[clang] 1df5c94 - [AArch64] Implement FP8 floating-point mode helper intrinsics (#100608)
via cfe-commits
cfe-commits at lists.llvm.org
Mon Oct 28 04:22:41 PDT 2024
Author: Momchil Velikov
Date: 2024-10-28T11:22:38Z
New Revision: 1df5c943439b050dbc34e7a1c88c4d8ec90d26b5
URL: https://github.com/llvm/llvm-project/commit/1df5c943439b050dbc34e7a1c88c4d8ec90d26b5
DIFF: https://github.com/llvm/llvm-project/commit/1df5c943439b050dbc34e7a1c88c4d8ec90d26b5.diff
LOG: [AArch64] Implement FP8 floating-point mode helper intrinsics (#100608)
Implement FP8 mode helper intrinsics (as inline functions) as
specified in ACLE 2024Q3 "14.2 Helper intrinsics"
https://github.com/ARM-software/acle/releases/download/r2024Q3/acle-2024Q3.pdf
Added:
clang/test/CodeGen/aarch64-fpm-helpers.c
Modified:
clang/utils/TableGen/NeonEmitter.cpp
Removed:
################################################################################
diff --git a/clang/test/CodeGen/aarch64-fpm-helpers.c b/clang/test/CodeGen/aarch64-fpm-helpers.c
new file mode 100644
index 00000000000000..4bced01d5c71fa
--- /dev/null
+++ b/clang/test/CodeGen/aarch64-fpm-helpers.c
@@ -0,0 +1,165 @@
+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --version 5
+
+// RUN: %clang_cc1 -O2 -triple aarch64 -emit-llvm -x c -DUSE_NEON_H %s -o - | FileCheck %s
+// RUN: %clang_cc1 -O2 -triple aarch64 -emit-llvm -x c -DUSE_SVE_H %s -o - | FileCheck %s
+// RUN: %clang_cc1 -O2 -triple aarch64 -emit-llvm -x c -DUSE_SME_H %s -o - | FileCheck %s
+// RUN: %clang_cc1 -O2 -triple aarch64 -emit-llvm -x c++ -DUSE_NEON_H %s -o - | FileCheck %s
+// RUN: %clang_cc1 -O2 -triple aarch64 -emit-llvm -x c++ -DUSE_SVE_H %s -o - | FileCheck %s
+// RUN: %clang_cc1 -O2 -triple aarch64 -emit-llvm -x c++ -DUSE_SME_H %s -o - | FileCheck %s
+
+// REQUIRES: aarch64-registered-target
+
+#ifdef USE_NEON_H
+#include "arm_neon.h"
+#endif
+
+#ifdef USE_SVE_H
+#include "arm_sve.h"
+#endif
+
+#ifdef USE_SME_H
+#include "arm_sme.h"
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#define INIT_ZERO 0
+#define INIT_ONES 0xffffffffffffffffU
+
+// CHECK-LABEL: define dso_local noundef i64 @test_init(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0:[0-9]+]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 0
+//
+fpm_t test_init() { return __arm_fpm_init(); }
+
+// CHECK-LABEL: define dso_local noundef i64 @test_src1_1(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 -8
+//
+fpm_t test_src1_1() {
+ return __arm_set_fpm_src1_format(INIT_ONES, __ARM_FPM_E5M2);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_src1_2(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 1
+//
+fpm_t test_src1_2() {
+ return __arm_set_fpm_src1_format(INIT_ZERO, __ARM_FPM_E4M3);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_src2_1(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 -57
+//
+fpm_t test_src2_1() {
+ return __arm_set_fpm_src2_format(INIT_ONES, __ARM_FPM_E5M2);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_src2_2(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 8
+//
+fpm_t test_src2_2() {
+ return __arm_set_fpm_src2_format(INIT_ZERO, __ARM_FPM_E4M3);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_dst1_1(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 -449
+//
+fpm_t test_dst1_1() {
+ return __arm_set_fpm_dst_format(INIT_ONES, __ARM_FPM_E5M2);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_dst2_2(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 64
+//
+fpm_t test_dst2_2() {
+ return __arm_set_fpm_dst_format(INIT_ZERO, __ARM_FPM_E4M3);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_of_mul_1(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 -16385
+//
+fpm_t test_of_mul_1() {
+ return __arm_set_fpm_overflow_mul(INIT_ONES, __ARM_FPM_INFNAN);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_of_mul_2(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 16384
+//
+fpm_t test_of_mul_2() {
