[llvm] [SLP]Do not include the cost of and -1, <v> and emit just <v> after MinBitWidth. (PR #90739)
Alexey Bataev via llvm-commits
llvm-commits at lists.llvm.org
Wed May 1 09:48:10 PDT 2024
https://github.com/alexey-bataev updated https://github.com/llvm/llvm-project/pull/90739
>From b373cc03cf34e624be6c8eb1b367ebc371685287 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev at outlook.com>
Date: Wed, 1 May 2024 15:31:01 +0000
Subject: [PATCH 1/2] =?UTF-8?q?[=F0=9D=98=80=F0=9D=97=BD=F0=9D=97=BF]=20in?=
=?UTF-8?q?itial=20version?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Created using spr 1.3.5
---
.../Transforms/Vectorize/SLPVectorizer.cpp | 24 +++++++++++++++++++
.../RISCV/minbw-with-and-and-scalar-trunc.ll | 3 +--
.../RISCV/trunc-to-large-than-bw.ll | 3 +--
3 files changed, 26 insertions(+), 4 deletions(-)
diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index c33d90d531bf53..914a7b205256c2 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -9484,6 +9484,16 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
Op1Info, Op2Info, Operands, VI);
};
auto GetVectorCost = [=](InstructionCost CommonCost) {
+ if (ShuffleOrOp == Instruction::And && It != MinBWs.end()) {
+ for (unsigned I : seq<unsigned>(0, E->getNumOperands())) {
+ ArrayRef<Value *> Ops = E->getOperand(I);
+ if (all_of(Ops, [&](Value *Op) {
+ auto *CI = dyn_cast<ConstantInt>(Op);
+ return CI && CI->getValue().countr_one() == It->second.first;
+ }))
+ return CommonCost;
+ }
+ }
unsigned OpIdx = isa<UnaryOperator>(VL0) ? 0 : 1;
TTI::OperandValueInfo Op1Info = getOperandInfo(E->getOperand(0));
TTI::OperandValueInfo Op2Info = getOperandInfo(E->getOperand(OpIdx));
@@ -12969,6 +12979,20 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
LLVM_DEBUG(dbgs() << "SLP: Diamond merged for " << *VL0 << ".\n");
return E->VectorizedValue;
}
+ if (ShuffleOrOp == Instruction::And && It != MinBWs.end()) {
+ for (unsigned I : seq<unsigned>(0, E->getNumOperands())) {
+ ArrayRef<Value *> Ops = E->getOperand(I);
+ if (all_of(Ops, [&](Value *Op) {
+ auto *CI = dyn_cast<ConstantInt>(Op);
+ return CI && CI->getValue().countr_one() == It->second.first;
+ })) {
+ V = FinalShuffle(I == 0 ? RHS : LHS, E, VecTy);
+ E->VectorizedValue = V;
+ ++NumVectorInstructions;
+ return V;
+ }
+ }
+ }
if (LHS->getType() != VecTy || RHS->getType() != VecTy) {
assert((It != MinBWs.end() ||
getOperandEntry(E, 0)->State == TreeEntry::NeedToGather ||
diff --git a/llvm/test/Transforms/SLPVectorizer/RISCV/minbw-with-and-and-scalar-trunc.ll b/llvm/test/Transforms/SLPVectorizer/RISCV/minbw-with-and-and-scalar-trunc.ll
index fc977585614baf..d6dc3bcc3354c1 100644
--- a/llvm/test/Transforms/SLPVectorizer/RISCV/minbw-with-and-and-scalar-trunc.ll
+++ b/llvm/test/Transforms/SLPVectorizer/RISCV/minbw-with-and-and-scalar-trunc.ll
