brintos

brintos / llvm-project-archived public Read only

0
0
Text · 23.7 KiB · fa07c51 Raw
254 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 <4 x double> @multiply_2x2(<4 x double> %a, <4 x double> %b) {6; CHECK-LABEL: @multiply_2x2(7; CHECK-NEXT:  entry:8; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <4 x double> [[A:%.*]], <4 x double> poison, <2 x i32> <i32 0, i32 1>9; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <4 x double> [[A]], <4 x double> poison, <2 x i32> <i32 2, i32 3>10; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <4 x double> [[B:%.*]], <4 x double> poison, <2 x i32> <i32 0, i32 1>11; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <4 x double> [[B]], <4 x double> poison, <2 x i32> <i32 2, i32 3>12; CHECK-NEXT:    [[BLOCK:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> zeroinitializer13; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 014; CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x double> poison, double [[TMP0]], i64 015; CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT]], <1 x double> poison, <1 x i32> zeroinitializer16; CHECK-NEXT:    [[TMP1:%.*]] = fmul <1 x double> [[BLOCK]], [[SPLAT_SPLAT]]17; CHECK-NEXT:    [[BLOCK4:%.*]] = shufflevector <2 x double> [[SPLIT1]], <2 x double> poison, <1 x i32> zeroinitializer18; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 119; CHECK-NEXT:    [[SPLAT_SPLATINSERT5:%.*]] = insertelement <1 x double> poison, double [[TMP2]], i64 020; CHECK-NEXT:    [[SPLAT_SPLAT6:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT5]], <1 x double> poison, <1 x i32> zeroinitializer21; CHECK-NEXT:    [[TMP3:%.*]] = fmul <1 x double> [[BLOCK4]], [[SPLAT_SPLAT6]]22; CHECK-NEXT:    [[TMP4:%.*]] = fadd <1 x double> [[TMP1]], [[TMP3]]23; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <1 x double> [[TMP4]], <1 x double> poison, <2 x i32> <i32 0, i32 poison>24; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x double> poison, <2 x double> [[TMP5]], <2 x i32> <i32 2, i32 1>25; CHECK-NEXT:    [[BLOCK7:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> <i32 1>26; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 027; CHECK-NEXT:    [[SPLAT_SPLATINSERT8:%.*]] = insertelement <1 x double> poison, double [[TMP7]], i64 028; CHECK-NEXT:    [[SPLAT_SPLAT9:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT8]], <1 x double> poison, <1 x i32> zeroinitializer29; CHECK-NEXT:    [[TMP8:%.*]] = fmul <1 x double> [[BLOCK7]], [[SPLAT_SPLAT9]]30; CHECK-NEXT:    [[BLOCK10:%.*]] = shufflevector <2 x double> [[SPLIT1]], <2 x double> poison, <1 x i32> <i32 1>31; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 132; CHECK-NEXT:    [[SPLAT_SPLATINSERT11:%.*]] = insertelement <1 x double> poison, double [[TMP9]], i64 033; CHECK-NEXT:    [[SPLAT_SPLAT12:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT11]], <1 x double> poison, <1 x i32> zeroinitializer34; CHECK-NEXT:    [[TMP10:%.*]] = fmul <1 x double> [[BLOCK10]], [[SPLAT_SPLAT12]]35; CHECK-NEXT:    [[TMP11:%.*]] = fadd <1 x double> [[TMP8]], [[TMP10]]36; CHECK-NEXT:    [[TMP12:%.*]] = shufflevector <1 x double> [[TMP11]], <1 x double> poison, <2 x i32> <i32 0, i32 poison>37; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x double> [[TMP6]], <2 x double> [[TMP12]], <2 x i32> <i32 0, i32 2>38; CHECK-NEXT:    [[BLOCK13:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> zeroinitializer39; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 040; CHECK-NEXT:    [[SPLAT_SPLATINSERT14:%.