114 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py2; RUN: opt -passes='lower-matrix-intrinsics' -S < %s | FileCheck %s3 4 5define <8 x float> @transpose(<8 x float> %a) {6; CHECK-LABEL: @transpose(7; CHECK-NEXT: entry:8; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <2 x i32> <i32 0, i32 1>9; CHECK-NEXT: [[SPLIT1:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <2 x i32> <i32 2, i32 3>10; CHECK-NEXT: [[SPLIT2:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <2 x i32> <i32 4, i32 5>11; CHECK-NEXT: [[SPLIT3:%.*]] = shufflevector <8 x float> [[A]], <8 x float> poison, <2 x i32> <i32 6, i32 7>12; CHECK-NEXT: [[TMP0:%.*]] = extractelement <2 x float> [[SPLIT]], i64 013; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i64 014; CHECK-NEXT: [[TMP2:%.*]] = extractelement <2 x float> [[SPLIT1]], i64 015; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x float> [[TMP1]], float [[TMP2]], i64 116; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x float> [[SPLIT2]], i64 017; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x float> [[TMP3]], float [[TMP4]], i64 218; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x float> [[SPLIT3]], i64 019; CHECK-NEXT: [[TMP7:%.*]] = insertelement <4 x float> [[TMP5]], float [[TMP6]], i64 320; CHECK-NEXT: [[TMP8:%.*]] = extractelement <2 x float> [[SPLIT]], i64 121; CHECK-NEXT: [[TMP9:%.*]] = insertelement <4 x float> poison, float [[TMP8]], i64 022; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x float> [[SPLIT1]], i64 123; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x float> [[TMP9]], float [[TMP10]], i64 124; CHECK-NEXT: [[TMP12:%.*]] = extractelement <2 x float> [[SPLIT2]], i64 125; CHECK-NEXT: [[TMP13:%.*]] = insertelement <4 x float> [[TMP11]], float [[TMP12]], i64 226; CHECK-NEXT: [[TMP14:%.*]] = extractelement <2 x float> [[SPLIT3]], i64 127; CHECK-NEXT: [[TMP15:%.*]] = insertelement <4 x float> [[TMP13]], float [[TMP14]], i64 328; CHECK-NEXT: [[TMP16:%.*]] = shufflevector <4 x float> [[TMP7]], <4 x float> [[TMP15]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>29; CHECK-NEXT: ret <8 x float> [[TMP16]]30;31entry:32 %c = call <8 x float> @llvm.matrix.transpose(<8 x float> %a, i32 2, i32 4)33 ret <8 x float> %c34}35 36declare <8 x float> @llvm.matrix.transpose(<8 x float>, i32, i32)37 38define <8 x float> @transpose_single_column(<8 x float> %a) {39; CHECK-LABEL: @transpose_single_column(40; CHECK-NEXT: entry:41; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <8 x float> [[A:%.*]], <8 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>42; CHECK-NEXT: [[TMP0:%.*]] = extractelement <8 x float> [[SPLIT]], i64 043; CHECK-NEXT: [[TMP1:%.*]] = insertelement <1 x float> poison, float [[TMP0]], i64 044; CHECK-NEXT: [[TMP2:%.*]] = extractelement <8 x float> [[SPLIT]], i64 145; CHECK-NEXT: [[TMP3:%.*]] = insertelement <1 x float> poison, float [[TMP2]], i64 046; CHECK-NEXT: [[TMP4:%.*]] = extractelement <8 x float> [[SPLIT]], i64 247; CHECK-NEXT: [[TMP5:%.*]] = insertelement <1 x float> poison, float [[TMP4]], i64 048; CHECK-NEXT: [[TMP6:%.*]] = extractelement <8 x float> [[SPLIT]], i64 349; CHECK-NEXT: [[TMP7:%.*]] = insertelement <1 x float> poison, float [[TMP6]], i64 050; CHECK-NEXT: [[TMP8:%.*]] = extractelement <8 x float> [[SPLIT]], i64 451; CHECK-NEXT: [[TMP9:%.*]] = insertelement <1 x float> poison, float [[TMP8]], i64 052; CHECK-NEXT: [[TMP10:%.*]] = extractelement <8 x float> [[SPLIT]], i64 553; CHECK-NEXT: [[TMP11:%.*]] = insertelement <1 x float> poison, float [[TMP10]], i64 054; CHECK-NEXT: [[TMP12:%.*]] = extractelement <8 x float> [[SPLIT]], i64 655; CHECK-NEXT: [[TMP13:%.*]] = insertelement <1 x float> poison, float [[TMP12]], i64 056; CHECK-NEXT: [[TMP14:%.*]] = extractelement <8 x float> [[SPLIT]], i64 757; CHECK-NEXT: [[TMP15:%.*]] = insertelement <1 x float> poison, float [[TMP14]], i64 058; CHECK-NEXT: [[TMP16:%.*]] = shufflevector <1 x float> [[TMP1]], <1 x float> [[TMP3]], <2 x i32> <i32 0, i32 1>59; CHECK-NEXT: [[TMP17:%.*]] = shufflevector <1 x float> [[TMP5]], <1 x float> [[TMP7]], <2 x i32> <i32 0, i32 1>60; CHECK-NEXT: [[TMP18:%.