diff options
| -rw-r--r-- | llvm/lib/Target/X86/X86TargetTransformInfo.cpp | 5 | ||||
| -rw-r--r-- | llvm/test/Analysis/CostModel/X86/uitofp.ll | 6 | ||||
| -rw-r--r-- | llvm/test/Transforms/SLPVectorizer/X86/uitofp.ll | 116 |
3 files changed, 81 insertions, 46 deletions
diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp index b77ac5c9953..2da069e44ed 100644 --- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp +++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp @@ -1291,6 +1291,8 @@ int X86TTIImpl::getCastInstrCost(unsigned Opcode, Type *Dst, Type *Src, { ISD::UINT_TO_FP, MVT::v4f64, MVT::v4i64, 12 }, { ISD::UINT_TO_FP, MVT::v8f64, MVT::v8i64, 26 }, + { ISD::UINT_TO_FP, MVT::f64, MVT::i64, 1 }, + { ISD::FP_TO_UINT, MVT::v2i32, MVT::v2f32, 1 }, { ISD::FP_TO_UINT, MVT::v4i32, MVT::v4f32, 1 }, { ISD::FP_TO_UINT, MVT::v4i32, MVT::v4f64, 1 }, @@ -1444,6 +1446,7 @@ int X86TTIImpl::getCastInstrCost(unsigned Opcode, Type *Dst, Type *Src, { ISD::TRUNCATE, MVT::v8i16, MVT::v8i32, 3 }, { ISD::TRUNCATE, MVT::v16i16, MVT::v16i32, 6 }, + { ISD::UINT_TO_FP, MVT::f64, MVT::i64, 4 }, }; static const TypeConversionCostTblEntry SSE2ConversionTbl[] = { @@ -1470,6 +1473,8 @@ int X86TTIImpl::getCastInstrCost(unsigned Opcode, Type *Dst, Type *Src, { ISD::FP_TO_SINT, MVT::v2i32, MVT::v2f64, 3 }, + { ISD::UINT_TO_FP, MVT::f64, MVT::i64, 6 }, + { ISD::ZERO_EXTEND, MVT::v4i16, MVT::v4i8, 1 }, { ISD::SIGN_EXTEND, MVT::v4i16, MVT::v4i8, 6 }, { ISD::ZERO_EXTEND, MVT::v4i32, MVT::v4i8, 2 }, diff --git a/llvm/test/Analysis/CostModel/X86/uitofp.ll b/llvm/test/Analysis/CostModel/X86/uitofp.ll index 9b8bd082923..621975509f2 100644 --- a/llvm/test/Analysis/CostModel/X86/uitofp.ll +++ b/llvm/test/Analysis/CostModel/X86/uitofp.ll @@ -120,14 +120,14 @@ define i32 @uitofp_i32_double() { define i32 @uitofp_i64_double() { ; SSE-LABEL: 'uitofp_i64_double' -; SSE-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %cvt_i64_f64 = uitofp i64 undef to double +; SSE-NEXT: Cost Model: Found an estimated cost of 6 for instruction: %cvt_i64_f64 = uitofp i64 undef to double ; SSE-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %cvt_v2i64_v2f64 = uitofp <2 x i64> undef to <2 x double> ; SSE-NEXT: Cost Model: Found an estimated cost of 40 for instruction: %cvt_v4i64_v4f64 = uitofp <4 x i64> undef to <4 x double> ; SSE-NEXT: Cost Model: Found an estimated cost of 80 for instruction: %cvt_v8i64_v8f64 = uitofp <8 x i64> undef to <8 x double> ; SSE-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef ; ; AVX-LABEL: 'uitofp_i64_double' -; AVX-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %cvt_i64_f64 = uitofp i64 undef to double +; AVX-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %cvt_i64_f64 = uitofp i64 undef to double ; AVX-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %cvt_v2i64_v2f64 = uitofp <2 x i64> undef to <2 x double> ; AVX-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %cvt_v4i64_v4f64 = uitofp <4 x i64> undef to <4 x double> ; AVX-NEXT: Cost Model: Found an estimated cost of 41 for instruction: %cvt_v8i64_v8f64 = uitofp <8 x i64> undef to <8 x double> @@ -148,7 +148,7 @@ define i32 @uitofp_i64_double() { ; AVX512DQ-NEXT: Cost Model: Found an estimated cost of 0 for instruction: ret i32 undef ; ; BTVER2-LABEL: 'uitofp_i64_double' -; BTVER2-NEXT: Cost Model: Found an estimated cost of 1 for instruction: %cvt_i64_f64 = uitofp i64 undef to double +; BTVER2-NEXT: Cost Model: Found an estimated cost of 4 for instruction: %cvt_i64_f64 = uitofp i64 undef to double ; BTVER2-NEXT: Cost Model: Found an estimated cost of 10 for instruction: %cvt_v2i64_v2f64 = uitofp <2 x i64> undef to <2 x double> ; BTVER2-NEXT: Cost Model: Found an estimated cost of 20 for instruction: %cvt_v4i64_v4f64 = uitofp <4 x i64> undef to <4 x double> ; BTVER2-NEXT: Cost Model: Found an estimated cost of 41 for instruction: %cvt_v8i64_v8f64 = uitofp <8 x i64> undef to <8 x double> diff --git a/llvm/test/Transforms/SLPVectorizer/X86/uitofp.ll b/llvm/test/Transforms/SLPVectorizer/X86/uitofp.ll index ff63fe35bdd..3356f560f35 100644 --- a/llvm/test/Transforms/SLPVectorizer/X86/uitofp.ll +++ b/llvm/test/Transforms/SLPVectorizer/X86/uitofp.ll @@ -1,8 +1,8 @@ ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py ; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE -; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256 -; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver1 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256 -; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256 +; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256 --check-prefix=AVX1 +; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver1 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256 --check-prefix=XOP +; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX256 --check-prefix=AVX2 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=skylake-avx512 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX512 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" @@ -69,20 +69,32 @@ define void @uitofp_4i64_4f64() #0 { ; SSE-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE-NEXT: ret void ; -; AVX256-LABEL: @uitofp_4i64_4f64( -; AVX256-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64 -; AVX256-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8 -; AVX256-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16 -; AVX256-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8 -; AVX256-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to double -; AVX256-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to double -; AVX256-NEXT: [[CVT2:%.*]] = uitofp i64 [[LD2]] to double -; AVX256-NEXT: [[CVT3:%.*]] = uitofp i64 [[LD3]] to double -; AVX256-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64 -; AVX256-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8 -; AVX256-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16 -; AVX256-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 -; AVX256-NEXT: ret void +; AVX1-LABEL: @uitofp_4i64_4f64( +; AVX1-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64 +; AVX1-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8 +; AVX1-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16 +; AVX1-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8 +; AVX1-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to double +; AVX1-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to double +; AVX1-NEXT: [[CVT2:%.*]] = uitofp i64 [[LD2]] to double +; AVX1-NEXT: [[CVT3:%.*]] = uitofp i64 [[LD3]] to double +; AVX1-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64 +; AVX1-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8 +; AVX1-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16 +; AVX1-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 +; AVX1-NEXT: ret void +; +; XOP-LABEL: @uitofp_4i64_4f64( +; XOP-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @src64 to <4 x i64>*), align 64 +; XOP-NEXT: [[TMP2:%.*]] = uitofp <4 x i64> [[TMP1]] to <4 x double> +; XOP-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64 +; XOP-NEXT: ret void +; +; AVX2-LABEL: @uitofp_4i64_4f64( +; AVX2-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @src64 to <4 x i64>*), align 64 +; AVX2-NEXT: [[TMP2:%.*]] = uitofp <4 x i64> [[TMP1]] to <4 x double> +; AVX2-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64 +; AVX2-NEXT: ret void ; ; AVX512-LABEL: @uitofp_4i64_4f64( ; AVX512-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @src64 to <4 x i64>*), align 64 @@ -133,32 +145,50 @@ define void @uitofp_8i64_8f64() #0 { ; SSE-NEXT: store double [[CVT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8 ; SSE-NEXT: ret void ; -; AVX256-LABEL: @uitofp_8i64_8f64( -; AVX256-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64 -; AVX256-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8 -; AVX256-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16 -; AVX256-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8 -; AVX256-NEXT: [[LD4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4), align 32 -; AVX256-NEXT: [[LD5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 5), align 8 -; AVX256-NEXT: [[LD6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 6), align 16 -; AVX256-NEXT: [[LD7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 7), align 8 -; AVX256-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to double -; AVX256-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to double -; AVX256-NEXT: [[CVT2:%.