Skip to content

[libc][math] Refactor atanf16 implementation to header-only in src/__support/math folder. #150868

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Conversation

bassiounix
Copy link
Contributor

Copy link
Contributor Author

bassiounix commented Jul 28, 2025

@llvmbot
Copy link
Member

llvmbot commented Jul 28, 2025

@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/150868.diff

9 Files Affected:

  • (modified) libc/shared/math.h (+1)
  • (added) libc/shared/math/atanf16.h (+28)
  • (modified) libc/src/__support/math/CMakeLists.txt (+15)
  • (added) libc/src/__support/math/atanf16.h (+119)
  • (modified) libc/src/math/generic/CMakeLists.txt (+1-11)
  • (modified) libc/src/math/generic/atanf16.cpp (+2-93)
  • (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 (+22)
diff --git a/libc/shared/math.h b/libc/shared/math.h
index 21536647948f4..bcbe0de56170a 100644
--- a/libc/shared/math.h
+++ b/libc/shared/math.h
@@ -24,6 +24,7 @@
 #include "math/asinhf16.h"
 #include "math/atan.h"
 #include "math/atanf.h"
+#include "math/atanf16.h"
 #include "math/erff.h"
 #include "math/exp.h"
 #include "math/exp10.h"
diff --git a/libc/shared/math/atanf16.h b/libc/shared/math/atanf16.h
new file mode 100644
index 0000000000000..f196907059e01
--- /dev/null
+++ b/libc/shared/math/atanf16.h
@@ -0,0 +1,28 @@
+//===-- Shared atanf16 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_ATANF16_H
+#define LLVM_LIBC_SHARED_MATH_ATANF16_H
+
+#include "shared/libc_common.h"
+
+#ifdef LIBC_TYPES_HAS_FLOAT16
+
+#include "src/__support/math/atanf16.h"
+
+namespace LIBC_NAMESPACE_DECL {
+namespace shared {
+
+using math::atanf16;
+
+} // namespace shared
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_TYPES_HAS_FLOAT16
+
+#endif // LLVM_LIBC_SHARED_MATH_ATANF16_H
diff --git a/libc/src/__support/math/CMakeLists.txt b/libc/src/__support/math/CMakeLists.txt
index 95acc962cc885..04cbd3fd1cc01 100644
--- a/libc/src/__support/math/CMakeLists.txt
+++ b/libc/src/__support/math/CMakeLists.txt
@@ -214,6 +214,21 @@ add_header_library(
     libc.src.__support.macros.optimization
 )
 
