; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes='lower-matrix-intrinsics' -S < %s | FileCheck %s

define void @transpose_multiply(ptr %A.Ptr, ptr %B.Ptr, ptr %C.Ptr) {
; CHECK-LABEL: @transpose_multiply(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[A_PTR:%.*]], align 8
; CHECK-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[A_PTR]], i64 3
; CHECK-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; CHECK-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[A_PTR]], i64 6
; CHECK-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; CHECK-NEXT:    [[COL_LOAD4:%.*]] = load <3 x double>, ptr [[B_PTR:%.*]], align 8
; CHECK-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[B_PTR]], i64 3
; CHECK-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; CHECK-NEXT:    [[VEC_GEP7:%.*]] = getelementptr double, ptr [[B_PTR]], i64 6
; CHECK-NEXT:    [[COL_LOAD8:%.*]] = load <3 x double>, ptr [[VEC_GEP7]], align 8
; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <3 x double> [[COL_LOAD]], i64 0
; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <3 x double> poison, double [[TMP0]], i64 0
; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <3 x double> [[COL_LOAD1]], i64 0
; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <3 x double> [[TMP1]], double [[TMP2]], i64 1
; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <3 x double> [[COL_LOAD3]], i64 0
; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <3 x double> [[TMP3]], double [[TMP4]], i64 2
; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <3 x double> [[COL_LOAD]], i64 1
; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <3 x double> poison, double [[TMP6]], i64 0
; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <3 x double> [[COL_LOAD1]], i64 1
; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <3 x double> [[TMP7]], double [[TMP8]], i64 1
; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <3 x double> [[COL_LOAD3]], i64 1
; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <3 x double> [[TMP9]], double [[TMP10]], i64 2
; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <3 x double> [[COL_LOAD]], i64 2
; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <3 x double> poison, double [[TMP12]], i64 0
; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <3 x double> [[COL_LOAD1]], i64 2
; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <3 x double> [[TMP13]], double [[TMP14]], i64 1
; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <3 x double> [[COL_LOAD3]], i64 2
; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <3 x double> [[TMP15]], double [[TMP16]], i64 2
; CHECK-NEXT:    [[BLOCK:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x double> poison, double [[TMP18]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP19:%.*]] = fmul <1 x double> [[BLOCK]], [[SPLAT_SPLAT]]
; CHECK-NEXT:    [[BLOCK9:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP20:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT10:%.*]] = insertelement <1 x double> poison, double [[TMP20]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT11:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT10]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP21:%.*]] = fmul <1 x double> [[BLOCK9]], [[SPLAT_SPLAT11]]
; CHECK-NEXT:    [[TMP22:%.*]] = fadd <1 x double> [[TMP19]], [[TMP21]]
; CHECK-NEXT:    [[BLOCK12:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP23:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT13:%.*]] = insertelement <1 x double> poison, double [[TMP23]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT14:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT13]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP24:%.*]] = fmul <1 x double> [[BLOCK12]], [[SPLAT_SPLAT14]]
; CHECK-NEXT:    [[TMP25:%.*]] = fadd <1 x double> [[TMP22]], [[TMP24]]
; CHECK-NEXT:    [[TMP26:%.*]] = shufflevector <1 x double> [[TMP25]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP27:%.*]] = shufflevector <3 x double> poison, <3 x double> [[TMP26]], <3 x i32> <i32 3, i32 1, i32 2>
; CHECK-NEXT:    [[BLOCK15:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP28:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT16:%.*]] = insertelement <1 x double> poison, double [[TMP28]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT17:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT16]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP29:%.*]] = fmul <1 x double> [[BLOCK15]], [[SPLAT_SPLAT17]]
; CHECK-NEXT:    [[BLOCK18:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP30:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT19:%.*]] = insertelement <1 x double> poison, double [[TMP30]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT20:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT19]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP31:%.*]] = fmul <1 x double> [[BLOCK18]], [[SPLAT_SPLAT20]]
; CHECK-NEXT:    [[TMP32:%.*]] = fadd <1 x double> [[TMP29]], [[TMP31]]
; CHECK-NEXT:    [[BLOCK21:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP33:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT22:%.*]] = insertelement <1 x double> poison, double [[TMP33]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT23:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT22]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP34:%.*]] = fmul <1 x double> [[BLOCK21]], [[SPLAT_SPLAT23]]
; CHECK-NEXT:    [[TMP35:%.*]] = fadd <1 x double> [[TMP32]], [[TMP34]]
; CHECK-NEXT:    [[TMP36:%.*]] = shufflevector <1 x double> [[TMP35]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP37:%.*]] = shufflevector <3 x double> [[TMP27]], <3 x double> [[TMP36]], <3 x i32> <i32 0, i32 3, i32 2>
