Skip to content

Commit 2477f82

Browse files
virginia-cangelosivirginia-cangelosi
authored
[clang] Update SVE load and store intrinsics to have FP8 variants (#126726)
1 parent 9e2eb95 commit 2477f82

22 files changed

+1125
-57
lines changed

‎clang/include/clang/Basic/arm_sve.td

Lines changed: 25 additions & 25 deletions
Large diffs are not rendered by default.

‎clang/lib/CodeGen/CGBuiltin.cpp

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -10223,6 +10223,7 @@ llvm::Type *CodeGenFunction::getEltType(const SVETypeFlags &TypeFlags) {
1022310223
default:
1022410224
llvm_unreachable("Invalid SVETypeFlag!");
1022510225

10226+
case SVETypeFlags::EltTyMFloat8:
1022610227
case SVETypeFlags::EltTyInt8:
1022710228
return Builder.getInt8Ty();
1022810229
case SVETypeFlags::EltTyInt16:
@@ -10651,7 +10652,7 @@ Value *CodeGenFunction::EmitSVEMaskedLoad(const CallExpr *E,
1065110652
unsigned IntrinsicID,
1065210653
bool IsZExtReturn) {
1065310654
QualType LangPTy = E->getArg(1)->getType();
10654-
llvm::Type *MemEltTy = CGM.getTypes().ConvertType(
10655+
llvm::Type *MemEltTy = CGM.getTypes().ConvertTypeForMem(
1065510656
LangPTy->castAs<PointerType>()->getPointeeType());
1065610657

1065710658
// The vector type that is returned may be different from the
@@ -10698,7 +10699,7 @@ Value *CodeGenFunction::EmitSVEMaskedStore(const CallExpr *E,
1069810699
SmallVectorImpl<Value *> &Ops,
1069910700
unsigned IntrinsicID) {
1070010701
QualType LangPTy = E->getArg(1)->getType();
10701-
llvm::Type *MemEltTy = CGM.getTypes().ConvertType(
10702+
llvm::Type *MemEltTy = CGM.getTypes().ConvertTypeForMem(
1070210703
LangPTy->castAs<PointerType>()->getPointeeType());
1070310704

1070410705
// The vector type that is stored may be different from the

‎clang/lib/CodeGen/CodeGenTypes.cpp

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -107,6 +107,9 @@ llvm::Type *CodeGenTypes::ConvertTypeForMem(QualType T) {
107107
MT->getNumRows() * MT->getNumColumns());
108108
}
109109

110+
if (T->isMFloat8Type())
111+
return llvm::Type::getInt8Ty(getLLVMContext());
112+
110113
llvm::Type *R = ConvertType(T);
111114

112115
// Check for the boolean vector case.

‎clang/test/CodeGen/AArch64/fp8-intrinsics/acle_sve2_fp8_fdot.c

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -49,8 +49,8 @@ svfloat32_t test_svdot_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t zm,
4949
// CHECK-LABEL: define dso_local <vscale x 4 x float> @test_svdot_n_f32_mf8(
5050
// CHECK-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
5151
// CHECK-NEXT: [[ENTRY:.*:]]
52-
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
5352
// CHECK-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
53+
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
5454
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
5555
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
5656
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fdot.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -59,8 +59,8 @@ svfloat32_t test_svdot_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t zm,
5959
// CHECK-CXX-LABEL: define dso_local <vscale x 4 x float> @_Z20test_svdot_n_f32_mf8u13__SVFloat32_tu13__SVMfloat8_tu6__mfp8m(
6060
// CHECK-CXX-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
6161
// CHECK-CXX-NEXT: [[ENTRY:.*:]]
62-
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
6362
// CHECK-CXX-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
63+
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
6464
// CHECK-CXX-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
6565
// CHECK-CXX-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
6666
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fdot.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -91,8 +91,8 @@ svfloat16_t test_svdot_f16_mf8(svfloat16_t zda, svmfloat8_t zn, svmfloat8_t zm,
9191
// CHECK-LABEL: define dso_local <vscale x 8 x half> @test_svdot_n_f16_mf8(
9292
// CHECK-SAME: <vscale x 8 x half> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
9393
// CHECK-NEXT: [[ENTRY:.*:]]
94-
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
9594
// CHECK-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
95+
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
9696
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
9797
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
9898
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fdot.nxv8f16(<vscale x 8 x half> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -101,8 +101,8 @@ svfloat16_t test_svdot_f16_mf8(svfloat16_t zda, svmfloat8_t zn, svmfloat8_t zm,
101101
// CHECK-CXX-LABEL: define dso_local <vscale x 8 x half> @_Z20test_svdot_n_f16_mf8u13__SVFloat16_tu13__SVMfloat8_tu6__mfp8m(
102102
// CHECK-CXX-SAME: <vscale x 8 x half> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
103103
// CHECK-CXX-NEXT: [[ENTRY:.*:]]
104-
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
105104
// CHECK-CXX-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
105+
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
106106
// CHECK-CXX-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
107107
// CHECK-CXX-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
108108
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fdot.nxv8f16(<vscale x 8 x half> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])