*]] = insertelement <1 x double> poison, double [[TMP14]], i64 041; CHECK-NEXT:    [[SPLAT_SPLAT15:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT14]], <1 x double> poison, <1 x i32> zeroinitializer42; CHECK-NEXT:    [[TMP15:%.*]] = fmul <1 x double> [[BLOCK13]], [[SPLAT_SPLAT15]]43; CHECK-NEXT:    [[BLOCK16:%.*]] = shufflevector <2 x double> [[SPLIT1]], <2 x double> poison, <1 x i32> zeroinitializer44; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 145; CHECK-NEXT:    [[SPLAT_SPLATINSERT17:%.*]] = insertelement <1 x double> poison, double [[TMP16]], i64 046; CHECK-NEXT:    [[SPLAT_SPLAT18:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT17]], <1 x double> poison, <1 x i32> zeroinitializer47; CHECK-NEXT:    [[TMP17:%.*]] = fmul <1 x double> [[BLOCK16]], [[SPLAT_SPLAT18]]48; CHECK-NEXT:    [[TMP18:%.*]] = fadd <1 x double> [[TMP15]], [[TMP17]]49; CHECK-NEXT:    [[TMP19:%.*]] = shufflevector <1 x double> [[TMP18]], <1 x double> poison, <2 x i32> <i32 0, i32 poison>50; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <2 x double> poison, <2 x double> [[TMP19]], <2 x i32> <i32 2, i32 1>51; CHECK-NEXT:    [[BLOCK19:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> <i32 1>52; CHECK-NEXT:    [[TMP21:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 053; CHECK-NEXT:    [[SPLAT_SPLATINSERT20:%.*]] = insertelement <1 x double> poison, double [[TMP21]], i64 054; CHECK-NEXT:    [[SPLAT_SPLAT21:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT20]], <1 x double> poison, <1 x i32> zeroinitializer55; CHECK-NEXT:    [[TMP22:%.*]] = fmul <1 x double> [[BLOCK19]], [[SPLAT_SPLAT21]]56; CHECK-NEXT:    [[BLOCK22:%.*]] = shufflevector <2 x double> [[SPLIT1]], <2 x double> poison, <1 x i32> <i32 1>57; CHECK-NEXT:    [[TMP23:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 158; CHECK-NEXT:    [[SPLAT_SPLATINSERT23:%.*]] = insertelement <1 x double> poison, double [[TMP23]], i64 059; CHECK-NEXT:    [[SPLAT_SPLAT24:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT23]], <1 x double> poison, <1 x i32> zeroinitializer60; CHECK-NEXT:    [[TMP24:%.*]] = fmul <1 x double> [[BLOCK22]], [[SPLAT_SPLAT24]]61; CHECK-NEXT:    [[TMP25:%.*]] = fadd <1 x double> [[TMP22]], [[TMP24]]62; CHECK-NEXT:    [[TMP26:%.*]] = shufflevector <1 x double> [[TMP25]], <1 x double> poison, <2 x i32> <i32 0, i32 poison>63; CHECK-NEXT:    [[TMP27:%.*]] = shufflevector <2 x double> [[TMP20]], <2 x double> [[TMP26]], <2 x i32> <i32 0, i32 2>64; CHECK-NEXT:    [[TMP28:%.*]] = shufflevector <2 x double> [[TMP13]], <2 x double> [[TMP27]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>65; CHECK-NEXT:    ret <4 x double> [[TMP28]]66;67entry:68  %c = call <4 x double> @llvm.matrix.multiply.v4f64.v4f64.v4f64(<4 x double> %a, <4 x double> %b, i32 2, i32 2, i32 2)69  ret <4 x double> %c70}71 72declare <4 x double> @llvm.matrix.multiply.v4f64.v4f64.v4f64(<4 x double>, <4 x double>, i32, i32, i32)73 74define <4 x double> @multiply_1x2(<2 x double> %a, <2 x double> %b) {75; CHECK-LABEL: @multiply_1x2(76; CHECK-NEXT:  entry:77; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <2 x double> [[A:%.*]], <2 x double> poison, <2 x i32> <i32 0, i32 1>78; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <2 x double> [[B:%.