 8, <vscale x 2 x i1> [[TMP17]])
+; TF-SCALABLE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP18]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
; TF-SCALABLE-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP20]]
; TF-SCALABLE-NEXT: [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[DOTSPLAT]]
; TF-SCALABLE-NEXT: [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -1344,12 +1335,9 @@ define void @conditional_uniform_store(ptr noalias nocapture %a, ptr noalias noc
; TF-FIXEDLEN-NEXT: [[TMP1:%.*]] = icmp ugt <4 x i64> [[VEC_IND]], <i64 10, i64 10, i64 10, i64 10>
; TF-FIXEDLEN-NEXT: [[TMP2:%.*]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i1> [[TMP1]], <4 x i1> zeroinitializer
; TF-FIXEDLEN-NEXT: call void @llvm.masked.scatter.v4i64.v4p0(<4 x i64> [[BROADCAST_SPLAT]], <4 x ptr> [[BROADCAST_SPLAT2]], i32 8, <4 x i1> [[TMP2]])
-; TF-FIXEDLEN-NEXT: [[TMP4:%.*]] = xor <4 x i1> [[TMP1]], <i1 true, i1 true, i1 true, i1 true>
-; TF-FIXEDLEN-NEXT: [[TMP5:%.*]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i1> [[TMP4]], <4 x i1> zeroinitializer
-; TF-FIXEDLEN-NEXT: [[TMP6:%.*]] = or <4 x i1> [[TMP2]], [[TMP5]]
; TF-FIXEDLEN-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[TMP0]]
; TF-FIXEDLEN-NEXT: [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; TF-FIXEDLEN-NEXT: call void @llvm.masked.store.v4i64.p0(<4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP7]], i32 8, <4 x i1> [[TMP6]])
+; TF-FIXEDLEN-NEXT: call void @llvm.masked.store.v4i64.p0(<4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP7]], i32 8, <4 x i1> [[ACTIVE_LANE_MASK]])
; TF-FIXEDLEN-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 4
; TF-FIXEDLEN-NEXT: [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], <i64 4, i64 4, i64 4, i64 4>
; TF-FIXEDLEN-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1028
diff --git a/llvm/test/Transforms/LoopVectorize/uniform-blend.ll b/llvm/test/Transforms/LoopVectorize/uniform-blend.ll
index 19cbcac6090c6..f33ec1419b114 100644
--- a/llvm/test/Transforms/LoopVectorize/uniform-blend.ll
+++ b/llvm/test/Transforms/LoopVectorize/uniform-blend.ll
@@ -86,11 +86,8 @@ define void @blend_chain_iv(i1 %c) {
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %vector.ph ], [ [[VEC_IND_NEXT:%.*]], %vector.body ]
; CHECK-NEXT: [[TMP6:%.*]] = select <4 x i1> [[MASK1]], <4 x i1> [[MASK1]], <4 x i1> zeroinitializer
-; CHECK-NEXT: [[TMP4:%.*]] = xor <4 x i1> [[MASK1]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT: [[TMP5:%.*]] = select <4 x i1> [[MASK1]], <4 x i1> [[TMP4]], <4 x i1> zeroinitializer
-; CHECK-NEXT: [[TMP8:%.*]] = or <4 x i1> [[TMP6]], [[TMP5]]
; CHECK-NEXT: [[PREDPHI:%.*]] = select <4 x i1> [[TMP6]], <4 x i64> [[VEC_IND]], <4 x i64> undef
-; CHECK-NEXT: [[PREDPHI1:%.*]] = select <4 x i1> [[TMP8]], <4 x i64> [[PREDPHI]], <4 x i64> undef
+; CHECK-NEXT: [[PREDPHI1:%.*]] = select <4 x i1> [[MASK1]], <4 x i64> [[PREDPHI]], <4 x i64> undef
; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i64> [[PREDPHI1]], i32 0
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds [32 x i16], ptr @dst, i16 0, i64 [[TMP9]]
