shylie/llvm-project
1
0
Fork 0
Code Issues Pull Requests Actions 6 Packages Projects Releases Wiki Activity
llvm-project/clang/test/CodeGenHLSL/BasicFeatures/MatrixSingleSubscriptSetter.hlsl
Farzon Lotfi 1bb916c9af
[HLSL] Represent Matrix as arrays of vectors in memory (#179861) 
fixes https://github.com/llvm/llvm-project/issues/179859

For matrix types we need to check the language mode so we can change the
matrix memory layout to arrays of vectors. To make this play nice with
how the rest of clang treats matrices we need to modify the
MaybeConvertMatrixAddress and the CreateMemTemp function to know how to
reconstruct a flattened vector.

Rest of changes is just test updates.
2026-02-06 16:25:20 -05:00

234 lines
16 KiB
HLSL
Raw Blame History

// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 6
// RUN: %clang_cc1 -triple dxil-pc-shadermodel6.7-library -disable-llvm-passes -emit-llvm -finclude-default-header -o - %s | FileCheck %s
// CHECK-LABEL: define hidden void @_Z9setMatrixRu11matrix_typeILm4ELm4EfEiDv4_f(
// CHECK-SAME: ptr noalias noundef nonnull align 4 dereferenceable(64) [[M:%.*]], i32 noundef [[INDEX:%.*]], <4 x float> noundef nofpclass(nan inf) [[V:%.*]]) #[[ATTR0:[0-9]+]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[M_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[INDEX_ADDR:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[V_ADDR:%.*]] = alloca <4 x float>, align 16
// CHECK-NEXT: store ptr [[M]], ptr [[M_ADDR]], align 4
// CHECK-NEXT: store i32 [[INDEX]], ptr [[INDEX_ADDR]], align 4
// CHECK-NEXT: store <4 x float> [[V]], ptr [[V_ADDR]], align 16
// CHECK-NEXT: [[TMP0:%.*]] = load <4 x float>, ptr [[V_ADDR]], align 16
// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 4, !nonnull [[META3:![0-9]+]], !align [[META4:![0-9]+]]
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[INDEX_ADDR]], align 4
// CHECK-NEXT: [[TMP3:%.*]] = add i32 0, [[TMP2]]
// CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP0]], i32 0
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <16 x float>, ptr [[TMP1]], i32 0, i32 [[TMP3]]
// CHECK-NEXT: store float [[TMP4]], ptr [[TMP5]], align 4
// CHECK-NEXT: [[TMP6:%.*]] = add i32 4, [[TMP2]]
// CHECK-NEXT: [[TMP7:%.*]] = extractelement <4 x float> [[TMP0]], i32 1
// CHECK-NEXT: [[TMP8:%.*]] = getelementptr <16 x float>, ptr [[TMP1]], i32 0, i32 [[TMP6]]
// CHECK-NEXT: store float [[TMP7]], ptr [[TMP8]], align 4
// CHECK-NEXT: [[TMP9:%.*]] = add i32 8, [[TMP2]]
// CHECK-NEXT: [[TMP10:%.*]] = extractelement <4 x float> [[TMP0]], i32 2
// CHECK-NEXT: [[TMP11:%.*]] = getelementptr <16 x float>, ptr [[TMP1]], i32 0, i32 [[TMP9]]
// CHECK-NEXT: store float [[TMP10]], ptr [[TMP11]], align 4
// CHECK-NEXT: [[TMP12:%.*]] = add i32 12, [[TMP2]]
// CHECK-NEXT: [[TMP13:%.*]] = extractelement <4 x float> [[TMP0]], i32 3
// CHECK-NEXT: [[TMP14:%.*]] = getelementptr <16 x float>, ptr [[TMP1]], i32 0, i32 [[TMP12]]
// CHECK-NEXT: store float [[TMP13]], ptr [[TMP14]], align 4
// CHECK-NEXT: ret void
//
void setMatrix(out float4x4 M, int index, float4 V) {
M[index] = V;
}
// CHECK-LABEL: define hidden void @_Z15setMatrixScalarRu11matrix_typeILm2ELm1EfEif(
// CHECK-SAME: ptr noalias noundef nonnull align 4 dereferenceable(8) [[M:%.*]], i32 noundef [[INDEX:%.*]], float noundef nofpclass(nan inf) [[S:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[M_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[INDEX_ADDR:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[S_ADDR:%.*]] = alloca float, align 4
