blob: 3dd04b23f272da15ce7b3732e30a1c11564c96be [file] [log] [blame]
; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes=lower-matrix-intrinsics,instcombine -fuse-matrix-use-loops=false -fuse-matrix-tile-size=2 -matrix-allow-contract -force-fuse-matrix -verify-dom-info %s -S | FileCheck %s
; REQUIRES: aarch64-registered-target
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
target triple = "aarch64-apple-ios"
; Test tiling without generating explicit loops.
define void @multiply(ptr %A, ptr %B, ptr %C) {
; CHECK-LABEL: @multiply(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[STORE_BEGIN:%.*]] = ptrtoint ptr [[C:%.*]] to i64
; CHECK-NEXT: [[STORE_END:%.*]] = add nuw nsw i64 [[STORE_BEGIN]], 128
; CHECK-NEXT: [[LOAD_BEGIN:%.*]] = ptrtoint ptr [[A:%.*]] to i64
; CHECK-NEXT: [[TMP0:%.*]] = icmp ugt i64 [[STORE_END]], [[LOAD_BEGIN]]
; CHECK-NEXT: br i1 [[TMP0]], label [[ALIAS_CONT:%.*]], label [[NO_ALIAS:%.*]]
; CHECK: alias_cont:
; CHECK-NEXT: [[LOAD_END:%.*]] = add nuw nsw i64 [[LOAD_BEGIN]], 128
; CHECK-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[LOAD_END]], [[STORE_BEGIN]]
; CHECK-NEXT: br i1 [[TMP1]], label [[COPY:%.*]], label [[NO_ALIAS]]
; CHECK: copy:
; CHECK-NEXT: [[TMP2:%.*]] = alloca [16 x double], align 8
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(128) [[TMP2]], ptr noundef nonnull align 8 dereferenceable(128) [[A]], i64 128, i1 false)
; CHECK-NEXT: br label [[NO_ALIAS]]
; CHECK: no_alias:
; CHECK-NEXT: [[TMP3:%.*]] = phi ptr [ [[A]], [[ENTRY:%.*]] ], [ [[A]], [[ALIAS_CONT]] ], [ [[TMP2]], [[COPY]] ]
; CHECK-NEXT: [[STORE_BEGIN4:%.*]] = ptrtoint ptr [[C]] to i64
; CHECK-NEXT: [[STORE_END5:%.*]] = add nuw nsw i64 [[STORE_BEGIN4]], 128
; CHECK-NEXT: [[LOAD_BEGIN6:%.*]] = ptrtoint ptr [[B:%.*]] to i64
; CHECK-NEXT: [[TMP4:%.*]] = icmp ugt i64 [[STORE_END5]], [[LOAD_BEGIN6]]
; CHECK-NEXT: br i1 [[TMP4]], label [[ALIAS_CONT1:%.*]], label [[NO_ALIAS3:%.*]]
; CHECK: alias_cont1:
; CHECK-NEXT: [[LOAD_END7:%.*]] = add nuw nsw i64 [[LOAD_BEGIN6]], 128
; CHECK-NEXT: [[TMP5:%.*]] = icmp ugt i64 [[LOAD_END7]], [[STORE_BEGIN4]]
; CHECK-NEXT: br i1 [[TMP5]], label [[COPY2:%.*]], label [[NO_ALIAS3]]
; CHECK: copy2:
; CHECK-NEXT: [[TMP6:%.*]] = alloca [16 x double], align 8
; CHECK-NEXT: call void @llvm.memcpy.p0.p0.i64(ptr noundef nonnull align 8 dereferenceable(128) [[TMP6]], ptr noundef nonnull align 8 dereferenceable(128) [[B]], i64 128, i1 false)
; CHECK-NEXT: br label [[NO_ALIAS3]]
; CHECK: no_alias3:
; CHECK-NEXT: [[TMP7:%.*]] = phi ptr [ [[B]], [[NO_ALIAS]] ], [ [[B]], [[ALIAS_CONT1]] ], [ [[TMP6]], [[COPY2]] ]
; CHECK-NEXT: [[COL_LOAD:%.*]] = load <2 x double>, ptr [[TMP3]], align 8
; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr double, ptr [[TMP3]], i64 4
; CHECK-NEXT: [[COL_LOAD8:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8
; CHECK-NEXT: [[COL_LOAD9:%.*]] = load <2 x double>, ptr [[TMP7]], align 8
; CHECK-NEXT: [[VEC_GEP10:%.*]] = getelementptr double, ptr [[TMP7]], i64 4
; CHECK-NEXT: [[COL_LOAD11:%.*]] = load <2 x double>, ptr [[VEC_GEP10]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <2 x double> [[COL_LOAD9]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP8:%.*]] = fmul contract <2 x double> [[COL_LOAD]], [[SPLAT_SPLAT]]
; CHECK-NEXT: [[SPLAT_SPLAT14:%.*]] = shufflevector <2 x double> [[COL_LOAD9]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP9:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD8]], <2 x double> [[SPLAT_SPLAT14]], <2 x double> [[TMP8]])
; CHECK-NEXT: [[SPLAT_SPLAT17:%.*]] = shufflevector <2 x double> [[COL_LOAD11]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP10:%.*]] = fmul contract <2 x double> [[COL_LOAD]], [[SPLAT_SPLAT17]]