; CHECK-NEXT: [[TMP11:%.*]] = extractelement <4 x i64> [[PREDPHI1]], i32 1
diff --git a/llvm/test/Transforms/LoopVectorize/unused-blend-mask-for-first-operand.ll b/llvm/test/Transforms/LoopVectorize/unused-blend-mask-for-first-operand.ll
index 0f7bd3d71feb4..d79b4a7cefc25 100644
--- a/llvm/test/Transforms/LoopVectorize/unused-blend-mask-for-first-operand.ll
+++ b/llvm/test/Transforms/LoopVectorize/unused-blend-mask-for-first-operand.ll
@@ -172,8 +172,6 @@ define void @test_not_first_lane_only_wide_compare_incoming_order_swapped(ptr %A
; CHECK: vector.ph:
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i16> poison, i16 [[X]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT]], <4 x i16> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT: [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <4 x i16> poison, i16 [[Y]], i64 0
-; CHECK-NEXT: [[BROADCAST_SPLAT2:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT1]], <4 x i16> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
@@ -184,14 +182,8 @@ define void @test_not_first_lane_only_wide_compare_incoming_order_swapped(ptr %A
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP2]], align 2
; CHECK-NEXT: [[TMP3:%.*]] = icmp ult <4 x i16> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
; CHECK-NEXT: [[TMP4:%.*]] = xor <4 x i1> [[TMP3]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT: [[TMP5:%.*]] = icmp ult <4 x i16> [[WIDE_LOAD]], [[BROADCAST_SPLAT2]]
-; CHECK-NEXT: [[TMP6:%.*]] = select <4 x i1> [[TMP4]], <4 x i1> [[TMP5]], <4 x i1> zeroinitializer
-; CHECK-NEXT: [[TMP7:%.*]] = xor <4 x i1> [[TMP5]], <i1 true, i1 true, i1 true, i1 true>
-; CHECK-NEXT: [[TMP8:%.*]] = select <4 x i1> [[TMP4]], <4 x i1> [[TMP7]], <4 x i1> zeroinitializer
-; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i1> [[TMP6]], i32 0
-; CHECK-NEXT: [[TMP10:%.*]] = extractelement <4 x i1> [[TMP8]], i32 0
-; CHECK-NEXT: [[TMP11:%.*]] = or i1 [[TMP9]], [[TMP10]]
-; CHECK-NEXT: [[PREDPHI:%.*]] = select i1 [[TMP11]], ptr [[B]], ptr poison
+; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i1> [[TMP4]], i32 0
+; CHECK-NEXT: [[PREDPHI:%.*]] = select i1 [[TMP9]], ptr [[B]], ptr poison
; CHECK-NEXT: [[TMP12:%.*]] = load i16, ptr [[PREDPHI]], align 2
; CHECK-NEXT: [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i16> poison, i16 [[TMP12]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT3]], <4 x i16> poison, <4 x i32> zeroinitializer
diff --git a/llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge.ll b/llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge.ll
index 1e60e57a5409d..ae5879bb2bae9 100644
--- a/llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge.ll
+++ b/llvm/test/Transforms/LoopVectorize/vplan-sink-scalars-and-merge.ll
@@ -361,15 +361,12 @@ define void @pred_cfg1(i32 %k, i32 %j) {
; CHECK-NEXT: Successor(s): then.0.0
; CHECK-EMPTY:
; CHECK-NEXT: then.0.0:
-; CHECK-NEXT: EMIT vp<[[NOT:%.+]]> = not ir<%c.1>
-; CHECK-NEXT: EMIT vp<[[MASK3:%.+]]> = logical-and vp<[[MASK1]]>, vp<[[NOT]]>
-; CHECK-NEXT: EMIT vp<[[OR:%.+]]> = or vp<[[MASK2]]>, vp<[[MASK3]]>
; CHECK-NEXT: BLEND ir<%p> = ir<0> vp<[[PRED]]>/vp<[[MASK2]]>
