Page MenuHomePhabricator
Differential D43582 Diff 137030 test/Transforms/SLPVectorizer/X86/fround.ll

Changeset View

Standalone View

View Options

test/Transforms/SLPVectorizer/X86/fround.ll

Show First 20 Lines • Show All 66 Lines • ▼ Show 20 Lines
; SSE2-NEXT: [[CEIL3:%.*]] = call double @llvm.ceil.f64(double [[LD3]]) ; SSE2-NEXT: [[CEIL3:%.*]] = call double @llvm.ceil.f64(double [[LD3]])
; SSE2-NEXT: store double [[CEIL0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8 ; SSE2-NEXT: store double [[CEIL0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
; SSE2-NEXT: store double [[CEIL1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8 ; SSE2-NEXT: store double [[CEIL1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
; SSE2-NEXT: store double [[CEIL2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8 ; SSE2-NEXT: store double [[CEIL2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8
; SSE2-NEXT: store double [[CEIL3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[CEIL3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @ceil_4f64( ; SSE41-LABEL: @ceil_4f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.ceil.v4f64(<4 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.ceil.v2f64(<2 x double> [[TMP1]]) ; SSE41-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = call <2 x double> @llvm.ceil.v2f64(<2 x double> [[TMP2]])
; SSE41-NEXT: store <2 x double> [[TMP3]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP4]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @ceil_4f64( ; AVX-LABEL: @ceil_4f64(
; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.ceil.v4f64(<4 x double> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.ceil.v4f64(<4 x double> [[TMP1]])
; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
Show All 36 Lines
; SSE2-NEXT: store double [[CEIL3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[CEIL3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: store double [[CEIL4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8 ; SSE2-NEXT: store double [[CEIL4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8
; SSE2-NEXT: store double [[CEIL5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8 ; SSE2-NEXT: store double [[CEIL5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
; SSE2-NEXT: store double [[CEIL6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8 ; SSE2-NEXT: store double [[CEIL6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8
; SSE2-NEXT: store double [[CEIL7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8 ; SSE2-NEXT: store double [[CEIL7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @ceil_8f64( ; SSE41-LABEL: @ceil_8f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.ceil.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <2 x double>*), align 8 ; SSE41-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6) to <2 x double>*), align 8 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <2 x double> @llvm.ceil.v2f64(<2 x double> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <2 x double> @llvm.ceil.v2f64(<2 x double> [[TMP2]]) ; AVX-LABEL: @ceil_8f64(
; SSE41-NEXT: [[TMP7:%.*]] = call <2 x double> @llvm.ceil.v2f64(<2 x double> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP8:%.*]] = call <2 x double> @llvm.ceil.v2f64(<2 x double> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.ceil.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: store <2 x double> [[TMP5]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8 ; AVX-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP6]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8 ; AVX-NEXT: ret void
; SSE41-NEXT: store <2 x double> [[TMP7]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP8]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6) to <2 x double>*), align 8
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @ceil_8f64(
; AVX1-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX1-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX1-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.ceil.v4f64(<4 x double> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.ceil.v4f64(<4 x double> [[TMP2]])
; AVX1-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX1-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 8
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @ceil_8f64(
; AVX2-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX2-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX2-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.ceil.v4f64(<4 x double> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.ceil.v4f64(<4 x double> [[TMP2]])
; AVX2-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; 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 8
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @ceil_8f64(
; AVX512-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; AVX512-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.ceil.v8f64(<8 x double> [[TMP1]])
; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; AVX512-NEXT: ret void
; ;
%ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8 %ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
%ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8 %ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
%ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8 %ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
%ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8 %ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
%ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8 %ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
%ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8 %ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
%ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8 %ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8
▲ Show 20 Lines • Show All 60 Lines • ▼ Show 20 Lines
; SSE2-NEXT: [[FLOOR3:%.*]] = call double @llvm.floor.f64(double [[LD3]]) ; SSE2-NEXT: [[FLOOR3:%.*]] = call double @llvm.floor.f64(double [[LD3]])
; SSE2-NEXT: store double [[FLOOR0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8 ; SSE2-NEXT: store double [[FLOOR0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
; SSE2-NEXT: store double [[FLOOR1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8 ; SSE2-NEXT: store double [[FLOOR1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
; SSE2-NEXT: store double [[FLOOR2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8 ; SSE2-NEXT: store double [[FLOOR2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8
; SSE2-NEXT: store double [[FLOOR3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[FLOOR3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @floor_4f64( ; SSE41-LABEL: @floor_4f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.floor.v4f64(<4 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.floor.v2f64(<2 x double> [[TMP1]]) ; SSE41-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = call <2 x double> @llvm.floor.v2f64(<2 x double> [[TMP2]])
; SSE41-NEXT: store <2 x double> [[TMP3]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP4]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @floor_4f64( ; AVX-LABEL: @floor_4f64(
; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.floor.v4f64(<4 x double> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.floor.v4f64(<4 x double> [[TMP1]])
; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
Show All 36 Lines
; SSE2-NEXT: store double [[FLOOR3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[FLOOR3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: store double [[FLOOR4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8 ; SSE2-NEXT: store double [[FLOOR4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8
; SSE2-NEXT: store double [[FLOOR5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8 ; SSE2-NEXT: store double [[FLOOR5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
; SSE2-NEXT: store double [[FLOOR6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8 ; SSE2-NEXT: store double [[FLOOR6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8
; SSE2-NEXT: store double [[FLOOR7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8 ; SSE2-NEXT: store double [[FLOOR7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @floor_8f64( ; SSE41-LABEL: @floor_8f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.floor.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <2 x double>*), align 8 ; SSE41-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6) to <2 x double>*), align 8 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <2 x double> @llvm.floor.v2f64(<2 x double> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <2 x double> @llvm.floor.v2f64(<2 x double> [[TMP2]]) ; AVX-LABEL: @floor_8f64(
; SSE41-NEXT: [[TMP7:%.*]] = call <2 x double> @llvm.floor.v2f64(<2 x double> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP8:%.*]] = call <2 x double> @llvm.floor.v2f64(<2 x double> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.floor.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: store <2 x double> [[TMP5]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8 ; AVX-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP6]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8 ; AVX-NEXT: ret void
; SSE41-NEXT: store <2 x double> [[TMP7]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP8]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6) to <2 x double>*), align 8
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @floor_8f64(
; AVX1-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX1-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX1-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.floor.v4f64(<4 x double> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.floor.v4f64(<4 x double> [[TMP2]])
; AVX1-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX1-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 8
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @floor_8f64(
; AVX2-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX2-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX2-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.floor.v4f64(<4 x double> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.floor.v4f64(<4 x double> [[TMP2]])
; AVX2-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; 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 8
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @floor_8f64(
; AVX512-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; AVX512-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.floor.v8f64(<8 x double> [[TMP1]])
; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; AVX512-NEXT: ret void
; ;
%ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8 %ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
%ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8 %ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
%ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8 %ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
%ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8 %ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
%ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8 %ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
%ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8 %ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
%ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8 %ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8
▲ Show 20 Lines • Show All 60 Lines • ▼ Show 20 Lines
; SSE2-NEXT: [[NEARBYINT3:%.*]] = call double @llvm.nearbyint.f64(double [[LD3]]) ; SSE2-NEXT: [[NEARBYINT3:%.*]] = call double @llvm.nearbyint.f64(double [[LD3]])
; SSE2-NEXT: store double [[NEARBYINT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8 ; SSE2-NEXT: store double [[NEARBYINT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
; SSE2-NEXT: store double [[NEARBYINT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8 ; SSE2-NEXT: store double [[NEARBYINT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
; SSE2-NEXT: store double [[NEARBYINT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8 ; SSE2-NEXT: store double [[NEARBYINT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8
; SSE2-NEXT: store double [[NEARBYINT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[NEARBYINT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @nearbyint_4f64( ; SSE41-LABEL: @nearbyint_4f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.nearbyint.v4f64(<4 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> [[TMP1]]) ; SSE41-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> [[TMP2]])
; SSE41-NEXT: store <2 x double> [[TMP3]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP4]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @nearbyint_4f64( ; AVX-LABEL: @nearbyint_4f64(
; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.nearbyint.v4f64(<4 x double> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.nearbyint.v4f64(<4 x double> [[TMP1]])
; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
Show All 36 Lines
; SSE2-NEXT: store double [[NEARBYINT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[NEARBYINT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: store double [[NEARBYINT4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8 ; SSE2-NEXT: store double [[NEARBYINT4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8
; SSE2-NEXT: store double [[NEARBYINT5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8 ; SSE2-NEXT: store double [[NEARBYINT5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
; SSE2-NEXT: store double [[NEARBYINT6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8 ; SSE2-NEXT: store double [[NEARBYINT6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8
; SSE2-NEXT: store double [[NEARBYINT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8 ; SSE2-NEXT: store double [[NEARBYINT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @nearbyint_8f64( ; SSE41-LABEL: @nearbyint_8f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.nearbyint.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <2 x double>*), align 8 ; SSE41-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6) to <2 x double>*), align 8 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> [[TMP2]]) ; AVX-LABEL: @nearbyint_8f64(
; SSE41-NEXT: [[TMP7:%.*]] = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP8:%.*]] = call <2 x double> @llvm.nearbyint.v2f64(<2 x double> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.nearbyint.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: store <2 x double> [[TMP5]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8 ; AVX-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP6]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8 ; AVX-NEXT: ret void
