[libc][math] Refactor exp10m1f16 implementation to header-only in src/__support/math folder. #161119

bassiounix · 2025-09-29T04:04:16Z

in preparation for: https://discourse.llvm.org/t/rfc-make-clang-builtin-math-functions-constexpr-with-llvm-libc-to-support-c-23-constexpr-math-functions/86450

bassiounix · 2025-09-29T04:05:02Z

[libc][math] Refactor expm1f16 implementation to header-only in src/__support/math folder. #162132
[libc][math] Refactor expm1f implementation to header-only in src/__support/math folder. #162131
[libc][math] Refactor expm1 implementation to header-only in src/__support/math folder. #162127
[libc][math] Refactor exp2m1f16 implementation to header-only in src/__support/math folder. #162019
[libc][math] Refactor exp2m1f implementation to header-only in src/__support/math folder. #162017
[libc][math] Refactor exp2f16 implementation to header-only in src/__support/math folder. #161993
[libc][math] Refactor exp2f implementation to header-only in src/__support/math folder. #161992
[libc][math] Refactor exp2 implementation to header-only in src/__support/math folder. #161297
[libc][math] Refactor exp10m1f16 implementation to header-only in src/__support/math folder. #161119 👈 (View in Graphite)
[libc][math] Refactor exp10m1f implementation to header-only in src/__support/math folder. #159897
[libc][math] Refactor dsqrtl implementation to header-only in src/__support/math folder. #154868
[libc][math] Refactor cospif16 implementation to header-only in src/__support/math folder. #154222
[libc][math] Refactor cospif implementation to header-only in src/__support/math folder. #154215
[libc][math] Refactor coshf16 implementation to header-only in src/__support/math folder. #153582
[libc][math] Refactor coshf implementation to header-only in src/__support/math folder. #153427
[libc][math] Refactor cosf16 implementation to header-only in src/__support/math folder. #152871
[libc][math] Refactor cosf implementation to header-only in src/__support/math folder. #152069
[libc][math] Refactor cos implementation to header-only in src/__support/math folder. #151883
[libc][math] Refactor cbrtf implementation to header-only in src/__support/math folder. #151846
[libc][math] Refactor cbrt implementation to header-only in src/__support/math folder. #151837
[libc][math] Refactor atanhf16 implementation to header-only in src/__support/math folder. #151779
[libc][math] Refactor atanhf implementation to header-only in src/__support/math folder. #151399
[libc][math] Refactor atan2f128 implementation to header-only in src/__support/math folder. #151012
[libc][math] Refactor atan2f implementation to header-only in src/__support/math folder. #150993
[libc][math] Refactor atan2 implementation to header-only in src/__support/math folder. #150968
[libc][math] Refactor atanf16 implementation to header-only in src/__support/math folder. #150868
[libc][math] Refactor atanf implementation to header-only in src/__support/math folder. #150854
[libc][math] Refactor atan implementation to header-only in src/__support/math folder. #150852
[libc][math] Refactor asinhf16 implementation to header-only in src/__support/math folder. #150849
[libc][math] Refactor asinhf implementation to header-only in src/__support/math folder. #150843
main

This stack of pull requests is managed by Graphite. Learn more about stacking.

llvmbot · 2025-09-29T04:05:12Z

@llvm/pr-subscribers-libc

Author: Muhammad Bassiouni (bassiounix)

Changes

Part of #147386

in preparation for: https://discourse.llvm.org/t/rfc-make-clang-builtin-math-functions-constexpr-with-llvm-libc-to-support-c-23-constexpr-math-functions/86450