; CHECK-NEXT: [[SPLAT_SPLAT20:%.*]] = shufflevector <2 x double> [[COL_LOAD11]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP11:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD8]], <2 x double> [[SPLAT_SPLAT20]], <2 x double> [[TMP10]])
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr double, ptr [[TMP3]], i64 8
; CHECK-NEXT: [[COL_LOAD21:%.*]] = load <2 x double>, ptr [[TMP12]], align 8
; CHECK-NEXT: [[VEC_GEP22:%.*]] = getelementptr double, ptr [[TMP3]], i64 12
; CHECK-NEXT: [[COL_LOAD23:%.*]] = load <2 x double>, ptr [[VEC_GEP22]], align 8
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr double, ptr [[TMP7]], i64 2
; CHECK-NEXT: [[COL_LOAD24:%.*]] = load <2 x double>, ptr [[TMP13]], align 8
; CHECK-NEXT: [[VEC_GEP25:%.*]] = getelementptr double, ptr [[TMP7]], i64 6
; CHECK-NEXT: [[COL_LOAD26:%.*]] = load <2 x double>, ptr [[VEC_GEP25]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT30:%.*]] = shufflevector <2 x double> [[COL_LOAD24]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP14:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD21]], <2 x double> [[SPLAT_SPLAT30]], <2 x double> [[TMP9]])
; CHECK-NEXT: [[SPLAT_SPLAT33:%.*]] = shufflevector <2 x double> [[COL_LOAD24]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP15:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD23]], <2 x double> [[SPLAT_SPLAT33]], <2 x double> [[TMP14]])
; CHECK-NEXT: [[SPLAT_SPLAT37:%.*]] = shufflevector <2 x double> [[COL_LOAD26]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP16:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD21]], <2 x double> [[SPLAT_SPLAT37]], <2 x double> [[TMP11]])
; CHECK-NEXT: [[SPLAT_SPLAT40:%.*]] = shufflevector <2 x double> [[COL_LOAD26]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP17:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD23]], <2 x double> [[SPLAT_SPLAT40]], <2 x double> [[TMP16]])
; CHECK-NEXT: store <2 x double> [[TMP15]], ptr [[C]], align 8
; CHECK-NEXT: [[VEC_GEP41:%.*]] = getelementptr double, ptr [[C]], i64 4
; CHECK-NEXT: store <2 x double> [[TMP17]], ptr [[VEC_GEP41]], align 8
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr double, ptr [[TMP3]], i64 2
; CHECK-NEXT: [[COL_LOAD42:%.*]] = load <2 x double>, ptr [[TMP18]], align 8
; CHECK-NEXT: [[VEC_GEP43:%.*]] = getelementptr double, ptr [[TMP3]], i64 6
; CHECK-NEXT: [[COL_LOAD44:%.*]] = load <2 x double>, ptr [[VEC_GEP43]], align 8
; CHECK-NEXT: [[COL_LOAD45:%.*]] = load <2 x double>, ptr [[TMP7]], align 8
; CHECK-NEXT: [[VEC_GEP46:%.*]] = getelementptr double, ptr [[TMP7]], i64 4
; CHECK-NEXT: [[COL_LOAD47:%.*]] = load <2 x double>, ptr [[VEC_GEP46]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT50:%.*]] = shufflevector <2 x double> [[COL_LOAD45]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP19:%.*]] = fmul contract <2 x double> [[COL_LOAD42]], [[SPLAT_SPLAT50]]
; CHECK-NEXT: [[SPLAT_SPLAT53:%.*]] = shufflevector <2 x double> [[COL_LOAD45]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP20:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD44]], <2 x double> [[SPLAT_SPLAT53]], <2 x double> [[TMP19]])
; CHECK-NEXT: [[SPLAT_SPLAT56:%.*]] = shufflevector <2 x double> [[COL_LOAD47]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP21:%.*]] = fmul contract <2 x double> [[COL_LOAD42]], [[SPLAT_SPLAT56]]
; CHECK-NEXT: [[SPLAT_SPLAT59:%.*]] = shufflevector <2 x double> [[COL_LOAD47]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP22:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD44]], <2 x double> [[SPLAT_SPLAT59]], <2 x double> [[TMP21]])
; CHECK-NEXT: [[TMP23:%.*]] = getelementptr double, ptr [[TMP3]], i64 10
; CHECK-NEXT: [[COL_LOAD60:%.*]] = load <2 x double>, ptr [[TMP23]], align 8
; CHECK-NEXT: [[VEC_GEP61:%.*]] = getelementptr double, ptr [[TMP3]], i64 14
