[clang] [llvm] [AArch64] Add intrinsics for 16-bit non-widening FMLA/FMLS (PR #88553)
via llvm-commits
llvm-commits at lists.llvm.org
Fri Apr 12 11:41:08 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-aarch64
Author: Momchil Velikov (momchil-velikov)
<details>
<summary>Changes</summary>
---
Patch is 108.24 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/88553.diff
6 Files Affected:
- (modified) clang/include/clang/Basic/arm_sme.td (+34)
- (added) clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_fmlas16.c (+592)
- (added) clang/test/Sema/aarch64-sme2-intrinsics/acle_sme2_fmlas16.c (+90)
- (modified) llvm/lib/Target/AArch64/AArch64SMEInstrInfo.td (+28-29)
- (modified) llvm/lib/Target/AArch64/SMEInstrFormats.td (+48-6)
- (added) llvm/test/CodeGen/AArch64/sme2-intrinsics-fmlas16.ll (+462)
``````````diff
diff --git a/clang/include/clang/Basic/arm_sme.td b/clang/include/clang/Basic/arm_sme.td
index 1ac6d5170ea283..29bc627e7481ab 100644
--- a/clang/include/clang/Basic/arm_sme.td
+++ b/clang/include/clang/Basic/arm_sme.td
@@ -458,6 +458,40 @@ let TargetGuard = "sme2,sme-f64f64" in {
def SVMLS_LANE_VG1x4_F64 : Inst<"svmls_lane_za64[_{d}]_vg1x4", "vm4di", "d", MergeNone, "aarch64_sme_fmls_lane_vg1x4", [IsStreaming, IsInOutZA], [ImmCheck<3, ImmCheck0_1>]>;
}
+let TargetGuard = "sme2p1,sme-f16f16" in {
+ def SVMLA_MULTI_VG1x2_F16 : Inst<"svmla_za16[_f16]_vg1x2", "vm22", "h", MergeNone, "aarch64_sme_fmla_vg1x2", [IsStreaming, IsInOutZA], []>;
+ def SVMLA_MULTI_VG1x4_F16 : Inst<"svmla_za16[_f16]_vg1x4", "vm44", "h", MergeNone, "aarch64_sme_fmla_vg1x4", [IsStreaming, IsInOutZA], []>;
+ def SVMLS_MULTI_VG1x2_F16 : Inst<"svmls_za16[_f16]_vg1x2", "vm22", "h", MergeNone, "aarch64_sme_fmls_vg1x2", [IsStreaming, IsInOutZA], []>;
+ def SVMLS_MULTI_VG1x4_F16 : Inst<"svmls_za16[_f16]_vg1x4", "vm44", "h", MergeNone, "aarch64_sme_fmls_vg1x4", [IsStreaming, IsInOutZA], []>;
+
+ def SVMLA_SINGLE_VG1x2_F16 : Inst<"svmla[_single]_za16[_f16]_vg1x2", "vm2d", "h", MergeNone, "aarch64_sme_fmla_single_vg1x2", [IsStreaming, IsInOutZA], []>;
+ def SVMLA_SINGLE_VG1x4_F16 : Inst<"svmla[_single]_za16[_f16]_vg1x4", "vm4d", "h", MergeNone, "aarch64_sme_fmla_single_vg1x4", [IsStreaming, IsInOutZA], []>;
+ def SVMLS_SINGLE_VG1x2_F16 : Inst<"svmls[_single]_za16[_f16]_vg1x2", "vm2d", "h", MergeNone, "aarch64_sme_fmls_single_vg1x2", [IsStreaming, IsInOutZA], []>;
+ def SVMLS_SINGLE_VG1x4_F16 : Inst<"svmls[_single]_za16[_f16]_vg1x4", "vm4d", "h", MergeNone, "aarch64_sme_fmls_single_vg1x4", [IsStreaming, IsInOutZA], []>;
+
+ def SVMLA_LANE_VG1x2_F16 : Inst<"svmla_lane_za16[_f16]_vg1x2", "vm2di", "h", MergeNone, "aarch64_sme_fmla_lane_vg1x2", [IsStreaming, IsInOutZA], [ImmCheck<3, ImmCheck0_7>]>;
+ def SVMLA_LANE_VG1x4_F16 : Inst<"svmla_lane_za16[_f16]_vg1x4", "vm4di", "h", MergeNone, "aarch64_sme_fmla_lane_vg1x4", [IsStreaming, IsInOutZA], [ImmCheck<3, ImmCheck0_7>]>;
+ def SVMLS_LANE_VG1x2_F16 : Inst<"svmls_lane_za16[_f16]_vg1x2", "vm2di", "h", MergeNone, "aarch64_sme_fmls_lane_vg1x2", [IsStreaming, IsInOutZA], [ImmCheck<3, ImmCheck0_7>]>;
+ def SVMLS_LANE_VG1x4_F16 : Inst<"svmls_lane_za16[_f16]_vg1x4", "vm4di", "h", MergeNone, "aarch64_sme_fmls_lane_vg1x4", [IsStreaming, IsInOutZA], [ImmCheck<3, ImmCheck0_7>]>;
+}
+
+let TargetGuard = "sme2,b16b16" in {
+ def SVMLA_MULTI_VG1x2_BF16 : Inst<"svmla_za16[_bf16]_vg1x2", "vm22", "b", MergeNone, "aarch64_sme_fmla_vg1x2", [IsStreaming, IsInOutZA], []>;
+ def SVMLA_MULTI_VG1x4_BF16 : Inst<"svmla_za16[_bf16]_vg1x4", "vm44", "b", MergeNone, "aarch64_sme_fmla_vg1x4", [IsStreaming, IsInOutZA], []>;