; CHECK-NEXT:    [[BLOCK24:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP38:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT25:%.*]] = insertelement <1 x double> poison, double [[TMP38]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT26:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT25]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP39:%.*]] = fmul <1 x double> [[BLOCK24]], [[SPLAT_SPLAT26]]
; CHECK-NEXT:    [[BLOCK27:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP40:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT28:%.*]] = insertelement <1 x double> poison, double [[TMP40]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT29:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT28]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP41:%.*]] = fmul <1 x double> [[BLOCK27]], [[SPLAT_SPLAT29]]
; CHECK-NEXT:    [[TMP42:%.*]] = fadd <1 x double> [[TMP39]], [[TMP41]]
; CHECK-NEXT:    [[BLOCK30:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP43:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT31:%.*]] = insertelement <1 x double> poison, double [[TMP43]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT32:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT31]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP44:%.*]] = fmul <1 x double> [[BLOCK30]], [[SPLAT_SPLAT32]]
; CHECK-NEXT:    [[TMP45:%.*]] = fadd <1 x double> [[TMP42]], [[TMP44]]
; CHECK-NEXT:    [[TMP46:%.*]] = shufflevector <1 x double> [[TMP45]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP47:%.*]] = shufflevector <3 x double> [[TMP37]], <3 x double> [[TMP46]], <3 x i32> <i32 0, i32 1, i32 3>
; CHECK-NEXT:    [[BLOCK33:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP48:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT34:%.*]] = insertelement <1 x double> poison, double [[TMP48]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT35:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT34]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP49:%.*]] = fmul <1 x double> [[BLOCK33]], [[SPLAT_SPLAT35]]
; CHECK-NEXT:    [[BLOCK36:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP50:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT37:%.*]] = insertelement <1 x double> poison, double [[TMP50]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT38:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT37]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP51:%.*]] = fmul <1 x double> [[BLOCK36]], [[SPLAT_SPLAT38]]
; CHECK-NEXT:    [[TMP52:%.*]] = fadd <1 x double> [[TMP49]], [[TMP51]]
; CHECK-NEXT:    [[BLOCK39:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP53:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT40:%.*]] = insertelement <1 x double> poison, double [[TMP53]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT41:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT40]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP54:%.*]] = fmul <1 x double> [[BLOCK39]], [[SPLAT_SPLAT41]]
; CHECK-NEXT:    [[TMP55:%.*]] = fadd <1 x double> [[TMP52]], [[TMP54]]
; CHECK-NEXT:    [[TMP56:%.*]] = shufflevector <1 x double> [[TMP55]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP57:%.*]] = shufflevector <3 x double> poison, <3 x double> [[TMP56]], <3 x i32> <i32 3, i32 1, i32 2>
; CHECK-NEXT:    [[BLOCK42:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP58:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT43:%.*]] = insertelement <1 x double> poison, double [[TMP58]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT44:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT43]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP59:%.*]] = fmul <1 x double> [[BLOCK42]], [[SPLAT_SPLAT44]]
; CHECK-NEXT:    [[BLOCK45:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP60:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT46:%.*]] = insertelement <1 x double> poison, double [[TMP60]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT47:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT46]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP61:%.*]] = fmul <1 x double> [[BLOCK45]], [[SPLAT_SPLAT47]]
; CHECK-NEXT:    [[TMP62:%.*]] = fadd <1 x double> [[TMP59]], [[TMP61]]
; CHECK-NEXT:    [[BLOCK48:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP63:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT49:%.*]] = insertelement <1 x double> poison, double [[TMP63]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT50:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT49]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP64:%.*]] = fmul <1 x double> [[BLOCK48]], [[SPLAT_SPLAT50]]
; CHECK-NEXT:    [[TMP65:%.*]] = fadd <1 x double> [[TMP62]], [[TMP64]]
; CHECK-NEXT:    [[TMP66:%.*]] = shufflevector <1 x double> [[TMP65]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP67:%.*]] = shufflevector <3 x double> [[TMP57]], <3 x double> [[TMP66]], <3 x i32> <i32 0, i32 3, i32 2>
; CHECK-NEXT:    [[BLOCK51:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP68:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT52:%.*]] = insertelement <1 x double> poison, double [[TMP68]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT53:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT52]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP69:%.*]] = fmul <1 x double> [[BLOCK51]], [[SPLAT_SPLAT53]]
; CHECK-NEXT:    [[BLOCK54:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP70:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT55:%.*]] = insertelement <1 x double> poison, double [[TMP70]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT56:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT55]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP71:%.*]] = fmul <1 x double> [[BLOCK54]], [[SPLAT_SPLAT56]]
; CHECK-NEXT:    [[TMP72:%.*]] = fadd <1 x double> [[TMP69]], [[TMP71]]