; CHECK-NEXT: [[COL_LOAD62:%.*]] = load <2 x double>, ptr [[VEC_GEP61]], align 8
; CHECK-NEXT: [[TMP24:%.*]] = getelementptr double, ptr [[TMP7]], i64 2
; CHECK-NEXT: [[COL_LOAD63:%.*]] = load <2 x double>, ptr [[TMP24]], align 8
; CHECK-NEXT: [[VEC_GEP64:%.*]] = getelementptr double, ptr [[TMP7]], i64 6
; CHECK-NEXT: [[COL_LOAD65:%.*]] = load <2 x double>, ptr [[VEC_GEP64]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT69:%.*]] = shufflevector <2 x double> [[COL_LOAD63]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP25:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD60]], <2 x double> [[SPLAT_SPLAT69]], <2 x double> [[TMP20]])
; CHECK-NEXT: [[SPLAT_SPLAT72:%.*]] = shufflevector <2 x double> [[COL_LOAD63]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP26:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD62]], <2 x double> [[SPLAT_SPLAT72]], <2 x double> [[TMP25]])
; CHECK-NEXT: [[SPLAT_SPLAT76:%.*]] = shufflevector <2 x double> [[COL_LOAD65]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP27:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD60]], <2 x double> [[SPLAT_SPLAT76]], <2 x double> [[TMP22]])
; CHECK-NEXT: [[SPLAT_SPLAT79:%.*]] = shufflevector <2 x double> [[COL_LOAD65]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP28:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD62]], <2 x double> [[SPLAT_SPLAT79]], <2 x double> [[TMP27]])
; CHECK-NEXT: [[TMP29:%.*]] = getelementptr double, ptr [[C]], i64 2
; CHECK-NEXT: store <2 x double> [[TMP26]], ptr [[TMP29]], align 8
; CHECK-NEXT: [[VEC_GEP80:%.*]] = getelementptr double, ptr [[C]], i64 6
; CHECK-NEXT: store <2 x double> [[TMP28]], ptr [[VEC_GEP80]], align 8
; CHECK-NEXT: [[COL_LOAD81:%.*]] = load <2 x double>, ptr [[TMP3]], align 8
; CHECK-NEXT: [[VEC_GEP82:%.*]] = getelementptr double, ptr [[TMP3]], i64 4
; CHECK-NEXT: [[COL_LOAD83:%.*]] = load <2 x double>, ptr [[VEC_GEP82]], align 8
; CHECK-NEXT: [[TMP30:%.*]] = getelementptr double, ptr [[TMP7]], i64 8
; CHECK-NEXT: [[COL_LOAD84:%.*]] = load <2 x double>, ptr [[TMP30]], align 8
; CHECK-NEXT: [[VEC_GEP85:%.*]] = getelementptr double, ptr [[TMP7]], i64 12
; CHECK-NEXT: [[COL_LOAD86:%.*]] = load <2 x double>, ptr [[VEC_GEP85]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT89:%.*]] = shufflevector <2 x double> [[COL_LOAD84]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP31:%.*]] = fmul contract <2 x double> [[COL_LOAD81]], [[SPLAT_SPLAT89]]
; CHECK-NEXT: [[SPLAT_SPLAT92:%.*]] = shufflevector <2 x double> [[COL_LOAD84]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP32:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD83]], <2 x double> [[SPLAT_SPLAT92]], <2 x double> [[TMP31]])
; CHECK-NEXT: [[SPLAT_SPLAT95:%.*]] = shufflevector <2 x double> [[COL_LOAD86]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP33:%.*]] = fmul contract <2 x double> [[COL_LOAD81]], [[SPLAT_SPLAT95]]
; CHECK-NEXT: [[SPLAT_SPLAT98:%.*]] = shufflevector <2 x double> [[COL_LOAD86]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP34:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD83]], <2 x double> [[SPLAT_SPLAT98]], <2 x double> [[TMP33]])
; CHECK-NEXT: [[TMP35:%.*]] = getelementptr double, ptr [[TMP3]], i64 8
; CHECK-NEXT: [[COL_LOAD99:%.*]] = load <2 x double>, ptr [[TMP35]], align 8
; CHECK-NEXT: [[VEC_GEP100:%.*]] = getelementptr double, ptr [[TMP3]], i64 12
; CHECK-NEXT: [[COL_LOAD101:%.*]] = load <2 x double>, ptr [[VEC_GEP100]], align 8
; CHECK-NEXT: [[TMP36:%.*]] = getelementptr double, ptr [[TMP7]], i64 10
; CHECK-NEXT: [[COL_LOAD102:%.*]] = load <2 x double>, ptr [[TMP36]], align 8
; CHECK-NEXT: [[VEC_GEP103:%.*]] = getelementptr double, ptr [[TMP7]], i64 14
