48 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; Currently we only lower stores with shape information, but need to embed the5; matrix in a flat vector for function calls and returns.6define <8 x double> @strided_load_4x4(<8 x double> %in, ptr %Ptr) {7; CHECK-LABEL: @strided_load_4x4(8; CHECK-NEXT: [[SPLIT:%.*]] = shufflevector <8 x double> [[IN:%.*]], <8 x double> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>9; CHECK-NEXT: [[SPLIT1:%.*]] = shufflevector <8 x double> [[IN]], <8 x double> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>10; CHECK-NEXT: [[TMP1:%.*]] = extractelement <4 x double> [[SPLIT]], i64 011; CHECK-NEXT: [[TMP2:%.*]] = insertelement <2 x double> poison, double [[TMP1]], i64 012; CHECK-NEXT: [[TMP3:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 013; CHECK-NEXT: [[TMP4:%.*]] = insertelement <2 x double> [[TMP2]], double [[TMP3]], i64 114; CHECK-NEXT: [[TMP5:%.*]] = extractelement <4 x double> [[SPLIT]], i64 115; CHECK-NEXT: [[TMP6:%.*]] = insertelement <2 x double> poison, double [[TMP5]], i64 016; CHECK-NEXT: [[TMP7:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 117; CHECK-NEXT: [[TMP8:%.*]] = insertelement <2 x double> [[TMP6]], double [[TMP7]], i64 118; CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x double> [[SPLIT]], i64 219; CHECK-NEXT: [[TMP10:%.*]] = insertelement <2 x double> poison, double [[TMP9]], i64 020; CHECK-NEXT: [[TMP11:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 221; CHECK-NEXT: [[TMP12:%.*]] = insertelement <2 x double> [[TMP10]], double [[TMP11]], i64 122; CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x double> [[SPLIT]], i64 323; CHECK-NEXT: [[TMP14:%.*]] = insertelement <2 x double> poison, double [[TMP13]], i64 024; CHECK-NEXT: [[TMP15:%.*]] = extractelement <4 x double> [[SPLIT1]], i64 325; CHECK-NEXT: [[TMP16:%.*]] = insertelement <2 x double> [[TMP14]], double [[TMP15]], i64 126; CHECK-NEXT: [[TMP17:%.*]] = shufflevector <2 x double> [[TMP4]], <2 x double> [[TMP8]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>27; CHECK-NEXT: [[TMP18:%.*]] = shufflevector <2 x double> [[TMP12]], <2 x double> [[TMP16]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>28; CHECK-NEXT: [[TMP19:%.*]] = shufflevector <4 x double> [[TMP17]], <4 x double> [[TMP18]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>29; CHECK-NEXT: store <2 x double> [[TMP4]], ptr [[PTR:%.*]], align 830; CHECK-NEXT: [[VEC_GEP:%.*]] = getelementptr double, ptr [[PTR]], i64 231; CHECK-NEXT: store <2 x double> [[TMP8]], ptr [[VEC_GEP]], align 832; CHECK-NEXT: [[VEC_GEP2:%.*]] = getelementptr double, ptr [[PTR]], i64 433; CHECK-NEXT: store <2 x double> [[TMP12]], ptr [[VEC_GEP2]], align 834; CHECK-NEXT: [[VEC_GEP3:%.*]] = getelementptr double, ptr [[PTR]], i64 635; CHECK-NEXT: store <2 x double> [[TMP16]], ptr [[VEC_GEP3]], align 836; CHECK-NEXT: call void @foo(<8 x double> [[TMP19]])37; CHECK-NEXT: ret <8 x double> [[TMP19]]38;39 %transposed = call <8 x double> @llvm.matrix.transpose(<8 x double> %in, i32 4, i32 2)40 store <8 x double> %transposed, ptr %Ptr, align 841 call void @foo(<8 x double> %transposed)42 ret <8 x double> %transposed43}44 45declare <8 x double> @llvm.matrix.transpose(<8 x double>, i32, i32)46 47declare void @foo(<8 x double>)48