; CHECK-NEXT:    [[BLOCK57:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP73:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT58:%.*]] = insertelement <1 x double> poison, double [[TMP73]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT59:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT58]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP74:%.*]] = fmul <1 x double> [[BLOCK57]], [[SPLAT_SPLAT59]]
; CHECK-NEXT:    [[TMP75:%.*]] = fadd <1 x double> [[TMP72]], [[TMP74]]
; CHECK-NEXT:    [[TMP76:%.*]] = shufflevector <1 x double> [[TMP75]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP77:%.*]] = shufflevector <3 x double> [[TMP67]], <3 x double> [[TMP76]], <3 x i32> <i32 0, i32 1, i32 3>
; CHECK-NEXT:    [[BLOCK60:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP78:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT61:%.*]] = insertelement <1 x double> poison, double [[TMP78]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT62:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT61]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP79:%.*]] = fmul <1 x double> [[BLOCK60]], [[SPLAT_SPLAT62]]
; CHECK-NEXT:    [[BLOCK63:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP80:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT64:%.*]] = insertelement <1 x double> poison, double [[TMP80]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT65:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT64]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP81:%.*]] = fmul <1 x double> [[BLOCK63]], [[SPLAT_SPLAT65]]
; CHECK-NEXT:    [[TMP82:%.*]] = fadd <1 x double> [[TMP79]], [[TMP81]]
; CHECK-NEXT:    [[BLOCK66:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP83:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT67:%.*]] = insertelement <1 x double> poison, double [[TMP83]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT68:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT67]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP84:%.*]] = fmul <1 x double> [[BLOCK66]], [[SPLAT_SPLAT68]]
; CHECK-NEXT:    [[TMP85:%.*]] = fadd <1 x double> [[TMP82]], [[TMP84]]
; CHECK-NEXT:    [[TMP86:%.*]] = shufflevector <1 x double> [[TMP85]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP87:%.*]] = shufflevector <3 x double> poison, <3 x double> [[TMP86]], <3 x i32> <i32 3, i32 1, i32 2>
; CHECK-NEXT:    [[BLOCK69:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP88:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT70:%.*]] = insertelement <1 x double> poison, double [[TMP88]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT71:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT70]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP89:%.*]] = fmul <1 x double> [[BLOCK69]], [[SPLAT_SPLAT71]]
; CHECK-NEXT:    [[BLOCK72:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP90:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT73:%.*]] = insertelement <1 x double> poison, double [[TMP90]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT74:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT73]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP91:%.*]] = fmul <1 x double> [[BLOCK72]], [[SPLAT_SPLAT74]]
; CHECK-NEXT:    [[TMP92:%.*]] = fadd <1 x double> [[TMP89]], [[TMP91]]
; CHECK-NEXT:    [[BLOCK75:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP93:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT76:%.*]] = insertelement <1 x double> poison, double [[TMP93]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT77:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT76]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP94:%.*]] = fmul <1 x double> [[BLOCK75]], [[SPLAT_SPLAT77]]
; CHECK-NEXT:    [[TMP95:%.*]] = fadd <1 x double> [[TMP92]], [[TMP94]]
; CHECK-NEXT:    [[TMP96:%.*]] = shufflevector <1 x double> [[TMP95]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP97:%.*]] = shufflevector <3 x double> [[TMP87]], <3 x double> [[TMP96]], <3 x i32> <i32 0, i32 3, i32 2>
; CHECK-NEXT:    [[BLOCK78:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP98:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT79:%.*]] = insertelement <1 x double> poison, double [[TMP98]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT80:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT79]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP99:%.*]] = fmul <1 x double> [[BLOCK78]], [[SPLAT_SPLAT80]]
; CHECK-NEXT:    [[BLOCK81:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP100:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT82:%.*]] = insertelement <1 x double> poison, double [[TMP100]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT83:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT82]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP101:%.*]] = fmul <1 x double> [[BLOCK81]], [[SPLAT_SPLAT83]]
; CHECK-NEXT:    [[TMP102:%.*]] = fadd <1 x double> [[TMP99]], [[TMP101]]
; CHECK-NEXT:    [[BLOCK84:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP103:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT85:%.*]] = insertelement <1 x double> poison, double [[TMP103]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT86:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT85]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP104:%.*]] = fmul <1 x double> [[BLOCK84]], [[SPLAT_SPLAT86]]
; CHECK-NEXT:    [[TMP105:%.*]] = fadd <1 x double> [[TMP102]], [[TMP104]]
; CHECK-NEXT:    [[TMP106:%.*]] = shufflevector <1 x double> [[TMP105]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP107:%.*]] = shufflevector <3 x double> [[TMP97]], <3 x double> [[TMP106]], <3 x i32> <i32 0, i32 1, i32 3>
; CHECK-NEXT:    store <3 x double> [[TMP47]], ptr [[C_PTR:%.*]], align 8
; CHECK-NEXT:    [[VEC_GEP87:%.*]] = getelementptr double, ptr [[C_PTR]], i64 3
; CHECK-NEXT:    store <3 x double> [[TMP77]], ptr [[VEC_GEP87]], align 8
; CHECK-NEXT:    [[VEC_GEP88:%.*]] = getelementptr double, ptr [[C_PTR]], i64 6
; CHECK-NEXT:    store <3 x double> [[TMP107]], ptr [[VEC_GEP88]], align 8
; CHECK-NEXT:    ret void
;