// CHECK-NEXT: store ptr [[M]], ptr [[M_ADDR]], align 4
// CHECK-NEXT: store i32 [[INDEX]], ptr [[INDEX_ADDR]], align 4
// CHECK-NEXT: store float [[S]], ptr [[S_ADDR]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load float, ptr [[S_ADDR]], align 4
// CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x float> poison, float [[TMP0]], i64 0
// CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x float> [[SPLAT_SPLATINSERT]], <1 x float> poison, <1 x i32> zeroinitializer
// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 4, !nonnull [[META3]], !align [[META4]]
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[INDEX_ADDR]], align 4
// CHECK-NEXT: [[TMP3:%.*]] = add i32 0, [[TMP2]]
// CHECK-NEXT: [[TMP4:%.*]] = extractelement <1 x float> [[SPLAT_SPLAT]], i32 0
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <2 x float>, ptr [[TMP1]], i32 0, i32 [[TMP3]]
// CHECK-NEXT: store float [[TMP4]], ptr [[TMP5]], align 4
// CHECK-NEXT: ret void
//
void setMatrixScalar(out float2x1 M, int index, float S) {
M[index] = S;
}
// CHECK-LABEL: define hidden void @_Z13setBoolMatrixRu11matrix_typeILm4ELm4EbEiDv4_b(
// CHECK-SAME: ptr noalias noundef nonnull align 4 dereferenceable(64) [[M:%.*]], i32 noundef [[INDEX:%.*]], <4 x i1> noundef [[V:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[M_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[INDEX_ADDR:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[V_ADDR:%.*]] = alloca <4 x i32>, align 16
// CHECK-NEXT: store ptr [[M]], ptr [[M_ADDR]], align 4
// CHECK-NEXT: store i32 [[INDEX]], ptr [[INDEX_ADDR]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = zext <4 x i1> [[V]] to <4 x i32>
// CHECK-NEXT: store <4 x i32> [[TMP0]], ptr [[V_ADDR]], align 16
// CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, ptr [[V_ADDR]], align 16
// CHECK-NEXT: [[LOADEDV:%.*]] = trunc <4 x i32> [[TMP1]] to <4 x i1>
// CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[M_ADDR]], align 4, !nonnull [[META3]], !align [[META4]]
// CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[INDEX_ADDR]], align 4
// CHECK-NEXT: [[TMP4:%.*]] = zext <4 x i1> [[LOADEDV]] to <4 x i32>
// CHECK-NEXT: [[TMP5:%.*]] = add i32 0, [[TMP3]]
// CHECK-NEXT: [[TMP6:%.*]] = extractelement <4 x i32> [[TMP4]], i32 0
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <16 x i32>, ptr [[TMP2]], i32 0, i32 [[TMP5]]
// CHECK-NEXT: store i32 [[TMP6]], ptr [[TMP7]], align 4
// CHECK-NEXT: [[TMP8:%.*]] = add i32 4, [[TMP3]]
// CHECK-NEXT: [[TMP9:%.*]] = extractelement <4 x i32> [[TMP4]], i32 1
// CHECK-NEXT: [[TMP10:%.*]] = getelementptr <16 x i32>, ptr [[TMP2]], i32 0, i32 [[TMP8]]
// CHECK-NEXT: store i32 [[TMP9]], ptr [[TMP10]], align 4
// CHECK-NEXT: [[TMP11:%.*]] = add i32 8, [[TMP3]]
// CHECK-NEXT: [[TMP12:%.*]] = extractelement <4 x i32> [[TMP4]], i32 2
// CHECK-NEXT: [[TMP13:%.*]] = getelementptr <16 x i32>, ptr [[TMP2]], i32 0, i32 [[TMP11]]
// CHECK-NEXT: store i32 [[TMP12]], ptr [[TMP13]], align 4
// CHECK-NEXT: [[TMP14:%.*]] = add i32 12, [[TMP3]]
// CHECK-NEXT: [[TMP15:%.*]] = extractelement <4 x i32> [[TMP4]], i32 3
// CHECK-NEXT: [[TMP16:%.*]] = getelementptr <16 x i32>, ptr [[TMP2]], i32 0, i32 [[TMP14]]