; SSE41-NEXT: store <2 x double> [[TMP7]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP8]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6) to <2 x double>*), align 8
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @nearbyint_8f64(
; AVX1-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX1-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX1-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.nearbyint.v4f64(<4 x double> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.nearbyint.v4f64(<4 x double> [[TMP2]])
; AVX1-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX1-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 8
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @nearbyint_8f64(
; AVX2-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX2-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX2-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.nearbyint.v4f64(<4 x double> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.nearbyint.v4f64(<4 x double> [[TMP2]])
; AVX2-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; 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 8
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @nearbyint_8f64(
; AVX512-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; AVX512-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.nearbyint.v8f64(<8 x double> [[TMP1]])
; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; AVX512-NEXT: ret void
; ;
%ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8 %ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
%ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8 %ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
%ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8 %ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
%ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8 %ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
%ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8 %ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
%ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8 %ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
%ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8 %ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8
▲ Show 20 Lines • Show All 60 Lines • ▼ Show 20 Lines
; SSE2-NEXT: [[RINT3:%.*]] = call double @llvm.rint.f64(double [[LD3]]) ; SSE2-NEXT: [[RINT3:%.*]] = call double @llvm.rint.f64(double [[LD3]])
; SSE2-NEXT: store double [[RINT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8 ; SSE2-NEXT: store double [[RINT0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
; SSE2-NEXT: store double [[RINT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8 ; SSE2-NEXT: store double [[RINT1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
; SSE2-NEXT: store double [[RINT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8 ; SSE2-NEXT: store double [[RINT2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8
; SSE2-NEXT: store double [[RINT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[RINT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @rint_4f64( ; SSE41-LABEL: @rint_4f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.rint.v4f64(<4 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.rint.v2f64(<2 x double> [[TMP1]]) ; SSE41-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = call <2 x double> @llvm.rint.v2f64(<2 x double> [[TMP2]])
; SSE41-NEXT: store <2 x double> [[TMP3]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP4]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @rint_4f64( ; AVX-LABEL: @rint_4f64(
; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.rint.v4f64(<4 x double> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.rint.v4f64(<4 x double> [[TMP1]])
; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
Show All 36 Lines
; SSE2-NEXT: store double [[RINT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[RINT3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: store double [[RINT4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8 ; SSE2-NEXT: store double [[RINT4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8
; SSE2-NEXT: store double [[RINT5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8 ; SSE2-NEXT: store double [[RINT5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
; SSE2-NEXT: store double [[RINT6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8 ; SSE2-NEXT: store double [[RINT6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8
; SSE2-NEXT: store double [[RINT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8 ; SSE2-NEXT: store double [[RINT7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @rint_8f64( ; SSE41-LABEL: @rint_8f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.rint.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <2 x double>*), align 8 ; SSE41-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6) to <2 x double>*), align 8 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <2 x double> @llvm.rint.v2f64(<2 x double> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <2 x double> @llvm.rint.v2f64(<2 x double> [[TMP2]]) ; AVX-LABEL: @rint_8f64(
; SSE41-NEXT: [[TMP7:%.*]] = call <2 x double> @llvm.rint.v2f64(<2 x double> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP8:%.*]] = call <2 x double> @llvm.rint.v2f64(<2 x double> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.rint.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: store <2 x double> [[TMP5]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8 ; AVX-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP6]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8 ; AVX-NEXT: ret void
; SSE41-NEXT: store <2 x double> [[TMP7]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP8]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6) to <2 x double>*), align 8
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @rint_8f64(
; AVX1-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX1-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX1-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.rint.v4f64(<4 x double> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.rint.v4f64(<4 x double> [[TMP2]])
; AVX1-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX1-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 8
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @rint_8f64(
; AVX2-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX2-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX2-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.rint.v4f64(<4 x double> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.rint.v4f64(<4 x double> [[TMP2]])
; AVX2-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; 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 8
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @rint_8f64(
; AVX512-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; AVX512-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.rint.v8f64(<8 x double> [[TMP1]])
; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; AVX512-NEXT: ret void