; Load columns of input matrixes %A and %B.


; Transpose %A.


; Lower multiply(transpose(%A), %B)


; Store result columns.


entry:
  %a = load <9 x double>, ptr %A.Ptr, align 8
  %b = load <9 x double>, ptr %B.Ptr, align 8
  %a.trans  = call <9 x double> @llvm.matrix.transpose(<9 x double> %a, i32 3, i32 3)
  %c = call <9 x double> @llvm.matrix.multiply.v9f64.v9f64.v9f64(<9 x double> %a.trans, <9 x double> %b, i32 3, i32 3, i32 3)
  store <9 x double> %c, ptr %C.Ptr, align 8
  ret void
}

declare <9 x double> @llvm.matrix.transpose(<9 x double>, i32, i32)
declare <9 x double> @llvm.matrix.multiply.v9f64.v9f64.v9f64(<9 x double>, <9 x double>, i32, i32, i32)

define void @transpose_multiply_add(ptr %A.Ptr, ptr %B.Ptr, ptr %C.Ptr) {
; CHECK-LABEL: @transpose_multiply_add(
; CHECK-NEXT:  entry:
; CHECK-NEXT:    [[COL_LOAD:%.*]] = load <3 x double>, ptr [[A_PTR:%.*]], align 8
; CHECK-NEXT:    [[VEC_GEP:%.*]] = getelementptr double, ptr [[A_PTR]], i64 3
; CHECK-NEXT:    [[COL_LOAD1:%.*]] = load <3 x double>, ptr [[VEC_GEP]], align 8
; CHECK-NEXT:    [[VEC_GEP2:%.*]] = getelementptr double, ptr [[A_PTR]], i64 6
; CHECK-NEXT:    [[COL_LOAD3:%.*]] = load <3 x double>, ptr [[VEC_GEP2]], align 8
; CHECK-NEXT:    [[COL_LOAD4:%.*]] = load <3 x double>, ptr [[B_PTR:%.*]], align 8
; CHECK-NEXT:    [[VEC_GEP5:%.*]] = getelementptr double, ptr [[B_PTR]], i64 3
; CHECK-NEXT:    [[COL_LOAD6:%.*]] = load <3 x double>, ptr [[VEC_GEP5]], align 8
; CHECK-NEXT:    [[VEC_GEP7:%.*]] = getelementptr double, ptr [[B_PTR]], i64 6
; CHECK-NEXT:    [[COL_LOAD8:%.*]] = load <3 x double>, ptr [[VEC_GEP7]], align 8
; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <3 x double> [[COL_LOAD]], i64 0
; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <3 x double> poison, double [[TMP0]], i64 0
; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <3 x double> [[COL_LOAD1]], i64 0
; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <3 x double> [[TMP1]], double [[TMP2]], i64 1
; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <3 x double> [[COL_LOAD3]], i64 0
; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <3 x double> [[TMP3]], double [[TMP4]], i64 2
; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <3 x double> [[COL_LOAD]], i64 1
; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <3 x double> poison, double [[TMP6]], i64 0
; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <3 x double> [[COL_LOAD1]], i64 1
; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <3 x double> [[TMP7]], double [[TMP8]], i64 1
; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <3 x double> [[COL_LOAD3]], i64 1
; CHECK-NEXT:    [[TMP11:%.*]] = insertelement <3 x double> [[TMP9]], double [[TMP10]], i64 2
; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <3 x double> [[COL_LOAD]], i64 2
; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <3 x double> poison, double [[TMP12]], i64 0
; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <3 x double> [[COL_LOAD1]], i64 2
; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <3 x double> [[TMP13]], double [[TMP14]], i64 1
; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <3 x double> [[COL_LOAD3]], i64 2
; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <3 x double> [[TMP15]], double [[TMP16]], i64 2
; CHECK-NEXT:    [[BLOCK:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x double> poison, double [[TMP18]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP19:%.*]] = fmul <1 x double> [[BLOCK]], [[SPLAT_SPLAT]]
; CHECK-NEXT:    [[BLOCK9:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP20:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT10:%.*]] = insertelement <1 x double> poison, double [[TMP20]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT11:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT10]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP21:%.*]] = fmul <1 x double> [[BLOCK9]], [[SPLAT_SPLAT11]]
; CHECK-NEXT:    [[TMP22:%.*]] = fadd <1 x double> [[TMP19]], [[TMP21]]