; CHECK-NEXT: [[COL_LOAD104:%.*]] = load <2 x double>, ptr [[VEC_GEP103]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT108:%.*]] = shufflevector <2 x double> [[COL_LOAD102]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP37:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD99]], <2 x double> [[SPLAT_SPLAT108]], <2 x double> [[TMP32]])
; CHECK-NEXT: [[SPLAT_SPLAT111:%.*]] = shufflevector <2 x double> [[COL_LOAD102]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP38:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD101]], <2 x double> [[SPLAT_SPLAT111]], <2 x double> [[TMP37]])
; CHECK-NEXT: [[SPLAT_SPLAT115:%.*]] = shufflevector <2 x double> [[COL_LOAD104]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP39:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD99]], <2 x double> [[SPLAT_SPLAT115]], <2 x double> [[TMP34]])
; CHECK-NEXT: [[SPLAT_SPLAT118:%.*]] = shufflevector <2 x double> [[COL_LOAD104]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP40:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD101]], <2 x double> [[SPLAT_SPLAT118]], <2 x double> [[TMP39]])
; CHECK-NEXT: [[TMP41:%.*]] = getelementptr double, ptr [[C]], i64 8
; CHECK-NEXT: store <2 x double> [[TMP38]], ptr [[TMP41]], align 8
; CHECK-NEXT: [[VEC_GEP119:%.*]] = getelementptr double, ptr [[C]], i64 12
; CHECK-NEXT: store <2 x double> [[TMP40]], ptr [[VEC_GEP119]], align 8
; CHECK-NEXT: [[TMP42:%.*]] = getelementptr double, ptr [[TMP3]], i64 2
; CHECK-NEXT: [[COL_LOAD120:%.*]] = load <2 x double>, ptr [[TMP42]], align 8
; CHECK-NEXT: [[VEC_GEP121:%.*]] = getelementptr double, ptr [[TMP3]], i64 6
; CHECK-NEXT: [[COL_LOAD122:%.*]] = load <2 x double>, ptr [[VEC_GEP121]], align 8
; CHECK-NEXT: [[TMP43:%.*]] = getelementptr double, ptr [[TMP7]], i64 8
; CHECK-NEXT: [[COL_LOAD123:%.*]] = load <2 x double>, ptr [[TMP43]], align 8
; CHECK-NEXT: [[VEC_GEP124:%.*]] = getelementptr double, ptr [[TMP7]], i64 12
; CHECK-NEXT: [[COL_LOAD125:%.*]] = load <2 x double>, ptr [[VEC_GEP124]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT128:%.*]] = shufflevector <2 x double> [[COL_LOAD123]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP44:%.*]] = fmul contract <2 x double> [[COL_LOAD120]], [[SPLAT_SPLAT128]]
; CHECK-NEXT: [[SPLAT_SPLAT131:%.*]] = shufflevector <2 x double> [[COL_LOAD123]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP45:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD122]], <2 x double> [[SPLAT_SPLAT131]], <2 x double> [[TMP44]])
; CHECK-NEXT: [[SPLAT_SPLAT134:%.*]] = shufflevector <2 x double> [[COL_LOAD125]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP46:%.*]] = fmul contract <2 x double> [[COL_LOAD120]], [[SPLAT_SPLAT134]]
; CHECK-NEXT: [[SPLAT_SPLAT137:%.*]] = shufflevector <2 x double> [[COL_LOAD125]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP47:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD122]], <2 x double> [[SPLAT_SPLAT137]], <2 x double> [[TMP46]])
; CHECK-NEXT: [[TMP48:%.*]] = getelementptr double, ptr [[TMP3]], i64 10
; CHECK-NEXT: [[COL_LOAD138:%.*]] = load <2 x double>, ptr [[TMP48]], align 8
; CHECK-NEXT: [[VEC_GEP139:%.*]] = getelementptr double, ptr [[TMP3]], i64 14
; CHECK-NEXT: [[COL_LOAD140:%.*]] = load <2 x double>, ptr [[VEC_GEP139]], align 8
; CHECK-NEXT: [[TMP49:%.*]] = getelementptr double, ptr [[TMP7]], i64 10
; CHECK-NEXT: [[COL_LOAD141:%.*]] = load <2 x double>, ptr [[TMP49]], align 8
; CHECK-NEXT: [[VEC_GEP142:%.*]] = getelementptr double, ptr [[TMP7]], i64 14
; CHECK-NEXT: [[COL_LOAD143:%.*]] = load <2 x double>, ptr [[VEC_GEP142]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT147:%.*]] = shufflevector <2 x double> [[COL_LOAD141]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP50:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD138]], <2 x double> [[SPLAT_SPLAT147]], <2 x double> [[TMP45]])
