[libc] [llvm] [libc][math][c23] Add fmaf16 C23 math function. (PR #130757)
Harrison Hao via llvm-commits
llvm-commits at lists.llvm.org
Wed Mar 19 07:17:10 PDT 2025
https://github.com/harrisonGPU updated https://github.com/llvm/llvm-project/pull/130757
>From 1c05bf007bbe436fef8171ed9ccd520c03b0ac0f Mon Sep 17 00:00:00 2001
From: Harrison Hao <tsworld1314 at gmail.com>
Date: Tue, 11 Mar 2025 19:47:07 +0800
Subject: [PATCH 1/5] [libc][math][c23] Add fmaf16 C23 math function.
---
libc/config/gpu/amdgpu/entrypoints.txt | 1 +
libc/config/gpu/nvptx/entrypoints.txt | 1 +
libc/config/linux/aarch64/entrypoints.txt | 1 +
libc/config/linux/x86_64/entrypoints.txt | 1 +
libc/docs/headers/math/index.rst | 2 +-
libc/include/math.yaml | 9 ++++++++
libc/src/math/CMakeLists.txt | 1 +
libc/src/math/fmaf16.h | 21 +++++++++++++++++++
libc/src/math/generic/CMakeLists.txt | 10 +++++++++
libc/src/math/generic/fmaf16.cpp | 20 ++++++++++++++++++
libc/test/src/math/CMakeLists.txt | 17 +++++++++++++++
libc/test/src/math/fmaf16_test.cpp | 13 ++++++++++++
libc/test/src/math/smoke/CMakeLists.txt | 12 +++++++++++
libc/test/src/math/smoke/fmaf16_test.cpp | 13 ++++++++++++
libc/utils/MPFRWrapper/MPFRUtils.cpp | 5 +++++
.../llvm-project-overlay/libc/BUILD.bazel | 2 ++
.../libc/test/src/math/smoke/BUILD.bazel | 5 +++++
17 files changed, 133 insertions(+), 1 deletion(-)
create mode 100644 libc/src/math/fmaf16.h
create mode 100644 libc/src/math/generic/fmaf16.cpp
create mode 100644 libc/test/src/math/fmaf16_test.cpp
create mode 100644 libc/test/src/math/smoke/fmaf16_test.cpp
diff --git a/libc/config/gpu/amdgpu/entrypoints.txt b/libc/config/gpu/amdgpu/entrypoints.txt
index 291d86b4dd587..a5d32c8eda39f 100644
--- a/libc/config/gpu/amdgpu/entrypoints.txt
+++ b/libc/config/gpu/amdgpu/entrypoints.txt
@@ -548,6 +548,7 @@ if(LIBC_TYPES_HAS_FLOAT16)
libc.src.math.fabsf16
libc.src.math.fdimf16
libc.src.math.floorf16
+ libc.src.math.fmaf16
libc.src.math.fmaxf16
libc.src.math.fmaximum_mag_numf16
libc.src.math.fmaximum_magf16
diff --git a/libc/config/gpu/nvptx/entrypoints.txt b/libc/config/gpu/nvptx/entrypoints.txt
index 1ea0d9b03b37e..f553f7cc2b210 100644
--- a/libc/config/gpu/nvptx/entrypoints.txt
+++ b/libc/config/gpu/nvptx/entrypoints.txt
@@ -550,6 +550,7 @@ if(LIBC_TYPES_HAS_FLOAT16)
libc.src.math.fabsf16
libc.src.math.fdimf16
libc.src.math.floorf16
+ libc.src.math.fmaf16
libc.src.math.fmaxf16
libc.src.math.fmaximum_mag_numf16
libc.src.math.fmaximum_magf16
diff --git a/libc/config/linux/aarch64/entrypoints.txt b/libc/config/linux/aarch64/entrypoints.txt
index c1e688ea7e86b..f359304c018d1 100644
--- a/libc/config/linux/aarch64/entrypoints.txt
+++ b/libc/config/linux/aarch64/entrypoints.txt
@@ -676,6 +676,7 @@ if(LIBC_TYPES_HAS_FLOAT16)
libc.src.math.ffma
libc.src.math.ffmal
libc.src.math.floorf16
+ libc.src.math.fmaf16
libc.src.math.fmaxf16
libc.src.math.fmaximum_mag_numf16
libc.src.math.fmaximum_magf16
diff --git a/libc/config/linux/x86_64/entrypoints.txt b/libc/config/linux/x86_64/entrypoints.txt
index 3cb9ee82752b3..5ff4a559cb2d4 100644
--- a/libc/config/linux/x86_64/entrypoints.txt
+++ b/libc/config/linux/x86_64/entrypoints.txt
@@ -686,6 +686,7 @@ if(LIBC_TYPES_HAS_FLOAT16)
libc.src.math.fabsf16
libc.src.math.fdimf16
libc.src.math.floorf16
+ libc.src.math.fmaf16
libc.src.math.fmaxf16
libc.src.math.fmaximum_mag_numf16
libc.src.math.fmaximum_magf16