*]], <2 x double> poison, <1 x i32> zeroinitializer79; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <2 x double> [[B]], <2 x double> poison, <1 x i32> <i32 1>80; CHECK-NEXT:    [[BLOCK:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> zeroinitializer81; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <1 x double> [[SPLIT1]], i64 082; CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x double> poison, double [[TMP0]], i64 083; CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT]], <1 x double> poison, <1 x i32> zeroinitializer84; CHECK-NEXT:    [[TMP1:%.*]] = fmul <1 x double> [[BLOCK]], [[SPLAT_SPLAT]]85; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <1 x double> [[TMP1]], <1 x double> poison, <2 x i32> <i32 0, i32 poison>86; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x double> poison, <2 x double> [[TMP2]], <2 x i32> <i32 2, i32 1>87; CHECK-NEXT:    [[BLOCK3:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> <i32 1>88; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <1 x double> [[SPLIT1]], i64 089; CHECK-NEXT:    [[SPLAT_SPLATINSERT4:%.*]] = insertelement <1 x double> poison, double [[TMP4]], i64 090; CHECK-NEXT:    [[SPLAT_SPLAT5:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT4]], <1 x double> poison, <1 x i32> zeroinitializer91; CHECK-NEXT:    [[TMP5:%.*]] = fmul <1 x double> [[BLOCK3]], [[SPLAT_SPLAT5]]92; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <1 x double> [[TMP5]], <1 x double> poison, <2 x i32> <i32 0, i32 poison>93; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <2 x double> [[TMP3]], <2 x double> [[TMP6]], <2 x i32> <i32 0, i32 2>94; CHECK-NEXT:    [[BLOCK6:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> zeroinitializer95; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <1 x double> [[SPLIT2]], i64 096; CHECK-NEXT:    [[SPLAT_SPLATINSERT7:%.*]] = insertelement <1 x double> poison, double [[TMP8]], i64 097; CHECK-NEXT:    [[SPLAT_SPLAT8:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT7]], <1 x double> poison, <1 x i32> zeroinitializer98; CHECK-NEXT:    [[TMP9:%.*]] = fmul <1 x double> [[BLOCK6]], [[SPLAT_SPLAT8]]99; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <1 x double> [[TMP9]], <1 x double> poison, <2 x i32> <i32 0, i32 poison>100; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <2 x double> poison, <2 x double> [[TMP10]], <2 x i32> <i32 2, i32 1>101; CHECK-NEXT:    [[BLOCK9:%.*]] = shufflevector <2 x double> [[SPLIT]], <2 x double> poison, <1 x i32> <i32 1>102; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <1 x double> [[SPLIT2]], i64 0103; CHECK-NEXT:    [[SPLAT_SPLATINSERT10:%.*]] = insertelement <1 x double> poison, double [[TMP12]], i64 0104; CHECK-NEXT:    [[SPLAT_SPLAT11:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT10]], <1 x double> poison, <1 x i32> zeroinitializer105; CHECK-NEXT:    [[TMP13:%.*]] = fmul <1 x double> [[BLOCK9]], [[SPLAT_SPLAT11]]106; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <1 x double> [[TMP13]], <1 x double> poison, <2 x i32> <i32 0, i32 poison>107; CHECK-NEXT:    [[TMP15:%.*]] = shufflevector <2 x double> [[TMP11]], <2 x double> [[TMP14]], <2 x i32> <i32 0, i32 2>108; CHECK-NEXT:    [[TMP16:%.*]] = shufflevector <2 x double> [[TMP7]], <2 x double> [[TMP15]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>109; CHECK-NEXT:    ret <4 x double> [[TMP16]]110;111entry:112  %c = call <4 x double> @llvm.matrix.multiply.v4f64.v2f64.v2f64(<2 x double> %a, <2 x double> %b, i32 2, i32 1, i32 2)113  ret <4 x double> %c114}115 116declare <4 x double> @llvm.matrix.multiply.v4f64.v2f64.v2f64(<2 x double>, <2 x double>, i32, i32, i32)117 118define <9 x double> @multiply_2x3(<6 x double> %a, <6 x double> %b) {119; CHECK-LABEL: @multiply_2x3(120; CHECK-NEXT:  entry:121; CHECK-NEXT:    [[SPLIT:%.