*]] = shufflevector <1 x float> [[TMP9]], <1 x float> [[TMP11]], <2 x i32> <i32 0, i32 1>61; CHECK-NEXT: [[TMP19:%.*]] = shufflevector <1 x float> [[TMP13]], <1 x float> [[TMP15]], <2 x i32> <i32 0, i32 1>62; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <2 x float> [[TMP16]], <2 x float> [[TMP17]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>63; CHECK-NEXT: [[TMP21:%.*]] = shufflevector <2 x float> [[TMP18]], <2 x float> [[TMP19]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>64; CHECK-NEXT: [[TMP22:%.*]] = shufflevector <4 x float> [[TMP20]], <4 x float> [[TMP21]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>65; CHECK-NEXT: ret <8 x float> [[TMP22]]66;67entry:68 %c = call <8 x float> @llvm.matrix.transpose(<8 x float> %a, i32 8, i32 1)69 ret <8 x float> %c70}71 72declare <12 x float> @llvm.matrix.transpose.v12f32(<12 x float>, i32, i32)73 74define <12 x float> @transpose_float_3x4(<12 x float> %a) {75; CHECK-LABEL: @transpose_float_3x4(76; CHECK-NEXT: entry:77; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <12 x float> [[A:%.*]], <12 x float> poison, <3 x i32> <i32 0, i32 1, i32 2>78; CHECK-NEXT: [[SPLIT1:%.*]] = shufflevector <12 x float> [[A]], <12 x float> poison, <3 x i32> <i32 3, i32 4, i32 5>79; CHECK-NEXT: [[SPLIT2:%.*]] = shufflevector <12 x float> [[A]], <12 x float> poison, <3 x i32> <i32 6, i32 7, i32 8>80; CHECK-NEXT: [[SPLIT3:%.*]] = shufflevector <12 x float> [[A]], <12 x float> poison, <3 x i32> <i32 9, i32 10, i32 11>81; CHECK-NEXT: [[TMP0:%.*]] = extractelement <3 x float> [[SPLIT]], i64 082; CHECK-NEXT: [[TMP1:%.*]] = insertelement <4 x float> poison, float [[TMP0]], i64 083; CHECK-NEXT: [[TMP2:%.*]] = extractelement <3 x float> [[SPLIT1]], i64 084; CHECK-NEXT: [[TMP3:%.*]] = insertelement <4 x float> [[TMP1]], float [[TMP2]], i64 185; CHECK-NEXT: [[TMP4:%.*]] = extractelement <3 x float> [[SPLIT2]], i64 086; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x float> [[TMP3]], float [[TMP4]], i64 287; CHECK-NEXT: [[TMP6:%.*]] = extractelement <3 x float> [[SPLIT3]], i64 088; CHECK-NEXT: [[TMP7:%.*]] = insertelement <4 x float> [[TMP5]], float [[TMP6]], i64 389; CHECK-NEXT: [[TMP8:%.*]] = extractelement <3 x float> [[SPLIT]], i64 190; CHECK-NEXT: [[TMP9:%.*]] = insertelement <4 x float> poison, float [[TMP8]], i64 091; CHECK-NEXT: [[TMP10:%.*]] = extractelement <3 x float> [[SPLIT1]], i64 192; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x float> [[TMP9]], float [[TMP10]], i64 193; CHECK-NEXT: [[TMP12:%.*]] = extractelement <3 x float> [[SPLIT2]], i64 194; CHECK-NEXT: [[TMP13:%.*]] = insertelement <4 x float> [[TMP11]], float [[TMP12]], i64 295; CHECK-NEXT: [[TMP14:%.*]] = extractelement <3 x float> [[SPLIT3]], i64 196; CHECK-NEXT: [[TMP15:%.*]] = insertelement <4 x float> [[TMP13]], float [[TMP14]], i64 397; CHECK-NEXT: [[TMP16:%.*]] = extractelement <3 x float> [[SPLIT]], i64 298; CHECK-NEXT: [[TMP17:%.*]] = insertelement <4 x float> poison, float [[TMP16]], i64 099; CHECK-NEXT: [[TMP18:%.*]] = extractelement <3 x float> [[SPLIT1]], i64 2100; CHECK-NEXT: [[TMP19:%.*]] = insertelement <4 x float> [[TMP17]], float [[TMP18]], i64 1101; CHECK-NEXT: [[TMP20:%.*]] = extractelement <3 x float> [[SPLIT2]], i64 2102; CHECK-NEXT: [[TMP21:%.*]] = insertelement <4 x float> [[TMP19]], float [[TMP20]], i64 2103; CHECK-NEXT: [[TMP22:%.*]] = extractelement <3 x float> [[SPLIT3]], i64 2104; CHECK-NEXT: [[TMP23:%.*]] = insertelement <4 x float> [[TMP21]], float [[TMP22]], i64 3105; CHECK-NEXT: [[TMP24:%.*]] = shufflevector <4 x float> [[TMP7]], <4 x float> [[TMP15]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>106; CHECK-NEXT: [[TMP25:%.*]] = shufflevector <4 x float> [[TMP23]], <4 x float> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>107; CHECK-NEXT: [[TMP26:%.*]] = shufflevector <8 x float> [[TMP24]], <8 x float> [[TMP25]], <12 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11>108; CHECK-NEXT: ret <12 x float> [[TMP26]]109;110entry:111 %c = call <12 x float> @llvm.matrix.transpose.v12f32(<12 x float> %a, i32 3, i32 4)112 ret <12 x float> %c113}114