; ;
%ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8 %ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
%ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8 %ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
%ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8 %ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
%ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8 %ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
%ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8 %ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
%ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8 %ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
%ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8 %ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8
▲ Show 20 Lines • Show All 60 Lines • ▼ Show 20 Lines
; SSE2-NEXT: [[TRUNC3:%.*]] = call double @llvm.trunc.f64(double [[LD3]]) ; SSE2-NEXT: [[TRUNC3:%.*]] = call double @llvm.trunc.f64(double [[LD3]])
; SSE2-NEXT: store double [[TRUNC0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8 ; SSE2-NEXT: store double [[TRUNC0]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 0), align 8
; SSE2-NEXT: store double [[TRUNC1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8 ; SSE2-NEXT: store double [[TRUNC1]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 1), align 8
; SSE2-NEXT: store double [[TRUNC2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8 ; SSE2-NEXT: store double [[TRUNC2]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2), align 8
; SSE2-NEXT: store double [[TRUNC3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[TRUNC3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @trunc_4f64( ; SSE41-LABEL: @trunc_4f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.trunc.v4f64(<4 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <2 x double> @llvm.trunc.v2f64(<2 x double> [[TMP1]]) ; SSE41-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = call <2 x double> @llvm.trunc.v2f64(<2 x double> [[TMP2]])
; SSE41-NEXT: store <2 x double> [[TMP3]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP4]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @trunc_4f64( ; AVX-LABEL: @trunc_4f64(
; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.trunc.v4f64(<4 x double> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <4 x double> @llvm.trunc.v4f64(<4 x double> [[TMP1]])
; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8 ; AVX-NEXT: store <4 x double> [[TMP2]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
Show All 36 Lines
; SSE2-NEXT: store double [[TRUNC3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8 ; SSE2-NEXT: store double [[TRUNC3]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 3), align 8
; SSE2-NEXT: store double [[TRUNC4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8 ; SSE2-NEXT: store double [[TRUNC4]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4), align 8
; SSE2-NEXT: store double [[TRUNC5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8 ; SSE2-NEXT: store double [[TRUNC5]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 5), align 8
; SSE2-NEXT: store double [[TRUNC6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8 ; SSE2-NEXT: store double [[TRUNC6]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6), align 8
; SSE2-NEXT: store double [[TRUNC7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8 ; SSE2-NEXT: store double [[TRUNC7]], double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 7), align 8
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @trunc_8f64( ; SSE41-LABEL: @trunc_8f64(
; SSE41-NEXT: [[TMP1:%.*]] = load <2 x double>, <2 x double>* bitcast ([8 x double]* @src64 to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP2:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2) to <2 x double>*), align 8 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.trunc.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <2 x double>*), align 8 ; SSE41-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP4:%.*]] = load <2 x double>, <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6) to <2 x double>*), align 8 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <2 x double> @llvm.trunc.v2f64(<2 x double> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <2 x double> @llvm.trunc.v2f64(<2 x double> [[TMP2]]) ; AVX-LABEL: @trunc_8f64(
; SSE41-NEXT: [[TMP7:%.*]] = call <2 x double> @llvm.trunc.v2f64(<2 x double> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; SSE41-NEXT: [[TMP8:%.*]] = call <2 x double> @llvm.trunc.v2f64(<2 x double> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.trunc.v8f64(<8 x double> [[TMP1]])
; SSE41-NEXT: store <2 x double> [[TMP5]], <2 x double>* bitcast ([8 x double]* @dst64 to <2 x double>*), align 8 ; AVX-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP6]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 2) to <2 x double>*), align 8 ; AVX-NEXT: ret void
; SSE41-NEXT: store <2 x double> [[TMP7]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 4) to <2 x double>*), align 8
; SSE41-NEXT: store <2 x double> [[TMP8]], <2 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @dst64, i32 0, i64 6) to <2 x double>*), align 8
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @trunc_8f64(
; AVX1-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX1-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX1-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.trunc.v4f64(<4 x double> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.trunc.v4f64(<4 x double> [[TMP2]])
; AVX1-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; AVX1-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 8
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @trunc_8f64(
; AVX2-NEXT: [[TMP1:%.*]] = load <4 x double>, <4 x double>* bitcast ([8 x double]* @src64 to <4 x double>*), align 8
; AVX2-NEXT: [[TMP2:%.*]] = load <4 x double>, <4 x double>* bitcast (double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4) to <4 x double>*), align 8
; AVX2-NEXT: [[TMP3:%.*]] = call <4 x double> @llvm.trunc.v4f64(<4 x double> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <4 x double> @llvm.trunc.v4f64(<4 x double> [[TMP2]])
; AVX2-NEXT: store <4 x double> [[TMP3]], <4 x double>* bitcast ([8 x double]* @dst64 to <4 x double>*), align 8
; 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 8
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @trunc_8f64(
; AVX512-NEXT: [[TMP1:%.*]] = load <8 x double>, <8 x double>* bitcast ([8 x double]* @src64 to <8 x double>*), align 8
; AVX512-NEXT: [[TMP2:%.*]] = call <8 x double> @llvm.trunc.v8f64(<8 x double> [[TMP1]])
; AVX512-NEXT: store <8 x double> [[TMP2]], <8 x double>* bitcast ([8 x double]* @dst64 to <8 x double>*), align 8
; AVX512-NEXT: ret void
; ;
%ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8 %ld0 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 0), align 8
%ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8 %ld1 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 1), align 8
%ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8 %ld2 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 2), align 8
%ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8 %ld3 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 3), align 8
%ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8 %ld4 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 4), align 8
%ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8 %ld5 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 5), align 8
%ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8 %ld6 = load double, double* getelementptr inbounds ([8 x double], [8 x double]* @src64, i32 0, i64 6), align 8
▲ Show 20 Lines • Show All 84 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[CEIL3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4 ; SSE2-NEXT: store float [[CEIL3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
; SSE2-NEXT: store float [[CEIL4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4 ; SSE2-NEXT: store float [[CEIL4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
; SSE2-NEXT: store float [[CEIL5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4 ; SSE2-NEXT: store float [[CEIL5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
; SSE2-NEXT: store float [[CEIL6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4 ; SSE2-NEXT: store float [[CEIL6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
; SSE2-NEXT: store float [[CEIL7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4 ; SSE2-NEXT: store float [[CEIL7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @ceil_8f32( ; SSE41-LABEL: @ceil_8f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.ceil.v8f32(<8 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.ceil.v4f32(<4 x float> [[TMP1]]) ; SSE41-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.ceil.v4f32(<4 x float> [[TMP2]])
; SSE41-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP4]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @ceil_8f32( ; AVX-LABEL: @ceil_8f32(
; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.ceil.v8f32(<8 x float> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.ceil.v8f32(<8 x float> [[TMP1]])
; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4 ; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
▲ Show 20 Lines • Show All 72 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[CEIL11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4 ; SSE2-NEXT: store float [[CEIL11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
; SSE2-NEXT: store float [[CEIL12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4 ; SSE2-NEXT: store float [[CEIL12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4
; SSE2-NEXT: store float [[CEIL13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4 ; SSE2-NEXT: store float [[CEIL13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
; SSE2-NEXT: store float [[CEIL14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4 ; SSE2-NEXT: store float [[CEIL14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4
; SSE2-NEXT: store float [[CEIL15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4 ; SSE2-NEXT: store float [[CEIL15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @ceil_16f32( ; SSE41-LABEL: @ceil_16f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.ceil.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <4 x float>*), align 4 ; SSE41-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 12) to <4 x float>*), align 4 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.ceil.v4f32(<4 x float> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.ceil.v4f32(<4 x float> [[TMP2]]) ; AVX-LABEL: @ceil_16f32(
; SSE41-NEXT: [[TMP7:%.*]] = call <4 x float> @llvm.ceil.v4f32(<4 x float> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.ceil.v4f32(<4 x float> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.ceil.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4 ; AVX-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4 ; AVX-NEXT: ret void
; SSE41-NEXT: store <4 x float> [[TMP7]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP8]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 4
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @ceil_16f32(
; AVX1-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX1-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.ceil.v8f32(<8 x float> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.ceil.v8f32(<8 x float> [[TMP2]])
; AVX1-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX1-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @ceil_16f32(
; AVX2-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX2-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.ceil.v8f32(<8 x float> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.ceil.v8f32(<8 x float> [[TMP2]])
; AVX2-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX2-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @ceil_16f32(
; AVX512-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; AVX512-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.ceil.v16f32(<16 x float> [[TMP1]])
; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; AVX512-NEXT: ret void
; ;
%ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4 %ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4
%ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4 %ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4
%ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4 %ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4
%ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4 %ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4
%ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4 %ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4
%ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4 %ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4
%ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4 %ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4
▲ Show 20 Lines • Show All 108 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[FLOOR3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4 ; SSE2-NEXT: store float [[FLOOR3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
; SSE2-NEXT: store float [[FLOOR4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4 ; SSE2-NEXT: store float [[FLOOR4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
; SSE2-NEXT: store float [[FLOOR5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4 ; SSE2-NEXT: store float [[FLOOR5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
; SSE2-NEXT: store float [[FLOOR6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4 ; SSE2-NEXT: store float [[FLOOR6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
; SSE2-NEXT: store float [[FLOOR7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4 ; SSE2-NEXT: store float [[FLOOR7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @floor_8f32( ; SSE41-LABEL: @floor_8f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.floor.v8f32(<8 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.floor.v4f32(<4 x float> [[TMP1]]) ; SSE41-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.floor.v4f32(<4 x float> [[TMP2]])
; SSE41-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP4]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @floor_8f32( ; AVX-LABEL: @floor_8f32(
; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.floor.v8f32(<8 x float> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.floor.v8f32(<8 x float> [[TMP1]])
; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4 ; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
▲ Show 20 Lines • Show All 72 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[FLOOR11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4 ; SSE2-NEXT: store float [[FLOOR11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
; SSE2-NEXT: store float [[FLOOR12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4 ; SSE2-NEXT: store float [[FLOOR12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4
; SSE2-NEXT: store float [[FLOOR13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4 ; SSE2-NEXT: store float [[FLOOR13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
; SSE2-NEXT: store float [[FLOOR14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4 ; SSE2-NEXT: store float [[FLOOR14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4
; SSE2-NEXT: store float [[FLOOR15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4 ; SSE2-NEXT: store float [[FLOOR15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @floor_16f32( ; SSE41-LABEL: @floor_16f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.floor.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <4 x float>*), align 4 ; SSE41-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 12) to <4 x float>*), align 4 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.floor.v4f32(<4 x float> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.floor.v4f32(<4 x float> [[TMP2]]) ; AVX-LABEL: @floor_16f32(
; SSE41-NEXT: [[TMP7:%.*]] = call <4 x float> @llvm.floor.v4f32(<4 x float> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.floor.v4f32(<4 x float> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.floor.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4 ; AVX-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4 ; AVX-NEXT: ret void
; SSE41-NEXT: store <4 x float> [[TMP7]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP8]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 4
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @floor_16f32(
; AVX1-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX1-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.floor.v8f32(<8 x float> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.floor.v8f32(<8 x float> [[TMP2]])
; AVX1-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX1-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @floor_16f32(
; AVX2-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX2-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.floor.v8f32(<8 x float> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.floor.v8f32(<8 x float> [[TMP2]])
; AVX2-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX2-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @floor_16f32(
; AVX512-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; AVX512-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.floor.v16f32(<16 x float> [[TMP1]])
; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; AVX512-NEXT: ret void
; ;
%ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4 %ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4
%ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4 %ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4
%ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4 %ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4
%ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4 %ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4
%ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4 %ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4
%ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4 %ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4
%ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4 %ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4
▲ Show 20 Lines • Show All 108 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[NEARBYINT3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4 ; SSE2-NEXT: store float [[NEARBYINT3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
; SSE2-NEXT: store float [[NEARBYINT4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4 ; SSE2-NEXT: store float [[NEARBYINT4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
; SSE2-NEXT: store float [[NEARBYINT5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4 ; SSE2-NEXT: store float [[NEARBYINT5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
; SSE2-NEXT: store float [[NEARBYINT6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4 ; SSE2-NEXT: store float [[NEARBYINT6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
; SSE2-NEXT: store float [[NEARBYINT7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4 ; SSE2-NEXT: store float [[NEARBYINT7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @nearbyint_8f32( ; SSE41-LABEL: @nearbyint_8f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.nearbyint.v8f32(<8 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.nearbyint.v4f32(<4 x float> [[TMP1]]) ; SSE41-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.nearbyint.v4f32(<4 x float> [[TMP2]])
; SSE41-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP4]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @nearbyint_8f32( ; AVX-LABEL: @nearbyint_8f32(
; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.nearbyint.v8f32(<8 x float> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.nearbyint.v8f32(<8 x float> [[TMP1]])
; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4 ; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
▲ Show 20 Lines • Show All 72 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[NEARBYINT11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4 ; SSE2-NEXT: store float [[NEARBYINT11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
; SSE2-NEXT: store float [[NEARBYINT12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4 ; SSE2-NEXT: store float [[NEARBYINT12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4
; SSE2-NEXT: store float [[NEARBYINT13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4 ; SSE2-NEXT: store float [[NEARBYINT13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
; SSE2-NEXT: store float [[NEARBYINT14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4 ; SSE2-NEXT: store float [[NEARBYINT14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4
; SSE2-NEXT: store float [[NEARBYINT15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4 ; SSE2-NEXT: store float [[NEARBYINT15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @nearbyint_16f32( ; SSE41-LABEL: @nearbyint_16f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.nearbyint.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <4 x float>*), align 4 ; SSE41-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 12) to <4 x float>*), align 4 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.nearbyint.v4f32(<4 x float> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.nearbyint.v4f32(<4 x float> [[TMP2]]) ; AVX-LABEL: @nearbyint_16f32(
; SSE41-NEXT: [[TMP7:%.*]] = call <4 x float> @llvm.nearbyint.v4f32(<4 x float> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.nearbyint.v4f32(<4 x float> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.nearbyint.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4 ; AVX-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4 ; AVX-NEXT: ret void