@@ -12,8 +12,7 @@ define i16 @test() {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TMP0:%.*]] = call <4 x i64> @llvm.experimental.vp.strided.load.v4i64.p0.i64(ptr align 8 @c, i64 24, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, i32 4)
; CHECK-NEXT: [[TMP1:%.*]] = trunc <4 x i64> [[TMP0]] to <4 x i16>
-; CHECK-NEXT: [[TMP2:%.*]] = and <4 x i16> [[TMP1]], <i16 -1, i16 -1, i16 -1, i16 -1>
-; CHECK-NEXT: [[TMP3:%.*]] = xor <4 x i16> [[TMP2]], <i16 -1, i16 -1, i16 -1, i16 -1>
+; CHECK-NEXT: [[TMP3:%.*]] = xor <4 x i16> [[TMP1]], <i16 -1, i16 -1, i16 -1, i16 -1>
; CHECK-NEXT: [[TMP4:%.*]] = call i16 @llvm.vector.reduce.umax.v4i16(<4 x i16> [[TMP3]])
; CHECK-NEXT: [[TMP5:%.*]] = zext i16 [[TMP4]] to i32
; CHECK-NEXT: [[T:%.*]] = trunc i32 [[TMP5]] to i16
diff --git a/llvm/test/Transforms/SLPVectorizer/RISCV/trunc-to-large-than-bw.ll b/llvm/test/Transforms/SLPVectorizer/RISCV/trunc-to-large-than-bw.ll
index 04d275742832ef..0c0c723e669961 100644
--- a/llvm/test/Transforms/SLPVectorizer/RISCV/trunc-to-large-than-bw.ll
+++ b/llvm/test/Transforms/SLPVectorizer/RISCV/trunc-to-large-than-bw.ll
@@ -9,8 +9,7 @@ define i32 @test() {
; CHECK-NEXT: entry:
; CHECK-NEXT: [[TMP0:%.*]] = call <4 x i64> @llvm.experimental.vp.strided.load.v4i64.p0.i64(ptr align 8 @c, i64 24, <4 x i1> <i1 true, i1 true, i1 true, i1 true>, i32 4)
; CHECK-NEXT: [[TMP1:%.*]] = trunc <4 x i64> [[TMP0]] to <4 x i16>
-; CHECK-NEXT: [[TMP2:%.*]] = and <4 x i16> [[TMP1]], <i16 -1, i16 -1, i16 -1, i16 -1>
-; CHECK-NEXT: [[TMP3:%.*]] = xor <4 x i16> [[TMP2]], <i16 -1, i16 -1, i16 -1, i16 -1>
+; CHECK-NEXT: [[TMP3:%.*]] = xor <4 x i16> [[TMP1]], <i16 -1, i16 -1, i16 -1, i16 -1>
; CHECK-NEXT: [[TMP4:%.*]] = call i16 @llvm.vector.reduce.umax.v4i16(<4 x i16> [[TMP3]])
; CHECK-NEXT: [[TMP5:%.*]] = zext i16 [[TMP4]] to i32
; CHECK-NEXT: [[TMP6:%.*]] = call i32 @llvm.umax.i32(i32 [[TMP5]], i32 1)
>From c3cac1932903629c8fd078cf71547d41bb293361 Mon Sep 17 00:00:00 2001
From: Alexey Bataev <a.bataev at outlook.com>
Date: Wed, 1 May 2024 16:48:01 +0000
Subject: [PATCH 2/2] Address comment
Created using spr 1.3.5
---
llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 914a7b205256c2..038bceb39cb8db 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -9489,7 +9489,7 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
ArrayRef<Value *> Ops = E->getOperand(I);
if (all_of(Ops, [&](Value *Op) {
auto *CI = dyn_cast<ConstantInt>(Op);
- return CI && CI->getValue().countr_one() == It->second.first;
+ return CI && CI->getValue().countr_one() >= It->second.first;
}))
return CommonCost;
}
@@ -12984,7 +12984,7 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
ArrayRef<Value *> Ops = E->getOperand(I);
if (all_of(Ops, [&](Value *Op) {
auto *CI = dyn_cast<ConstantInt>(Op);
- return CI && CI->getValue().countr_one() == It->second.first;
+ return CI && CI->getValue().countr_one() >= It->second.first;
})) {
V = FinalShuffle(I == 0 ? RHS : LHS, E, VecTy);
E->VectorizedValue = V;
More information about the llvm-commits
mailing list