[AArch64] Implement FP8 SVE/SME reinterpret intrinsics

[momchil-velikov]

No description provided.

[llvmbot]

@llvm/pr-subscribers-clang

Author: Momchil Velikov (momchil-velikov)

Changes

---

Patch is 199.05 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/121063.diff

2 Files Affected:

• (added) clang/test/CodeGen/AArch64/fp8-intrinsics/acle_sve2_fp8_reinterpret.c (+2279)
• (modified) clang/utils/TableGen/SveEmitter.cpp (+3-2)

diff --git a/clang/test/CodeGen/AArch64/fp8-intrinsics/acle_sve2_fp8_reinterpret.c b/clang/test/CodeGen/AArch64/fp8-intrinsics/acle_sve2_fp8_reinterpret.c
new file mode 100644
index 00000000000000..f99d77108c6089
--- /dev/null
+++ b/clang/test/CodeGen/AArch64/fp8-intrinsics/acle_sve2_fp8_reinterpret.c
@@ -0,0 +1,2279 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
+// RUN: %clang_cc1        -triple aarch64-none-linux-gnu -target-feature +sve -target-feature +fp8 -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
+// RUN: %clang_cc1 -x c++ -triple aarch64-none-linux-gnu -target-feature +sme -target-feature +fp8 -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CHECK-CXX
+
+// RUN: %clang_cc1        -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -target-feature +fp8 -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
+// RUN: %clang_cc1 -x c++ -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sme -target-feature +fp8 -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CHECK-CXX
+
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve -target-feature +fp8 -S -disable-O0-optnone -Werror -Wall -o /dev/null %s
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sme -target-feature +fp8 -S -disable-O0-optnone -Werror -Wall -o /dev/null %s
+
+// REQUIRES: aarch64-registered-target
+
+#ifdef __ARM_FEATURE_SME
+#include <arm_sme.h>
+#else
+#include <arm_sve.h>
+#endif
+
+#ifdef SVE_OVERLOADED_FORMS
+#define SVE_ACLE_FUNC(A1, A2_UNUSED) A1
+#else
+#define SVE_ACLE_FUNC(A1, A2) A1##A2
+#endif
+
+#ifdef __ARM_FEATURE_SME
+#define STREAMING __arm_streaming
+#else
+#define STREAMING
+#endif
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_s8_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z25test_svreinterpret_s8_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0:[0-9]+]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+svint8_t test_svreinterpret_s8_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_s8, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_u8_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z25test_svreinterpret_u8_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+svuint8_t test_svreinterpret_u8_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_u8, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_s8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z25test_svreinterpret_mf8_s8u10__SVInt8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+svmfloat8_t test_svreinterpret_mf8_s8(svint8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _s8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_u8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z25test_svreinterpret_mf8_u8u11__SVUint8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+svmfloat8_t test_svreinterpret_mf8_u8(svuint8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _u8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[OP]]
+//
+svmfloat8_t test_svreinterpret_mf8_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_s16(
+// CHECK-SAME: <vscale x 8 x i16> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 8 x i16> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_s16u11__SVInt16_t(
+// CHECK-CXX-SAME: <vscale x 8 x i16> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 8 x i16> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_s16(svint16_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _s16)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_u16(
+// CHECK-SAME: <vscale x 8 x i16> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 8 x i16> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_u16u12__SVUint16_t(
+// CHECK-CXX-SAME: <vscale x 8 x i16> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 8 x i16> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_u16(svuint16_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _u16)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_s32(
+// CHECK-SAME: <vscale x 4 x i32> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 4 x i32> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_s32u11__SVInt32_t(
+// CHECK-CXX-SAME: <vscale x 4 x i32> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 4 x i32> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_s32(svint32_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _s32)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_u32(
+// CHECK-SAME: <vscale x 4 x i32> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 4 x i32> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_u32u12__SVUint32_t(
+// CHECK-CXX-SAME: <vscale x 4 x i32> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 4 x i32> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_u32(svuint32_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _u32)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_s64(
+// CHECK-SAME: <vscale x 2 x i64> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 2 x i64> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_s64u11__SVInt64_t(
+// CHECK-CXX-SAME: <vscale x 2 x i64> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 2 x i64> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_s64(svint64_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _s64)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_u64(
+// CHECK-SAME: <vscale x 2 x i64> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 2 x i64> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_u64u12__SVUint64_t(
+// CHECK-CXX-SAME: <vscale x 2 x i64> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 2 x i64> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_u64(svuint64_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _u64)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_f16(
+// CHECK-SAME: <vscale x 8 x half> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 8 x half> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_f16u13__SVFloat16_t(