; CHECK-NEXT: [[SPLAT_SPLAT150:%.*]] = shufflevector <2 x double> [[COL_LOAD141]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP51:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD140]], <2 x double> [[SPLAT_SPLAT150]], <2 x double> [[TMP50]])
; CHECK-NEXT: [[SPLAT_SPLAT154:%.*]] = shufflevector <2 x double> [[COL_LOAD143]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP52:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD138]], <2 x double> [[SPLAT_SPLAT154]], <2 x double> [[TMP47]])
; CHECK-NEXT: [[SPLAT_SPLAT157:%.*]] = shufflevector <2 x double> [[COL_LOAD143]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP53:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD140]], <2 x double> [[SPLAT_SPLAT157]], <2 x double> [[TMP52]])
; CHECK-NEXT: [[TMP54:%.*]] = getelementptr double, ptr [[C]], i64 10
; CHECK-NEXT: store <2 x double> [[TMP51]], ptr [[TMP54]], align 8
; CHECK-NEXT: [[VEC_GEP158:%.*]] = getelementptr double, ptr [[C]], i64 14
; CHECK-NEXT: store <2 x double> [[TMP53]], ptr [[VEC_GEP158]], align 8
; CHECK-NEXT: ret void
;
;; np.dot(a[0:2, 0:2], b[0:2, 0:2])
;; + np.dot(a[0:2, 2:4], b[2:4, 0:2])
;; -> c[0:2, 0:2]
;; np.dot(a[2:4, 0:2], b[0:2, 0:2])
;; + np.dot(a[2:4, 2:4], b[2:4, 0:2])
;; -> c[2:4, 0:2]
;; np.dot(a[0:2, 0:2], b[0:2, 2:4])
;; + np.dot(a[0:2, 2:4], b[2:4, 2:4])
;; -> c[0:2, 2:4]
;; np.dot(a[2:4, 0:2], b[2:4, 0:2])
;; + np.dot(a[2:4, 2:4], b[2:4, 2:4])
;; -> c[2:4, 2:4]
entry:
%a = load <16 x double>, ptr %A, align 8
%b = load <16 x double>, ptr %B, align 8
%c = call <16 x double> @llvm.matrix.multiply(<16 x double> %a, <16 x double> %b, i32 4, i32 4, i32 4)
store <16 x double> %c, ptr %C, align 8
ret void
}
; The same load is used for both operands of the multiply.
define void @multiply_reuse_load(ptr noalias %A, ptr noalias %B, ptr noalias %C) {
; CHECK-LABEL: @multiply_reuse_load(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[COL_LOAD:%.*]] = load <2 x double>, ptr [[A:%.*]], align 8
; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr double, ptr [[A]], i64 4
; CHECK-NEXT: [[COL_LOAD1:%.*]] = load <2 x double>, ptr [[VEC_GEP]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <2 x double> [[COL_LOAD]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP0:%.*]] = fmul contract <2 x double> [[COL_LOAD]], [[SPLAT_SPLAT]]
; CHECK-NEXT: [[SPLAT_SPLAT7:%.*]] = shufflevector <2 x double> [[COL_LOAD]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP1:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD1]], <2 x double> [[SPLAT_SPLAT7]], <2 x double> [[TMP0]])
; CHECK-NEXT: [[SPLAT_SPLAT10:%.*]] = shufflevector <2 x double> [[COL_LOAD1]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP2:%.*]] = fmul contract <2 x double> [[COL_LOAD]], [[SPLAT_SPLAT10]]
; CHECK-NEXT: [[SPLAT_SPLAT13:%.*]] = shufflevector <2 x double> [[COL_LOAD1]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP3:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD1]], <2 x double> [[SPLAT_SPLAT13]], <2 x double> [[TMP2]])
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr double, ptr [[A]], i64 8
; CHECK-NEXT: [[COL_LOAD14:%.*]] = load <2 x double>, ptr [[TMP4]], align 8
; CHECK-NEXT: [[VEC_GEP15:%.*]] = getelementptr double, ptr [[A]], i64 12
; CHECK-NEXT: [[COL_LOAD16:%.*]] = load <2 x double>, ptr [[VEC_GEP15]], align 8
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr double, ptr [[A]], i64 2
; CHECK-NEXT: [[COL_LOAD17:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
; CHECK-NEXT: [[VEC_GEP18:%.*]] = getelementptr double, ptr [[A]], i64 6
; CHECK-NEXT: [[COL_LOAD19:%.*]] = load <2 x double>, ptr [[VEC_GEP18]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT23:%.*]] = shufflevector <2 x double> [[COL_LOAD17]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP6:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD14]], <2 x double> [[SPLAT_SPLAT23]], <2 x double> [[TMP1]])