; CHECK-NEXT: Successor(s): pred.store
; CHECK-EMPTY:
; CHECK-NEXT: <xVFxUF> pred.store: {
; CHECK-NEXT: pred.store.entry:
-; CHECK-NEXT: BRANCH-ON-MASK vp<[[OR]]>
+; CHECK-NEXT: BRANCH-ON-MASK vp<[[MASK1]]>
; CHECK-NEXT: Successor(s): pred.store.if, pred.store.continue
; CHECK-EMPTY:
; CHECK-NEXT: pred.store.if:
@@ -462,16 +459,13 @@ define void @pred_cfg2(i32 %k, i32 %j) {
; CHECK-NEXT: Successor(s): then.0.0
; CHECK-EMPTY:
; CHECK-NEXT: then.0.0:
-; CHECK-NEXT: EMIT vp<[[NOT:%.+]]> = not ir<%c.0>
-; CHECK-NEXT: EMIT vp<[[MASK3:%.+]]> = logical-and vp<[[MASK1]]>, vp<[[NOT]]>
-; CHECK-NEXT: EMIT vp<[[OR:%.+]]> = or vp<[[MASK2]]>, vp<[[MASK3]]>
; CHECK-NEXT: BLEND ir<%p> = ir<0> vp<[[PRED]]>/vp<[[MASK2]]>
-; CHECK-NEXT: EMIT vp<[[MASK4:%.+]]> = logical-and vp<[[OR]]>, ir<%c.1>
+; CHECK-NEXT: EMIT vp<[[MASK3:%.+]]> = logical-and vp<[[MASK1]]>, ir<%c.1>
; CHECK-NEXT: Successor(s): pred.store
; CHECK-EMPTY:
; CHECK-NEXT: <xVFxUF> pred.store: {
; CHECK-NEXT: pred.store.entry:
-; CHECK-NEXT: BRANCH-ON-MASK vp<[[MASK4]]>
+; CHECK-NEXT: BRANCH-ON-MASK vp<[[MASK3]]>
; CHECK-NEXT: Successor(s): pred.store.if, pred.store.continue
; CHECK-EMPTY:
; CHECK-NEXT: pred.store.if:
@@ -570,16 +564,13 @@ define void @pred_cfg3(i32 %k, i32 %j) {
; CHECK-NEXT: Successor(s): then.0.0
; CHECK-EMPTY:
; CHECK-NEXT: then.0.0:
-; CHECK-NEXT: EMIT vp<[[NOT:%.+]]> = not ir<%c.0>
-; CHECK-NEXT: EMIT vp<[[MASK3:%.+]]> = logical-and vp<[[MASK1]]>, vp<[[NOT]]>
-; CHECK-NEXT: EMIT vp<[[MASK4:%.+]]> = or vp<[[MASK2]]>, vp<[[MASK3]]>
; CHECK-NEXT: BLEND ir<%p> = ir<0> vp<[[PRED]]>/vp<[[MASK2]]>
-; CHECK-NEXT: EMIT vp<[[MASK5:%.+]]> = logical-and vp<[[MASK4]]>, ir<%c.0>
+; CHECK-NEXT: EMIT vp<[[MASK3:%.+]]> = logical-and vp<[[MASK1]]>, ir<%c.0>
; CHECK-NEXT: Successor(s): pred.store
; CHECK-EMPTY:
; CHECK-NEXT: <xVFxUF> pred.store: {
; CHECK-NEXT: pred.store.entry:
-; CHECK-NEXT: BRANCH-ON-MASK vp<[[MASK5]]>
+; CHECK-NEXT: BRANCH-ON-MASK vp<[[MASK3]]>
; CHECK-NEXT: Successor(s): pred.store.if, pred.store.continue
; CHECK-EMPTY:
; CHECK-NEXT: pred.store.if:
>From 077555cf8580a28c42d9c506458b9c28d41912a4 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo at fhahn.com>
Date: Fri, 17 May 2024 13:59:34 +0100
Subject: [PATCH 2/2] [VPlan] Add commutative binary OR matcher, use in
transform.
Split off from https://github.com/llvm/llvm-project/pull/89386, this
extends the binary matcher to support matching commuative operations.
This is used for a new m_c_BinaryOr matcher, used in simplifyRecipe.
---
.../Transforms/Vectorize/VPlanPatternMatch.h | 39 ++++++++++++-------
.../Transforms/Vectorize/VPlanTransforms.cpp | 4 +-
.../LoopVectorize/AArch64/masked-call.ll | 9 ++---
.../LoopVectorize/AArch64/sve-tail-folding.ll | 3 +-
4 files changed, 32 insertions(+), 23 deletions(-)
diff --git a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
index 56cbaa4201297..21f395f62ca60 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanPatternMatch.h
@@ -157,7 +157,7 @@ using AllUnaryRecipe_match =
UnaryRecipe_match<Op0_t, Opcode, VPWidenRecipe, VPReplicateRecipe,
VPWidenCastRecipe, VPInstruction>;
-template <typename Op0_t, typename Op1_t, unsigned Opcode,