+ def SVMLS_MULTI_VG1x2_BF16 : Inst<"svmls_za16[_bf16]_vg1x2", "vm22", "b", MergeNone, "aarch64_sme_fmls_vg1x2", [IsStreaming, IsInOutZA], []>;
+ def SVMLS_MULTI_VG1x4_BF16 : Inst<"svmls_za16[_bf16]_vg1x4", "vm44", "b", MergeNone, "aarch64_sme_fmls_vg1x4", [IsStreaming, IsInOutZA], []>;
+
+ def SVMLA_SINGLE_VG1x2_BF16 : Inst<"svmla[_single]_za16[_bf16]_vg1x2", "vm2d", "b", MergeNone, "aarch64_sme_fmla_single_vg1x2", [IsStreaming, IsInOutZA], []>;
+ def SVMLA_SINGLE_VG1x4_BF16 : Inst<"svmla[_single]_za16[_bf16]_vg1x4", "vm4d", "b", MergeNone, "aarch64_sme_fmla_single_vg1x4", [IsStreaming, IsInOutZA], []>;
+ def SVMLS_SINGLE_VG1x2_BF16 : Inst<"svmls[_single]_za16[_bf16]_vg1x2", "vm2d", "b", MergeNone, "aarch64_sme_fmls_single_vg1x2", [IsStreaming, IsInOutZA], []>;
+ def SVMLS_SINGLE_VG1x4_BF16 : Inst<"svmls[_single]_za16[_bf16]_vg1x4", "vm4d", "b", MergeNone, "aarch64_sme_fmls_single_vg1x4", [IsStreaming, IsInOutZA], []>;
+
+ def SVMLA_LANE_VG1x2_BF16 : Inst<"svmla_lane_za16[_bf16]_vg1x2", "vm2di", "b", MergeNone, "aarch64_sme_fmla_lane_vg1x2", [IsStreaming, IsInOutZA], [ImmCheck<3, ImmCheck0_7>]>;
+ def SVMLA_LANE_VG1x4_BF16 : Inst<"svmla_lane_za16[_bf16]_vg1x4", "vm4di", "b", MergeNone, "aarch64_sme_fmla_lane_vg1x4", [IsStreaming, IsInOutZA], [ImmCheck<3, ImmCheck0_7>]>;
+ def SVMLS_LANE_VG1x2_BF16 : Inst<"svmls_lane_za16[_bf16]_vg1x2", "vm2di", "b", MergeNone, "aarch64_sme_fmls_lane_vg1x2", [IsStreaming, IsInOutZA], [ImmCheck<3, ImmCheck0_7>]>;
+ def SVMLS_LANE_VG1x4_BF16 : Inst<"svmls_lane_za16[_bf16]_vg1x4", "vm4di", "b", MergeNone, "aarch64_sme_fmls_lane_vg1x4", [IsStreaming, IsInOutZA], [ImmCheck<3, ImmCheck0_7>]>;
+}
+
// FMLAL/FMLSL/UMLAL/SMLAL
// SMLALL/UMLALL/USMLALL/SUMLALL
let TargetGuard = "sme2" in {
diff --git a/clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_fmlas16.c b/clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_fmlas16.c
new file mode 100644
index 00000000000000..81ab3e62276b2b
--- /dev/null
+++ b/clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_fmlas16.c
@@ -0,0 +1,592 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 4
+// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64-none-linux-gnu -target-feature +sme2p1 -target-feature +sme-f16f16 -target-feature +b16b16 -O2 -S -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -x c++ -fclang-abi-compat=latest -triple aarch64-none-linux-gnu -target-feature +sme2p1 -target-feature +sme-f16f16 -target-feature +b16b16 -O2 -S -Werror -Wall -emit-llvm -o - %s | FileCheck %s --check-prefix CHECK-CXX
+// RUN: %clang_cc1 -DSME_OVERLOADED_FORMS -fclang-abi-compat=latest -triple aarch64-none-linux-gnu -target-feature +sme2p1 -target-feature +sme-f16f16 -target-feature +b16b16 -O2 -S -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -DSME_OVERLOADED_FORMS -x c++ -fclang-abi-compat=latest -triple aarch64-none-linux-gnu -target-feature +sme2p1 -target-feature +sme-f16f16 -target-feature +b16b16 -O2 -S -Werror -Wall -emit-llvm -o - %s | FileCheck %s --check-prefix CHECK-CXX
+
+// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64-none-linux-gnu -target-feature +sme2p1 -target-feature +sme-f16f16 -target-feature +b16b16 -O2 -S -Werror -Wall %s -o /dev/null
+
+// REQUIRES: aarch64-registered-target
+#include <arm_sme.h>
+
+#ifdef SME_OVERLOADED_FORMS
+#define SME_ACLE_FUNC(A1, A2_UNUSED, A3, A4_UNUSED, A5) A1##A3##A5
+#else
+#define SME_ACLE_FUNC(A1, A2, A3, A4, A5) A1##A2##A3##A4##A5
+#endif
+
+// CHECK-LABEL: define dso_local void @test_svmla_single_za16_f16_vg1x2(
+// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
+// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
+// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]])