; CHECK-NEXT:    [[BLOCK12:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP23:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT13:%.*]] = insertelement <1 x double> poison, double [[TMP23]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT14:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT13]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP24:%.*]] = fmul <1 x double> [[BLOCK12]], [[SPLAT_SPLAT14]]
; CHECK-NEXT:    [[TMP25:%.*]] = fadd <1 x double> [[TMP22]], [[TMP24]]
; CHECK-NEXT:    [[TMP26:%.*]] = shufflevector <1 x double> [[TMP25]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP27:%.*]] = shufflevector <3 x double> poison, <3 x double> [[TMP26]], <3 x i32> <i32 3, i32 1, i32 2>
; CHECK-NEXT:    [[BLOCK15:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP28:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT16:%.*]] = insertelement <1 x double> poison, double [[TMP28]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT17:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT16]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP29:%.*]] = fmul <1 x double> [[BLOCK15]], [[SPLAT_SPLAT17]]
; CHECK-NEXT:    [[BLOCK18:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP30:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT19:%.*]] = insertelement <1 x double> poison, double [[TMP30]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT20:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT19]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP31:%.*]] = fmul <1 x double> [[BLOCK18]], [[SPLAT_SPLAT20]]
; CHECK-NEXT:    [[TMP32:%.*]] = fadd <1 x double> [[TMP29]], [[TMP31]]
; CHECK-NEXT:    [[BLOCK21:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP33:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT22:%.*]] = insertelement <1 x double> poison, double [[TMP33]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT23:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT22]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP34:%.*]] = fmul <1 x double> [[BLOCK21]], [[SPLAT_SPLAT23]]
; CHECK-NEXT:    [[TMP35:%.*]] = fadd <1 x double> [[TMP32]], [[TMP34]]
; CHECK-NEXT:    [[TMP36:%.*]] = shufflevector <1 x double> [[TMP35]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP37:%.*]] = shufflevector <3 x double> [[TMP27]], <3 x double> [[TMP36]], <3 x i32> <i32 0, i32 3, i32 2>
; CHECK-NEXT:    [[BLOCK24:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP38:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT25:%.*]] = insertelement <1 x double> poison, double [[TMP38]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT26:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT25]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP39:%.*]] = fmul <1 x double> [[BLOCK24]], [[SPLAT_SPLAT26]]
; CHECK-NEXT:    [[BLOCK27:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP40:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT28:%.*]] = insertelement <1 x double> poison, double [[TMP40]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT29:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT28]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP41:%.*]] = fmul <1 x double> [[BLOCK27]], [[SPLAT_SPLAT29]]
; CHECK-NEXT:    [[TMP42:%.*]] = fadd <1 x double> [[TMP39]], [[TMP41]]
; CHECK-NEXT:    [[BLOCK30:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP43:%.*]] = extractelement <3 x double> [[COL_LOAD4]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT31:%.*]] = insertelement <1 x double> poison, double [[TMP43]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT32:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT31]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP44:%.*]] = fmul <1 x double> [[BLOCK30]], [[SPLAT_SPLAT32]]
; CHECK-NEXT:    [[TMP45:%.*]] = fadd <1 x double> [[TMP42]], [[TMP44]]
; CHECK-NEXT:    [[TMP46:%.*]] = shufflevector <1 x double> [[TMP45]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP47:%.*]] = shufflevector <3 x double> [[TMP37]], <3 x double> [[TMP46]], <3 x i32> <i32 0, i32 1, i32 3>