+template <typename Op0_t, typename Op1_t, unsigned Opcode, bool Commutative,
typename... RecipeTys>
struct BinaryRecipe_match {
Op0_t Op0;
@@ -179,18 +179,23 @@ struct BinaryRecipe_match {
return false;
assert(R->getNumOperands() == 2 &&
"recipe with matched opcode does not have 2 operands");
- return Op0.match(R->getOperand(0)) && Op1.match(R->getOperand(1));
+ if (Op0.match(R->getOperand(0)) && Op1.match(R->getOperand(1)))
+ return true;
+ return Commutative && Op0.match(R->getOperand(1)) &&
+ Op1.match(R->getOperand(0));
}
};
template <typename Op0_t, typename Op1_t, unsigned Opcode>
using BinaryVPInstruction_match =
- BinaryRecipe_match<Op0_t, Op1_t, Opcode, VPInstruction>;
+ BinaryRecipe_match<Op0_t, Op1_t, Opcode, /*Commutative*/ false,
+ VPInstruction>;
-template <typename Op0_t, typename Op1_t, unsigned Opcode>
+template <typename Op0_t, typename Op1_t, unsigned Opcode,
+ bool Commutative = false>
using AllBinaryRecipe_match =
- BinaryRecipe_match<Op0_t, Op1_t, Opcode, VPWidenRecipe, VPReplicateRecipe,
- VPWidenCastRecipe, VPInstruction>;
+ BinaryRecipe_match<Op0_t, Op1_t, Opcode, Commutative, VPWidenRecipe,
+ VPReplicateRecipe, VPWidenCastRecipe, VPInstruction>;
template <unsigned Opcode, typename Op0_t>
inline UnaryVPInstruction_match<Op0_t, Opcode>
@@ -256,10 +261,11 @@ m_ZExtOrSExt(const Op0_t &Op0) {
return m_CombineOr(m_ZExt(Op0), m_SExt(Op0));
}
-template <unsigned Opcode, typename Op0_t, typename Op1_t>
-inline AllBinaryRecipe_match<Op0_t, Op1_t, Opcode> m_Binary(const Op0_t &Op0,
- const Op1_t &Op1) {
- return AllBinaryRecipe_match<Op0_t, Op1_t, Opcode>(Op0, Op1);
+template <unsigned Opcode, typename Op0_t, typename Op1_t,
+ bool Commutative = false>
+inline AllBinaryRecipe_match<Op0_t, Op1_t, Opcode, Commutative>
+m_Binary(const Op0_t &Op0, const Op1_t &Op1) {
+ return AllBinaryRecipe_match<Op0_t, Op1_t, Opcode, Commutative>(Op0, Op1);
}
template <typename Op0_t, typename Op1_t>
@@ -268,10 +274,17 @@ m_Mul(const Op0_t &Op0, const Op1_t &Op1) {
return m_Binary<Instruction::Mul, Op0_t, Op1_t>(Op0, Op1);
}
-template <typename Op0_t, typename Op1_t>
-inline AllBinaryRecipe_match<Op0_t, Op1_t, Instruction::Or>
+template <typename Op0_t, typename Op1_t, bool Commutative = false>
+inline AllBinaryRecipe_match<Op0_t, Op1_t, Instruction::Or, Commutative>
m_BinaryOr(const Op0_t &Op0, const Op1_t &Op1) {
- return m_Binary<Instruction::Or, Op0_t, Op1_t>(Op0, Op1);
+ return m_Binary<Instruction::Or, Op0_t, Op1_t, Commutative>(Op0, Op1);
+}
+
+template <typename Op0_t, typename Op1_t>
+inline AllBinaryRecipe_match<Op0_t, Op1_t, Instruction::Or,
+ /*Commutative*/ true>
+m_c_BinaryOr(const Op0_t &Op0, const Op1_t &Op1) {
+ return m_BinaryOr<Op0_t, Op1_t, /*Commutative*/ true>(Op0, Op1);
}
template <typename Op0_t, typename Op1_t>
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 5927345546e23..2e7e0fe1b1ab3 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -944,8 +944,8 @@ static void simplifyRecipe(VPRecipeBase &R, VPTypeAnalysis &TypeInfo) {
// && (Y || Z) and (X || !X) into true. This requires queuing newly created
// recipes to be visited during simplification.