+// CHECK-NEXT: ret void
+//
+// CHECK-CXX-LABEL: define dso_local void @_Z32test_svmla_single_za16_f16_vg1x2j13svfloat16x2_tu13__SVFloat16_t(
+// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// CHECK-CXX-NEXT: entry:
+// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
+// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
+// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]])
+// CHECK-CXX-NEXT: ret void
+//
+void test_svmla_single_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
+ SME_ACLE_FUNC(svmla,_single,_za16,_f16,_vg1x2)(slice, zn, zm);
+}
+
+// CHECK-LABEL: define dso_local void @test_svmla_single_za16_f16_vg1x4(
+// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
+// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
+// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
+// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
+// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]])
+// CHECK-NEXT: ret void
+//
+// CHECK-CXX-LABEL: define dso_local void @_Z32test_svmla_single_za16_f16_vg1x4j13svfloat16x4_tu13__SVFloat16_t(
+// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-CXX-NEXT: entry:
+// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
+// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
+// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
+// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
+// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]])
+// CHECK-CXX-NEXT: ret void
+//
+void test_svmla_single_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
+ SME_ACLE_FUNC(svmla,_single,_za16,_f16,_vg1x4)(slice, zn, zm);
+}
+
+// CHECK-LABEL: define dso_local void @test_svmls_single_za16_f16_vg1x2(
+// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
+// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
+// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]])
+// CHECK-NEXT: ret void
+//
+// CHECK-CXX-LABEL: define dso_local void @_Z32test_svmls_single_za16_f16_vg1x2j13svfloat16x2_tu13__SVFloat16_t(
+// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-CXX-NEXT: entry:
+// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
+// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
+// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[ZM]])
+// CHECK-CXX-NEXT: ret void
+//
+void test_svmls_single_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
+ SME_ACLE_FUNC(svmls,_single,_za16,_f16,_vg1x2)(slice, zn, zm);
+}
+
+// CHECK-LABEL: define dso_local void @test_svmls_single_za16_f16_vg1x4(
+// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
+// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
+// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
+// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
+// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]])
+// CHECK-NEXT: ret void
+//
+// CHECK-CXX-LABEL: define dso_local void @_Z32test_svmls_single_za16_f16_vg1x4j13svfloat16x4_tu13__SVFloat16_t(
+// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 8 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-CXX-NEXT: entry:
+// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
+// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
+// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
+// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
+// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmls.single.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[ZM]])
+// CHECK-CXX-NEXT: ret void
+//
+void test_svmls_single_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16_t zm) __arm_streaming __arm_inout("za") {