Full diff: https://github.com/llvm/llvm-project/pull/161119.diff

9 Files Affected:

(modified) libc/shared/math.h (+1)
(added) libc/shared/math/exp10m1f16.h (+29)
(modified) libc/src/__support/math/CMakeLists.txt (+17)
(added) libc/src/__support/math/exp10m1f16.h (+185)
(modified) libc/src/math/generic/CMakeLists.txt (+1-12)
(modified) libc/src/math/generic/exp10m1f16.cpp (+2-156)
(modified) libc/test/shared/CMakeLists.txt (+1)
(modified) libc/test/shared/shared_math_test.cpp (+1)
(modified) utils/bazel/llvm-project-overlay/libc/BUILD.bazel (+17-1)

diff --git a/libc/shared/math.h b/libc/shared/math.h
index cccd6a375930e..4b2a0d8c712ad 100644
--- a/libc/shared/math.h
+++ b/libc/shared/math.h
@@ -46,6 +46,7 @@
 #include "math/exp10f.h"
 #include "math/exp10f16.h"
 #include "math/exp10m1f.h"
+#include "math/exp10m1f16.h"
 #include "math/expf.h"
 #include "math/expf16.h"
 #include "math/frexpf.h"
diff --git a/libc/shared/math/exp10m1f16.h b/libc/shared/math/exp10m1f16.h
new file mode 100644
index 0000000000000..5f18f2986207e
--- /dev/null
+++ b/libc/shared/math/exp10m1f16.h
@@ -0,0 +1,29 @@
+//===-- Shared exp10m1f16 function ------------------------------*- 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_SHARED_MATH_EXP10M1F16_H
+#define LLVM_LIBC_SHARED_MATH_EXP10M1F16_H
+
+#include "include/llvm-libc-macros/float16-macros.h"
+#include "shared/libc_common.h"
+
+#ifdef LIBC_TYPES_HAS_FLOAT16
+
+#include "src/__support/math/exp10m1f16.h"
+
+namespace LIBC_NAMESPACE_DECL {
+namespace shared {
+
+using math::exp10m1f16;
+
+} // namespace shared
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_TYPES_HAS_FLOAT16
+
+#endif // LLVM_LIBC_SHARED_MATH_EXP10M1F16_H
diff --git a/libc/src/__support/math/CMakeLists.txt b/libc/src/__support/math/CMakeLists.txt
index 84c1b15498672..98f9bb42f91f4 100644
--- a/libc/src/__support/math/CMakeLists.txt
+++ b/libc/src/__support/math/CMakeLists.txt
@@ -498,6 +498,23 @@ add_header_library(
     libc.src.__support.macros.optimization
 )
 