diff --git a/libc/docs/headers/math/index.rst b/libc/docs/headers/math/index.rst
index 5b855ce4881c3..dda8855df1a5a 100644
--- a/libc/docs/headers/math/index.rst
+++ b/libc/docs/headers/math/index.rst
@@ -299,7 +299,7 @@ Higher Math Functions
+-----------+------------------+-----------------+------------------------+----------------------+------------------------+------------------------+----------------------------+
| expm1 | |check| | |check| | | |check| | | 7.12.6.6 | F.10.3.6 |
+-----------+------------------+-----------------+------------------------+----------------------+------------------------+------------------------+----------------------------+
-| fma | |check| | |check| | | | | 7.12.13.1 | F.10.10.1 |
+| fma | |check| | |check| | | |check| | | 7.12.13.1 | F.10.10.1 |
+-----------+------------------+-----------------+------------------------+----------------------+------------------------+------------------------+----------------------------+
| f16sqrt | |check|\* | |check|\* | |check|\* | N/A | |check| | 7.12.14.6 | F.10.11 |
+-----------+------------------+-----------------+------------------------+----------------------+------------------------+------------------------+----------------------------+
diff --git a/libc/include/math.yaml b/libc/include/math.yaml
index a66f981030864..08a28263a186d 100644
--- a/libc/include/math.yaml
+++ b/libc/include/math.yaml
@@ -736,6 +736,15 @@ functions:
- type: float
- type: float
- type: float
+ - name: fmaf16
+ standards:
+ - stdc
+ return_type: _Float16
+ arguments:
+ - type: _Float16
+ - type: _Float16
+ - type: _Float16
+ guard: LIBC_TYPES_HAS_FLOAT16
- name: fmax
standards:
- stdc
diff --git a/libc/src/math/CMakeLists.txt b/libc/src/math/CMakeLists.txt
index f18a73d46f9aa..a42421614661a 100644
--- a/libc/src/math/CMakeLists.txt
+++ b/libc/src/math/CMakeLists.txt
@@ -209,6 +209,7 @@ add_math_entrypoint_object(floorf128)
add_math_entrypoint_object(fma)
add_math_entrypoint_object(fmaf)
+add_math_entrypoint_object(fmaf16)
add_math_entrypoint_object(fmax)
add_math_entrypoint_object(fmaxf)
diff --git a/libc/src/math/fmaf16.h b/libc/src/math/fmaf16.h
new file mode 100644
index 0000000000000..1c4d468e67a7a
--- /dev/null
+++ b/libc/src/math/fmaf16.h
@@ -0,0 +1,21 @@
+//===-- Implementation header for fmaf16 ------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_MATH_FMAF16_H
+#define LLVM_LIBC_SRC_MATH_FMAF16_H
+
+#include "src/__support/macros/config.h"
+#include "src/__support/macros/properties/types.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+float16 fmaf16(float16 x, float16 y, float16 z);
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SRC_MATH_FMAF16_H
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index 3114289bad486..781e6e7e04906 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -4241,6 +4241,16 @@ add_entrypoint_object(
libc.src.__support.FPUtil.fma
)
+add_entrypoint_object(
+ fmaf16
+ SRCS
+ fmaf16.cpp
+ HDRS
+ ../fmaf16.h
+ DEPENDS
+ libc.src.__support.FPUtil.fma
+)
+
add_entrypoint_object(
fma
SRCS
diff --git a/libc/src/math/generic/fmaf16.cpp b/libc/src/math/generic/fmaf16.cpp
new file mode 100644
index 0000000000000..4f712f5de764f
--- /dev/null
+++ b/libc/src/math/generic/fmaf16.cpp
@@ -0,0 +1,20 @@
+//===-- Implementation of fmaf16 function ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/math/fmaf16.h"
+#include "src/__support/FPUtil/FMA.h"