+ SME_ACLE_FUNC(svmls,_single,_za16,_f16,_vg1x4)(slice, zn, zm);
+}
+
+// CHECK-LABEL: define dso_local void @test_svmla_za16_f16_vg1x2(
+// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 16 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
+// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
+// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 0)
+// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
+// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
+// CHECK-NEXT: ret void
+//
+// CHECK-CXX-LABEL: define dso_local void @_Z25test_svmla_za16_f16_vg1x2j13svfloat16x2_tS_(
+// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 16 x half> [[ZN:%.*]], <vscale x 16 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-CXX-NEXT: entry:
+// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 0)
+// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZN]], i64 8)
+// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 0)
+// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv16f16(<vscale x 16 x half> [[ZM]], i64 8)
+// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x2.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]])
+// CHECK-CXX-NEXT: ret void
+//
+void test_svmla_za16_f16_vg1x2(uint32_t slice, svfloat16x2_t zn, svfloat16x2_t zm) __arm_streaming __arm_inout("za") {
+ SME_ACLE_FUNC(svmla,,_za16,_f16,_vg1x2)(slice, zn, zm);
+}
+
+// CHECK-LABEL: define dso_local void @test_svmla_za16_f16_vg1x4(
+// CHECK-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 32 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
+// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
+// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
+// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
+// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 0)
+// CHECK-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
+// CHECK-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
+// CHECK-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
+// CHECK-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
+// CHECK-NEXT: ret void
+//
+// CHECK-CXX-LABEL: define dso_local void @_Z25test_svmla_za16_f16_vg1x4j13svfloat16x4_tS_(
+// CHECK-CXX-SAME: i32 noundef [[SLICE:%.*]], <vscale x 32 x half> [[ZN:%.*]], <vscale x 32 x half> [[ZM:%.*]]) local_unnamed_addr #[[ATTR0]] {
+// CHECK-CXX-NEXT: entry:
+// CHECK-CXX-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 0)
+// CHECK-CXX-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 8)
+// CHECK-CXX-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 16)
+// CHECK-CXX-NEXT: [[TMP3:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZN]], i64 24)
+// CHECK-CXX-NEXT: [[TMP4:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 0)
+// CHECK-CXX-NEXT: [[TMP5:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 8)
+// CHECK-CXX-NEXT: [[TMP6:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 16)
+// CHECK-CXX-NEXT: [[TMP7:%.*]] = tail call <vscale x 8 x half> @llvm.vector.extract.nxv8f16.nxv32f16(<vscale x 32 x half> [[ZM]], i64 24)
+// CHECK-CXX-NEXT: tail call void @llvm.aarch64.sme.fmla.vg1x4.nxv8f16(i32 [[SLICE]], <vscale x 8 x half> [[TMP0]], <vscale x 8 x half> [[TMP1]], <vscale x 8 x half> [[TMP2]], <vscale x 8 x half> [[TMP3]], <vscale x 8 x half> [[TMP4]], <vscale x 8 x half> [[TMP5]], <vscale x 8 x half> [[TMP6]], <vscale x 8 x half> [[TMP7]])
+// CHECK-CXX-NEXT: ret void
+//
+void test_svmla_za16_f16_vg1x4(uint32_t slice, svfloat16x4_t zn, svfloat16x4_t zm) __arm_streaming __arm_inout("za") {
+ SME_AC...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/88553
More information about the llvm-commits
mailing list