; CHECK-NEXT: [[SPLAT_SPLAT26:%.*]] = shufflevector <2 x double> [[COL_LOAD17]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP7:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD16]], <2 x double> [[SPLAT_SPLAT26]], <2 x double> [[TMP6]])
; CHECK-NEXT: [[SPLAT_SPLAT30:%.*]] = shufflevector <2 x double> [[COL_LOAD19]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP8:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD14]], <2 x double> [[SPLAT_SPLAT30]], <2 x double> [[TMP3]])
; CHECK-NEXT: [[SPLAT_SPLAT33:%.*]] = shufflevector <2 x double> [[COL_LOAD19]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP9:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD16]], <2 x double> [[SPLAT_SPLAT33]], <2 x double> [[TMP8]])
; CHECK-NEXT: store <2 x double> [[TMP7]], ptr [[C:%.*]], align 8
; CHECK-NEXT: [[VEC_GEP34:%.*]] = getelementptr double, ptr [[C]], i64 4
; CHECK-NEXT: store <2 x double> [[TMP9]], ptr [[VEC_GEP34]], align 8
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr double, ptr [[A]], i64 2
; CHECK-NEXT: [[COL_LOAD35:%.*]] = load <2 x double>, ptr [[TMP10]], align 8
; CHECK-NEXT: [[VEC_GEP36:%.*]] = getelementptr double, ptr [[A]], i64 6
; CHECK-NEXT: [[COL_LOAD37:%.*]] = load <2 x double>, ptr [[VEC_GEP36]], align 8
; CHECK-NEXT: [[COL_LOAD38:%.*]] = load <2 x double>, ptr [[A]], align 8
; CHECK-NEXT: [[VEC_GEP39:%.*]] = getelementptr double, ptr [[A]], i64 4
; CHECK-NEXT: [[COL_LOAD40:%.*]] = load <2 x double>, ptr [[VEC_GEP39]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT43:%.*]] = shufflevector <2 x double> [[COL_LOAD38]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP11:%.*]] = fmul contract <2 x double> [[COL_LOAD35]], [[SPLAT_SPLAT43]]
; CHECK-NEXT: [[SPLAT_SPLAT46:%.*]] = shufflevector <2 x double> [[COL_LOAD38]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP12:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD37]], <2 x double> [[SPLAT_SPLAT46]], <2 x double> [[TMP11]])
; CHECK-NEXT: [[SPLAT_SPLAT49:%.*]] = shufflevector <2 x double> [[COL_LOAD40]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP13:%.*]] = fmul contract <2 x double> [[COL_LOAD35]], [[SPLAT_SPLAT49]]
; CHECK-NEXT: [[SPLAT_SPLAT52:%.*]] = shufflevector <2 x double> [[COL_LOAD40]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP14:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD37]], <2 x double> [[SPLAT_SPLAT52]], <2 x double> [[TMP13]])
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr double, ptr [[A]], i64 10
; CHECK-NEXT: [[COL_LOAD53:%.*]] = load <2 x double>, ptr [[TMP15]], align 8
; CHECK-NEXT: [[VEC_GEP54:%.*]] = getelementptr double, ptr [[A]], i64 14
; CHECK-NEXT: [[COL_LOAD55:%.*]] = load <2 x double>, ptr [[VEC_GEP54]], align 8
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr double, ptr [[A]], i64 2
; CHECK-NEXT: [[COL_LOAD56:%.*]] = load <2 x double>, ptr [[TMP16]], align 8
; CHECK-NEXT: [[VEC_GEP57:%.*]] = getelementptr double, ptr [[A]], i64 6
; CHECK-NEXT: [[COL_LOAD58:%.*]] = load <2 x double>, ptr [[VEC_GEP57]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT62:%.*]] = shufflevector <2 x double> [[COL_LOAD56]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP17:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD53]], <2 x double> [[SPLAT_SPLAT62]], <2 x double> [[TMP12]])
; CHECK-NEXT: [[SPLAT_SPLAT65:%.*]] = shufflevector <2 x double> [[COL_LOAD56]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP18:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD55]], <2 x double> [[SPLAT_SPLAT65]], <2 x double> [[TMP17]])
; CHECK-NEXT: [[SPLAT_SPLAT69:%.*]] = shufflevector <2 x double> [[COL_LOAD58]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP19:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD53]], <2 x double> [[SPLAT_SPLAT69]], <2 x double> [[TMP14]])