; CHECK-NEXT:    [[BLOCK33:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP48:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT34:%.*]] = insertelement <1 x double> poison, double [[TMP48]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT35:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT34]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP49:%.*]] = fmul <1 x double> [[BLOCK33]], [[SPLAT_SPLAT35]]
; CHECK-NEXT:    [[BLOCK36:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP50:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT37:%.*]] = insertelement <1 x double> poison, double [[TMP50]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT38:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT37]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP51:%.*]] = fmul <1 x double> [[BLOCK36]], [[SPLAT_SPLAT38]]
; CHECK-NEXT:    [[TMP52:%.*]] = fadd <1 x double> [[TMP49]], [[TMP51]]
; CHECK-NEXT:    [[BLOCK39:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP53:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT40:%.*]] = insertelement <1 x double> poison, double [[TMP53]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT41:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT40]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP54:%.*]] = fmul <1 x double> [[BLOCK39]], [[SPLAT_SPLAT41]]
; CHECK-NEXT:    [[TMP55:%.*]] = fadd <1 x double> [[TMP52]], [[TMP54]]
; CHECK-NEXT:    [[TMP56:%.*]] = shufflevector <1 x double> [[TMP55]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP57:%.*]] = shufflevector <3 x double> poison, <3 x double> [[TMP56]], <3 x i32> <i32 3, i32 1, i32 2>
; CHECK-NEXT:    [[BLOCK42:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP58:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT43:%.*]] = insertelement <1 x double> poison, double [[TMP58]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT44:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT43]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP59:%.*]] = fmul <1 x double> [[BLOCK42]], [[SPLAT_SPLAT44]]
; CHECK-NEXT:    [[BLOCK45:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP60:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT46:%.*]] = insertelement <1 x double> poison, double [[TMP60]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT47:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT46]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP61:%.*]] = fmul <1 x double> [[BLOCK45]], [[SPLAT_SPLAT47]]
; CHECK-NEXT:    [[TMP62:%.*]] = fadd <1 x double> [[TMP59]], [[TMP61]]
; CHECK-NEXT:    [[BLOCK48:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP63:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT49:%.*]] = insertelement <1 x double> poison, double [[TMP63]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT50:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT49]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP64:%.*]] = fmul <1 x double> [[BLOCK48]], [[SPLAT_SPLAT50]]
; CHECK-NEXT:    [[TMP65:%.*]] = fadd <1 x double> [[TMP62]], [[TMP64]]
; CHECK-NEXT:    [[TMP66:%.*]] = shufflevector <1 x double> [[TMP65]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP67:%.*]] = shufflevector <3 x double> [[TMP57]], <3 x double> [[TMP66]], <3 x i32> <i32 0, i32 3, i32 2>
; CHECK-NEXT:    [[BLOCK51:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP68:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT52:%.*]] = insertelement <1 x double> poison, double [[TMP68]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT53:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT52]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP69:%.*]] = fmul <1 x double> [[BLOCK51]], [[SPLAT_SPLAT53]]