+add_header_library(
+  exp10m1f16
+  HDRS
+    exp10m1f16.h
+  DEPENDS
+    .exp10f16_utils
+    libc.src.__support.FPUtil.cast
+    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.polyeval
+    libc.src.__support.FPUtil.rounding_mode
+    libc.src.__support.macros.optimization
+    libc.src.__support.macros.properties.cpu_features
+)
+
 add_header_library(
   erff
   HDRS
diff --git a/libc/src/__support/math/exp10m1f16.h b/libc/src/__support/math/exp10m1f16.h
new file mode 100644
index 0000000000000..6367a857fa98a
--- /dev/null
+++ b/libc/src/__support/math/exp10m1f16.h
@@ -0,0 +1,185 @@
+//===-- Implementation header for exp10m1f16 --------------------*- 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___SUPPORT_MATH_EXP10M1F16_H
+#define LLVM_LIBC_SRC___SUPPORT_MATH_EXP10M1F16_H
+
+#include "include/llvm-libc-macros/float16-macros.h"
+
+#ifdef LIBC_TYPES_HAS_FLOAT16
+
+#include "exp10f16_utils.h"
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/__support/FPUtil/PolyEval.h"
+#include "src/__support/FPUtil/cast.h"
+#include "src/__support/FPUtil/except_value_utils.h"
+#include "src/__support/FPUtil/multiply_add.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 {
+
+namespace math {
+
+LIBC_INLINE static constexpr float16 exp10m1f16(float16 x) {
+
+#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+  constexpr fputil::ExceptValues<float16, 3> EXP10M1F16_EXCEPTS_LO = {{
+      // (input, RZ output, RU offset, RD offset, RN offset)
+      // x = 0x1.5c4p-4, exp10m1f16(x) = 0x1.bacp-3 (RZ)
+      {0x2d71U, 0x32ebU, 1U, 0U, 0U},
+      // x = -0x1.5ep-13, exp10m1f16(x) = -0x1.92cp-12 (RZ)
+      {0x8978U, 0x8e4bU, 0U, 1U, 0U},
+      // x = -0x1.e2p-10, exp10m1f16(x) = -0x1.14cp-8 (RZ)
+      {0x9788U, 0x9c53U, 0U, 1U, 0U},
+  }};
+
+#ifdef LIBC_TARGET_CPU_HAS_FMA_FLOAT
+  constexpr size_t N_EXP10M1F16_EXCEPTS_HI = 3;
+#else
+  constexpr size_t N_EXP10M1F16_EXCEPTS_HI = 6;
+#endif
+
+  constexpr fputil::ExceptValues<float16, N_EXP10M1F16_EXCEPTS_HI>
+      EXP10M1F16_EXCEPTS_HI = {{
+          // (input, RZ output, RU offset, RD offset, RN offset)
+          // x = 0x1.8f4p-2, exp10m1f16(x) = 0x1.744p+0 (RZ)
+          {0x363dU, 0x3dd1U, 1U, 0U, 0U},
+          // x = 0x1.95cp-2, exp10m1f16(x) = 0x1.7d8p+0 (RZ)
+          {0x3657U, 0x3df6U, 1U, 0U, 0U},
+          // x = 0x1.d04p-2, exp10m1f16(x) = 0x1.d7p+0 (RZ)
+          {0x3741U, 0x3f5cU, 1U, 0U, 1U},
+#ifndef LIBC_TARGET_CPU_HAS_FMA_FLOAT
+          // x = 0x1.0cp+1, exp10m1f16(x) = 0x1.ec4p+6 (RZ)
+          {0x4030U, 0x57b1U, 1U, 0U, 1U},
+          // x = 0x1.1b8p+1, exp10m1f16(x) = 0x1.45cp+7 (RZ)
+          {0x406eU, 0x5917U, 1U, 0U, 1U},
+          // x = 0x1.2f4p+2, exp10m1f16(x) = 0x1.ab8p+15 (RZ)
+          {0x44bdU, 0x7aaeU, 1U, 0U, 1U},
+#endif
+      }};
+#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+
+  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| <= 2^(-3), or |x| >= 11 * log10(2), or x is NaN.
+  if (LIBC_UNLIKELY(x_abs <= 0x3000U || x_abs >= 0x429fU)) {
+    // exp10m1(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 >= 16 * log10(2).
+    if (x_u >= 0x44d1U && x_bits.is_pos()) {
+      // exp10m1(+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 | FE_INEXACT);
+        return FPBits::inf().get_val();
+      default:
+        return FPBits::max_normal().get_val();
+      }
+    }
+
+    // When x < -11 * log10(2).
+    if (x_u > 0xc29fU) {
+      // exp10m1(-inf) = -1
+      if (x_bits.is_inf())
+        return FPBits::one(Sign::NEG).get_val();
+
+      // When x >= -0x1.ce4p+1, round(10^x - 1, HP, RN) = -0x1.ffcp-1.
+      if (x_u <= 0xc339U) {
+        return fputil::round_result_slightly_down(
+            fputil::cast<float16>(-0x1.ffcp-1));
+      }
+
+      // When x < -0x1.ce4p+1, round(10^x - 1, HP, RN) = -1.
+      switch (fputil::quick_get_round()) {
+      case FE_TONEAREST:
+      case FE_DOWNWARD:
+        return FPBits::one(Sign::NEG).get_val();
+      default:
+        return fputil::cast<float16>(-0x1.ffcp-1);
+      }
+    }
+
+    // When |x| <= 2^(-3).
+    if (x_abs <= 0x3000U) {
+      if (LIBC_UNLIKELY(x_abs == 0))
+        return x;
+
+#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+      if (auto r = EXP10M1F16_EXCEPTS_LO.lookup(x_u);
+          LIBC_UNLIKELY(r.has_value()))
+        return r.value();
+#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+
+      float xf = x;
+      // Degree-5 minimax polynomial generated by Sollya with the following
+      // commands:
+      //   > display = hexadecimal;
+      //   > P = fpminimax((10^x - 1)/x, 4, [|SG...|], [-2^-3, 2^-3]);
+      //   > x * P;
+      return fputil::cast<float16>(
+          xf * fputil::polyeval(xf, 0x1.26bb1cp+1f, 0x1.5351c8p+1f,
+                                0x1.04704p+1f, 0x1.2ce084p+0f, 0x1.14a6bep-1f));
+    }
+  }
+
+  // When x is 1, 2, or 3. These are hard-to-round cases with exact results.
+  // 10^4 - 1 = 9'999 is not exactly representable as a float16, but luckily the
+  // polynomial approximation gives the correct result for x = 4 in all
+  // rounding modes.
+  if (LIBC_UNLIKELY((x_u & ~(0x3c00U | 0x4000U | 0x4200U | 0x4400U)) == 0)) {
+    switch (x_u) {
+    case 0x3c00U: // x = 1.0f16
+      return fputil::cast<float16>(9.0);
+    case 0x4000U: // x = 2.0f16
+      return fputil::cast<float16>(99.0);
+    case 0x4200U: // x = 3.0f16
+      return fputil::cast<float16>(999.0);
+    }
+  }
+
+#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+  if (auto r = EXP10M1F16_EXCEPTS_HI.lookup(x_u); LIBC_UNLIKELY(r.has_value()))
+    return r.value();
+#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+
+  // exp10(x) = exp2((hi + mid) * log2(10)) * exp10(lo)
+  auto [exp2_hi_mid, exp10_lo] = exp10_range_reduction(x);
+  // exp10m1(x) = exp2((hi + mid) * log2(lo)) * exp10(lo) - 1
+  return fputil::cast<float16>(
+      fputil::multiply_add(exp2_hi_mid, exp10_lo, -1.0f));
+}
+
+} // namespace math
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_TYPES_HAS_FLOAT16
+
+#endif // LLVM_LIBC_SRC___SUPPORT_MATH_EXP10M1F16_H
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index 8074a3925626c..99c1b08326d53 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -1603,18 +1603,7 @@ add_entrypoint_object(
   HDRS
     ../exp10m1f16.h
   DEPENDS
-    libc.hdr.errno_macros
-    libc.hdr.fenv_macros
-    libc.src.__support.FPUtil.cast
-    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.polyeval
-    libc.src.__support.FPUtil.rounding_mode
-    libc.src.__support.macros.optimization
-    libc.src.__support.macros.properties.cpu_features
-    libc.src.__support.math.exp10f16_utils
+    libc.src.__support.math.exp10m1f16
 )
 