; CHECK-NEXT: [[SPLAT_SPLAT72:%.*]] = shufflevector <2 x double> [[COL_LOAD58]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP20:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD55]], <2 x double> [[SPLAT_SPLAT72]], <2 x double> [[TMP19]])
; CHECK-NEXT: [[TMP21:%.*]] = getelementptr double, ptr [[C]], i64 2
; CHECK-NEXT: store <2 x double> [[TMP18]], ptr [[TMP21]], align 8
; CHECK-NEXT: [[VEC_GEP73:%.*]] = getelementptr double, ptr [[C]], i64 6
; CHECK-NEXT: store <2 x double> [[TMP20]], ptr [[VEC_GEP73]], align 8
; CHECK-NEXT: [[COL_LOAD74:%.*]] = load <2 x double>, ptr [[A]], align 8
; CHECK-NEXT: [[VEC_GEP75:%.*]] = getelementptr double, ptr [[A]], i64 4
; CHECK-NEXT: [[COL_LOAD76:%.*]] = load <2 x double>, ptr [[VEC_GEP75]], align 8
; CHECK-NEXT: [[TMP22:%.*]] = getelementptr double, ptr [[A]], i64 8
; CHECK-NEXT: [[COL_LOAD77:%.*]] = load <2 x double>, ptr [[TMP22]], align 8
; CHECK-NEXT: [[VEC_GEP78:%.*]] = getelementptr double, ptr [[A]], i64 12
; CHECK-NEXT: [[COL_LOAD79:%.*]] = load <2 x double>, ptr [[VEC_GEP78]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT82:%.*]] = shufflevector <2 x double> [[COL_LOAD77]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP23:%.*]] = fmul contract <2 x double> [[COL_LOAD74]], [[SPLAT_SPLAT82]]
; CHECK-NEXT: [[SPLAT_SPLAT85:%.*]] = shufflevector <2 x double> [[COL_LOAD77]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP24:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD76]], <2 x double> [[SPLAT_SPLAT85]], <2 x double> [[TMP23]])
; CHECK-NEXT: [[SPLAT_SPLAT88:%.*]] = shufflevector <2 x double> [[COL_LOAD79]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP25:%.*]] = fmul contract <2 x double> [[COL_LOAD74]], [[SPLAT_SPLAT88]]
; CHECK-NEXT: [[SPLAT_SPLAT91:%.*]] = shufflevector <2 x double> [[COL_LOAD79]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP26:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD76]], <2 x double> [[SPLAT_SPLAT91]], <2 x double> [[TMP25]])
; CHECK-NEXT: [[TMP27:%.*]] = getelementptr double, ptr [[A]], i64 8
; CHECK-NEXT: [[COL_LOAD92:%.*]] = load <2 x double>, ptr [[TMP27]], align 8
; CHECK-NEXT: [[VEC_GEP93:%.*]] = getelementptr double, ptr [[A]], i64 12
; CHECK-NEXT: [[COL_LOAD94:%.*]] = load <2 x double>, ptr [[VEC_GEP93]], align 8
; CHECK-NEXT: [[TMP28:%.*]] = getelementptr double, ptr [[A]], i64 10
; CHECK-NEXT: [[COL_LOAD95:%.*]] = load <2 x double>, ptr [[TMP28]], align 8
; CHECK-NEXT: [[VEC_GEP96:%.*]] = getelementptr double, ptr [[A]], i64 14
; CHECK-NEXT: [[COL_LOAD97:%.*]] = load <2 x double>, ptr [[VEC_GEP96]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT101:%.*]] = shufflevector <2 x double> [[COL_LOAD95]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP29:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD92]], <2 x double> [[SPLAT_SPLAT101]], <2 x double> [[TMP24]])
; CHECK-NEXT: [[SPLAT_SPLAT104:%.*]] = shufflevector <2 x double> [[COL_LOAD95]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP30:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD94]], <2 x double> [[SPLAT_SPLAT104]], <2 x double> [[TMP29]])
; CHECK-NEXT: [[SPLAT_SPLAT108:%.*]] = shufflevector <2 x double> [[COL_LOAD97]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP31:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD92]], <2 x double> [[SPLAT_SPLAT108]], <2 x double> [[TMP26]])
; CHECK-NEXT: [[SPLAT_SPLAT111:%.*]] = shufflevector <2 x double> [[COL_LOAD97]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP32:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD94]], <2 x double> [[SPLAT_SPLAT111]], <2 x double> [[TMP31]])
; CHECK-NEXT: [[TMP33:%.*]] = getelementptr double, ptr [[C]], i64 8