; CHECK-NEXT:    [[BLOCK54:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP70:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT55:%.*]] = insertelement <1 x double> poison, double [[TMP70]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT56:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT55]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP71:%.*]] = fmul <1 x double> [[BLOCK54]], [[SPLAT_SPLAT56]]
; CHECK-NEXT:    [[TMP72:%.*]] = fadd <1 x double> [[TMP69]], [[TMP71]]
; CHECK-NEXT:    [[BLOCK57:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP73:%.*]] = extractelement <3 x double> [[COL_LOAD6]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT58:%.*]] = insertelement <1 x double> poison, double [[TMP73]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT59:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT58]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP74:%.*]] = fmul <1 x double> [[BLOCK57]], [[SPLAT_SPLAT59]]
; CHECK-NEXT:    [[TMP75:%.*]] = fadd <1 x double> [[TMP72]], [[TMP74]]
; CHECK-NEXT:    [[TMP76:%.*]] = shufflevector <1 x double> [[TMP75]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP77:%.*]] = shufflevector <3 x double> [[TMP67]], <3 x double> [[TMP76]], <3 x i32> <i32 0, i32 1, i32 3>
; CHECK-NEXT:    [[BLOCK60:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP78:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT61:%.*]] = insertelement <1 x double> poison, double [[TMP78]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT62:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT61]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP79:%.*]] = fmul <1 x double> [[BLOCK60]], [[SPLAT_SPLAT62]]
; CHECK-NEXT:    [[BLOCK63:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP80:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT64:%.*]] = insertelement <1 x double> poison, double [[TMP80]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT65:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT64]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP81:%.*]] = fmul <1 x double> [[BLOCK63]], [[SPLAT_SPLAT65]]
; CHECK-NEXT:    [[TMP82:%.*]] = fadd <1 x double> [[TMP79]], [[TMP81]]
; CHECK-NEXT:    [[BLOCK66:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP83:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT67:%.*]] = insertelement <1 x double> poison, double [[TMP83]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT68:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT67]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP84:%.*]] = fmul <1 x double> [[BLOCK66]], [[SPLAT_SPLAT68]]
; CHECK-NEXT:    [[TMP85:%.*]] = fadd <1 x double> [[TMP82]], [[TMP84]]
; CHECK-NEXT:    [[TMP86:%.*]] = shufflevector <1 x double> [[TMP85]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP87:%.*]] = shufflevector <3 x double> poison, <3 x double> [[TMP86]], <3 x i32> <i32 3, i32 1, i32 2>
; CHECK-NEXT:    [[BLOCK69:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP88:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT70:%.*]] = insertelement <1 x double> poison, double [[TMP88]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT71:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT70]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP89:%.*]] = fmul <1 x double> [[BLOCK69]], [[SPLAT_SPLAT71]]
; CHECK-NEXT:    [[BLOCK72:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP90:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT73:%.*]] = insertelement <1 x double> poison, double [[TMP90]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT74:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT73]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP91:%.*]] = fmul <1 x double> [[BLOCK72]], [[SPLAT_SPLAT74]]
; CHECK-NEXT:    [[TMP92:%.*]] = fadd <1 x double> [[TMP89]], [[TMP91]]
; CHECK-NEXT:    [[BLOCK75:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 1>
; CHECK-NEXT:    [[TMP93:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT76:%.*]] = insertelement <1 x double> poison, double [[TMP93]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT77:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT76]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP94:%.*]] = fmul <1 x double> [[BLOCK75]], [[SPLAT_SPLAT77]]