+add_header_library(
+  atanf16
+  HDRS
+    atanf16.h
+  DEPENDS
+    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.sqrt
+    libc.src.__support.macros.optimization
+)
+
 add_header_library(
   asinf
   HDRS
diff --git a/libc/src/__support/math/atanf16.h b/libc/src/__support/math/atanf16.h
new file mode 100644
index 0000000000000..f75d145f36852
--- /dev/null
+++ b/libc/src/__support/math/atanf16.h
@@ -0,0 +1,119 @@
+//===-- Implementation header for atanf16 -----------------------*- 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_ATANF16_H
+#define LLVM_LIBC_SRC___SUPPORT_MATH_ATANF16_H
+
+#include "include/llvm-libc-macros/float16-macros.h"
+
+#ifdef LIBC_TYPES_HAS_FLOAT16
+
+#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/sqrt.h"
+#include "src/__support/macros/optimization.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+namespace math {
+
+LIBC_INLINE static constexpr float16 atanf16(float16 x) {
+  // Generated by Solly using the following command:
+  // > round(pi/2, SG, RN);
+  constexpr float PI_2 = 0x1.921fb6p0;
+
+#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+  constexpr size_t N_EXCEPTS = 6;
+
+  constexpr fputil::ExceptValues<float16, N_EXCEPTS> ATANF16_EXCEPTS{{
+      // (input, RZ output, RU offset, RD offset, RN offset)
+      {0x2745, 0x2744, 1, 0, 1},
+      {0x3099, 0x3090, 1, 0, 1},
+      {0x3c6c, 0x3aae, 1, 0, 1},
+      {0x466e, 0x3daa, 1, 0, 1},
+      {0x48ae, 0x3ddb, 1, 0, 0},
+      {0x5619, 0x3e3d, 1, 0, 1},
+  }};
+#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+
+  using FPBits = fputil::FPBits<float16>;
+  FPBits xbits(x);
+
+  uint16_t x_u = xbits.uintval();
+  uint16_t x_abs = x_u & 0x7fff;
+  bool x_sign = x_u >> 15;
+  float sign = (x_sign ? -1.0 : 1.0);
+
+  // |x| >= +/-inf
+  if (LIBC_UNLIKELY(x_abs >= 0x7c00)) {
+    if (xbits.is_nan()) {
+      if (xbits.is_signaling_nan()) {
+        fputil::raise_except_if_required(FE_INVALID);
+        return FPBits::quiet_nan().get_val();
+      }
+      return x;
+    }
+
+    // atanf16(+/-inf) = +/-pi/2
+    return fputil::cast<float16>(sign * PI_2);
+  }
+
+  float xf = x;
+  float xsq = xf * xf;
+#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+  // Handle exceptional values
+  if (auto r = ATANF16_EXCEPTS.lookup_odd(x_abs, x_sign);
+      LIBC_UNLIKELY(r.has_value()))
+    return r.value();
+#endif
+
+  // |x| <= 0x1p0, |x| <= 1
+  if (x_abs <= 0x3c00) {
+    // atanf16(+/-0) = +/-0
+    if (LIBC_UNLIKELY(x_abs == 0))
+      return x;
+
+    // Degree-14 minimax odd polynomial of atan(x) generated by Sollya with:
+    // > P = fpminimax(atan(x)/x, [|0, 2, 4, 6, 8, 10, 12, 14|], [|SG...|],
+    // [0, 1]);
+    float result = fputil::polyeval(
+        xsq, 0x1.fffffcp-1f, -0x1.55519ep-2f, 0x1.98f6a8p-3f, -0x1.1f0a92p-3f,
+        0x1.95b654p-4f, -0x1.e65492p-5f, 0x1.8c0c36p-6f, -0x1.32316ep-8f);
+    return fputil::cast<float16>(xf * result);
+  }
+
+  // If |x| > 1
+  // y = atan(x) = sign(x) * atan(|x|)
+  // atan(|x|) = pi/2 - atan(1/|x|)
+  // Recall, 1/|x| < 1
+  float x_inv_sq = 1.0f / xsq;
+  float x_inv = fputil::sqrt<float>(x_inv_sq);
+
+  // Degree-14 minimax odd polynomial of atan(x) generated by Sollya with:
+  // > P = fpminimax(atan(x)/x, [|0, 2, 4, 6, 8, 10, 12, 14|], [|SG...|],
+  // [0, 1]);
+  float interm =
+      fputil::polyeval(x_inv_sq, 0x1.fffffcp-1f, -0x1.55519ep-2f,
+                       0x1.98f6a8p-3f, -0x1.1f0a92p-3f, 0x1.95b654p-4f,
+                       -0x1.e65492p-5f, 0x1.8c0c36p-6f, -0x1.32316ep-8f);
+
+  return fputil::cast<float16>(sign *
+                               fputil::multiply_add(x_inv, -interm, PI_2));
+}
+
+} // namespace math
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_TYPES_HAS_FLOAT16
+
+#endif // LLVM_LIBC_SRC___SUPPORT_MATH_ATANF16_H
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index 196e00841f2d1..007e5701b7538 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -4024,17 +4024,7 @@ add_entrypoint_object(
   HDRS
     ../atanf16.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.sqrt
-    libc.src.__support.macros.optimization
-    libc.src.__support.macros.properties.types
+    libc.src.__support.math.atanf16
 )
 