+#include "src/__support/common.h"
+#include "src/__support/macros/config.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+LLVM_LIBC_FUNCTION(float16, fmaf16, (float16 x, float16 y, float16 z)) {
+ return fputil::fma<float16>(x, y, z);
+}
+
+} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/test/src/math/CMakeLists.txt b/libc/test/src/math/CMakeLists.txt
index 032050bb06ec3..79f45d286f95e 100644
--- a/libc/test/src/math/CMakeLists.txt
+++ b/libc/test/src/math/CMakeLists.txt
@@ -1776,6 +1776,23 @@ add_fp_unittest(
FMA_OPT__ONLY
)
+add_fp_unittest(
+ fmaf16_test
+ NEED_MPFR
+ SUITE
+ libc-math-unittests
+ SRCS
+ fmaf16_test.cpp
+ HDRS
+ FmaTest.h
+ DEPENDS
+ libc.src.math.fmaf16
+ libc.src.stdlib.rand
+ libc.src.stdlib.srand
+ FLAGS
+ FMA_OPT__ONLY
+)
+
add_fp_unittest(
fma_test
NEED_MPFR
diff --git a/libc/test/src/math/fmaf16_test.cpp b/libc/test/src/math/fmaf16_test.cpp
new file mode 100644
index 0000000000000..233d3a7c54cf4
--- /dev/null
+++ b/libc/test/src/math/fmaf16_test.cpp
@@ -0,0 +1,13 @@
+//===-- Unittests for fmaf16 ----------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "FmaTest.h"
+
+#include "src/math/fmaf16.h"
+
+LIST_FMA_TESTS(float16, LIBC_NAMESPACE::fmaf16)
diff --git a/libc/test/src/math/smoke/CMakeLists.txt b/libc/test/src/math/smoke/CMakeLists.txt
index aef7c83ba0215..57d72b46f8b93 100644
--- a/libc/test/src/math/smoke/CMakeLists.txt
+++ b/libc/test/src/math/smoke/CMakeLists.txt
@@ -3562,6 +3562,18 @@ add_fp_unittest(
libc.src.__support.macros.properties.types
)
+add_fp_unittest(
+ fmaf16_test
+ SUITE
+ libc-math-smoke-tests
+ SRCS
+ fmaf16_test.cpp
+ HDRS
+ FmaTest.h
+ DEPENDS
+ libc.src.math.fmaf16
+)
+
add_fp_unittest(
fma_test
SUITE
diff --git a/libc/test/src/math/smoke/fmaf16_test.cpp b/libc/test/src/math/smoke/fmaf16_test.cpp
new file mode 100644
index 0000000000000..cc1f01e2f9541
--- /dev/null
+++ b/libc/test/src/math/smoke/fmaf16_test.cpp
@@ -0,0 +1,13 @@
+//===-- Unittests for fmaxf16 --------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "FmaTest.h"
+
+#include "src/math/fmaf16.h"
+
+LIST_FMA_TESTS(float16, LIBC_NAMESPACE::fmaf16)
diff --git a/libc/utils/MPFRWrapper/MPFRUtils.cpp b/libc/utils/MPFRWrapper/MPFRUtils.cpp
index fc260f4abed49..4a033dbc2049b 100644
--- a/libc/utils/MPFRWrapper/MPFRUtils.cpp
+++ b/libc/utils/MPFRWrapper/MPFRUtils.cpp
@@ -454,6 +454,8 @@ explain_ternary_operation_one_output_error(Operation,
long double, double, RoundingMode);
#ifdef LIBC_TYPES_HAS_FLOAT16
+template void explain_ternary_operation_one_output_error(
+ Operation, const TernaryInput<float16> &, float16, double, RoundingMode);
template void explain_ternary_operation_one_output_error(
Operation, const TernaryInput<float> &, float16, double, RoundingMode);
template void explain_ternary_operation_one_output_error(
@@ -672,6 +674,9 @@ compare_ternary_operation_one_output(Operation,
long double, double, RoundingMode);
#ifdef LIBC_TYPES_HAS_FLOAT16
+template bool
+compare_ternary_operation_one_output(Operation, const TernaryInput<float16> &,
+ float16, double, RoundingMode);
template bool compare_ternary_operation_one_output(Operation,
const TernaryInput<float> &,
float16, double,
diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index f3620186d776f..a747da3492552 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -2828,6 +2828,8 @@ libc_math_function(name = "floorf16")
# TODO: Add fma, fmaf, fmal, fmaf128 functions.
+libc_math_function(name = "fmaf16")
+
libc_math_function(name = "fmax")
libc_math_function(name = "fmaxf")
diff --git a/utils/bazel/llvm-project-overlay/libc/test/src/math/smoke/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/test/src/math/smoke/BUILD.bazel
index 803548178b0e3..6b959fe559e45 100644
--- a/utils/bazel/llvm-project-overlay/libc/test/src/math/smoke/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/test/src/math/smoke/BUILD.bazel
@@ -474,6 +474,11 @@ math_test(
# TODO: Add fma, fmaf, fmal, fmaf128 tests.
+math_test(
+ name = "fmaf16",
+ hdrs = ["FMaxTest.h"],
+)
+
math_test(
name = "fmax",
hdrs = ["FMaxTest.h"],
>From e45c9d9eddba66181a8eee9f7b433966883127b7 Mon Sep 17 00:00:00 2001
From: Harrison Hao <tsworld1314 at gmail.com>
Date: Wed, 12 Mar 2025 10:03:43 +0800
Subject: [PATCH 2/5] [libc][math][c23] Remove FMA_OPT flag.
---
libc/test/src/math/CMakeLists.txt | 2 --
1 file changed, 2 deletions(-)
diff --git a/libc/test/src/math/CMakeLists.txt b/libc/test/src/math/CMakeLists.txt
index 79f45d286f95e..2c4d35c0c3c96 100644
--- a/libc/test/src/math/CMakeLists.txt
+++ b/libc/test/src/math/CMakeLists.txt
@@ -1789,8 +1789,6 @@ add_fp_unittest(
libc.src.math.fmaf16
libc.src.stdlib.rand
libc.src.stdlib.srand
- FLAGS
- FMA_OPT__ONLY
)
add_fp_unittest(
>From 65c399b76b56757878639ca62a7e5b65bf5632ce Mon Sep 17 00:00:00 2001
From: Harrison Hao <tsworld1314 at gmail.com>
Date: Thu, 13 Mar 2025 11:12:25 +0800
Subject: [PATCH 3/5] [libc][math][c23] Do some change for comments.
---
libc/src/math/generic/CMakeLists.txt | 1 +
libc/test/src/math/smoke/CMakeLists.txt | 1 +
libc/test/src/math/smoke/fmaf16_test.cpp | 2 +-
.../llvm-project-overlay/libc/test/src/math/smoke/BUILD.bazel | 2 +-
4 files changed, 4 insertions(+), 2 deletions(-)
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index 781e6e7e04906..a929a8d9d06cd 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -4249,6 +4249,7 @@ add_entrypoint_object(
../fmaf16.h
DEPENDS
libc.src.__support.FPUtil.fma
+ libc.src.__support.macros.properties.types
)
add_entrypoint_object(
diff --git a/libc/test/src/math/smoke/CMakeLists.txt b/libc/test/src/math/smoke/CMakeLists.txt
index 57d72b46f8b93..33ce74d44a3a1 100644
--- a/libc/test/src/math/smoke/CMakeLists.txt
+++ b/libc/test/src/math/smoke/CMakeLists.txt
@@ -3572,6 +3572,7 @@ add_fp_unittest(
FmaTest.h
DEPENDS
libc.src.math.fmaf16
+ libc.src.__support.macros.properties.types
)
add_fp_unittest(
diff --git a/libc/test/src/math/smoke/fmaf16_test.cpp b/libc/test/src/math/smoke/fmaf16_test.cpp
index cc1f01e2f9541..233d3a7c54cf4 100644
--- a/libc/test/src/math/smoke/fmaf16_test.cpp
+++ b/libc/test/src/math/smoke/fmaf16_test.cpp
@@ -1,4 +1,4 @@
-//===-- Unittests for fmaxf16 --------------------------------------------===//
+//===-- Unittests for fmaf16 ----------------------------------------------===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
diff --git a/utils/bazel/llvm-project-overlay/libc/test/src/math/smoke/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/test/src/math/smoke/BUILD.bazel
index 6b959fe559e45..29b3c5be157d7 100644
--- a/utils/bazel/llvm-project-overlay/libc/test/src/math/smoke/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/test/src/math/smoke/BUILD.bazel
@@ -476,7 +476,7 @@ math_test(
math_test(
name = "fmaf16",
- hdrs = ["FMaxTest.h"],
+ hdrs = ["FmaTest.h"],
)
math_test(
>From 5f2b38438ce8d7db58d3b421084a127913dbb4d9 Mon Sep 17 00:00:00 2001
From: Harrison Hao <tsworld1314 at gmail.com>
Date: Wed, 19 Mar 2025 18:10:27 +0800
Subject: [PATCH 4/5] [libc][math][c23] Update cmake.
---
libc/test/src/math/smoke/CMakeLists.txt | 2 ++
libc/test/src/math/smoke/FmaTest.h | 6 ------
2 files changed, 2 insertions(+), 6 deletions(-)
diff --git a/libc/test/src/math/smoke/CMakeLists.txt b/libc/test/src/math/smoke/CMakeLists.txt
index 33ce74d44a3a1..10c03d5ad3505 100644
--- a/libc/test/src/math/smoke/CMakeLists.txt
+++ b/libc/test/src/math/smoke/CMakeLists.txt
@@ -3572,6 +3572,8 @@ add_fp_unittest(
FmaTest.h
DEPENDS
libc.src.math.fmaf16
+ libc.src.__support.CPP.type_traits
+ libc.src.__support.FPUtil.cast
libc.src.__support.macros.properties.types
)
diff --git a/libc/test/src/math/smoke/FmaTest.h b/libc/test/src/math/smoke/FmaTest.h
index 41093422d51b2..d044a9818a4ae 100644
--- a/libc/test/src/math/smoke/FmaTest.h
+++ b/libc/test/src/math/smoke/FmaTest.h
@@ -90,12 +90,6 @@ class FmaTestTemplate : public LIBC_NAMESPACE::testing::FEnvSafeTest {
// Test overflow.
OutType z = out.max_normal;
InType in_z = LIBC_NAMESPACE::fputil::cast<InType>(out.max_normal);
-#if defined(LIBC_TYPES_HAS_FLOAT16) && !defined(__LIBC_USE_FLOAT16_CONVERSION)
- // Rounding modes other than the default might not be usable with float16.
- if constexpr (LIBC_NAMESPACE::cpp::is_same_v<OutType, float16>)
- EXPECT_FP_EQ(OutType(0.75) * z, func(InType(1.75), in_z, -in_z));
- else
-#endif
EXPECT_FP_EQ_ALL_ROUNDING(OutType(0.75) * z,
func(InType(1.75), in_z, -in_z));
>From 450d6bb4f75d49f89dc24bdd4817e7d726e9a1e9 Mon Sep 17 00:00:00 2001
From: Harrison Hao <tsworld1314 at gmail.com>
Date: Wed, 19 Mar 2025 14:16:05 +0000
Subject: [PATCH 5/5] [libc][math][c23] Fix clang format.
---
libc/test/src/math/smoke/FmaTest.h | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/libc/test/src/math/smoke/FmaTest.h b/libc/test/src/math/smoke/FmaTest.h
index d044a9818a4ae..9752e07f4ca51 100644
--- a/libc/test/src/math/smoke/FmaTest.h
+++ b/libc/test/src/math/smoke/FmaTest.h
@@ -90,8 +90,8 @@ class FmaTestTemplate : public LIBC_NAMESPACE::testing::FEnvSafeTest {
// Test overflow.
OutType z = out.max_normal;
InType in_z = LIBC_NAMESPACE::fputil::cast<InType>(out.max_normal);
- EXPECT_FP_EQ_ALL_ROUNDING(OutType(0.75) * z,
- func(InType(1.75), in_z, -in_z));
+ EXPECT_FP_EQ_ALL_ROUNDING(OutType(0.75) * z,
+ func(InType(1.75), in_z, -in_z));
// Exact cancellation.
EXPECT_FP_EQ_ROUNDING_NEAREST(
More information about the llvm-commits
mailing list