// CHECK-NEXT: store i32 [[TMP15]], ptr [[TMP16]], align 4
// CHECK-NEXT: ret void
//
void setBoolMatrix(out bool4x4 M, int index, bool4 V) {
M[index] = V;
}
// CHECK-LABEL: define hidden void @_Z19setBoolMatrixScalarRu11matrix_typeILm2ELm1EbEib(
// CHECK-SAME: ptr noalias noundef nonnull align 4 dereferenceable(8) [[M:%.*]], i32 noundef [[INDEX:%.*]], i1 noundef [[S:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[M_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[INDEX_ADDR:%.*]] = alloca i32, align 4
// CHECK-NEXT: [[S_ADDR:%.*]] = alloca i32, align 4
// CHECK-NEXT: store ptr [[M]], ptr [[M_ADDR]], align 4
// CHECK-NEXT: store i32 [[INDEX]], ptr [[INDEX_ADDR]], align 4
// CHECK-NEXT: [[STOREDV:%.*]] = zext i1 [[S]] to i32
// CHECK-NEXT: store i32 [[STOREDV]], ptr [[S_ADDR]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load i32, ptr [[S_ADDR]], align 4
// CHECK-NEXT: [[LOADEDV:%.*]] = trunc i32 [[TMP0]] to i1
// CHECK-NEXT: [[SPLAT_SPLATINSERT:%.*]] = insertelement <1 x i1> poison, i1 [[LOADEDV]], i64 0
// CHECK-NEXT: [[SPLAT_SPLAT:%.*]] = shufflevector <1 x i1> [[SPLAT_SPLATINSERT]], <1 x i1> poison, <1 x i32> zeroinitializer
// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 4, !nonnull [[META3]], !align [[META4]]
// CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[INDEX_ADDR]], align 4
// CHECK-NEXT: [[TMP3:%.*]] = zext <1 x i1> [[SPLAT_SPLAT]] to <1 x i32>
// CHECK-NEXT: [[TMP4:%.*]] = add i32 0, [[TMP2]]
// CHECK-NEXT: [[TMP5:%.*]] = extractelement <1 x i32> [[TMP3]], i32 0
// CHECK-NEXT: [[TMP6:%.*]] = getelementptr <2 x i32>, ptr [[TMP1]], i32 0, i32 [[TMP4]]
// CHECK-NEXT: store i32 [[TMP5]], ptr [[TMP6]], align 4
// CHECK-NEXT: ret void
//
void setBoolMatrixScalar(out bool2x1 M, int index, bool S) {
M[index] = S;
}
// CHECK-LABEL: define hidden void @_Z19setMatrixConstIndexRu11matrix_typeILm4ELm4EiES_(
// CHECK-SAME: ptr noalias noundef nonnull align 4 dereferenceable(64) [[M:%.*]], <16 x i32> noundef [[N:%.*]]) #[[ATTR0]] {
// CHECK-NEXT: [[ENTRY:.*:]]
// CHECK-NEXT: [[M_ADDR:%.*]] = alloca ptr, align 4
// CHECK-NEXT: [[N_ADDR:%.*]] = alloca [4 x <4 x i32>], align 4
// CHECK-NEXT: store ptr [[M]], ptr [[M_ADDR]], align 4
// CHECK-NEXT: store <16 x i32> [[N]], ptr [[N_ADDR]], align 4
// CHECK-NEXT: [[TMP0:%.*]] = load <16 x i32>, ptr [[N_ADDR]], align 4
// CHECK-NEXT: [[MATRIX_ELEM:%.*]] = extractelement <16 x i32> [[TMP0]], i32 3
// CHECK-NEXT: [[MATRIX_ROW_INS:%.*]] = insertelement <4 x i32> poison, i32 [[MATRIX_ELEM]], i32 0
// CHECK-NEXT: [[MATRIX_ELEM1:%.*]] = extractelement <16 x i32> [[TMP0]], i32 7
// CHECK-NEXT: [[MATRIX_ROW_INS2:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS]], i32 [[MATRIX_ELEM1]], i32 1
// CHECK-NEXT: [[MATRIX_ELEM3:%.*]] = extractelement <16 x i32> [[TMP0]], i32 11
// CHECK-NEXT: [[MATRIX_ROW_INS4:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS2]], i32 [[MATRIX_ELEM3]], i32 2
// CHECK-NEXT: [[MATRIX_ELEM5:%.*]] = extractelement <16 x i32> [[TMP0]], i32 15
// CHECK-NEXT: [[MATRIX_ROW_INS6:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS4]], i32 [[MATRIX_ELEM5]], i32 3
// CHECK-NEXT: [[TMP1:%.*]] = load ptr, ptr [[M_ADDR]], align 4, !nonnull [[META3]], !align [[META4]]
// CHECK-NEXT: [[TMP2:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS6]], i32 0
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr <16 x i32>, ptr [[TMP1]], i32 0, i32 0