*]] = shufflevector <6 x double> [[A:%.*]], <6 x double> poison, <3 x i32> <i32 0, i32 1, i32 2>122; CHECK-NEXT:    [[SPLIT1:%.*]] = shufflevector <6 x double> [[A]], <6 x double> poison, <3 x i32> <i32 3, i32 4, i32 5>123; CHECK-NEXT:    [[SPLIT2:%.*]] = shufflevector <6 x double> [[B:%.*]], <6 x double> poison, <2 x i32> <i32 0, i32 1>124; CHECK-NEXT:    [[SPLIT3:%.*]] = shufflevector <6 x double> [[B]], <6 x double> poison, <2 x i32> <i32 2, i32 3>125; CHECK-NEXT:    [[SPLIT4:%.*]] = shufflevector <6 x double> [[B]], <6 x double> poison, <2 x i32> <i32 4, i32 5>126; CHECK-NEXT:    [[BLOCK:%.*]] = shufflevector <3 x double> [[SPLIT]], <3 x double> poison, <1 x i32> zeroinitializer127; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 0128; CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x double> poison, double [[TMP0]], i64 0129; CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT]], <1 x double> poison, <1 x i32> zeroinitializer130; CHECK-NEXT:    [[TMP1:%.*]] = fmul <1 x double> [[BLOCK]], [[SPLAT_SPLAT]]131; CHECK-NEXT:    [[BLOCK5:%.*]] = shufflevector <3 x double> [[SPLIT1]], <3 x double> poison, <1 x i32> zeroinitializer132; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 1133; CHECK-NEXT:    [[SPLAT_SPLATINSERT6:%.*]] = insertelement <1 x double> poison, double [[TMP2]], i64 0134; CHECK-NEXT:    [[SPLAT_SPLAT7:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT6]], <1 x double> poison, <1 x i32> zeroinitializer135; CHECK-NEXT:    [[TMP3:%.*]] = fmul <1 x double> [[BLOCK5]], [[SPLAT_SPLAT7]]136; CHECK-NEXT:    [[TMP4:%.*]] = fadd <1 x double> [[TMP1]], [[TMP3]]137; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <1 x double> [[TMP4]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>138; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <3 x double> poison, <3 x double> [[TMP5]], <3 x i32> <i32 3, i32 1, i32 2>139; CHECK-NEXT:    [[BLOCK8:%.*]] = shufflevector <3 x double> [[SPLIT]], <3 x double> poison, <1 x i32> <i32 1>140; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 0141; CHECK-NEXT:    [[SPLAT_SPLATINSERT9:%.*]] = insertelement <1 x double> poison, double [[TMP7]], i64 0142; CHECK-NEXT:    [[SPLAT_SPLAT10:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT9]], <1 x double> poison, <1 x i32> zeroinitializer143; CHECK-NEXT:    [[TMP8:%.*]] = fmul <1 x double> [[BLOCK8]], [[SPLAT_SPLAT10]]144; CHECK-NEXT:    [[BLOCK11:%.*]] = shufflevector <3 x double> [[SPLIT1]], <3 x double> poison, <1 x i32> <i32 1>145; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 1146; CHECK-NEXT:    [[SPLAT_SPLATINSERT12:%.*]] = insertelement <1 x double> poison, double [[TMP9]], i64 0147; CHECK-NEXT:    [[SPLAT_SPLAT13:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT12]], <1 x double> poison, <1 x i32> zeroinitializer148; CHECK-NEXT:    [[TMP10:%.*]] = fmul <1 x double> [[BLOCK11]], [[SPLAT_SPLAT13]]149; CHECK-NEXT:    [[TMP11:%.*]] = fadd <1 x double> [[TMP8]], [[TMP10]]150; CHECK-NEXT:    [[TMP12:%.*]] = shufflevector <1 x double> [[TMP11]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>151; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <3 x double> [[TMP6]], <3 x double> [[TMP12]], <3 x i32> <i32 0, i32 3, i32 2>152; CHECK-NEXT:    [[BLOCK14:%.