+ return __arm_set_fpm_overflow_mul(INIT_ZERO, __ARM_FPM_SATURATE);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_of_cvt_1(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 -32769
+//
+fpm_t test_of_cvt_1() {
+ return __arm_set_fpm_overflow_cvt(INIT_ONES, __ARM_FPM_INFNAN);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_of_cvt_2(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 32768
+//
+fpm_t test_of_cvt_2() {
+ return __arm_set_fpm_overflow_cvt(INIT_ZERO, __ARM_FPM_SATURATE);
+}
+
+// CHECK-LABEL: define dso_local noundef i64 @test_lscale(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 8323072
+//
+fpm_t test_lscale() { return __arm_set_fpm_lscale(INIT_ZERO, 127); }
+
+// CHECK-LABEL: define dso_local noundef i64 @test_lscale2(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 270582939648
+//
+fpm_t test_lscale2() { return __arm_set_fpm_lscale2(INIT_ZERO, 63); }
+
+// CHECK-LABEL: define dso_local noundef range(i64 0, 4294967296) i64 @test_nscale_1(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 2147483648
+//
+fpm_t test_nscale_1() { return __arm_set_fpm_nscale(INIT_ZERO, -128); }
+
+// CHECK-LABEL: define dso_local noundef range(i64 0, 4294967296) i64 @test_nscale_2(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 2130706432
+//
+fpm_t test_nscale_2() { return __arm_set_fpm_nscale(INIT_ZERO, 127); }
+
+// CHECK-LABEL: define dso_local noundef range(i64 0, 4294967296) i64 @test_nscale_3(
+// CHECK-SAME: ) local_unnamed_addr #[[ATTR0]] {
+// CHECK-NEXT: [[ENTRY:.*:]]
+// CHECK-NEXT: ret i64 4278190080
+//
+fpm_t test_nscale_3() { return __arm_set_fpm_nscale(INIT_ZERO, -1); }
+
+#ifdef __cplusplus
+}
+#endif
diff --git a/clang/utils/TableGen/NeonEmitter.cpp b/clang/utils/TableGen/NeonEmitter.cpp
index 8caba774649a2b..c6d82646b40de2 100644
--- a/clang/utils/TableGen/NeonEmitter.cpp
+++ b/clang/utils/TableGen/NeonEmitter.cpp
@@ -2594,6 +2594,60 @@ void NeonEmitter::runVectorTypes(raw_ostream &OS) {
OS << "typedef double float64_t;\n";
OS << "#endif\n\n";
+ OS << R"(
+typedef uint64_t fpm_t;
+
+enum __ARM_FPM_FORMAT { __ARM_FPM_E5M2, __ARM_FPM_E4M3 };
+
+enum __ARM_FPM_OVERFLOW { __ARM_FPM_INFNAN, __ARM_FPM_SATURATE };
+
+static __inline__ fpm_t __attribute__((__always_inline__, __nodebug__))
+__arm_fpm_init(void) {
+ return 0;
+}
+
+static __inline__ fpm_t __attribute__((__always_inline__, __nodebug__))
+__arm_set_fpm_src1_format(fpm_t __fpm, enum __ARM_FPM_FORMAT __format) {
+ return (__fpm & ~7ull) | (fpm_t)__format;
+}
+
+static __inline__ fpm_t __attribute__((__always_inline__, __nodebug__))
+__arm_set_fpm_src2_format(fpm_t __fpm, enum __ARM_FPM_FORMAT __format) {
+ return (__fpm & ~0x38ull) | ((fpm_t)__format << 3u);
+}
+
+static __inline__ fpm_t __attribute__((__always_inline__, __nodebug__))
+__arm_set_fpm_dst_format(fpm_t __fpm, enum __ARM_FPM_FORMAT __format) {
+ return (__fpm & ~0x1c0ull) | ((fpm_t)__format << 6u);
+}
+
+static __inline__ fpm_t __attribute__((__always_inline__, __nodebug__))
+__arm_set_fpm_overflow_mul(fpm_t __fpm, enum __ARM_FPM_OVERFLOW __behaviour) {
+ return (__fpm & ~0x4000ull) | ((fpm_t)__behaviour << 14u);
+}
+
+static __inline__ fpm_t __attribute__((__always_inline__, __nodebug__))
+__arm_set_fpm_overflow_cvt(fpm_t __fpm, enum __ARM_FPM_OVERFLOW __behaviour) {
+ return (__fpm & ~0x8000ull) | ((fpm_t)__behaviour << 15u);
+}
+
+static __inline__ fpm_t __attribute__((__always_inline__, __nodebug__))
+__arm_set_fpm_lscale(fpm_t __fpm, uint64_t __scale) {
+ return (__fpm & ~0x7f0000ull) | (__scale << 16u);
+}
+
+static __inline__ fpm_t __attribute__((__always_inline__, __nodebug__))
+__arm_set_fpm_nscale(fpm_t __fpm, int64_t __scale) {
+ return (__fpm & ~0xff000000ull) | (((fpm_t)__scale & 0xffu) << 24u);
+}
+
+static __inline__ fpm_t __attribute__((__always_inline__, __nodebug__))
+__arm_set_fpm_lscale2(fpm_t __fpm, uint64_t __scale) {
+ return (uint32_t)__fpm | (__scale << 32u);
+}
+
+)";
+
emitNeonTypeDefs("cQcsQsiQilQlUcQUcUsQUsUiQUiUlQUlhQhfQfdQd", OS);
emitNeonTypeDefs("bQb", OS);
More information about the cfe-commits
mailing list