// CHECK-NEXT: store i32 [[TMP2]], ptr [[TMP3]], align 4
// CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS6]], i32 1
// CHECK-NEXT: [[TMP5:%.*]] = getelementptr <16 x i32>, ptr [[TMP1]], i32 0, i32 4
// CHECK-NEXT: store i32 [[TMP4]], ptr [[TMP5]], align 4
// CHECK-NEXT: [[TMP6:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS6]], i32 2
// CHECK-NEXT: [[TMP7:%.*]] = getelementptr <16 x i32>, ptr [[TMP1]], i32 0, i32 8
// CHECK-NEXT: store i32 [[TMP6]], ptr [[TMP7]], align 4
// CHECK-NEXT: [[TMP8:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS6]], i32 3
// CHECK-NEXT: [[TMP9:%.*]] = getelementptr <16 x i32>, ptr [[TMP1]], i32 0, i32 12
// CHECK-NEXT: store i32 [[TMP8]], ptr [[TMP9]], align 4
// CHECK-NEXT: [[TMP10:%.*]] = load <16 x i32>, ptr [[N_ADDR]], align 4
// CHECK-NEXT: [[MATRIX_ELEM7:%.*]] = extractelement <16 x i32> [[TMP10]], i32 2
// CHECK-NEXT: [[MATRIX_ROW_INS8:%.*]] = insertelement <4 x i32> poison, i32 [[MATRIX_ELEM7]], i32 0
// CHECK-NEXT: [[MATRIX_ELEM9:%.*]] = extractelement <16 x i32> [[TMP10]], i32 6
// CHECK-NEXT: [[MATRIX_ROW_INS10:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS8]], i32 [[MATRIX_ELEM9]], i32 1
// CHECK-NEXT: [[MATRIX_ELEM11:%.*]] = extractelement <16 x i32> [[TMP10]], i32 10
// CHECK-NEXT: [[MATRIX_ROW_INS12:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS10]], i32 [[MATRIX_ELEM11]], i32 2
// CHECK-NEXT: [[MATRIX_ELEM13:%.*]] = extractelement <16 x i32> [[TMP10]], i32 14
// CHECK-NEXT: [[MATRIX_ROW_INS14:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS12]], i32 [[MATRIX_ELEM13]], i32 3
// CHECK-NEXT: [[TMP11:%.*]] = load ptr, ptr [[M_ADDR]], align 4, !nonnull [[META3]], !align [[META4]]
// CHECK-NEXT: [[TMP12:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS14]], i32 0
// CHECK-NEXT: [[TMP13:%.*]] = getelementptr <16 x i32>, ptr [[TMP11]], i32 0, i32 1
// CHECK-NEXT: store i32 [[TMP12]], ptr [[TMP13]], align 4
// CHECK-NEXT: [[TMP14:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS14]], i32 1
// CHECK-NEXT: [[TMP15:%.*]] = getelementptr <16 x i32>, ptr [[TMP11]], i32 0, i32 5
// CHECK-NEXT: store i32 [[TMP14]], ptr [[TMP15]], align 4
// CHECK-NEXT: [[TMP16:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS14]], i32 2
// CHECK-NEXT: [[TMP17:%.*]] = getelementptr <16 x i32>, ptr [[TMP11]], i32 0, i32 9
// CHECK-NEXT: store i32 [[TMP16]], ptr [[TMP17]], align 4
// CHECK-NEXT: [[TMP18:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS14]], i32 3
// CHECK-NEXT: [[TMP19:%.*]] = getelementptr <16 x i32>, ptr [[TMP11]], i32 0, i32 13
// CHECK-NEXT: store i32 [[TMP18]], ptr [[TMP19]], align 4
// CHECK-NEXT: [[TMP20:%.*]] = load <16 x i32>, ptr [[N_ADDR]], align 4
// CHECK-NEXT: [[MATRIX_ELEM15:%.*]] = extractelement <16 x i32> [[TMP20]], i32 1
// CHECK-NEXT: [[MATRIX_ROW_INS16:%.*]] = insertelement <4 x i32> poison, i32 [[MATRIX_ELEM15]], i32 0
// CHECK-NEXT: [[MATRIX_ELEM17:%.*]] = extractelement <16 x i32> [[TMP20]], i32 5
// CHECK-NEXT: [[MATRIX_ROW_INS18:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS16]], i32 [[MATRIX_ELEM17]], i32 1
// CHECK-NEXT: [[MATRIX_ELEM19:%.*]] = extractelement <16 x i32> [[TMP20]], i32 9
// CHECK-NEXT: [[MATRIX_ROW_INS20:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS18]], i32 [[MATRIX_ELEM19]], i32 2