; SSE41-NEXT: store <4 x float> [[TMP7]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP8]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 4
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @nearbyint_16f32(
; AVX1-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX1-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.nearbyint.v8f32(<8 x float> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.nearbyint.v8f32(<8 x float> [[TMP2]])
; AVX1-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX1-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @nearbyint_16f32(
; AVX2-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX2-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.nearbyint.v8f32(<8 x float> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.nearbyint.v8f32(<8 x float> [[TMP2]])
; AVX2-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX2-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @nearbyint_16f32(
; AVX512-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; AVX512-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.nearbyint.v16f32(<16 x float> [[TMP1]])
; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; AVX512-NEXT: ret void
; ;
%ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4 %ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4
%ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4 %ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4
%ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4 %ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4
%ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4 %ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4
%ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4 %ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4
%ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4 %ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4
%ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4 %ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4
▲ Show 20 Lines • Show All 108 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[RINT3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4 ; SSE2-NEXT: store float [[RINT3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
; SSE2-NEXT: store float [[RINT4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4 ; SSE2-NEXT: store float [[RINT4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
; SSE2-NEXT: store float [[RINT5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4 ; SSE2-NEXT: store float [[RINT5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
; SSE2-NEXT: store float [[RINT6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4 ; SSE2-NEXT: store float [[RINT6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
; SSE2-NEXT: store float [[RINT7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4 ; SSE2-NEXT: store float [[RINT7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @rint_8f32( ; SSE41-LABEL: @rint_8f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.rint.v8f32(<8 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.rint.v4f32(<4 x float> [[TMP1]]) ; SSE41-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.rint.v4f32(<4 x float> [[TMP2]])
; SSE41-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP4]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @rint_8f32( ; AVX-LABEL: @rint_8f32(
; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.rint.v8f32(<8 x float> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.rint.v8f32(<8 x float> [[TMP1]])
; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4 ; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
▲ Show 20 Lines • Show All 72 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[RINT11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4 ; SSE2-NEXT: store float [[RINT11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
; SSE2-NEXT: store float [[RINT12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4 ; SSE2-NEXT: store float [[RINT12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4
; SSE2-NEXT: store float [[RINT13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4 ; SSE2-NEXT: store float [[RINT13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
; SSE2-NEXT: store float [[RINT14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4 ; SSE2-NEXT: store float [[RINT14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4
; SSE2-NEXT: store float [[RINT15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4 ; SSE2-NEXT: store float [[RINT15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @rint_16f32( ; SSE41-LABEL: @rint_16f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.rint.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <4 x float>*), align 4 ; SSE41-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 12) to <4 x float>*), align 4 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.rint.v4f32(<4 x float> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.rint.v4f32(<4 x float> [[TMP2]]) ; AVX-LABEL: @rint_16f32(
; SSE41-NEXT: [[TMP7:%.*]] = call <4 x float> @llvm.rint.v4f32(<4 x float> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.rint.v4f32(<4 x float> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.rint.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4 ; AVX-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4 ; AVX-NEXT: ret void
; SSE41-NEXT: store <4 x float> [[TMP7]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP8]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 4
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @rint_16f32(
; AVX1-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX1-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.rint.v8f32(<8 x float> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.rint.v8f32(<8 x float> [[TMP2]])
; AVX1-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX1-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @rint_16f32(
; AVX2-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX2-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.rint.v8f32(<8 x float> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.rint.v8f32(<8 x float> [[TMP2]])
; AVX2-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX2-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @rint_16f32(
; AVX512-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; AVX512-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.rint.v16f32(<16 x float> [[TMP1]])
; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; AVX512-NEXT: ret void
; ;
%ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4 %ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4
%ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4 %ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4
%ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4 %ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4
%ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4 %ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4
%ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4 %ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4
%ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4 %ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4
%ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4 %ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4