; CHECK-NEXT: store <2 x double> [[TMP30]], ptr [[TMP33]], align 8
; CHECK-NEXT: [[VEC_GEP112:%.*]] = getelementptr double, ptr [[C]], i64 12
; CHECK-NEXT: store <2 x double> [[TMP32]], ptr [[VEC_GEP112]], align 8
; CHECK-NEXT: [[TMP34:%.*]] = getelementptr double, ptr [[A]], i64 2
; CHECK-NEXT: [[COL_LOAD113:%.*]] = load <2 x double>, ptr [[TMP34]], align 8
; CHECK-NEXT: [[VEC_GEP114:%.*]] = getelementptr double, ptr [[A]], i64 6
; CHECK-NEXT: [[COL_LOAD115:%.*]] = load <2 x double>, ptr [[VEC_GEP114]], align 8
; CHECK-NEXT: [[TMP35:%.*]] = getelementptr double, ptr [[A]], i64 8
; CHECK-NEXT: [[COL_LOAD116:%.*]] = load <2 x double>, ptr [[TMP35]], align 8
; CHECK-NEXT: [[VEC_GEP117:%.*]] = getelementptr double, ptr [[A]], i64 12
; CHECK-NEXT: [[COL_LOAD118:%.*]] = load <2 x double>, ptr [[VEC_GEP117]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT121:%.*]] = shufflevector <2 x double> [[COL_LOAD116]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP36:%.*]] = fmul contract <2 x double> [[COL_LOAD113]], [[SPLAT_SPLAT121]]
; CHECK-NEXT: [[SPLAT_SPLAT124:%.*]] = shufflevector <2 x double> [[COL_LOAD116]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP37:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD115]], <2 x double> [[SPLAT_SPLAT124]], <2 x double> [[TMP36]])
; CHECK-NEXT: [[SPLAT_SPLAT127:%.*]] = shufflevector <2 x double> [[COL_LOAD118]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP38:%.*]] = fmul contract <2 x double> [[COL_LOAD113]], [[SPLAT_SPLAT127]]
; CHECK-NEXT: [[SPLAT_SPLAT130:%.*]] = shufflevector <2 x double> [[COL_LOAD118]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP39:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD115]], <2 x double> [[SPLAT_SPLAT130]], <2 x double> [[TMP38]])
; CHECK-NEXT: [[TMP40:%.*]] = getelementptr double, ptr [[A]], i64 10
; CHECK-NEXT: [[COL_LOAD131:%.*]] = load <2 x double>, ptr [[TMP40]], align 8
; CHECK-NEXT: [[VEC_GEP132:%.*]] = getelementptr double, ptr [[A]], i64 14
; CHECK-NEXT: [[COL_LOAD133:%.*]] = load <2 x double>, ptr [[VEC_GEP132]], align 8
; CHECK-NEXT: [[SPLAT_SPLAT140:%.*]] = shufflevector <2 x double> [[COL_LOAD131]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP41:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD131]], <2 x double> [[SPLAT_SPLAT140]], <2 x double> [[TMP37]])
; CHECK-NEXT: [[SPLAT_SPLAT143:%.*]] = shufflevector <2 x double> [[COL_LOAD131]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP42:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD133]], <2 x double> [[SPLAT_SPLAT143]], <2 x double> [[TMP41]])
; CHECK-NEXT: [[SPLAT_SPLAT147:%.*]] = shufflevector <2 x double> [[COL_LOAD133]], <2 x double> poison, <2 x i32> zeroinitializer
; CHECK-NEXT: [[TMP43:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD131]], <2 x double> [[SPLAT_SPLAT147]], <2 x double> [[TMP39]])
; CHECK-NEXT: [[SPLAT_SPLAT150:%.*]] = shufflevector <2 x double> [[COL_LOAD133]], <2 x double> undef, <2 x i32> <i32 1, i32 1>
; CHECK-NEXT: [[TMP44:%.*]] = call contract <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[COL_LOAD133]], <2 x double> [[SPLAT_SPLAT150]], <2 x double> [[TMP43]])
; CHECK-NEXT: [[TMP45:%.*]] = getelementptr double, ptr [[C]], i64 10
; CHECK-NEXT: store <2 x double> [[TMP42]], ptr [[TMP45]], align 8
; CHECK-NEXT: [[VEC_GEP151:%.*]] = getelementptr double, ptr [[C]], i64 14
; CHECK-NEXT: store <2 x double> [[TMP44]], ptr [[VEC_GEP151]], align 8
; CHECK-NEXT: ret void
;
entry:
%a = load <16 x double>, ptr %A, align 8
%c = call <16 x double> @llvm.matrix.multiply(<16 x double> %a, <16 x double> %a, i32 4, i32 4, i32 4)
store <16 x double> %c, ptr %C, align 8
ret void
}
declare <16 x double> @llvm.matrix.multiply(<16 x double>, <16 x double>, i32, i32, i32)