// CHECK-NEXT: [[MATRIX_ELEM21:%.*]] = extractelement <16 x i32> [[TMP20]], i32 13
// CHECK-NEXT: [[MATRIX_ROW_INS22:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS20]], i32 [[MATRIX_ELEM21]], i32 3
// CHECK-NEXT: [[TMP21:%.*]] = load ptr, ptr [[M_ADDR]], align 4, !nonnull [[META3]], !align [[META4]]
// CHECK-NEXT: [[TMP22:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS22]], i32 0
// CHECK-NEXT: [[TMP23:%.*]] = getelementptr <16 x i32>, ptr [[TMP21]], i32 0, i32 2
// CHECK-NEXT: store i32 [[TMP22]], ptr [[TMP23]], align 4
// CHECK-NEXT: [[TMP24:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS22]], i32 1
// CHECK-NEXT: [[TMP25:%.*]] = getelementptr <16 x i32>, ptr [[TMP21]], i32 0, i32 6
// CHECK-NEXT: store i32 [[TMP24]], ptr [[TMP25]], align 4
// CHECK-NEXT: [[TMP26:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS22]], i32 2
// CHECK-NEXT: [[TMP27:%.*]] = getelementptr <16 x i32>, ptr [[TMP21]], i32 0, i32 10
// CHECK-NEXT: store i32 [[TMP26]], ptr [[TMP27]], align 4
// CHECK-NEXT: [[TMP28:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS22]], i32 3
// CHECK-NEXT: [[TMP29:%.*]] = getelementptr <16 x i32>, ptr [[TMP21]], i32 0, i32 14
// CHECK-NEXT: store i32 [[TMP28]], ptr [[TMP29]], align 4
// CHECK-NEXT: [[TMP30:%.*]] = load <16 x i32>, ptr [[N_ADDR]], align 4
// CHECK-NEXT: [[MATRIX_ELEM23:%.*]] = extractelement <16 x i32> [[TMP30]], i32 0
// CHECK-NEXT: [[MATRIX_ROW_INS24:%.*]] = insertelement <4 x i32> poison, i32 [[MATRIX_ELEM23]], i32 0
// CHECK-NEXT: [[MATRIX_ELEM25:%.*]] = extractelement <16 x i32> [[TMP30]], i32 4
// CHECK-NEXT: [[MATRIX_ROW_INS26:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS24]], i32 [[MATRIX_ELEM25]], i32 1
// CHECK-NEXT: [[MATRIX_ELEM27:%.*]] = extractelement <16 x i32> [[TMP30]], i32 8
// CHECK-NEXT: [[MATRIX_ROW_INS28:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS26]], i32 [[MATRIX_ELEM27]], i32 2
// CHECK-NEXT: [[MATRIX_ELEM29:%.*]] = extractelement <16 x i32> [[TMP30]], i32 12
// CHECK-NEXT: [[MATRIX_ROW_INS30:%.*]] = insertelement <4 x i32> [[MATRIX_ROW_INS28]], i32 [[MATRIX_ELEM29]], i32 3
// CHECK-NEXT: [[TMP31:%.*]] = load ptr, ptr [[M_ADDR]], align 4, !nonnull [[META3]], !align [[META4]]
// CHECK-NEXT: [[TMP32:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS30]], i32 0
// CHECK-NEXT: [[TMP33:%.*]] = getelementptr <16 x i32>, ptr [[TMP31]], i32 0, i32 3
// CHECK-NEXT: store i32 [[TMP32]], ptr [[TMP33]], align 4
// CHECK-NEXT: [[TMP34:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS30]], i32 1
// CHECK-NEXT: [[TMP35:%.*]] = getelementptr <16 x i32>, ptr [[TMP31]], i32 0, i32 7
// CHECK-NEXT: store i32 [[TMP34]], ptr [[TMP35]], align 4
// CHECK-NEXT: [[TMP36:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS30]], i32 2
// CHECK-NEXT: [[TMP37:%.*]] = getelementptr <16 x i32>, ptr [[TMP31]], i32 0, i32 11
// CHECK-NEXT: store i32 [[TMP36]], ptr [[TMP37]], align 4
// CHECK-NEXT: [[TMP38:%.*]] = extractelement <4 x i32> [[MATRIX_ROW_INS30]], i32 3
// CHECK-NEXT: [[TMP39:%.*]] = getelementptr <16 x i32>, ptr [[TMP31]], i32 0, i32 15
// CHECK-NEXT: store i32 [[TMP38]], ptr [[TMP39]], align 4
// CHECK-NEXT: ret void
//
void setMatrixConstIndex(out int4x4 M, int4x4 N ) {
M[0] = N[3];
M[1] = N[2];
M[2] = N[1];
M[3] = N[0];
}
//.
// CHECK: [[META3]] = !{}
// CHECK: [[META4]] = !{i64 4}
//.
Reference in New Issue View Git Blame Copy Permalink