if (match(&R,
- m_BinaryOr(m_LogicalAnd(m_VPValue(X), m_VPValue(Y)),
- m_LogicalAnd(m_VPValue(X1), m_Not(m_VPValue(Y1))))) &&
+ m_c_BinaryOr(m_LogicalAnd(m_VPValue(X), m_VPValue(Y)),
+ m_LogicalAnd(m_VPValue(X1), m_Not(m_VPValue(Y1))))) &&
X == X1 && Y == Y1) {
R.getVPSingleValue()->replaceAllUsesWith(X);
return;
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll b/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll
index d335ac4b69709..200c2adcf0e66 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/masked-call.ll
@@ -402,10 +402,9 @@ define void @test_widen_if_then_else(ptr noalias %a, ptr readnone %b) #4 {
; TFCOMMON-NEXT: [[TMP11:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> zeroinitializer, <vscale x 2 x i1> [[TMP10]])
; TFCOMMON-NEXT: [[TMP12:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i1> [[TMP8]], <vscale x 2 x i1> zeroinitializer
; TFCOMMON-NEXT: [[TMP13:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[TMP12]])
-; TFCOMMON-NEXT: [[TMP14:%.*]] = or <vscale x 2 x i1> [[TMP10]], [[TMP12]]
; TFCOMMON-NEXT: [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP10]], <vscale x 2 x i64> [[TMP11]], <vscale x 2 x i64> [[TMP13]]
; TFCOMMON-NEXT: [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
-; TFCOMMON-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP15]], i32 8, <vscale x 2 x i1> [[TMP14]])
+; TFCOMMON-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP15]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
; TFCOMMON-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
; TFCOMMON-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
; TFCOMMON-NEXT: [[TMP16:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 2 x i1> insertelement (<vscale x 2 x i1> poison, i1 true, i64 0), <vscale x 2 x i1> poison, <vscale x 2 x i32> zeroinitializer)
@@ -453,16 +452,14 @@ define void @test_widen_if_then_else(ptr noalias %a, ptr readnone %b) #4 {
; TFA_INTERLEAVE-NEXT: [[TMP22:%.*]] = select <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]], <vscale x 2 x i1> [[TMP14]], <vscale x 2 x i1> zeroinitializer
; TFA_INTERLEAVE-NEXT: [[TMP23:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD]], <vscale x 2 x i1> [[TMP21]])
; TFA_INTERLEAVE-NEXT: [[TMP24:%.*]] = call <vscale x 2 x i64> @foo_vector(<vscale x 2 x i64> [[WIDE_MASKED_LOAD3]], <vscale x 2 x i1> [[TMP22]])
-; TFA_INTERLEAVE-NEXT: [[TMP25:%.*]] = or <vscale x 2 x i1> [[TMP17]], [[TMP21]]
-; TFA_INTERLEAVE-NEXT: [[TMP26:%.*]] = or <vscale x 2 x i1> [[TMP18]], [[TMP22]]
; TFA_INTERLEAVE-NEXT: [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP17]], <vscale x 2 x i64> [[TMP19]], <vscale x 2 x i64> [[TMP23]]
; TFA_INTERLEAVE-NEXT: [[PREDPHI4:%.*]] = select <vscale x 2 x i1> [[TMP18]], <vscale x 2 x i64> [[TMP20]], <vscale x 2 x i64> [[TMP24]]
; TFA_INTERLEAVE-NEXT: [[TMP27:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
; TFA_INTERLEAVE-NEXT: [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
; TFA_INTERLEAVE-NEXT: [[TMP29:%.*]] = mul i64 [[TMP28]], 2
; TFA_INTERLEAVE-NEXT: [[TMP30:%.*]] = getelementptr inbounds i64, ptr [[TMP27]], i64 [[TMP29]]
-; TFA_INTERLEAVE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP27]], i32 8, <vscale x 2 x i1> [[TMP25]])
-; TFA_INTERLEAVE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI4]], ptr [[TMP30]], i32 8, <vscale x 2 x i1> [[TMP26]])
+; TFA_INTERLEAVE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI]], ptr [[TMP27]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; TFA_INTERLEAVE-NEXT: call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[PREDPHI4]], ptr [[TMP30]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK2]])
; TFA_INTERLEAVE-NEXT: [[INDEX_NEXT:%.*]] = add i64 [[INDEX]], [[TMP6]]
; TFA_INTERLEAVE-NEXT: [[TMP31:%.*]] = call i64 @llvm.vscale.i64()
; TFA_INTERLEAVE-NEXT: [[TMP32:%.*]] = mul i64 [[TMP31]], 2
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding.ll
index 2b2742ca7ccbc..63ad98b2d8ab2 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding.ll
@@ -480,11 +480,10 @@ define void @cond_uniform_load(ptr noalias %dst, ptr noalias readonly %src, ptr
; CHECK-NEXT: [[TMP15:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP14]], <vscale x 4 x i1> zeroinitializer
; CHECK-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[BROADCAST_SPLAT]], i32 4, <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> poison)
; CHECK-NEXT: [[TMP16:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> zeroinitializer
-; CHECK-NEXT: [[TMP18:%.*]] = or <vscale x 4 x i1> [[TMP15]], [[TMP16]]
; CHECK-NEXT: [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP16]], <vscale x 4 x i32> zeroinitializer, <vscale x 4 x i32> [[WIDE_MASKED_GATHER]]
; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 [[TMP10]]
; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
-; CHECK-NEXT: call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[PREDPHI]], ptr [[TMP19]], i32 4, <vscale x 4 x i1> [[TMP18]])
+; CHECK-NEXT: call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[PREDPHI]], ptr [[TMP19]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
; CHECK-NEXT: [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP21]]
; CHECK-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[TMP9]])
; CHECK-NEXT: [[TMP22:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer)
More information about the llvm-commits
mailing list