130 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py2; RUN: opt -matrix-allow-contract=false -passes='lower-matrix-intrinsics' -S < %s | FileCheck %s3 4define <8 x double> @fadd_transpose(<8 x double> %a, <8 x double> %b) {5; CHECK-LABEL: @fadd_transpose(6; CHECK-NEXT: entry:7; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <8 x double> [[A:%.*]], <8 x double> poison, <2 x i32> <i32 0, i32 1>8; CHECK-NEXT: [[SPLIT1:%.*]] = shufflevector <8 x double> [[A]], <8 x double> poison, <2 x i32> <i32 2, i32 3>9; CHECK-NEXT: [[SPLIT2:%.*]] = shufflevector <8 x double> [[A]], <8 x double> poison, <2 x i32> <i32 4, i32 5>10; CHECK-NEXT: [[SPLIT3:%.*]] = shufflevector <8 x double> [[A]], <8 x double> poison, <2 x i32> <i32 6, i32 7>11; CHECK-NEXT: [[SPLIT4:%.*]] = shufflevector <8 x double> [[B:%.*]], <8 x double> poison, <2 x i32> <i32 0, i32 1>12; CHECK-NEXT: [[SPLIT5:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 2, i32 3>13; CHECK-NEXT: [[SPLIT6:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 4, i32 5>14; CHECK-NEXT: [[SPLIT7:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 6, i32 7>15; CHECK-NEXT: [[TMP0:%.*]] = fadd <2 x double> [[SPLIT]], [[SPLIT4]]16; CHECK-NEXT: [[TMP1:%.*]] = fadd <2 x double> [[SPLIT1]], [[SPLIT5]]17; CHECK-NEXT: [[TMP2:%.*]] = fadd <2 x double> [[SPLIT2]], [[SPLIT6]]18; CHECK-NEXT: [[TMP3:%.*]] = fadd <2 x double> [[SPLIT3]], [[SPLIT7]]19; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x double> [[TMP0]], i64 020; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x double> poison, double [[TMP4]], i64 021; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x double> [[TMP1]], i64 022; CHECK-NEXT: [[TMP7:%.*]] = insertelement <4 x double> [[TMP5]], double [[TMP6]], i64 123; CHECK-NEXT: [[TMP8:%.*]] = extractelement <2 x double> [[TMP2]], i64 024; CHECK-NEXT: [[TMP9:%.*]] = insertelement <4 x double> [[TMP7]], double [[TMP8]], i64 225; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x double> [[TMP3]], i64 026; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x double> [[TMP9]], double [[TMP10]], i64 327; CHECK-NEXT: [[TMP12:%.*]] = extractelement <2 x double> [[TMP0]], i64 128; CHECK-NEXT: [[TMP13:%.*]] = insertelement <4 x double> poison, double [[TMP12]], i64 029; CHECK-NEXT: [[TMP14:%.*]] = extractelement <2 x double> [[TMP1]], i64 130; CHECK-NEXT: [[TMP15:%.*]] = insertelement <4 x double> [[TMP13]], double [[TMP14]], i64 131; CHECK-NEXT: [[TMP16:%.*]] = extractelement <2 x double> [[TMP2]], i64 132; CHECK-NEXT: [[TMP17:%.*]] = insertelement <4 x double> [[TMP15]], double [[TMP16]], i64 233; CHECK-NEXT: [[TMP18:%.*]] = extractelement <2 x double> [[TMP3]], i64 134; CHECK-NEXT: [[TMP19:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP18]], i64 335; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <4 x double> [[TMP11]], <4 x double> [[TMP19]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>36; CHECK-NEXT: ret <8 x double> [[TMP20]]37;38entry:39 %add = fadd <8 x double> %a, %b40 %c = call <8 x double> @llvm.matrix.transpose(<8 x double> %add, i32 2, i32 4)41 ret <8 x double> %c42}43 44define <8 x double> @load_fadd_transpose(ptr %A.Ptr, <8 x double> %b) {45; CHECK-LABEL: @load_fadd_transpose(46; CHECK-NEXT: entry:47; CHECK-NEXT: [[COL_LOAD:%.