▲ Show 20 Lines • Show All 108 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[TRUNC3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4 ; SSE2-NEXT: store float [[TRUNC3]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 3), align 4
; SSE2-NEXT: store float [[TRUNC4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4 ; SSE2-NEXT: store float [[TRUNC4]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4), align 4
; SSE2-NEXT: store float [[TRUNC5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4 ; SSE2-NEXT: store float [[TRUNC5]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 5), align 4
; SSE2-NEXT: store float [[TRUNC6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4 ; SSE2-NEXT: store float [[TRUNC6]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 6), align 4
; SSE2-NEXT: store float [[TRUNC7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4 ; SSE2-NEXT: store float [[TRUNC7]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 7), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @trunc_8f32( ; SSE41-LABEL: @trunc_8f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.trunc.v8f32(<8 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = call <4 x float> @llvm.trunc.v4f32(<4 x float> [[TMP1]]) ; SSE41-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = call <4 x float> @llvm.trunc.v4f32(<4 x float> [[TMP2]])
; SSE41-NEXT: store <4 x float> [[TMP3]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP4]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4
; SSE41-NEXT: ret void ; SSE41-NEXT: ret void
; ;
; AVX-LABEL: @trunc_8f32( ; AVX-LABEL: @trunc_8f32(
; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.trunc.v8f32(<8 x float> [[TMP1]]) ; AVX-NEXT: [[TMP2:%.*]] = call <8 x float> @llvm.trunc.v8f32(<8 x float> [[TMP1]])
; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4 ; AVX-NEXT: store <8 x float> [[TMP2]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX-NEXT: ret void ; AVX-NEXT: ret void
; ;
▲ Show 20 Lines • Show All 72 Lines • ▼ Show 20 Lines
; SSE2-NEXT: store float [[TRUNC11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4 ; SSE2-NEXT: store float [[TRUNC11]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 11), align 4
; SSE2-NEXT: store float [[TRUNC12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4 ; SSE2-NEXT: store float [[TRUNC12]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12), align 4
; SSE2-NEXT: store float [[TRUNC13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4 ; SSE2-NEXT: store float [[TRUNC13]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 13), align 4
; SSE2-NEXT: store float [[TRUNC14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4 ; SSE2-NEXT: store float [[TRUNC14]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 14), align 4
; SSE2-NEXT: store float [[TRUNC15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4 ; SSE2-NEXT: store float [[TRUNC15]], float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 15), align 4
; SSE2-NEXT: ret void ; SSE2-NEXT: ret void
; ;
; SSE41-LABEL: @trunc_16f32( ; SSE41-LABEL: @trunc_16f32(
; SSE41-NEXT: [[TMP1:%.*]] = load <4 x float>, <4 x float>* bitcast ([16 x float]* @src32 to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP2:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4) to <4 x float>*), align 4 ; SSE41-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.trunc.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: [[TMP3:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <4 x float>*), align 4 ; SSE41-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP4:%.*]] = load <4 x float>, <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 12) to <4 x float>*), align 4 ; SSE41-NEXT: ret void
; SSE41-NEXT: [[TMP5:%.*]] = call <4 x float> @llvm.trunc.v4f32(<4 x float> [[TMP1]]) ;
; SSE41-NEXT: [[TMP6:%.*]] = call <4 x float> @llvm.trunc.v4f32(<4 x float> [[TMP2]]) ; AVX-LABEL: @trunc_16f32(
; SSE41-NEXT: [[TMP7:%.*]] = call <4 x float> @llvm.trunc.v4f32(<4 x float> [[TMP3]]) ; AVX-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; SSE41-NEXT: [[TMP8:%.*]] = call <4 x float> @llvm.trunc.v4f32(<4 x float> [[TMP4]]) ; AVX-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.trunc.v16f32(<16 x float> [[TMP1]])
; SSE41-NEXT: store <4 x float> [[TMP5]], <4 x float>* bitcast ([16 x float]* @dst32 to <4 x float>*), align 4 ; AVX-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP6]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 4) to <4 x float>*), align 4 ; AVX-NEXT: ret void
; SSE41-NEXT: store <4 x float> [[TMP7]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <4 x float>*), align 4
; SSE41-NEXT: store <4 x float> [[TMP8]], <4 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 12) to <4 x float>*), align 4
; SSE41-NEXT: ret void
;
; AVX1-LABEL: @trunc_16f32(
; AVX1-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX1-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.trunc.v8f32(<8 x float> [[TMP1]])
; AVX1-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.trunc.v8f32(<8 x float> [[TMP2]])
; AVX1-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX1-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX1-NEXT: ret void
;
; AVX2-LABEL: @trunc_16f32(
; AVX2-NEXT: [[TMP1:%.*]] = load <8 x float>, <8 x float>* bitcast ([16 x float]* @src32 to <8 x float>*), align 4
; AVX2-NEXT: [[TMP2:%.*]] = load <8 x float>, <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: [[TMP3:%.*]] = call <8 x float> @llvm.trunc.v8f32(<8 x float> [[TMP1]])
; AVX2-NEXT: [[TMP4:%.*]] = call <8 x float> @llvm.trunc.v8f32(<8 x float> [[TMP2]])
; AVX2-NEXT: store <8 x float> [[TMP3]], <8 x float>* bitcast ([16 x float]* @dst32 to <8 x float>*), align 4
; AVX2-NEXT: store <8 x float> [[TMP4]], <8 x float>* bitcast (float* getelementptr inbounds ([16 x float], [16 x float]* @dst32, i32 0, i64 8) to <8 x float>*), align 4
; AVX2-NEXT: ret void
;
; AVX512-LABEL: @trunc_16f32(
; AVX512-NEXT: [[TMP1:%.*]] = load <16 x float>, <16 x float>* bitcast ([16 x float]* @src32 to <16 x float>*), align 4
; AVX512-NEXT: [[TMP2:%.*]] = call <16 x float> @llvm.trunc.v16f32(<16 x float> [[TMP1]])
; AVX512-NEXT: store <16 x float> [[TMP2]], <16 x float>* bitcast ([16 x float]* @dst32 to <16 x float>*), align 4
; AVX512-NEXT: ret void
; ;
%ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4 %ld0 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 0 ), align 4
%ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4 %ld1 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 1 ), align 4
%ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4 %ld2 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 2 ), align 4
%ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4 %ld3 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 3 ), align 4
%ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4 %ld4 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 4 ), align 4
%ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4 %ld5 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 5 ), align 4
%ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4 %ld6 = load float, float* getelementptr inbounds ([16 x float], [16 x float]* @src32, i32 0, i64 6 ), align 4
▲ Show 20 Lines • Show All 46 Lines • Show Last 20 Lines