+// CHECK-CXX-SAME: <vscale x 8 x half> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 8 x half> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_f16(svfloat16_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _f16)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_bf16(
+// CHECK-SAME: <vscale x 8 x bfloat> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 8 x bfloat> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z27test_svreinterpret_mf8_bf16u14__SVBfloat16_t(
+// CHECK-CXX-SAME: <vscale x 8 x bfloat> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 8 x bfloat> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_bf16(svbfloat16_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _bf16)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_f32(
+// CHECK-SAME: <vscale x 4 x float> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 4 x float> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_f32u13__SVFloat32_t(
+// CHECK-CXX-SAME: <vscale x 4 x float> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 4 x float> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_f32(svfloat32_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _f32)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 16 x i8> @test_svreinterpret_mf8_f64(
+// CHECK-SAME: <vscale x 2 x double> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 2 x double> [[OP]] to <vscale x 16 x i8>
+// CHECK-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 16 x i8> @_Z26test_svreinterpret_mf8_f64u13__SVFloat64_t(
+// CHECK-CXX-SAME: <vscale x 2 x double> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 2 x double> [[OP]] to <vscale x 16 x i8>
+// CHECK-CXX-NEXT:    ret <vscale x 16 x i8> [[TMP0]]
+//
+svmfloat8_t test_svreinterpret_mf8_f64(svfloat64_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_mf8, _f64)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 8 x i16> @test_svreinterpret_s16_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 8 x i16>
+// CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 8 x i16> @_Z26test_svreinterpret_s16_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 8 x i16>
+// CHECK-CXX-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
+//
+svint16_t test_svreinterpret_s16_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_s16, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 8 x i16> @test_svreinterpret_u16_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 8 x i16>
+// CHECK-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 8 x i16> @_Z26test_svreinterpret_u16_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 8 x i16>
+// CHECK-CXX-NEXT:    ret <vscale x 8 x i16> [[TMP0]]
+//
+svuint16_t test_svreinterpret_u16_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_u16, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 4 x i32> @test_svreinterpret_s32_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 4 x i32>
+// CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 4 x i32> @_Z26test_svreinterpret_s32_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 4 x i32>
+// CHECK-CXX-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+svint32_t test_svreinterpret_s32_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_s32, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 4 x i32> @test_svreinterpret_u32_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 4 x i32>
+// CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 4 x i32> @_Z26test_svreinterpret_u32_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 4 x i32>
+// CHECK-CXX-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+svuint32_t test_svreinterpret_u32_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_u32, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 2 x i64> @test_svreinterpret_s64_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 2 x i64>
+// CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 2 x i64> @_Z26test_svreinterpret_s64_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 2 x i64>
+// CHECK-CXX-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
+//
+svint64_t test_svreinterpret_s64_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_s64, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 2 x i64> @test_svreinterpret_u64_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 2 x i64>
+// CHECK-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 2 x i64> @_Z26test_svreinterpret_u64_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 2 x i64>
+// CHECK-CXX-NEXT:    ret <vscale x 2 x i64> [[TMP0]]
+//
+svuint64_t test_svreinterpret_u64_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_u64, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 8 x half> @test_svreinterpret_f16_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 8 x half>
+// CHECK-NEXT:    ret <vscale x 8 x half> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 8 x half> @_Z26test_svreinterpret_f16_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 8 x half>
+// CHECK-CXX-NEXT:    ret <vscale x 8 x half> [[TMP0]]
+//
+svfloat16_t test_svreinterpret_f16_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_f16, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 8 x bfloat> @test_svreinterpret_bf16_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 8 x bfloat>
+// CHECK-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 8 x bfloat> @_Z27test_svreinterpret_bf16_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 8 x bfloat>
+// CHECK-CXX-NEXT:    ret <vscale x 8 x bfloat> [[TMP0]]
+//
+svbfloat16_t test_svreinterpret_bf16_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_bf16, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 4 x float> @test_svreinterpret_f32_mf8(
+// CHECK-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-NEXT:  [[ENTRY:.*:]]
+// CHECK-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 4 x float>
+// CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
+//
+// CHECK-CXX-LABEL: define dso_local <vscale x 4 x float> @_Z26test_svreinterpret_f32_mf8u13__SVMfloat8_t(
+// CHECK-CXX-SAME: <vscale x 16 x i8> [[OP:%.*]]) #[[ATTR0]] {
+// CHECK-CXX-NEXT:  [[ENTRY:.*:]]
+// CHECK-CXX-NEXT:    [[TMP0:%.*]] = bitcast <vscale x 16 x i8> [[OP]] to <vscale x 4 x float>
+// CHECK-CXX-NEXT:    ret <vscale x 4 x float> [[TMP0]]
+//
+svfloat32_t test_svreinterpret_f32_mf8(svmfloat8_t op) STREAMING {
+  return SVE_ACLE_FUNC(svreinterpret_f32, _mf8)(op);
+}
+
+// CHECK-LABEL: define dso_local <vscale x 2 x double> @test_svreinterpret_f64_mf8(
+// CH...
[truncated]

[jthackray]

LGTM