[SLP] Fix PR35047: Fix default cost model for cast op in X86.

Summary:
The cost calculation for default case on X86 target does not always
follow correct wayt because of missing 4-th argument in
`BaseT::getCastInstrCost()` call. Added this missing parameter.

Reviewers: hfinkel, mkuper, RKSimon, spatel

Subscribers: llvm-commits

Differential Revision: https://reviews.llvm.org/D39687

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@317576 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/X86/X86TargetTransformInfo.cpp b/lib/Target/X86/X86TargetTransformInfo.cpp
index 8bddf574554f537b66f159b1ede1ee3c82b441fd..6772d96c7997a5156717be68599472a750460206 100644 (file)
--- a/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -1437,7 +1437,7 @@ int X86TTIImpl::getCastInstrCost(unsigned Opcode, Type *Dst, Type *Src,
       return Entry->Cost;
   }
 
-  return BaseT::getCastInstrCost(Opcode, Dst, Src);
+  return BaseT::getCastInstrCost(Opcode, Dst, Src, I);
 }
 
 int X86TTIImpl::getCmpSelInstrCost(unsigned Opcode, Type *ValTy, Type *CondTy,

diff --git a/test/Transforms/SLPVectorizer/X86/cast.ll b/test/Transforms/SLPVectorizer/X86/cast.ll
index 5d7118753e9281aa29f97c0b2ff30e73507cbb0c..2f9f84948eaf0f9b25e48ac6e69dd25e54a378de 100644 (file)
--- a/test/Transforms/SLPVectorizer/X86/cast.ll
+++ b/test/Transforms/SLPVectorizer/X86/cast.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 -basicaa -slp-vectorizer -dce -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE
-; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -basicaa -slp-vectorizer -dce -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7 -basicaa -slp-vectorizer -dce -S | FileCheck %s
+; RUN: opt < %s -mtriple=x86_64-apple-macosx10.8.0 -mcpu=corei7-avx -basicaa -slp-vectorizer -dce -S | FileCheck %s
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
 
@@ -14,10 +14,10 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
 define i32 @test_sext_4i8_to_4i32(i32* noalias nocapture %A, i8* noalias nocapture %B) {
 ; CHECK-LABEL: @test_sext_4i8_to_4i32(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* %B to <4 x i8>*
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i8* [[B:%.*]] to <4 x i8>*
 ; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i8>, <4 x i8>* [[TMP0]], align 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = sext <4 x i8> [[TMP1]] to <4 x i32>
-; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* %A to <4 x i32>*
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to <4 x i32>*
 ; CHECK-NEXT:    store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 4
 ; CHECK-NEXT:    ret i32 undef
 ;
@@ -46,10 +46,10 @@ entry:
 define i32 @test_zext_4i16_to_4i32(i32* noalias nocapture %A, i16* noalias nocapture %B) {
 ; CHECK-LABEL: @test_zext_4i16_to_4i32(
 ; CHECK-NEXT:  entry:
-; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* %B to <4 x i16>*
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[B:%.*]] to <4 x i16>*
 ; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = zext <4 x i16> [[TMP1]] to <4 x i32>
-; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* %A to <4 x i32>*
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i32* [[A:%.*]] to <4 x i32>*
 ; CHECK-NEXT:    store <4 x i32> [[TMP2]], <4 x i32>* [[TMP3]], align 4
 ; CHECK-NEXT:    ret i32 undef
 ;
@@ -76,30 +76,21 @@ entry:
 }
 