‎clang/test/CodeGen/AArch64/fp8-intrinsics/acle_sve2_fp8_fmla.c

Lines changed: 12 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -49,8 +49,8 @@ svfloat16_t test_svmlalb_f16_mf8(svfloat16_t zda, svmfloat8_t zn, svmfloat8_t zm
4949
// CHECK-LABEL: define dso_local <vscale x 8 x half> @test_svmlalb_n_f16_mf8(
5050
// CHECK-SAME: <vscale x 8 x half> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
5151
// CHECK-NEXT: [[ENTRY:.*:]]
52-
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
5352
// CHECK-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
53+
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
5454
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
5555
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
5656
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fmlalb.nxv8f16(<vscale x 8 x half> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -59,8 +59,8 @@ svfloat16_t test_svmlalb_f16_mf8(svfloat16_t zda, svmfloat8_t zn, svmfloat8_t zm
5959
// CHECK-CXX-LABEL: define dso_local <vscale x 8 x half> @_Z22test_svmlalb_n_f16_mf8u13__SVFloat16_tu13__SVMfloat8_tu6__mfp8m(
6060
// CHECK-CXX-SAME: <vscale x 8 x half> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
6161
// CHECK-CXX-NEXT: [[ENTRY:.*:]]
62-
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
6362
// CHECK-CXX-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
63+
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
6464
// CHECK-CXX-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
6565
// CHECK-CXX-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
6666
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fmlalb.nxv8f16(<vscale x 8 x half> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -91,8 +91,8 @@ svfloat16_t test_svmlalt_f16_mf8(svfloat16_t zda, svmfloat8_t zn, svmfloat8_t zm
9191
// CHECK-LABEL: define dso_local <vscale x 8 x half> @test_svmlalt_n_f16_mf8(
9292
// CHECK-SAME: <vscale x 8 x half> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
9393
// CHECK-NEXT: [[ENTRY:.*:]]
94-
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
9594
// CHECK-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
95+
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
9696
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
9797
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
9898
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fmlalt.nxv8f16(<vscale x 8 x half> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -101,8 +101,8 @@ svfloat16_t test_svmlalt_f16_mf8(svfloat16_t zda, svmfloat8_t zn, svmfloat8_t zm
101101
// CHECK-CXX-LABEL: define dso_local <vscale x 8 x half> @_Z22test_svmlalt_n_f16_mf8u13__SVFloat16_tu13__SVMfloat8_tu6__mfp8m(
102102
// CHECK-CXX-SAME: <vscale x 8 x half> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
103103
// CHECK-CXX-NEXT: [[ENTRY:.*:]]
104-
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
105104
// CHECK-CXX-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
105+
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
106106
// CHECK-CXX-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
107107
// CHECK-CXX-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
108108
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.fp8.fmlalt.nxv8f16(<vscale x 8 x half> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -169,8 +169,8 @@ svfloat32_t test_svmlallbb_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t
169169
// CHECK-LABEL: define dso_local <vscale x 4 x float> @test_svmlallbb_n_f32_mf8(
170170
// CHECK-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
171171
// CHECK-NEXT: [[ENTRY:.*:]]
172-
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
173172
// CHECK-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
173+
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
174174
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
175175
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
176176
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlallbb.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -179,8 +179,8 @@ svfloat32_t test_svmlallbb_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t
179179
// CHECK-CXX-LABEL: define dso_local <vscale x 4 x float> @_Z24test_svmlallbb_n_f32_mf8u13__SVFloat32_tu13__SVMfloat8_tu6__mfp8m(
180180
// CHECK-CXX-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
181181
// CHECK-CXX-NEXT: [[ENTRY:.*:]]
182-
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
183182
// CHECK-CXX-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
183+
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
184184
// CHECK-CXX-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
185185
// CHECK-CXX-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
186186
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlallbb.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -211,8 +211,8 @@ svfloat32_t test_svmlallbt_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t