*]] = shufflevector <3 x double> [[SPLIT]], <3 x double> poison, <1 x i32> <i32 2>153; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 0154; CHECK-NEXT:    [[SPLAT_SPLATINSERT15:%.*]] = insertelement <1 x double> poison, double [[TMP14]], i64 0155; CHECK-NEXT:    [[SPLAT_SPLAT16:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT15]], <1 x double> poison, <1 x i32> zeroinitializer156; CHECK-NEXT:    [[TMP15:%.*]] = fmul <1 x double> [[BLOCK14]], [[SPLAT_SPLAT16]]157; CHECK-NEXT:    [[BLOCK17:%.*]] = shufflevector <3 x double> [[SPLIT1]], <3 x double> poison, <1 x i32> <i32 2>158; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <2 x double> [[SPLIT2]], i64 1159; CHECK-NEXT:    [[SPLAT_SPLATINSERT18:%.*]] = insertelement <1 x double> poison, double [[TMP16]], i64 0160; CHECK-NEXT:    [[SPLAT_SPLAT19:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT18]], <1 x double> poison, <1 x i32> zeroinitializer161; CHECK-NEXT:    [[TMP17:%.*]] = fmul <1 x double> [[BLOCK17]], [[SPLAT_SPLAT19]]162; CHECK-NEXT:    [[TMP18:%.*]] = fadd <1 x double> [[TMP15]], [[TMP17]]163; CHECK-NEXT:    [[TMP19:%.*]] = shufflevector <1 x double> [[TMP18]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>164; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <3 x double> [[TMP13]], <3 x double> [[TMP19]], <3 x i32> <i32 0, i32 1, i32 3>165; CHECK-NEXT:    [[BLOCK20:%.*]] = shufflevector <3 x double> [[SPLIT]], <3 x double> poison, <1 x i32> zeroinitializer166; CHECK-NEXT:    [[TMP21:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 0167; CHECK-NEXT:    [[SPLAT_SPLATINSERT21:%.*]] = insertelement <1 x double> poison, double [[TMP21]], i64 0168; CHECK-NEXT:    [[SPLAT_SPLAT22:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT21]], <1 x double> poison, <1 x i32> zeroinitializer169; CHECK-NEXT:    [[TMP22:%.*]] = fmul <1 x double> [[BLOCK20]], [[SPLAT_SPLAT22]]170; CHECK-NEXT:    [[BLOCK23:%.*]] = shufflevector <3 x double> [[SPLIT1]], <3 x double> poison, <1 x i32> zeroinitializer171; CHECK-NEXT:    [[TMP23:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 1172; CHECK-NEXT:    [[SPLAT_SPLATINSERT24:%.*]] = insertelement <1 x double> poison, double [[TMP23]], i64 0173; CHECK-NEXT:    [[SPLAT_SPLAT25:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT24]], <1 x double> poison, <1 x i32> zeroinitializer174; CHECK-NEXT:    [[TMP24:%.*]] = fmul <1 x double> [[BLOCK23]], [[SPLAT_SPLAT25]]175; CHECK-NEXT:    [[TMP25:%.*]] = fadd <1 x double> [[TMP22]], [[TMP24]]176; CHECK-NEXT:    [[TMP26:%.*]] = shufflevector <1 x double> [[TMP25]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>177; CHECK-NEXT:    [[TMP27:%.*]] = shufflevector <3 x double> poison, <3 x double> [[TMP26]], <3 x i32> <i32 3, i32 1, i32 2>178; CHECK-NEXT:    [[BLOCK26:%.*]] = shufflevector <3 x double> [[SPLIT]], <3 x double> poison, <1 x i32> <i32 1>179; CHECK-NEXT:    [[TMP28:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 0180; CHECK-NEXT:    [[SPLAT_SPLATINSERT27:%.*]] = insertelement <1 x double> poison, double [[TMP28]], i64 0181; CHECK-NEXT:    [[SPLAT_SPLAT28:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT27]], <1 x double> poison, <1 x i32> zeroinitializer182; CHECK-NEXT:    [[TMP29:%.*]] = fmul <1 x double> [[BLOCK26]], [[SPLAT_SPLAT28]]183; CHECK-NEXT:    [[BLOCK29:%.*]] = shufflevector <3 x double> [[SPLIT1]], <3 x double> poison, <1 x i32> <i32 1>184; CHECK-NEXT:    [[TMP30:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 1185; CHECK-NEXT:    [[SPLAT_SPLATINSERT30:%.