 add_entrypoint_object(
diff --git a/libc/src/math/generic/exp10m1f16.cpp b/libc/src/math/generic/exp10m1f16.cpp
index 6c2fdbea418df..8a3c4abf1f10e 100644
--- a/libc/src/math/generic/exp10m1f16.cpp
+++ b/libc/src/math/generic/exp10m1f16.cpp
@@ -7,166 +7,12 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/math/exp10m1f16.h"
-#include "hdr/errno_macros.h"
-#include "hdr/fenv_macros.h"
-#include "src/__support/FPUtil/FEnvImpl.h"
-#include "src/__support/FPUtil/FPBits.h"
-#include "src/__support/FPUtil/PolyEval.h"
-#include "src/__support/FPUtil/cast.h"
-#include "src/__support/FPUtil/except_value_utils.h"
-#include "src/__support/FPUtil/multiply_add.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"
-#include "src/__support/math/exp10f16_utils.h"
+#include "src/__support/math/exp10m1f16.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-static constexpr fputil::ExceptValues<float16, 3> EXP10M1F16_EXCEPTS_LO = {{
-    // (input, RZ output, RU offset, RD offset, RN offset)
-    // x = 0x1.5c4p-4, exp10m1f16(x) = 0x1.bacp-3 (RZ)
-    {0x2d71U, 0x32ebU, 1U, 0U, 0U},
-    // x = -0x1.5ep-13, exp10m1f16(x) = -0x1.92cp-12 (RZ)
-    {0x8978U, 0x8e4bU, 0U, 1U, 0U},
-    // x = -0x1.e2p-10, exp10m1f16(x) = -0x1.14cp-8 (RZ)
-    {0x9788U, 0x9c53U, 0U, 1U, 0U},
-}};
-
-#ifdef LIBC_TARGET_CPU_HAS_FMA_FLOAT
-static constexpr size_t N_EXP10M1F16_EXCEPTS_HI = 3;
-#else
-static constexpr size_t N_EXP10M1F16_EXCEPTS_HI = 6;
-#endif
-
-static constexpr fputil::ExceptValues<float16, N_EXP10M1F16_EXCEPTS_HI>
-    EXP10M1F16_EXCEPTS_HI = {{
-        // (input, RZ output, RU offset, RD offset, RN offset)
-        // x = 0x1.8f4p-2, exp10m1f16(x) = 0x1.744p+0 (RZ)
-        {0x363dU, 0x3dd1U, 1U, 0U, 0U},
-        // x = 0x1.95cp-2, exp10m1f16(x) = 0x1.7d8p+0 (RZ)
-        {0x3657U, 0x3df6U, 1U, 0U, 0U},
-        // x = 0x1.d04p-2, exp10m1f16(x) = 0x1.d7p+0 (RZ)
-        {0x3741U, 0x3f5cU, 1U, 0U, 1U},
-#ifndef LIBC_TARGET_CPU_HAS_FMA_FLOAT
-        // x = 0x1.0cp+1, exp10m1f16(x) = 0x1.ec4p+6 (RZ)
-        {0x4030U, 0x57b1U, 1U, 0U, 1U},
-        // x = 0x1.1b8p+1, exp10m1f16(x) = 0x1.45cp+7 (RZ)
-        {0x406eU, 0x5917U, 1U, 0U, 1U},
-        // x = 0x1.2f4p+2, exp10m1f16(x) = 0x1.ab8p+15 (RZ)
-        {0x44bdU, 0x7aaeU, 1U, 0U, 1U},
-#endif
-    }};
-#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-
 LLVM_LIBC_FUNCTION(float16, exp10m1f16, (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| <= 2^(-3), or |x| >= 11 * log10(2), or x is NaN.
-  if (LIBC_UNLIKELY(x_abs <= 0x3000U || x_abs >= 0x429fU)) {
-    // exp10m1(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 >= 16 * log10(2).
-    if (x_u >= 0x44d1U && x_bits.is_pos()) {
-      // exp10m1(+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 | FE_INEXACT);
-        return FPBits::inf().get_val();
-      default:
-        return FPBits::max_normal().get_val();
-      }
-    }
-
-    // When x < -11 * log10(2).
-    if (x_u > 0xc29fU) {
-      // exp10m1(-inf) = -1
-      if (x_bits.is_inf())
-        return FPBits::one(Sign::NEG).get_val();
-
-      // When x >= -0x1.ce4p+1, round(10^x - 1, HP, RN) = -0x1.ffcp-1.
-      if (x_u <= 0xc339U) {
-        return fputil::round_result_slightly_down(
-            fputil::cast<float16>(-0x1.ffcp-1));
-      }
-