211211
// CHECK-LABEL: define dso_local <vscale x 4 x float> @test_svmlallbt_n_f32_mf8(
212212
// CHECK-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
213213
// CHECK-NEXT: [[ENTRY:.*:]]
214-
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
215214
// CHECK-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
215+
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
216216
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
217217
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
218218
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlallbt.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -221,8 +221,8 @@ svfloat32_t test_svmlallbt_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t
221221
// CHECK-CXX-LABEL: define dso_local <vscale x 4 x float> @_Z24test_svmlallbt_n_f32_mf8u13__SVFloat32_tu13__SVMfloat8_tu6__mfp8m(
222222
// CHECK-CXX-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
223223
// CHECK-CXX-NEXT: [[ENTRY:.*:]]
224-
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
225224
// CHECK-CXX-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
225+
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
226226
// CHECK-CXX-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
227227
// CHECK-CXX-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
228228
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlallbt.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -253,8 +253,8 @@ svfloat32_t test_svmlalltb_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t
253253
// CHECK-LABEL: define dso_local <vscale x 4 x float> @test_svmlalltb_n_f32_mf8(
254254
// CHECK-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
255255
// CHECK-NEXT: [[ENTRY:.*:]]
256-
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
257256
// CHECK-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
257+
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
258258
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
259259
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
260260
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlalltb.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -263,8 +263,8 @@ svfloat32_t test_svmlalltb_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t
263263
// CHECK-CXX-LABEL: define dso_local <vscale x 4 x float> @_Z24test_svmlalltb_n_f32_mf8u13__SVFloat32_tu13__SVMfloat8_tu6__mfp8m(
264264
// CHECK-CXX-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
265265
// CHECK-CXX-NEXT: [[ENTRY:.*:]]
266-
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
267266
// CHECK-CXX-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
267+
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
268268
// CHECK-CXX-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
269269
// CHECK-CXX-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
270270
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlalltb.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -295,8 +295,8 @@ svfloat32_t test_svmlalltt_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t
295295
// CHECK-LABEL: define dso_local <vscale x 4 x float> @test_svmlalltt_n_f32_mf8(
296296
// CHECK-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
297297
// CHECK-NEXT: [[ENTRY:.*:]]
298-
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
299298
// CHECK-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
299+
// CHECK-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
300300
// CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
301301
// CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
302302
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlalltt.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])
@@ -305,8 +305,8 @@ svfloat32_t test_svmlalltt_f32_mf8(svfloat32_t zda, svmfloat8_t zn, svmfloat8_t
305305
// CHECK-CXX-LABEL: define dso_local <vscale x 4 x float> @_Z24test_svmlalltt_n_f32_mf8u13__SVFloat32_tu13__SVMfloat8_tu6__mfp8m(
306306
// CHECK-CXX-SAME: <vscale x 4 x float> [[ZDA:%.*]], <vscale x 16 x i8> [[ZN:%.*]], <1 x i8> [[ZM:%.*]], i64 noundef [[FPM:%.*]]) #[[ATTR0]] {
307307
// CHECK-CXX-NEXT: [[ENTRY:.*:]]
308-
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
309308
// CHECK-CXX-NEXT: [[TMP0:%.*]] = extractelement <1 x i8> [[ZM]], i64 0
309+
// CHECK-CXX-NEXT: tail call void @llvm.aarch64.set.fpmr(i64 [[FPM]])
310310
// CHECK-CXX-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 16 x i8> poison, i8 [[TMP0]], i64 0
311311
// CHECK-CXX-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 16 x i8> [[DOTSPLATINSERT]], <vscale x 16 x i8> poison, <vscale x 16 x i32> zeroinitializer
312312
// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fp8.fmlalltt.nxv4f32(<vscale x 4 x float> [[ZDA]], <vscale x 16 x i8> [[ZN]], <vscale x 16 x i8> [[DOTSPLAT]])

0 commit comments

Comments
 (0)