*]] = load <2 x double>, ptr [[A_PTR:%.*]], align 848; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr double, ptr [[A_PTR]], i64 249; CHECK-NEXT: [[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 850; CHECK-NEXT: [[VEC_GEP2:%.*]] = getelementptr double, ptr [[A_PTR]], i64 451; CHECK-NEXT: [[COL_LOAD3:%.*]] = load <2 x double>, ptr [[VEC_GEP2]], align 852; CHECK-NEXT: [[VEC_GEP4:%.*]] = getelementptr double, ptr [[A_PTR]], i64 653; CHECK-NEXT: [[COL_LOAD5:%.*]] = load <2 x double>, ptr [[VEC_GEP4]], align 854; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <8 x double> [[B:%.*]], <8 x double> poison, <2 x i32> <i32 0, i32 1>55; CHECK-NEXT: [[SPLIT6:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 2, i32 3>56; CHECK-NEXT: [[SPLIT7:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 4, i32 5>57; CHECK-NEXT: [[SPLIT8:%.*]] = shufflevector <8 x double> [[B]], <8 x double> poison, <2 x i32> <i32 6, i32 7>58; CHECK-NEXT: [[TMP0:%.*]] = fadd <2 x double> [[COL_LOAD]], [[SPLIT]]59; CHECK-NEXT: [[TMP1:%.*]] = fadd <2 x double> [[COL_LOAD1]], [[SPLIT6]]60; CHECK-NEXT: [[TMP2:%.*]] = fadd <2 x double> [[COL_LOAD3]], [[SPLIT7]]61; CHECK-NEXT: [[TMP3:%.*]] = fadd <2 x double> [[COL_LOAD5]], [[SPLIT8]]62; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x double> [[TMP0]], i64 063; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x double> poison, double [[TMP4]], i64 064; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x double> [[TMP1]], i64 065; CHECK-NEXT: [[TMP7:%.*]] = insertelement <4 x double> [[TMP5]], double [[TMP6]], i64 166; CHECK-NEXT: [[TMP8:%.*]] = extractelement <2 x double> [[TMP2]], i64 067; CHECK-NEXT: [[TMP9:%.*]] = insertelement <4 x double> [[TMP7]], double [[TMP8]], i64 268; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x double> [[TMP3]], i64 069; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x double> [[TMP9]], double [[TMP10]], i64 370; CHECK-NEXT: [[TMP12:%.*]] = extractelement <2 x double> [[TMP0]], i64 171; CHECK-NEXT: [[TMP13:%.*]] = insertelement <4 x double> poison, double [[TMP12]], i64 072; CHECK-NEXT: [[TMP14:%.*]] = extractelement <2 x double> [[TMP1]], i64 173; CHECK-NEXT: [[TMP15:%.*]] = insertelement <4 x double> [[TMP13]], double [[TMP14]], i64 174; CHECK-NEXT: [[TMP16:%.*]] = extractelement <2 x double> [[TMP2]], i64 175; CHECK-NEXT: [[TMP17:%.*]] = insertelement <4 x double> [[TMP15]], double [[TMP16]], i64 276; CHECK-NEXT: [[TMP18:%.