-      // When x < -0x1.ce4p+1, round(10^x - 1, HP, RN) = -1.
-      switch (fputil::quick_get_round()) {
-      case FE_TONEAREST:
-      case FE_DOWNWARD:
-        return FPBits::one(Sign::NEG).get_val();
-      default:
-        return fputil::cast<float16>(-0x1.ffcp-1);
-      }
-    }
-
-    // When |x| <= 2^(-3).
-    if (x_abs <= 0x3000U) {
-      if (LIBC_UNLIKELY(x_abs == 0))
-        return x;
-
-#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-      if (auto r = EXP10M1F16_EXCEPTS_LO.lookup(x_u);
-          LIBC_UNLIKELY(r.has_value()))
-        return r.value();
-#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-
-      float xf = x;
-      // Degree-5 minimax polynomial generated by Sollya with the following
-      // commands:
-      //   > display = hexadecimal;
-      //   > P = fpminimax((10^x - 1)/x, 4, [|SG...|], [-2^-3, 2^-3]);
-      //   > x * P;
-      return fputil::cast<float16>(
-          xf * fputil::polyeval(xf, 0x1.26bb1cp+1f, 0x1.5351c8p+1f,
-                                0x1.04704p+1f, 0x1.2ce084p+0f, 0x1.14a6bep-1f));
-    }
-  }
-
-  // When x is 1, 2, or 3. These are hard-to-round cases with exact results.
-  // 10^4 - 1 = 9'999 is not exactly representable as a float16, but luckily the
-  // polynomial approximation gives the correct result for x = 4 in all
-  // rounding modes.
-  if (LIBC_UNLIKELY((x_u & ~(0x3c00U | 0x4000U | 0x4200U | 0x4400U)) == 0)) {
-    switch (x_u) {
-    case 0x3c00U: // x = 1.0f16
-      return fputil::cast<float16>(9.0);
-    case 0x4000U: // x = 2.0f16
-      return fputil::cast<float16>(99.0);
-    case 0x4200U: // x = 3.0f16
-      return fputil::cast<float16>(999.0);
-    }
-  }
-
-#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-  if (auto r = EXP10M1F16_EXCEPTS_HI.lookup(x_u); LIBC_UNLIKELY(r.has_value()))
-    return r.value();
-#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-
-  // exp10(x) = exp2((hi + mid) * log2(10)) * exp10(lo)
-  auto [exp2_hi_mid, exp10_lo] = exp10_range_reduction(x);
-  // exp10m1(x) = exp2((hi + mid) * log2(lo)) * exp10(lo) - 1
-  return fputil::cast<float16>(
-      fputil::multiply_add(exp2_hi_mid, exp10_lo, -1.0f));
+  return math::exp10m1f16(x);
 }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/test/shared/CMakeLists.txt b/libc/test/shared/CMakeLists.txt
index 13a0aae5d4c67..ea4634cbe7f9f 100644
--- a/libc/test/shared/CMakeLists.txt
+++ b/libc/test/shared/CMakeLists.txt
@@ -37,6 +37,7 @@ add_fp_unittest(
     libc.src.__support.math.cospif16
     libc.src.__support.math.dsqrtl
     libc.src.__support.math.exp10m1f
+    libc.src.__support.math.exp10m1f16
     libc.src.__support.math.erff
     libc.src.__support.math.exp
     libc.src.__support.math.exp10
diff --git a/libc/test/shared/shared_math_test.cpp b/libc/test/shared/shared_math_test.cpp
index 25bf5ad8ae411..17221932927b0 100644
--- a/libc/test/shared/shared_math_test.cpp
+++ b/libc/test/shared/shared_math_test.cpp
@@ -27,6 +27,7 @@ TEST(LlvmLibcSharedMathTest, AllFloat16) {
   EXPECT_FP_EQ(0x1p+0f16, LIBC_NAMESPACE::shared::coshf16(0.0f16));
   EXPECT_FP_EQ(0x1p+0f16, LIBC_NAMESPACE::shared::cospif16(0.0f16));
   EXPECT_FP_EQ(0x1p+0f16, LIBC_NAMESPACE::shared::exp10f16(0.0f16));
+  EXPECT_FP_EQ(0x0p+0f16, LIBC_NAMESPACE::shared::exp10m1f16(0.0f16));
 
