[llvm] [RISCV] Refine cost on Min/Max reduction with i1 type (PR #79401)
Shih-Po Hung via llvm-commits
llvm-commits at lists.llvm.org
Thu Jan 25 01:00:55 PST 2024
https://github.com/arcbbb updated https://github.com/llvm/llvm-project/pull/79401
>From fcf73ae93903dff1b55efe82224f6323defe3f3a Mon Sep 17 00:00:00 2001
From: ShihPo Hung <shihpo.hung at sifive.com>
Date: Wed, 24 Jan 2024 19:03:52 -0800
Subject: [PATCH 1/2] [RISCV] Refine cost on Min/Max reduction with i1 type
InstCombine transform min/max reduction with i1 into arithmetic
reduction.
This patch reuse the cost logic in arithmetic reduction cost function.
---
.../Target/RISCV/RISCVTargetTransformInfo.cpp | 13 +++++---
.../Analysis/CostModel/RISCV/reduce-max.ll | 32 +++++++++----------
.../Analysis/CostModel/RISCV/reduce-min.ll | 32 +++++++++----------
3 files changed, 41 insertions(+), 36 deletions(-)
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
index 0799af0a7bad550..87b36b15a900562 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
@@ -936,10 +936,15 @@ RISCVTTIImpl::getMinMaxReductionCost(Intrinsic::ID IID, VectorType *Ty,
return BaseT::getMinMaxReductionCost(IID, Ty, FMF, CostKind);
std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(Ty);
- if (Ty->getElementType()->isIntegerTy(1))
- // vcpop sequences, see vreduction-mask.ll. umax, smin actually only
- // cost 2, but we don't have enough info here so we slightly over cost.
- return (LT.first - 1) + 3;
+ if (Ty->getElementType()->isIntegerTy(1)) {
+ // InstCombine does following transform:
+ // vector_reduce_u{min,max}(<n x i1>) --> vector_reduce_{and,or}(<n x i1>)
+ // vector_reduce_s{min,max}(<n x i1>) --> vector_reduce_{or,and}(<n x i1>)
+ if ((IID == Intrinsic::umax) || (IID == Intrinsic::smin))
+ return getArithmeticReductionCost(Instruction::Or, Ty, FMF, CostKind);
+ else
+ return getArithmeticReductionCost(Instruction::And, Ty, FMF, CostKind);
+ }
// IR Reduction is composed by two vmv and one rvv reduction instruction.
InstructionCost BaseCost = 2;
diff --git a/llvm/test/Analysis/CostModel/RISCV/reduce-max.ll b/llvm/test/Analysis/CostModel/RISCV/reduce-max.ll
index c21b8520112cf88..f3436b54df90ab2 100644
--- a/llvm/test/Analysis/CostModel/RISCV/reduce-max.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/reduce-max.ll
@@ -6,25 +6,25 @@
define i32 @reduce_umin_i1(i32 %arg) {
; CHECK-LABEL: 'reduce_umin_i1'
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V1 = call i1 @llvm.vector.reduce.umax.v1i1(<1 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i1 @llvm.vector.reduce.umax.v2i1(<2 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i1 @llvm.vector.reduce.umax.v4i1(<4 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8 = call i1 @llvm.vector.reduce.umax.v8i1(<8 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V16 = call i1 @llvm.vector.reduce.umax.v16i1(<16 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V32 = call i1 @llvm.vector.reduce.umax.v32i1(<32 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V64 = call i1 @llvm.vector.reduce.umax.v64i1(<64 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V128 = call i1 @llvm.vector.reduce.umax.v128i1(<128 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V1 = call i1 @llvm.vector.reduce.umax.v1i1(<1 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2 = call i1 @llvm.vector.reduce.umax.v2i1(<2 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4 = call i1 @llvm.vector.reduce.umax.v4i1(<4 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i1 @llvm.vector.reduce.umax.v8i1(<8 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16 = call i1 @llvm.vector.reduce.umax.v16i1(<16 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32 = call i1 @llvm.vector.reduce.umax.v32i1(<32 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i1 @llvm.vector.reduce.umax.v64i1(<64 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V128 = call i1 @llvm.vector.reduce.umax.v128i1(<128 x i1> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
;
; SIZE-LABEL: 'reduce_umin_i1'
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V1 = call i1 @llvm.vector.reduce.umax.v1i1(<1 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i1 @llvm.vector.reduce.umax.v2i1(<2 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i1 @llvm.vector.reduce.umax.v4i1(<4 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8 = call i1 @llvm.vector.reduce.umax.v8i1(<8 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V16 = call i1 @llvm.vector.reduce.umax.v16i1(<16 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V32 = call i1 @llvm.vector.reduce.umax.v32i1(<32 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V64 = call i1 @llvm.vector.reduce.umax.v64i1(<64 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V128 = call i1 @llvm.vector.reduce.umax.v128i1(<128 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V1 = call i1 @llvm.vector.reduce.umax.v1i1(<1 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2 = call i1 @llvm.vector.reduce.umax.v2i1(<2 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4 = call i1 @llvm.vector.reduce.umax.v4i1(<4 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i1 @llvm.vector.reduce.umax.v8i1(<8 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16 = call i1 @llvm.vector.reduce.umax.v16i1(<16 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32 = call i1 @llvm.vector.reduce.umax.v32i1(<32 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i1 @llvm.vector.reduce.umax.v64i1(<64 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V128 = call i1 @llvm.vector.reduce.umax.v128i1(<128 x i1> undef)
; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
%V1 = call i1 @llvm.vector.reduce.umax.v1i1(<1 x i1> undef)
diff --git a/llvm/test/Analysis/CostModel/RISCV/reduce-min.ll b/llvm/test/Analysis/CostModel/RISCV/reduce-min.ll
index 941c3cf8a999c87..1964e35b81b1a65 100644
--- a/llvm/test/Analysis/CostModel/RISCV/reduce-min.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/reduce-min.ll
@@ -176,25 +176,25 @@ define i32 @reduce_umin_i64(i32 %arg) {
define i32 @reduce_smin_i1(i32 %arg) {
; CHECK-LABEL: 'reduce_smin_i1'
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V1 = call i1 @llvm.vector.reduce.smin.v1i1(<1 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i1 @llvm.vector.reduce.smin.v2i1(<2 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i1 @llvm.vector.reduce.smin.v4i1(<4 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8 = call i1 @llvm.vector.reduce.smin.v8i1(<8 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V16 = call i1 @llvm.vector.reduce.smin.v16i1(<16 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V32 = call i1 @llvm.vector.reduce.smin.v32i1(<32 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V64 = call i1 @llvm.vector.reduce.smin.v64i1(<64 x i1> undef)
-; CHECK-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V128 = call i1 @llvm.vector.reduce.smin.v128i1(<128 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V1 = call i1 @llvm.vector.reduce.smin.v1i1(<1 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2 = call i1 @llvm.vector.reduce.smin.v2i1(<2 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4 = call i1 @llvm.vector.reduce.smin.v4i1(<4 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i1 @llvm.vector.reduce.smin.v8i1(<8 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16 = call i1 @llvm.vector.reduce.smin.v16i1(<16 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32 = call i1 @llvm.vector.reduce.smin.v32i1(<32 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i1 @llvm.vector.reduce.smin.v64i1(<64 x i1> undef)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V128 = call i1 @llvm.vector.reduce.smin.v128i1(<128 x i1> undef)
; CHECK-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef
;
; SIZE-LABEL: 'reduce_smin_i1'
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V1 = call i1 @llvm.vector.reduce.smin.v1i1(<1 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V2 = call i1 @llvm.vector.reduce.smin.v2i1(<2 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V4 = call i1 @llvm.vector.reduce.smin.v4i1(<4 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V8 = call i1 @llvm.vector.reduce.smin.v8i1(<8 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V16 = call i1 @llvm.vector.reduce.smin.v16i1(<16 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V32 = call i1 @llvm.vector.reduce.smin.v32i1(<32 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V64 = call i1 @llvm.vector.reduce.smin.v64i1(<64 x i1> undef)
-; SIZE-NEXT: Cost Model: Found an estimated cost of 3 for instruction: %V128 = call i1 @llvm.vector.reduce.smin.v128i1(<128 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V1 = call i1 @llvm.vector.reduce.smin.v1i1(<1 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V2 = call i1 @llvm.vector.reduce.smin.v2i1(<2 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V4 = call i1 @llvm.vector.reduce.smin.v4i1(<4 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8 = call i1 @llvm.vector.reduce.smin.v8i1(<8 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16 = call i1 @llvm.vector.reduce.smin.v16i1(<16 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32 = call i1 @llvm.vector.reduce.smin.v32i1(<32 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V64 = call i1 @llvm.vector.reduce.smin.v64i1(<64 x i1> undef)
+; SIZE-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V128 = call i1 @llvm.vector.reduce.smin.v128i1(<128 x i1> undef)
; SIZE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret i32 undef
;
%V1 = call i1 @llvm.vector.reduce.smin.v1i1(<1 x i1> undef)
>From 0e866364a1fb4c4671727a4617d067ff1f7d55ac Mon Sep 17 00:00:00 2001
From: ShihPo Hung <shihpo.hung at sifive.com>
Date: Thu, 25 Jan 2024 00:58:30 -0800
Subject: [PATCH 2/2] Address comments
---
llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
index 87b36b15a900562..23a0ec623fe6824 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
@@ -937,10 +937,10 @@ RISCVTTIImpl::getMinMaxReductionCost(Intrinsic::ID IID, VectorType *Ty,
std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(Ty);
if (Ty->getElementType()->isIntegerTy(1)) {
- // InstCombine does following transform:
- // vector_reduce_u{min,max}(<n x i1>) --> vector_reduce_{and,or}(<n x i1>)
- // vector_reduce_s{min,max}(<n x i1>) --> vector_reduce_{or,and}(<n x i1>)
- if ((IID == Intrinsic::umax) || (IID == Intrinsic::smin))
+ // SelectionDAGBuilder does following transforms:
+ // vector_reduce_{smin,umax}(<n x i1>) --> vector_reduce_or(<n x i1>)
+ // vector_reduce_{smax,umin}(<n x i1>) --> vector_reduce_and(<n x i1>)
+ if (IID == Intrinsic::umax || IID == Intrinsic::smin)
return getArithmeticReductionCost(Instruction::Or, Ty, FMF, CostKind);
else
return getArithmeticReductionCost(Instruction::And, Ty, FMF, CostKind);
More information about the llvm-commits
mailing list