*]] = insertelement <1 x double> poison, double [[TMP30]], i64 0186; CHECK-NEXT:    [[SPLAT_SPLAT31:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT30]], <1 x double> poison, <1 x i32> zeroinitializer187; CHECK-NEXT:    [[TMP31:%.*]] = fmul <1 x double> [[BLOCK29]], [[SPLAT_SPLAT31]]188; CHECK-NEXT:    [[TMP32:%.*]] = fadd <1 x double> [[TMP29]], [[TMP31]]189; CHECK-NEXT:    [[TMP33:%.*]] = shufflevector <1 x double> [[TMP32]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>190; CHECK-NEXT:    [[TMP34:%.*]] = shufflevector <3 x double> [[TMP27]], <3 x double> [[TMP33]], <3 x i32> <i32 0, i32 3, i32 2>191; CHECK-NEXT:    [[BLOCK32:%.*]] = shufflevector <3 x double> [[SPLIT]], <3 x double> poison, <1 x i32> <i32 2>192; CHECK-NEXT:    [[TMP35:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 0193; CHECK-NEXT:    [[SPLAT_SPLATINSERT33:%.*]] = insertelement <1 x double> poison, double [[TMP35]], i64 0194; CHECK-NEXT:    [[SPLAT_SPLAT34:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT33]], <1 x double> poison, <1 x i32> zeroinitializer195; CHECK-NEXT:    [[TMP36:%.*]] = fmul <1 x double> [[BLOCK32]], [[SPLAT_SPLAT34]]196; CHECK-NEXT:    [[BLOCK35:%.*]] = shufflevector <3 x double> [[SPLIT1]], <3 x double> poison, <1 x i32> <i32 2>197; CHECK-NEXT:    [[TMP37:%.*]] = extractelement <2 x double> [[SPLIT3]], i64 1198; CHECK-NEXT:    [[SPLAT_SPLATINSERT36:%.*]] = insertelement <1 x double> poison, double [[TMP37]], i64 0199; CHECK-NEXT:    [[SPLAT_SPLAT37:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT36]], <1 x double> poison, <1 x i32> zeroinitializer200; CHECK-NEXT:    [[TMP38:%.*]] = fmul <1 x double> [[BLOCK35]], [[SPLAT_SPLAT37]]201; CHECK-NEXT:    [[TMP39:%.*]] = fadd <1 x double> [[TMP36]], [[TMP38]]202; CHECK-NEXT:    [[TMP40:%.*]] = shufflevector <1 x double> [[TMP39]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>203; CHECK-NEXT:    [[TMP41:%.*]] = shufflevector <3 x double> [[TMP34]], <3 x double> [[TMP40]], <3 x i32> <i32 0, i32 1, i32 3>204; CHECK-NEXT:    [[BLOCK38:%.*]] = shufflevector <3 x double> [[SPLIT]], <3 x double> poison, <1 x i32> zeroinitializer205; CHECK-NEXT:    [[TMP42:%.*]] = extractelement <2 x double> [[SPLIT4]], i64 0206; CHECK-NEXT:    [[SPLAT_SPLATINSERT39:%.*]] = insertelement <1 x double> poison, double [[TMP42]], i64 0207; CHECK-NEXT:    [[SPLAT_SPLAT40:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT39]], <1 x double> poison, <1 x i32> zeroinitializer208; CHECK-NEXT:    [[TMP43:%.*]] = fmul <1 x double> [[BLOCK38]], [[SPLAT_SPLAT40]]209; CHECK-NEXT:    [[BLOCK41:%.*]] = shufflevector <3 x double> [[SPLIT1]], <3 x double> poison, <1 x i32> zeroinitializer210; CHECK-NEXT:    [[TMP44:%.*]] = extractelement <2 x double> [[SPLIT4]], i64 1211; CHECK-NEXT:    [[SPLAT_SPLATINSERT42:%.*]] = insertelement <1 x double> poison, double [[TMP44]], i64 0212; CHECK-NEXT:    [[SPLAT_SPLAT43:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT42]], <1 x double> poison, <1 x i32> zeroinitializer213; CHECK-NEXT:    [[TMP45:%.*]] = fmul <1 x double> [[BLOCK41]], [[SPLAT_SPLAT43]]214; CHECK-NEXT:    [[TMP46:%.*]] = fadd <1 x double> [[TMP43]], [[TMP45]]215; CHECK-NEXT:    [[TMP47:%.*]] = shufflevector <1 x double> [[TMP46]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>216; CHECK-NEXT:    [[TMP48:%.*]] = shufflevector <3 x double> poison, <3 x double> [[TMP47]], <3 x i32> <i32 3, i32 1, i32 2>217; CHECK-NEXT:    [[BLOCK44:%.*]] = shufflevector <3 x double> [[SPLIT]], <3 x double> poison, <1 x i32> <i32 1>218; CHECK-NEXT:    [[TMP49:%.