   EXPECT_FP_EQ(0x1p+0f16, LIBC_NAMESPACE::shared::expf16(0.0f16));
 
diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index 8d9e80393bf20..e57d9dea036dd 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -2673,6 +2673,22 @@ libc_support_library(
     ],
 )
 
+libc_support_library(
+    name = "__support_math_exp10m1f16",
+    hdrs = ["src/__support/math/exp10m1f16.h"],
+    deps = [
+        ":__support_fputil_except_value_utils",
+        ":__support_fputil_fenv_impl",
+        ":__support_fputil_fp_bits",
+        ":__support_fputil_multiply_add",
+        ":__support_fputil_polyeval",
+        ":__support_fputil_rounding_mode",
+        ":__support_macros_optimization",
+        ":__support_math_exp10f16_utils",
+        ":errno",
+    ],
+)
+
 libc_support_library(
     name = "__support_math_erff",
     hdrs = ["src/__support/math/erff.h"],
@@ -3622,7 +3638,7 @@ libc_math_function(
 libc_math_function(
     name = "exp10m1f16",
     additional_deps = [
-        ":__support_math_exp10f16_utils",
+        ":__support_math_exp10m1f16",
     ],
 )

bassiounix · 2025-10-01T08:09:30Z

Merge activity

Oct 1, 8:09 AM UTC: A user started a stack merge that includes this pull request via Graphite.
Oct 1, 8:12 AM UTC: Graphite rebased this pull request as part of a merge.
Oct 1, 8:15 AM UTC: Graphite rebased this pull request as part of a merge.
Oct 1, 8:21 AM UTC: @bassiounix merged this pull request with Graphite.