 define i64 @test_sext_4i16_to_4i64(i64* noalias nocapture %A, i16* noalias nocapture %B) {
-; SSE-LABEL: @test_sext_4i16_to_4i64(
-; SSE-NEXT:  entry:
-; SSE-NEXT:    [[TMP0:%.*]] = bitcast i16* %B to <2 x i16>*
-; SSE-NEXT:    [[TMP1:%.*]] = load <2 x i16>, <2 x i16>* [[TMP0]], align 1
-; SSE-NEXT:    [[TMP2:%.*]] = sext <2 x i16> [[TMP1]] to <2 x i64>
-; SSE-NEXT:    [[TMP3:%.*]] = bitcast i64* %A to <2 x i64>*
-; SSE-NEXT:    store <2 x i64> [[TMP2]], <2 x i64>* [[TMP3]], align 4
-; SSE-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* %B, i64 2
-; SSE-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i64, i64* %A, i64 2
-; SSE-NEXT:    [[TMP4:%.*]] = bitcast i16* [[ARRAYIDX5]] to <2 x i16>*
-; SSE-NEXT:    [[TMP5:%.*]] = load <2 x i16>, <2 x i16>* [[TMP4]], align 1
-; SSE-NEXT:    [[TMP6:%.*]] = sext <2 x i16> [[TMP5]] to <2 x i64>
-; SSE-NEXT:    [[TMP7:%.*]] = bitcast i64* [[ARRAYIDX7]] to <2 x i64>*
-; SSE-NEXT:    store <2 x i64> [[TMP6]], <2 x i64>* [[TMP7]], align 4
-; SSE-NEXT:    ret i64 undef
-;
-; AVX-LABEL: @test_sext_4i16_to_4i64(
-; AVX-NEXT:  entry:
-; AVX-NEXT:    [[TMP0:%.*]] = bitcast i16* %B to <4 x i16>*
-; AVX-NEXT:    [[TMP1:%.*]] = load <4 x i16>, <4 x i16>* [[TMP0]], align 1
-; AVX-NEXT:    [[TMP2:%.*]] = sext <4 x i16> [[TMP1]] to <4 x i64>
-; AVX-NEXT:    [[TMP3:%.*]] = bitcast i64* %A to <4 x i64>*
-; AVX-NEXT:    store <4 x i64> [[TMP2]], <4 x i64>* [[TMP3]], align 4
-; AVX-NEXT:    ret i64 undef
+; CHECK-LABEL: @test_sext_4i16_to_4i64(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i16* [[B:%.*]] to <2 x i16>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <2 x i16>, <2 x i16>* [[TMP0]], align 1
+; CHECK-NEXT:    [[TMP2:%.*]] = sext <2 x i16> [[TMP1]] to <2 x i64>
+; CHECK-NEXT:    [[TMP3:%.*]] = bitcast i64* [[A:%.*]] to <2 x i64>*
+; CHECK-NEXT:    store <2 x i64> [[TMP2]], <2 x i64>* [[TMP3]], align 4
+; CHECK-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds i16, i16* [[B]], i64 2
+; CHECK-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds i64, i64* [[A]], i64 2
+; CHECK-NEXT:    [[TMP4:%.*]] = bitcast i16* [[ARRAYIDX5]] to <2 x i16>*
+; CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i16>, <2 x i16>* [[TMP4]], align 1
+; CHECK-NEXT:    [[TMP6:%.*]] = sext <2 x i16> [[TMP5]] to <2 x i64>
+; CHECK-NEXT:    [[TMP7:%.*]] = bitcast i64* [[ARRAYIDX7]] to <2 x i64>*
+; CHECK-NEXT:    store <2 x i64> [[TMP6]], <2 x i64>* [[TMP7]], align 4
+; CHECK-NEXT:    ret i64 undef
 ;
 entry:
   %0 = load i16, i16* %B, align 1

diff --git a/test/Transforms/SLPVectorizer/X86/load-merge.ll b/test/Transforms/SLPVectorizer/X86/load-merge.ll
index 57b5db5ce98372a13bc8a02dc154ed655055c81e..df990be073b15421121bf8ce0114086812366e92 100644 (file)
--- a/test/Transforms/SLPVectorizer/X86/load-merge.ll
+++ b/test/Transforms/SLPVectorizer/X86/load-merge.ll
@@ -17,15 +17,15 @@ define i32 @_Z9load_le32Ph(i8* nocapture readonly %data) {
 ; CHECK-NEXT:    [[SHL3:%.*]] = shl nuw nsw i32 [[CONV2]], 8
 ; CHECK-NEXT:    [[OR:%.*]] = or i32 [[SHL3]], [[CONV]]
 ; CHECK-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[DATA]], i64 2
+; CHECK-NEXT:    [[TMP2:%.*]] = load i8, i8* [[ARRAYIDX4]], align 1
+; CHECK-NEXT:    [[CONV5:%.*]] = zext i8 [[TMP2]] to i32
+; CHECK-NEXT:    [[SHL6:%.*]] = shl nuw nsw i32 [[CONV5]], 16
+; CHECK-NEXT:    [[OR7:%.*]] = or i32 [[OR]], [[SHL6]]
 ; CHECK-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[DATA]], i64 3
-; CHECK-NEXT:    [[TMP2:%.*]] = bitcast i8* [[ARRAYIDX4]] to <2 x i8>*
-; CHECK-NEXT:    [[TMP3:%.*]] = load <2 x i8>, <2 x i8>* [[TMP2]], align 1
-; CHECK-NEXT:    [[TMP4:%.*]] = zext <2 x i8> [[TMP3]] to <2 x i32>
-; CHECK-NEXT:    [[TMP5:%.*]] = shl nuw <2 x i32> [[TMP4]], <i32 16, i32 24>
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x i32> [[TMP5]], i32 0
-; CHECK-NEXT:    [[OR7:%.*]] = or i32 [[OR]], [[TMP6]]
-; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x i32> [[TMP5]], i32 1
-; CHECK-NEXT:    [[OR11:%.*]] = or i32 [[OR7]], [[TMP7]]
+; CHECK-NEXT:    [[TMP3:%.*]] = load i8, i8* [[ARRAYIDX8]], align 1
+; CHECK-NEXT:    [[CONV9:%.*]] = zext i8 [[TMP3]] to i32
+; CHECK-NEXT:    [[SHL10:%.*]] = shl nuw i32 [[CONV9]], 24
+; CHECK-NEXT:    [[OR11:%.*]] = or i32 [[OR7]], [[SHL10]]
 ; CHECK-NEXT:    ret i32 [[OR11]]
 ;
 entry: