[libc-commits] [libc] [libc][math][c23] Add exp10f16 C23 math function (PR #101588)
via libc-commits
libc-commits at lists.llvm.org
Wed Aug 7 05:58:52 PDT 2024
https://github.com/overmighty updated https://github.com/llvm/llvm-project/pull/101588
>From 2295725e6c6e2dd5d9c2e790890703e119c3e88c Mon Sep 17 00:00:00 2001
From: OverMighty <its.overmighty at gmail.com>
Date: Fri, 2 Aug 2024 02:10:46 +0200
Subject: [PATCH 1/4] [libc][math][c23] Add exp10f16 C23 math function
Part of #95250.
---
libc/config/linux/x86_64/entrypoints.txt | 1 +
libc/docs/math/index.rst | 2 +-
libc/spec/stdc.td | 1 +
libc/src/math/CMakeLists.txt | 1 +
libc/src/math/exp10f16.h | 21 +++
libc/src/math/generic/CMakeLists.txt | 23 +++
libc/src/math/generic/exp10f16.cpp | 177 +++++++++++++++++++++
libc/test/src/math/CMakeLists.txt | 31 ++--
libc/test/src/math/exp10f16_test.cpp | 40 +++++
libc/test/src/math/smoke/CMakeLists.txt | 30 ++--
libc/test/src/math/smoke/exp10f16_test.cpp | 65 ++++++++
11 files changed, 372 insertions(+), 20 deletions(-)
create mode 100644 libc/src/math/exp10f16.h
create mode 100644 libc/src/math/generic/exp10f16.cpp
create mode 100644 libc/test/src/math/exp10f16_test.cpp
create mode 100644 libc/test/src/math/smoke/exp10f16_test.cpp
diff --git a/libc/config/linux/x86_64/entrypoints.txt b/libc/config/linux/x86_64/entrypoints.txt
index 9f36c24ec61ec..6702d856afe00 100644
--- a/libc/config/linux/x86_64/entrypoints.txt
+++ b/libc/config/linux/x86_64/entrypoints.txt
@@ -576,6 +576,7 @@ if(LIBC_TYPES_HAS_FLOAT16)
libc.src.math.canonicalizef16
libc.src.math.ceilf16
libc.src.math.copysignf16
+ libc.src.math.exp10f16
libc.src.math.exp2f16
libc.src.math.expf16
libc.src.math.f16add
diff --git a/libc/docs/math/index.rst b/libc/docs/math/index.rst
index 193e0675357ca..ab7fdefd30abd 100644
--- a/libc/docs/math/index.rst
+++ b/libc/docs/math/index.rst
@@ -286,7 +286,7 @@ Higher Math Functions
+-----------+------------------+-----------------+------------------------+----------------------+------------------------+------------------------+----------------------------+
| exp | |check| | |check| | | |check| | | 7.12.6.1 | F.10.3.1 |
+-----------+------------------+-----------------+------------------------+----------------------+------------------------+------------------------+----------------------------+
-| exp10 | |check| | |check| | | | | 7.12.6.2 | F.10.3.2 |
+| exp10 | |check| | |check| | | |check| | | 7.12.6.2 | F.10.3.2 |
+-----------+------------------+-----------------+------------------------+----------------------+------------------------+------------------------+----------------------------+
| exp10m1 | | | | | | 7.12.6.3 | F.10.3.3 |
+-----------+------------------+-----------------+------------------------+----------------------+------------------------+------------------------+----------------------------+
diff --git a/libc/spec/stdc.td b/libc/spec/stdc.td
index 400b673953a27..24de4c5e01ba3 100644
--- a/libc/spec/stdc.td
+++ b/libc/spec/stdc.td
@@ -591,6 +591,7 @@ def StdC : StandardSpec<"stdc"> {
FunctionSpec<"exp10", RetValSpec<DoubleType>, [ArgSpec<DoubleType>]>,
FunctionSpec<"exp10f", RetValSpec<FloatType>, [ArgSpec<FloatType>]>,
+ GuardedFunctionSpec<"exp10f16", RetValSpec<Float16Type>, [ArgSpec<Float16Type>], "LIBC_TYPES_HAS_FLOAT16">,
FunctionSpec<"remainder", RetValSpec<DoubleType>, [ArgSpec<DoubleType>, ArgSpec<DoubleType>]>,
FunctionSpec<"remainderf", RetValSpec<FloatType>, [ArgSpec<FloatType>, ArgSpec<FloatType>]>,
diff --git a/libc/src/math/CMakeLists.txt b/libc/src/math/CMakeLists.txt
index 1730eec181877..b9cecc7e1f130 100644
--- a/libc/src/math/CMakeLists.txt
+++ b/libc/src/math/CMakeLists.txt
@@ -116,6 +116,7 @@ add_math_entrypoint_object(exp2m1f)
add_math_entrypoint_object(exp10)
add_math_entrypoint_object(exp10f)
+add_math_entrypoint_object(exp10f16)
add_math_entrypoint_object(expm1)
add_math_entrypoint_object(expm1f)
diff --git a/libc/src/math/exp10f16.h b/libc/src/math/exp10f16.h
new file mode 100644
index 0000000000000..62a62f7da59ed
--- /dev/null
+++ b/libc/src/math/exp10f16.h
@@ -0,0 +1,21 @@
+//===-- Implementation header for exp10f16 ----------------------*- 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_EXP10F16_H
+#define LLVM_LIBC_SRC_MATH_EXP10F16_H
+
+#include "src/__support/macros/config.h"
+#include "src/__support/macros/properties/types.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+float16 exp10f16(float16 x);
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SRC_MATH_EXP10F16_H
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index 7574f6c22bc17..d17a17596454d 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -1497,6 +1497,29 @@ add_entrypoint_object(
-O3
)
+add_entrypoint_object(
+ exp10f16
+ SRCS
+ exp10f16.cpp
+ HDRS
+ ../exp10f16.h
+ DEPENDS
+ libc.hdr.errno_macros
+ libc.hdr.fenv_macros
+ libc.src.__support.CPP.array
+ libc.src.__support.FPUtil.except_value_utils
+ libc.src.__support.FPUtil.fenv_impl
+ libc.src.__support.FPUtil.fp_bits
+ libc.src.__support.FPUtil.multiply_add
+ libc.src.__support.FPUtil.nearest_integer
+ libc.src.__support.FPUtil.polyeval
+ libc.src.__support.FPUtil.rounding_mode
+ libc.src.__support.macros.optimization
+ libc.src.__support.macros.properties.cpu_features
+ COMPILE_OPTIONS
+ -O3
+)
+
add_entrypoint_object(
expm1
SRCS
diff --git a/libc/src/math/generic/exp10f16.cpp b/libc/src/math/generic/exp10f16.cpp
new file mode 100644
index 0000000000000..62349ee14d9c2
--- /dev/null
+++ b/libc/src/math/generic/exp10f16.cpp
@@ -0,0 +1,177 @@
+//===-- Half-precision 10^x 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/exp10f16.h"
+#include "hdr/errno_macros.h"
+#include "hdr/fenv_macros.h"
+#include "src/__support/CPP/array.h"
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/__support/FPUtil/PolyEval.h"
+#include "src/__support/FPUtil/except_value_utils.h"
+#include "src/__support/FPUtil/multiply_add.h"
+#include "src/__support/FPUtil/nearest_integer.h"
+#include "src/__support/FPUtil/rounding_mode.h"
+#include "src/__support/common.h"
+#include "src/__support/macros/config.h"
+#include "src/__support/macros/optimization.h"
+#include "src/__support/macros/properties/cpu_features.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+static constexpr size_t N_EXP10F16_EXCEPTS = 5
+#ifndef LIBC_TARGET_CPU_HAS_FMA
+ + 3
+#endif
+ ;
+
+static constexpr fputil::ExceptValues<float16, N_EXP10F16_EXCEPTS>
+ EXP10F16_EXCEPTS = {{
+ // x = 0x1.8f4p-2, exp10f16(x) = 0x1.3ap+1 (RZ)
+ {0x363dU, 0x40e8U, 1U, 0U, 1U},
+ // x = 0x1.95cp-2, exp10f16(x) = 0x1.3ecp+1 (RZ)
+ {0x3657U, 0x40fbU, 1U, 0U, 0U},
+ // x = -0x1.018p-4, exp10f16(x) = 0x1.bbp-1 (RZ)
+ {0xac06U, 0x3aecU, 1U, 0U, 0U},
+ // x = -0x1.c28p+0, exp10f16(x) = 0x1.1ccp-6 (RZ)
+ {0xbf0aU, 0x2473U, 1U, 0U, 0U},
+ // x = -0x1.e1cp+1, exp10f16(x) = 0x1.694p-13 (RZ)
+ {0xc387U, 0x09a5U, 1U, 0U, 0U},
+#ifndef LIBC_TARGET_CPU_HAS_FMA
+ // x = 0x1.0cp+1, exp10f16(x) = 0x1.f04p+6 (RZ)
+ {0x4030U, 0x57c1U, 1U, 0U, 1U},
+ // x = 0x1.1b8p+1, exp10f16(x) = 0x1.47cp+7 (RZ)
+ {0x406eU, 0x591fU, 1U, 0U, 1U},
+ // x = 0x1.1b8p+2, exp10f16(x) = 0x1.a4p+14 (RZ)
+ {0x446eU, 0x7690U, 1U, 0U, 1U},
+#endif
+ }};
+
+// Generated by Sollya with the following commands:
+// > display = hexadecimal;
+// > round(log2(10), SG, RN);
+static constexpr float LOG2F_10 = 0x1.a934fp+1f;
+
+// Generated by Sollya with the following commands:
+// > display = hexadecimal;
+// > round(log10(2), SG, RN);
+static constexpr float LOG10F_2 = 0x1.344136p-2f;
+
+// Generated by Sollya with the following commands:
+// > display = hexadecimal;
+// > for i from 0 to 7 do printsingle(round(2^(i * 2^-3), SG, RN));
+static constexpr cpp::array<uint32_t, 8> EXP2_MID_BITS = {
+ 0x3f80'0000U, 0x3f8b'95c2U, 0x3f98'37f0U, 0x3fa5'fed7U,
+ 0x3fb5'04f3U, 0x3fc5'672aU, 0x3fd7'44fdU, 0x3fea'c0c7U,
+};
+
+LLVM_LIBC_FUNCTION(float16, exp10f16, (float16 x)) {
+ using FPBits = fputil::FPBits<float16>;
+ FPBits x_bits(x);
+
+ uint16_t x_u = x_bits.uintval();
+ uint16_t x_abs = x_u & 0x7fffU;
+
+ // When |x| >= 5, or x is NaN.
+ if (LIBC_UNLIKELY(x_abs >= 0x4500U)) {
+ // exp10(NaN) = NaN
+ if (x_bits.is_nan()) {
+ if (x_bits.is_signaling_nan()) {
+ fputil::raise_except_if_required(FE_INVALID);
+ return FPBits::quiet_nan().get_val();
+ }
+
+ return x;
+ }
+
+ // When x >= 5.
+ if (x_bits.is_pos()) {
+ // exp10(+inf) = +inf
+ if (x_bits.is_inf())
+ return FPBits::inf().get_val();
+
+ switch (fputil::quick_get_round()) {
+ case FE_TONEAREST:
+ case FE_UPWARD:
+ fputil::set_errno_if_required(ERANGE);
+ fputil::raise_except_if_required(FE_OVERFLOW);
+ return FPBits::inf().get_val();
+ default:
+ return FPBits::max_normal().get_val();
+ }
+ }
+
+ // When x <= -8.
+ if (x_u >= 0xc800U) {
+ // exp10(-inf) = +0
+ if (x_bits.is_inf())
+ return FPBits::zero().get_val();
+
+ fputil::set_errno_if_required(ERANGE);
+ fputil::raise_except_if_required(FE_UNDERFLOW | FE_INEXACT);
+
+ if (fputil::fenv_is_round_up())
+ return FPBits::min_subnormal().get_val();
+ return FPBits::zero().get_val();
+ }
+ }
+
+ // When x is 1, 2, 3, or 4. These are hard-to-round cases with exact results.
+ if (LIBC_UNLIKELY((x_u & ~(0x3c00U | 0x4000U | 0x4200U | 0x4400U)) == 0)) {
+ switch (x_u) {
+ case 0x3c00U: // x = 1.0f16
+ return static_cast<float16>(10.0);
+ case 0x4000U: // x = 2.0f16
+ return static_cast<float16>(100.0);
+ case 0x4200U: // x = 3.0f16
+ return static_cast<float16>(1'000.0);
+ case 0x4400U: // x = 4.0f16
+ return static_cast<float16>(10'000.0);
+ }
+ }
+
+ if (auto r = EXP10F16_EXCEPTS.lookup(x_u); LIBC_UNLIKELY(r.has_value()))
+ return r.value();
+
+ // For -8 < x < 5, to compute 10^x, we perform the following range reduction:
+ // find hi, mid, lo, such that:
+ // x = (hi + mid) * log2(10) + lo, in which
+ // hi is an integer,
+ // mid * 2^3 is an integer,
+ // -2^(-4) <= lo < 2^(-4).
+ // In particular,
+ // hi + mid = round(x * 2^3) * 2^(-3).
+ // Then,
+ // 10^x = 10^(hi + mid + lo) = 2^((hi + mid) * log2(10)) + 10^lo
+ // We store 2^mid in the lookup table EXP2_MID_BITS, and compute 2^hi * 2^mid
+ // by adding hi to the exponent field of 2^mid. 10^lo is computed using a
+ // degree-4 minimax polynomial generated by Sollya.
+
+ float xf = x;
+ float kf = fputil::nearest_integer(xf * (LOG2F_10 * 0x1.0p+3f));
+ int x_hi_mid = static_cast<int>(kf);
+ int x_hi = x_hi_mid >> 3;
+ int x_mid = x_hi_mid & 0x7;
+ // lo = x - (hi + mid) = round(x * 2^3 * log2(10)) * log10(2) * (-2^(-3)) + x
+ float lo = fputil::multiply_add(kf, LOG10F_2 * -0x1.0p-3f, xf);
+
+ uint32_t exp2_hi_mid_bits =
+ EXP2_MID_BITS[x_mid] +
+ static_cast<uint32_t>(x_hi << fputil::FPBits<float>::FRACTION_LEN);
+ float exp2_hi_mid = fputil::FPBits<float>(exp2_hi_mid_bits).get_val();
+ // Degree-4 minimax polynomial generated by Sollya with the following
+ // commands:
+ // > display = hexadecimal;
+ // > P = fpminimax((10^x - 1)/x, 3, [|SG...|], [-2^-4, 2^-4]);
+ // > 1 + x * P;
+ float exp10_lo = fputil::polyeval(lo, 0x1p+0f, 0x1.26bb14p+1f, 0x1.53526p+1f,
+ 0x1.04b434p+1f, 0x1.2bcf9ep+0f);
+ return static_cast<float16>(exp2_hi_mid * exp10_lo);
+}
+
+} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/test/src/math/CMakeLists.txt b/libc/test/src/math/CMakeLists.txt
index 7cb24f85170bc..2eb5abd4f9d7e 100644
--- a/libc/test/src/math/CMakeLists.txt
+++ b/libc/test/src/math/CMakeLists.txt
@@ -990,6 +990,19 @@ add_fp_unittest(
libc.src.__support.FPUtil.fp_bits
)
+add_fp_unittest(
+ exp10_test
+ NEED_MPFR
+ SUITE
+ libc-math-unittests
+ SRCS
+ exp10_test.cpp
+ DEPENDS
+ libc.src.errno.errno
+ libc.src.math.exp10
+ libc.src.__support.FPUtil.fp_bits
+)
+
add_fp_unittest(
exp10f_test
NEED_MPFR
@@ -1004,16 +1017,14 @@ add_fp_unittest(
)
add_fp_unittest(
- exp10_test
- NEED_MPFR
- SUITE
- libc-math-unittests
- SRCS
- exp10_test.cpp
- DEPENDS
- libc.src.errno.errno
- libc.src.math.exp10
- libc.src.__support.FPUtil.fp_bits
+ exp10f16_test
+ NEED_MPFR
+ SUITE
+ libc-math-unittests
+ SRCS
+ exp10f16_test.cpp
+ DEPENDS
+ libc.src.math.exp10f16
)
add_fp_unittest(
diff --git a/libc/test/src/math/exp10f16_test.cpp b/libc/test/src/math/exp10f16_test.cpp
new file mode 100644
index 0000000000000..fc49331b5917c
--- /dev/null
+++ b/libc/test/src/math/exp10f16_test.cpp
@@ -0,0 +1,40 @@
+//===-- Exhaustive test for exp10f16 --------------------------------------===//
+//
+// 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/exp10f16.h"
+#include "test/UnitTest/FPMatcher.h"
+#include "test/UnitTest/Test.h"
+#include "utils/MPFRWrapper/MPFRUtils.h"
+
+using LlvmLibcExp10f16Test = LIBC_NAMESPACE::testing::FPTest<float16>;
+
+namespace mpfr = LIBC_NAMESPACE::testing::mpfr;
+
+// Range: [0, Inf];
+static constexpr uint16_t POS_START = 0x0000U;
+static constexpr uint16_t POS_STOP = 0x7c00U;
+
+// Range: [-Inf, 0];
+static constexpr uint16_t NEG_START = 0x8000U;
+static constexpr uint16_t NEG_STOP = 0xfc00U;
+
+TEST_F(LlvmLibcExp10f16Test, PositiveRange) {
+ for (uint16_t v = POS_START; v <= POS_STOP; ++v) {
+ float16 x = FPBits(v).get_val();
+ EXPECT_MPFR_MATCH_ALL_ROUNDING(mpfr::Operation::Exp10, x,
+ LIBC_NAMESPACE::exp10f16(x), 0.5);
+ }
+}
+
+TEST_F(LlvmLibcExp10f16Test, NegativeRange) {
+ for (uint16_t v = NEG_START; v <= NEG_STOP; ++v) {
+ float16 x = FPBits(v).get_val();
+ EXPECT_MPFR_MATCH_ALL_ROUNDING(mpfr::Operation::Exp10, x,
+ LIBC_NAMESPACE::exp10f16(x), 0.5);
+ }
+}
diff --git a/libc/test/src/math/smoke/CMakeLists.txt b/libc/test/src/math/smoke/CMakeLists.txt
index 07d5d78d57996..86357688404e6 100644
--- a/libc/test/src/math/smoke/CMakeLists.txt
+++ b/libc/test/src/math/smoke/CMakeLists.txt
@@ -1076,6 +1076,18 @@ add_fp_unittest(
libc.src.math.exp2m1f
)
+add_fp_unittest(
+ exp10_test
+ SUITE
+ libc-math-smoke-tests
+ SRCS
+ exp10_test.cpp
+ DEPENDS
+ libc.src.errno.errno
+ libc.src.math.exp10
+ libc.src.__support.FPUtil.fp_bits
+)
+
add_fp_unittest(
exp10f_test
SUITE
@@ -1089,15 +1101,15 @@ add_fp_unittest(
)
add_fp_unittest(
- exp10_test
- SUITE
- libc-math-smoke-tests
- SRCS
- exp10_test.cpp
- DEPENDS
- libc.src.errno.errno
- libc.src.math.exp10
- libc.src.__support.FPUtil.fp_bits
+ exp10f16_test
+ SUITE
+ libc-math-smoke-tests
+ SRCS
+ exp10f16_test.cpp
+ DEPENDS
+ libc.hdr.fenv_macros
+ libc.src.errno.errno
+ libc.src.math.exp10f16
)
add_fp_unittest(
diff --git a/libc/test/src/math/smoke/exp10f16_test.cpp b/libc/test/src/math/smoke/exp10f16_test.cpp
new file mode 100644
index 0000000000000..006dfafa8aa14
--- /dev/null
+++ b/libc/test/src/math/smoke/exp10f16_test.cpp
@@ -0,0 +1,65 @@
+//===-- Unittests for exp10f16 --------------------------------------------===//
+//
+// 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 "hdr/fenv_macros.h"
+#include "src/errno/libc_errno.h"
+#include "src/math/exp10f16.h"
+#include "test/UnitTest/FPMatcher.h"
+#include "test/UnitTest/Test.h"
+
+using LlvmLibcExp10f16Test = LIBC_NAMESPACE::testing::FPTest<float16>;
+
+TEST_F(LlvmLibcExp10f16Test, SpecialNumbers) {
+ LIBC_NAMESPACE::libc_errno = 0;
+
+ EXPECT_FP_EQ_ALL_ROUNDING(aNaN, LIBC_NAMESPACE::exp10f16(aNaN));
+ EXPECT_MATH_ERRNO(0);
+
+ EXPECT_FP_EQ_WITH_EXCEPTION(aNaN, LIBC_NAMESPACE::exp10f16(sNaN), FE_INVALID);
+ EXPECT_MATH_ERRNO(0);
+
+ EXPECT_FP_EQ_ALL_ROUNDING(inf, LIBC_NAMESPACE::exp10f16(inf));
+ EXPECT_MATH_ERRNO(0);
+
+ EXPECT_FP_EQ_ALL_ROUNDING(static_cast<float16>(zero),
+ LIBC_NAMESPACE::exp10f16(neg_inf));
+ EXPECT_MATH_ERRNO(0);
+
+ EXPECT_FP_EQ_ALL_ROUNDING(static_cast<float16>(1.0f),
+ LIBC_NAMESPACE::exp10f16(zero));
+ EXPECT_MATH_ERRNO(0);
+
+ EXPECT_FP_EQ_ALL_ROUNDING(static_cast<float16>(1.0f),
+ LIBC_NAMESPACE::exp10f16(neg_zero));
+ EXPECT_MATH_ERRNO(0);
+}
+
+TEST_F(LlvmLibcExp10f16Test, Overflow) {
+ LIBC_NAMESPACE::libc_errno = 0;
+
+ EXPECT_FP_EQ_WITH_EXCEPTION(inf, LIBC_NAMESPACE::exp10f16(max_normal),
+ FE_OVERFLOW);
+ EXPECT_MATH_ERRNO(ERANGE);
+
+ EXPECT_FP_EQ_WITH_EXCEPTION(
+ inf, LIBC_NAMESPACE::exp10f16(static_cast<float16>(5.0)), FE_OVERFLOW);
+ EXPECT_MATH_ERRNO(ERANGE);
+}
+
+TEST_F(LlvmLibcExp10f16Test, Underflow) {
+ LIBC_NAMESPACE::libc_errno = 0;
+
+ EXPECT_FP_EQ_WITH_EXCEPTION(zero, LIBC_NAMESPACE::exp10f16(neg_max_normal),
+ FE_UNDERFLOW | FE_INEXACT);
+ EXPECT_MATH_ERRNO(ERANGE);
+
+ EXPECT_FP_EQ_WITH_EXCEPTION(
+ zero, LIBC_NAMESPACE::exp10f16(static_cast<float16>(-8.0)),
+ FE_UNDERFLOW | FE_INEXACT);
+ EXPECT_MATH_ERRNO(ERANGE);
+}
>From 933cb759738cd5283b4b2b3cdfcab518e6309e81 Mon Sep 17 00:00:00 2001
From: OverMighty <its.overmighty at gmail.com>
Date: Fri, 2 Aug 2024 20:26:51 +0200
Subject: [PATCH 2/4] Move EXP2_MID_BITS to a new header shared between exp2f16
and exp10f16
---
libc/src/math/generic/CMakeLists.txt | 10 ++++++++++
libc/src/math/generic/exp10f16.cpp | 9 +--------
libc/src/math/generic/exp2f16.cpp | 9 +--------
libc/src/math/generic/expxf16.h | 28 ++++++++++++++++++++++++++++
4 files changed, 40 insertions(+), 16 deletions(-)
create mode 100644 libc/src/math/generic/expxf16.h
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index d17a17596454d..63af0439b1503 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -1402,6 +1402,7 @@ add_entrypoint_object(
HDRS
../exp2f16.h
DEPENDS
+ .expxf16
libc.hdr.errno_macros
libc.hdr.fenv_macros
libc.src.__support.CPP.array
@@ -1504,6 +1505,7 @@ add_entrypoint_object(
HDRS
../exp10f16.h
DEPENDS
+ .expxf16
libc.hdr.errno_macros
libc.hdr.fenv_macros
libc.src.__support.CPP.array
@@ -4783,3 +4785,11 @@ add_entrypoint_object(
COMPILE_OPTIONS
-O3
)
+
+add_header_library(
+ expxf16
+ HDRS
+ expxf16.h
+ DEPENDS
+ libc.src.__support.CPP.array
+)
diff --git a/libc/src/math/generic/exp10f16.cpp b/libc/src/math/generic/exp10f16.cpp
index 62349ee14d9c2..207237fbe414d 100644
--- a/libc/src/math/generic/exp10f16.cpp
+++ b/libc/src/math/generic/exp10f16.cpp
@@ -7,6 +7,7 @@
//===----------------------------------------------------------------------===//
#include "src/math/exp10f16.h"
+#include "expxf16.h"
#include "hdr/errno_macros.h"
#include "hdr/fenv_macros.h"
#include "src/__support/CPP/array.h"
@@ -62,14 +63,6 @@ static constexpr float LOG2F_10 = 0x1.a934fp+1f;
// > round(log10(2), SG, RN);
static constexpr float LOG10F_2 = 0x1.344136p-2f;
-// Generated by Sollya with the following commands:
-// > display = hexadecimal;
-// > for i from 0 to 7 do printsingle(round(2^(i * 2^-3), SG, RN));
-static constexpr cpp::array<uint32_t, 8> EXP2_MID_BITS = {
- 0x3f80'0000U, 0x3f8b'95c2U, 0x3f98'37f0U, 0x3fa5'fed7U,
- 0x3fb5'04f3U, 0x3fc5'672aU, 0x3fd7'44fdU, 0x3fea'c0c7U,
-};
-
LLVM_LIBC_FUNCTION(float16, exp10f16, (float16 x)) {
using FPBits = fputil::FPBits<float16>;
FPBits x_bits(x);
diff --git a/libc/src/math/generic/exp2f16.cpp b/libc/src/math/generic/exp2f16.cpp
index 3f18ab9166a85..66b7956704005 100644
--- a/libc/src/math/generic/exp2f16.cpp
+++ b/libc/src/math/generic/exp2f16.cpp
@@ -7,6 +7,7 @@
//===----------------------------------------------------------------------===//
#include "src/math/exp2f16.h"
+#include "expxf16.h"
#include "hdr/errno_macros.h"
#include "hdr/fenv_macros.h"
#include "src/__support/CPP/array.h"
@@ -33,14 +34,6 @@ static constexpr fputil::ExceptValues<float16, 3> EXP2F16_EXCEPTS = {{
{0xaf57U, 0x3b63U, 1U, 0U, 0U},
}};
-// Generated by Sollya with the following commands:
-// > display = hexadecimal;
-// > for i from 0 to 7 do printsingle(round(2^(i * 2^-3), SG, RN));
-static constexpr cpp::array<uint32_t, 8> EXP2_MID_BITS = {
- 0x3f80'0000U, 0x3f8b'95c2U, 0x3f98'37f0U, 0x3fa5'fed7U,
- 0x3fb5'04f3U, 0x3fc5'672aU, 0x3fd7'44fdU, 0x3fea'c0c7U,
-};
-
LLVM_LIBC_FUNCTION(float16, exp2f16, (float16 x)) {
using FPBits = fputil::FPBits<float16>;
FPBits x_bits(x);
diff --git a/libc/src/math/generic/expxf16.h b/libc/src/math/generic/expxf16.h
new file mode 100644
index 0000000000000..c33aca337b98d
--- /dev/null
+++ b/libc/src/math/generic/expxf16.h
@@ -0,0 +1,28 @@
+//===-- Common utilities for half-precision exponential functions ---------===//
+//
+// 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_GENERIC_EXPXF16_H
+#define LLVM_LIBC_SRC_MATH_GENERIC_EXPXF16_H
+
+#include "src/__support/CPP/array.h"
+#include "src/__support/macros/config.h"
+#include <stdint.h>
+
+namespace LIBC_NAMESPACE_DECL {
+
+// Generated by Sollya with the following commands:
+// > display = hexadecimal;
+// > for i from 0 to 7 do printsingle(round(2^(i * 2^-3), SG, RN));
+constexpr cpp::array<uint32_t, 8> EXP2_MID_BITS = {
+ 0x3f80'0000U, 0x3f8b'95c2U, 0x3f98'37f0U, 0x3fa5'fed7U,
+ 0x3fb5'04f3U, 0x3fc5'672aU, 0x3fd7'44fdU, 0x3fea'c0c7U,
+};
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SRC_MATH_GENERIC_EXPXF16_H
>From 81599193828358cb395b53f1e333b0a269383f22 Mon Sep 17 00:00:00 2001
From: OverMighty <its.overmighty at gmail.com>
Date: Tue, 6 Aug 2024 14:48:35 +0200
Subject: [PATCH 3/4] Add performance test
---
.../math/performance_testing/CMakeLists.txt | 11 ++++++++++
.../performance_testing/exp10f16_perf.cpp | 22 +++++++++++++++++++
2 files changed, 33 insertions(+)
create mode 100644 libc/test/src/math/performance_testing/exp10f16_perf.cpp
diff --git a/libc/test/src/math/performance_testing/CMakeLists.txt b/libc/test/src/math/performance_testing/CMakeLists.txt
index 09211537d93f4..8e529ca09ed79 100644
--- a/libc/test/src/math/performance_testing/CMakeLists.txt
+++ b/libc/test/src/math/performance_testing/CMakeLists.txt
@@ -153,6 +153,17 @@ add_perf_binary(
-fno-builtin
)
+add_perf_binary(
+ exp10f16_perf
+ SRCS
+ exp10f16_perf.cpp
+ DEPENDS
+ .single_input_single_output_diff
+ libc.src.math.exp10f16
+ COMPILE_OPTIONS
+ -fno-builtin
+)
+
add_perf_binary(
exp2f_perf
SRCS
diff --git a/libc/test/src/math/performance_testing/exp10f16_perf.cpp b/libc/test/src/math/performance_testing/exp10f16_perf.cpp
new file mode 100644
index 0000000000000..b9e76d420afce
--- /dev/null
+++ b/libc/test/src/math/performance_testing/exp10f16_perf.cpp
@@ -0,0 +1,22 @@
+//===-- Performance test for exp10f16 -------------------------------------===//
+//
+// 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 "SingleInputSingleOutputPerf.h"
+
+#include "src/math/exp10f16.h"
+
+// LLVM libc might be the only libc implementation with support for float16 math
+// functions currently. We can't compare our float16 functions against the
+// system libc, so we compare them against this placeholder function.
+static float16 placeholderf16(float16 x) { return x; }
+
+int main() {
+ SINGLE_INPUT_SINGLE_OUTPUT_PERF_EX(float16, LIBC_NAMESPACE::exp10f16,
+ ::placeholderf16, 20'000,
+ "exp10f16_perf.log")
+}
>From 0ebe2b4748845de4785e387dbe06a680d5a00159 Mon Sep 17 00:00:00 2001
From: OverMighty <its.overmighty at gmail.com>
Date: Wed, 7 Aug 2024 14:57:19 +0200
Subject: [PATCH 4/4] Make initialization of N_EXP10F16_EXCEPTS more readable
---
libc/src/math/generic/exp10f16.cpp | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/libc/src/math/generic/exp10f16.cpp b/libc/src/math/generic/exp10f16.cpp
index 207237fbe414d..9959f7450b591 100644
--- a/libc/src/math/generic/exp10f16.cpp
+++ b/libc/src/math/generic/exp10f16.cpp
@@ -25,11 +25,11 @@
namespace LIBC_NAMESPACE_DECL {
-static constexpr size_t N_EXP10F16_EXCEPTS = 5
-#ifndef LIBC_TARGET_CPU_HAS_FMA
- + 3
+#ifdef LIBC_TARGET_CPU_HAS_FMA
+static constexpr size_t N_EXP10F16_EXCEPTS = 5;
+#else
+static constexpr size_t N_EXP10F16_EXCEPTS = 8;
#endif
- ;
static constexpr fputil::ExceptValues<float16, N_EXP10F16_EXCEPTS>
EXP10F16_EXCEPTS = {{
More information about the libc-commits
mailing list