…/__support/math folder.

…/__support/math folder. (llvm#161119) Part of llvm#147386 in preparation for: https://discourse.llvm.org/t/rfc-make-clang-builtin-math-functions-constexpr-with-llvm-libc-to-support-c-23-constexpr-math-functions/86450

bassiounix requested a review from lntue September 29, 2025 04:04

bassiounix added bazel "Peripheral" support tier build system: utils/bazel libc labels Sep 29, 2025 — with Graphite App

bassiounix marked this pull request as ready for review September 29, 2025 04:04

bassiounix requested review from rupprecht, keith and aaronmondal September 29, 2025 04:04

bassiounix removed request for keith, rupprecht and aaronmondal September 29, 2025 04:05

bassiounix mentioned this pull request Sep 29, 2025

[libc][math] Tracking Issue - Refactor Higher Math Functions to Header Only #147386

Open

bassiounix force-pushed the users/bassiounix/spr/09-20-_libc_math_refactor_exp10m1f_implementation_to_header-only_in_src___support_math_folder branch from e09e702 to 9f306b4 Compare September 29, 2025 04:44

Base automatically changed from users/bassiounix/spr/09-20-_libc_math_refactor_exp10m1f_implementation_to_header-only_in_src___support_math_folder to main September 29, 2025 04:58

bassiounix force-pushed the users/bassiounix/spr/09-29-_libc_math_refactor_exp10m1f16_implementation_to_header-only_in_src___support_math_folder branch 2 times, most recently from 7d75309 to 6667925 Compare September 29, 2025 05:08

bassiounix mentioned this pull request Sep 29, 2025

[libc][math] Refactor exp2 implementation to header-only in src/__support/math folder. #161297

Merged

lntue approved these changes Oct 1, 2025

View reviewed changes

bassiounix force-pushed the users/bassiounix/spr/09-29-_libc_math_refactor_exp10m1f16_implementation_to_header-only_in_src___support_math_folder branch from 6667925 to 4843827 Compare October 1, 2025 08:11

[libc][math] Refactor exp10m1f16 implementation to header-only in src…

7b6d71c

…/__support/math folder.

bassiounix force-pushed the users/bassiounix/spr/09-29-_libc_math_refactor_exp10m1f16_implementation_to_header-only_in_src___support_math_folder branch from 4843827 to 7b6d71c Compare October 1, 2025 08:15

bassiounix merged commit 6c032fa into main Oct 1, 2025
19 of 20 checks passed

bassiounix deleted the users/bassiounix/spr/09-29-_libc_math_refactor_exp10m1f16_implementation_to_header-only_in_src___support_math_folder branch October 1, 2025 08:21

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[libc][math] Refactor exp10m1f16 implementation to header-only in src/__support/math folder. #161119

[libc][math] Refactor exp10m1f16 implementation to header-only in src/__support/math folder. #161119

bassiounix commented Sep 29, 2025 •

edited

Loading

Uh oh!

bassiounix commented Sep 29, 2025 •

edited

Loading

Uh oh!

llvmbot commented Sep 29, 2025

Uh oh!

bassiounix commented Oct 1, 2025 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

[libc][math] Refactor exp10m1f16 implementation to header-only in src/__support/math folder. #161119

[libc][math] Refactor exp10m1f16 implementation to header-only in src/__support/math folder. #161119

Conversation

bassiounix commented Sep 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bassiounix commented Sep 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

llvmbot commented Sep 29, 2025

Uh oh!

bassiounix commented Oct 1, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merge activity

Uh oh!

Uh oh!

Uh oh!

bassiounix commented Sep 29, 2025 •

edited

Loading

bassiounix commented Sep 29, 2025 •

edited

Loading

bassiounix commented Oct 1, 2025 •

edited

Loading