*]] = extractelement <2 x double> [[SPLIT4]], i64 0219; CHECK-NEXT:    [[SPLAT_SPLATINSERT45:%.*]] = insertelement <1 x double> poison, double [[TMP49]], i64 0220; CHECK-NEXT:    [[SPLAT_SPLAT46:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT45]], <1 x double> poison, <1 x i32> zeroinitializer221; CHECK-NEXT:    [[TMP50:%.*]] = fmul <1 x double> [[BLOCK44]], [[SPLAT_SPLAT46]]222; CHECK-NEXT:    [[BLOCK47:%.*]] = shufflevector <3 x double> [[SPLIT1]], <3 x double> poison, <1 x i32> <i32 1>223; CHECK-NEXT:    [[TMP51:%.*]] = extractelement <2 x double> [[SPLIT4]], i64 1224; CHECK-NEXT:    [[SPLAT_SPLATINSERT48:%.*]] = insertelement <1 x double> poison, double [[TMP51]], i64 0225; CHECK-NEXT:    [[SPLAT_SPLAT49:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT48]], <1 x double> poison, <1 x i32> zeroinitializer226; CHECK-NEXT:    [[TMP52:%.*]] = fmul <1 x double> [[BLOCK47]], [[SPLAT_SPLAT49]]227; CHECK-NEXT:    [[TMP53:%.*]] = fadd <1 x double> [[TMP50]], [[TMP52]]228; CHECK-NEXT:    [[TMP54:%.*]] = shufflevector <1 x double> [[TMP53]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>229; CHECK-NEXT:    [[TMP55:%.*]] = shufflevector <3 x double> [[TMP48]], <3 x double> [[TMP54]], <3 x i32> <i32 0, i32 3, i32 2>230; CHECK-NEXT:    [[BLOCK50:%.*]] = shufflevector <3 x double> [[SPLIT]], <3 x double> poison, <1 x i32> <i32 2>231; CHECK-NEXT:    [[TMP56:%.*]] = extractelement <2 x double> [[SPLIT4]], i64 0232; CHECK-NEXT:    [[SPLAT_SPLATINSERT51:%.*]] = insertelement <1 x double> poison, double [[TMP56]], i64 0233; CHECK-NEXT:    [[SPLAT_SPLAT52:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT51]], <1 x double> poison, <1 x i32> zeroinitializer234; CHECK-NEXT:    [[TMP57:%.*]] = fmul <1 x double> [[BLOCK50]], [[SPLAT_SPLAT52]]235; CHECK-NEXT:    [[BLOCK53:%.*]] = shufflevector <3 x double> [[SPLIT1]], <3 x double> poison, <1 x i32> <i32 2>236; CHECK-NEXT:    [[TMP58:%.*]] = extractelement <2 x double> [[SPLIT4]], i64 1237; CHECK-NEXT:    [[SPLAT_SPLATINSERT54:%.*]] = insertelement <1 x double> poison, double [[TMP58]], i64 0238; CHECK-NEXT:    [[SPLAT_SPLAT55:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT54]], <1 x double> poison, <1 x i32> zeroinitializer239; CHECK-NEXT:    [[TMP59:%.*]] = fmul <1 x double> [[BLOCK53]], [[SPLAT_SPLAT55]]240; CHECK-NEXT:    [[TMP60:%.*]] = fadd <1 x double> [[TMP57]], [[TMP59]]241; CHECK-NEXT:    [[TMP61:%.*]] = shufflevector <1 x double> [[TMP60]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>242; CHECK-NEXT:    [[TMP62:%.*]] = shufflevector <3 x double> [[TMP55]], <3 x double> [[TMP61]], <3 x i32> <i32 0, i32 1, i32 3>243; CHECK-NEXT:    [[TMP63:%.*]] = shufflevector <3 x double> [[TMP20]], <3 x double> [[TMP41]], <6 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5>244; CHECK-NEXT:    [[TMP64:%.*]] = shufflevector <3 x double> [[TMP62]], <3 x double> poison, <6 x i32> <i32 0, i32 1, i32 2, i32 poison, i32 poison, i32 poison>245; CHECK-NEXT:    [[TMP65:%.*]] = shufflevector <6 x double> [[TMP63]], <6 x double> [[TMP64]], <9 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8>246; CHECK-NEXT:    ret <9 x double> [[TMP65]]247;248entry:249  %c = call <9 x double> @llvm.matrix.multiply.v9f64.v6f64.v6f64(<6 x double> %a, <6 x double> %b, i32 3, i32 2, i32 3)250  ret <9 x double> %c251}252 253declare <9 x double> @llvm.matrix.multiply.v9f64.v6f64.v6f64(<6 x double>, <6 x double>, i32, i32, i32)254