; CHECK-NEXT:    [[TMP95:%.*]] = fadd <1 x double> [[TMP92]], [[TMP94]]
; CHECK-NEXT:    [[TMP96:%.*]] = shufflevector <1 x double> [[TMP95]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP97:%.*]] = shufflevector <3 x double> [[TMP87]], <3 x double> [[TMP96]], <3 x i32> <i32 0, i32 3, i32 2>
; CHECK-NEXT:    [[BLOCK78:%.*]] = shufflevector <3 x double> [[TMP5]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP98:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLATINSERT79:%.*]] = insertelement <1 x double> poison, double [[TMP98]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT80:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT79]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP99:%.*]] = fmul <1 x double> [[BLOCK78]], [[SPLAT_SPLAT80]]
; CHECK-NEXT:    [[BLOCK81:%.*]] = shufflevector <3 x double> [[TMP11]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP100:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 1
; CHECK-NEXT:    [[SPLAT_SPLATINSERT82:%.*]] = insertelement <1 x double> poison, double [[TMP100]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT83:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT82]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP101:%.*]] = fmul <1 x double> [[BLOCK81]], [[SPLAT_SPLAT83]]
; CHECK-NEXT:    [[TMP102:%.*]] = fadd <1 x double> [[TMP99]], [[TMP101]]
; CHECK-NEXT:    [[BLOCK84:%.*]] = shufflevector <3 x double> [[TMP17]], <3 x double> poison, <1 x i32> <i32 2>
; CHECK-NEXT:    [[TMP103:%.*]] = extractelement <3 x double> [[COL_LOAD8]], i64 2
; CHECK-NEXT:    [[SPLAT_SPLATINSERT85:%.*]] = insertelement <1 x double> poison, double [[TMP103]], i64 0
; CHECK-NEXT:    [[SPLAT_SPLAT86:%.*]] = shufflevector <1 x double> [[SPLAT_SPLATINSERT85]], <1 x double> poison, <1 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP104:%.*]] = fmul <1 x double> [[BLOCK84]], [[SPLAT_SPLAT86]]
; CHECK-NEXT:    [[TMP105:%.*]] = fadd <1 x double> [[TMP102]], [[TMP104]]
; CHECK-NEXT:    [[TMP106:%.*]] = shufflevector <1 x double> [[TMP105]], <1 x double> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
; CHECK-NEXT:    [[TMP107:%.*]] = shufflevector <3 x double> [[TMP97]], <3 x double> [[TMP106]], <3 x i32> <i32 0, i32 1, i32 3>
; CHECK-NEXT:    [[COL_LOAD87:%.*]] = load <3 x double>, ptr [[C_PTR:%.*]], align 8
; CHECK-NEXT:    [[VEC_GEP88:%.*]] = getelementptr double, ptr [[C_PTR]], i64 3
; CHECK-NEXT:    [[COL_LOAD89:%.*]] = load <3 x double>, ptr [[VEC_GEP88]], align 8
; CHECK-NEXT:    [[VEC_GEP90:%.*]] = getelementptr double, ptr [[C_PTR]], i64 6
; CHECK-NEXT:    [[COL_LOAD91:%.*]] = load <3 x double>, ptr [[VEC_GEP90]], align 8
; CHECK-NEXT:    [[TMP108:%.*]] = fadd <3 x double> [[COL_LOAD87]], [[TMP47]]
; CHECK-NEXT:    [[TMP109:%.*]] = fadd <3 x double> [[COL_LOAD89]], [[TMP77]]
; CHECK-NEXT:    [[TMP110:%.*]] = fadd <3 x double> [[COL_LOAD91]], [[TMP107]]
; CHECK-NEXT:    store <3 x double> [[TMP108]], ptr [[C_PTR]], align 8
; CHECK-NEXT:    [[VEC_GEP92:%.*]] = getelementptr double, ptr [[C_PTR]], i64 3
; CHECK-NEXT:    store <3 x double> [[TMP109]], ptr [[VEC_GEP92]], align 8
; CHECK-NEXT:    [[VEC_GEP93:%.*]] = getelementptr double, ptr [[C_PTR]], i64 6
; CHECK-NEXT:    store <3 x double> [[TMP110]], ptr [[VEC_GEP93]], align 8
; CHECK-NEXT:    ret void
;


; Transpose %A.


; Lower multiply(transpose(%A), %B)


;  Embed result of multiply into flat vector.


; Load %C.


; Add column vectors.


; Store result columns.


entry:
  %a = load <9 x double>, ptr %A.Ptr, align 8
  %b = load <9 x double>, ptr %B.Ptr, align 8
  %a.trans  = call <9 x double> @llvm.matrix.transpose(<9 x double> %a, i32 3, i32 3)
  %mult = call <9 x double> @llvm.matrix.multiply.v9f64.v9f64.v9f64(<9 x double> %a.trans, <9 x double> %b, i32 3, i32 3, i32 3)
  %c = load <9 x double>, ptr %C.Ptr, align 8
  %res = fadd <9 x double> %c, %mult

  store <9 x double> %res, ptr %C.Ptr, align 8
  ret void
}