*]] = extractelement <2 x double> [[TMP3]], i64 177; CHECK-NEXT: [[TMP19:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP18]], i64 378; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <4 x double> [[TMP11]], <4 x double> [[TMP19]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>79; CHECK-NEXT: ret <8 x double> [[TMP20]]80;81 82 83entry:84 %a = load <8 x double>, ptr %A.Ptr, align 885 %add = fadd <8 x double> %a, %b86 %c = call <8 x double> @llvm.matrix.transpose(<8 x double> %add, i32 2, i32 4)87 ret <8 x double> %c88}89 90define <8 x double> @load_fneg_transpose(ptr %A.Ptr) {91; CHECK-LABEL: @load_fneg_transpose(92; CHECK-NEXT: entry:93; CHECK-NEXT: [[COL_LOAD:%.*]] = load <2 x double>, ptr [[A_PTR:%.*]], align 894; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr double, ptr [[A_PTR]], i64 295; CHECK-NEXT: [[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 896; CHECK-NEXT: [[VEC_GEP2:%.*]] = getelementptr double, ptr [[A_PTR]], i64 497; CHECK-NEXT: [[COL_LOAD3:%.*]] = load <2 x double>, ptr [[VEC_GEP2]], align 898; CHECK-NEXT: [[VEC_GEP4:%.*]] = getelementptr double, ptr [[A_PTR]], i64 699; CHECK-NEXT: [[COL_LOAD5:%.*]] = load <2 x double>, ptr [[VEC_GEP4]], align 8100; CHECK-NEXT: [[TMP0:%.*]] = fneg <2 x double> [[COL_LOAD]]101; CHECK-NEXT: [[TMP1:%.*]] = fneg <2 x double> [[COL_LOAD1]]102; CHECK-NEXT: [[TMP2:%.*]] = fneg <2 x double> [[COL_LOAD3]]103; CHECK-NEXT: [[TMP3:%.*]] = fneg <2 x double> [[COL_LOAD5]]104; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x double> [[TMP0]], i64 0105; CHECK-NEXT: [[TMP5:%.*]] = insertelement <4 x double> poison, double [[TMP4]], i64 0106; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x double> [[TMP1]], i64 0107; CHECK-NEXT: [[TMP7:%.*]] = insertelement <4 x double> [[TMP5]], double [[TMP6]], i64 1108; CHECK-NEXT: [[TMP8:%.*]] = extractelement <2 x double> [[TMP2]], i64 0109; CHECK-NEXT: [[TMP9:%.*]] = insertelement <4 x double> [[TMP7]], double [[TMP8]], i64 2110; CHECK-NEXT: [[TMP10:%.*]] = extractelement <2 x double> [[TMP3]], i64 0111; CHECK-NEXT: [[TMP11:%.*]] = insertelement <4 x double> [[TMP9]], double [[TMP10]], i64 3112; CHECK-NEXT: [[TMP12:%.*]] = extractelement <2 x double> [[TMP0]], i64 1113; CHECK-NEXT: [[TMP13:%.*]] = insertelement <4 x double> poison, double [[TMP12]], i64 0114; CHECK-NEXT: [[TMP14:%.*]] = extractelement <2 x double> [[TMP1]], i64 1115; CHECK-NEXT: [[TMP15:%.*]] = insertelement <4 x double> [[TMP13]], double [[TMP14]], i64 1116; CHECK-NEXT: [[TMP16:%.*]] = extractelement <2 x double> [[TMP2]], i64 1117; CHECK-NEXT: [[TMP17:%.*]] = insertelement <4 x double> [[TMP15]], double [[TMP16]], i64 2118; CHECK-NEXT: [[TMP18:%.*]] = extractelement <2 x double> [[TMP3]], i64 1119; CHECK-NEXT: [[TMP19:%.*]] = insertelement <4 x double> [[TMP17]], double [[TMP18]], i64 3120; CHECK-NEXT: [[TMP20:%.*]] = shufflevector <4 x double> [[TMP11]], <4 x double> [[TMP19]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>121; CHECK-NEXT: ret <8 x double> [[TMP20]]122;123entry:124 %a = load <8 x double>, ptr %A.Ptr, align 8125 %neg = fneg <8 x double> %a126 %c = call <8 x double> @llvm.matrix.transpose(<8 x double> %neg, i32 2, i32 4)127 ret <8 x double> %c128}129declare <8 x double> @llvm.matrix.transpose(<8 x double>, i32, i32)130