*]] = uitofp i64 [[LD2]] to double -; AVX256-NEXT: [[CVT3:%.*]] = uitofp i64 [[LD3]] to double -; AVX256-NEXT: [[CVT4:%.*]] = uitofp i64 [[LD4]] to double -; AVX256-NEXT: [[CVT5:%.*]] = uitofp i64 [[LD5]] to double -; AVX256-NEXT: [[CVT6:%.*]] = uitofp i64 [[LD6]] to double -; AVX256-NEXT: [[CVT7:%.*]] = uitofp i64 [[LD7]] to double -; AVX256-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64 -; AVX256-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8 -; AVX256-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16 -; AVX256-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 -; AVX256-NEXT: store double [[CVT4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 32 -; AVX256-NEXT: store double [[CVT5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8 -; AVX256-NEXT: store double [[CVT6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 16 -; AVX256-NEXT: store double [[CVT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8 -; AVX256-NEXT: ret void +; AVX1-LABEL: @uitofp_8i64_8f64( +; AVX1-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 0), align 64 +; AVX1-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 1), align 8 +; AVX1-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 2), align 16 +; AVX1-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 3), align 8 +; AVX1-NEXT: [[LD4:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4), align 32 +; AVX1-NEXT: [[LD5:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 5), align 8 +; AVX1-NEXT: [[LD6:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 6), align 16 +; AVX1-NEXT: [[LD7:%.*]] = load i64, i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 7), align 8 +; AVX1-NEXT: [[CVT0:%.*]] = uitofp i64 [[LD0]] to double +; AVX1-NEXT: [[CVT1:%.*]] = uitofp i64 [[LD1]] to double +; AVX1-NEXT: [[CVT2:%.*]] = uitofp i64 [[LD2]] to double +; AVX1-NEXT: [[CVT3:%.*]] = uitofp i64 [[LD3]] to double +; AVX1-NEXT: [[CVT4:%.*]] = uitofp i64 [[LD4]] to double +; AVX1-NEXT: [[CVT5:%.*]] = uitofp i64 [[LD5]] to double +; AVX1-NEXT: [[CVT6:%.*]] = uitofp i64 [[LD6]] to double +; AVX1-NEXT: [[CVT7:%.*]] = uitofp i64 [[LD7]] to double +; AVX1-NEXT: store double [[CVT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 64 +; AVX1-NEXT: store double [[CVT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8 +; AVX1-NEXT: store double [[CVT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 16 +; AVX1-NEXT: store double [[CVT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 +; AVX1-NEXT: store double [[CVT4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 32 +; AVX1-NEXT: store double [[CVT5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8 +; AVX1-NEXT: store double [[CVT6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 16 +; AVX1-NEXT: store double [[CVT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8 +; AVX1-NEXT: ret void +; +; XOP-LABEL: @uitofp_8i64_8f64( +; XOP-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @src64 to <4 x i64>*), align 64 +; XOP-NEXT: [[TMP2:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4) to <4 x i64>*), align 32 +; XOP-NEXT: [[TMP3:%.*]] = uitofp <4 x i64> [[TMP1]] to <4 x double> +; XOP-NEXT: [[TMP4:%.*]] = uitofp <4 x i64> [[TMP2]] to <4 x double> +; XOP-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64 +; XOP-NEXT: store <4 x double> [[TMP4]], <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <4 x double>*), align 32 +; XOP-NEXT: ret void +; +; AVX2-LABEL: @uitofp_8i64_8f64( +; AVX2-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([8 x i64]* @src64 to <4 x i64>*), align 64 +; AVX2-NEXT: [[TMP2:%.*]] = load <4 x i64>, <4 x i64>* bitcast (i64* getelementptr inbounds ([8 x i64], [8 x i64]* @src64, i32 0, i64 4) to <4 x i64>*), align 32 +; AVX2-NEXT: [[TMP3:%.*]] = uitofp <4 x i64> [[TMP1]] to <4 x double> +; AVX2-NEXT: [[TMP4:%.*]] = uitofp <4 x i64> [[TMP2]] to <4 x double> +; AVX2-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 64 +; AVX2-NEXT: store <4 x double> [[TMP4]], <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <4 x double>*), align 32 +; AVX2-NEXT: ret void ; ; AVX512-LABEL: @uitofp_8i64_8f64( ; AVX512-NEXT: [[TMP1:%.*]] = load <8 x i64>, <8 x i64>* bitcast ([8 x i64]* @src64 to <8 x i64>*), align 64 |