 add_entrypoint_object(
diff --git a/libc/src/math/generic/atanf16.cpp b/libc/src/math/generic/atanf16.cpp
index 9b6ec657417a5..7191c4242160a 100644
--- a/libc/src/math/generic/atanf16.cpp
+++ b/libc/src/math/generic/atanf16.cpp
@@ -7,101 +7,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/math/atanf16.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/sqrt.h"
-#include "src/__support/macros/optimization.h"
+#include "src/__support/math/atanf16.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-// Generated by Solly using the following command:
-// > round(pi/2, SG, RN);
-static constexpr float PI_2 = 0x1.921fb6p0;
-
-#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-static constexpr size_t N_EXCEPTS = 6;
-
-static constexpr fputil::ExceptValues<float16, N_EXCEPTS> ATANF16_EXCEPTS{{
-    // (input, RZ output, RU offset, RD offset, RN offset)
-    {0x2745, 0x2744, 1, 0, 1},
-    {0x3099, 0x3090, 1, 0, 1},
-    {0x3c6c, 0x3aae, 1, 0, 1},
-    {0x466e, 0x3daa, 1, 0, 1},
-    {0x48ae, 0x3ddb, 1, 0, 0},
-    {0x5619, 0x3e3d, 1, 0, 1},
-}};
-#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-
-LLVM_LIBC_FUNCTION(float16, atanf16, (float16 x)) {
-  using FPBits = fputil::FPBits<float16>;
-  FPBits xbits(x);
-
-  uint16_t x_u = xbits.uintval();
-  uint16_t x_abs = x_u & 0x7fff;
-  bool x_sign = x_u >> 15;
-  float sign = (x_sign ? -1.0 : 1.0);
-
-  // |x| >= +/-inf
-  if (LIBC_UNLIKELY(x_abs >= 0x7c00)) {
-    if (xbits.is_nan()) {
-      if (xbits.is_signaling_nan()) {
-        fputil::raise_except_if_required(FE_INVALID);
-        return FPBits::quiet_nan().get_val();
-      }
-      return x;
-    }
-
-    // atanf16(+/-inf) = +/-pi/2
-    return fputil::cast<float16>(sign * PI_2);
-  }
-
-  float xf = x;
-  float xsq = xf * xf;
-#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-  // Handle exceptional values
-  if (auto r = ATANF16_EXCEPTS.lookup_odd(x_abs, x_sign);
-      LIBC_UNLIKELY(r.has_value()))
-    return r.value();
-#endif
-
-  // |x| <= 0x1p0, |x| <= 1
-  if (x_abs <= 0x3c00) {
-    // atanf16(+/-0) = +/-0
-    if (LIBC_UNLIKELY(x_abs == 0))
-      return x;
-
-    // Degree-14 minimax odd polynomial of atan(x) generated by Sollya with:
-    // > P = fpminimax(atan(x)/x, [|0, 2, 4, 6, 8, 10, 12, 14|], [|SG...|],
-    // [0, 1]);
-    float result = fputil::polyeval(
-        xsq, 0x1.fffffcp-1f, -0x1.55519ep-2f, 0x1.98f6a8p-3f, -0x1.1f0a92p-3f,
-        0x1.95b654p-4f, -0x1.e65492p-5f, 0x1.8c0c36p-6f, -0x1.32316ep-8f);
-    return fputil::cast<float16>(xf * result);
-  }
-
-  // If |x| > 1
-  // y = atan(x) = sign(x) * atan(|x|)
-  // atan(|x|) = pi/2 - atan(1/|x|)
-  // Recall, 1/|x| < 1
-  float x_inv_sq = 1.0f / xsq;
-  float x_inv = fputil::sqrt<float>(x_inv_sq);
-
-  // Degree-14 minimax odd polynomial of atan(x) generated by Sollya with:
-  // > P = fpminimax(atan(x)/x, [|0, 2, 4, 6, 8, 10, 12, 14|], [|SG...|],
-  // [0, 1]);
-  float interm =
-      fputil::polyeval(x_inv_sq, 0x1.fffffcp-1f, -0x1.55519ep-2f,
-                       0x1.98f6a8p-3f, -0x1.1f0a92p-3f, 0x1.95b654p-4f,
-                       -0x1.e65492p-5f, 0x1.8c0c36p-6f, -0x1.32316ep-8f);
-
-  return fputil::cast<float16>(sign *
-                               fputil::multiply_add(x_inv, -interm, PI_2));
-}
+LLVM_LIBC_FUNCTION(float16, atanf16, (float16 x)) { return math::atanf16(x); }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/test/shared/CMakeLists.txt b/libc/test/shared/CMakeLists.txt
index 6d0601feda138..2dd5bdafe24f6 100644
--- a/libc/test/shared/CMakeLists.txt
+++ b/libc/test/shared/CMakeLists.txt
@@ -20,6 +20,7 @@ add_fp_unittest(
     libc.src.__support.math.asinhf16
     libc.src.__support.math.atan
     libc.src.__support.math.atanf
+    libc.src.__support.math.atanf16
     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 228fa42ec834e..2b80068f991be 100644
--- a/libc/test/shared/shared_math_test.cpp
+++ b/libc/test/shared/shared_math_test.cpp
@@ -19,6 +19,7 @@ TEST(LlvmLibcSharedMathTest, AllFloat16) {
   EXPECT_FP_EQ(0x0p+0f16, LIBC_NAMESPACE::shared::acospif16(1.0f16));
   EXPECT_FP_EQ(0x0p+0f16, LIBC_NAMESPACE::shared::asinf16(0.0f16));
   EXPECT_FP_EQ(0x0p+0f16, LIBC_NAMESPACE::shared::asinhf16(0.0f16));
+  EXPECT_FP_EQ(0x0p+0f16, LIBC_NAMESPACE::shared::atanf16(0.0f16));
 
   EXPECT_FP_EQ(0x1p+0f16, LIBC_NAMESPACE::shared::exp10f16(0.0f16));
 
diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index e8c59fa87d4c3..77e4829f44758 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -2288,6 +2288,21 @@ libc_support_library(
     ],
 )
 
+libc_support_library(
+    name = "__support_math_atanf16",
+    hdrs = ["src/__support/math/atanf16.h"],
+    deps = [
+        ":__support_fputil_cast",
+        ":__support_fputil_except_value_utils",
+        ":__support_fputil_fenv_impl",
+        ":__support_fputil_fp_bits",
+        ":__support_fputil_polyeval",
+        ":__support_fputil_multiply_add",
+        ":__support_fputil_sqrt",
+        ":__support_macros_optimization",
+    ],
+)
+
 libc_support_library(
     name = "__support_math_asinf",
     hdrs = ["src/__support/math/asinf.h"],
@@ -2904,6 +2919,13 @@ libc_math_function(
     ],
 )
 
+libc_math_function(
+    name = "atanf16",
+    additional_deps = [
+        ":__support_math_atanf16"
+    ],
+)
+
 libc_math_function(
     name = "atan",
     additional_deps = [

@bassiounix bassiounix requested review from lntue and removed request for keith, rupprecht and aaronmondal July 28, 2025 02:29
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf_implementation_to_header-only_in_src___support_math_folder branch from 19f366d to ca83f9d Compare July 28, 2025 15:17
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf16_implementation_to_header-only_in_src___support_math_folder branch from 65ffa96 to e361d72 Compare July 28, 2025 15:18
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf_implementation_to_header-only_in_src___support_math_folder branch from ca83f9d to 9cac4fc Compare July 28, 2025 16:45
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf16_implementation_to_header-only_in_src___support_math_folder branch from e361d72 to 5077f3f Compare July 28, 2025 16:46
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf16_implementation_to_header-only_in_src___support_math_folder branch from 5077f3f to bd94b9d Compare July 28, 2025 17:33
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf_implementation_to_header-only_in_src___support_math_folder branch from 9cac4fc to 9fc891c Compare July 28, 2025 17:33
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf16_implementation_to_header-only_in_src___support_math_folder branch from bd94b9d to f4dade3 Compare July 28, 2025 18:21
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf_implementation_to_header-only_in_src___support_math_folder branch 2 times, most recently from ec6b90c to 64249f7 Compare July 29, 2025 17:21
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf16_implementation_to_header-only_in_src___support_math_folder branch 2 times, most recently from 72b3b89 to a491096 Compare July 29, 2025 17:51
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf_implementation_to_header-only_in_src___support_math_folder branch from 64249f7 to 3930b0f Compare July 29, 2025 17:51
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf16_implementation_to_header-only_in_src___support_math_folder branch from a491096 to 0f24379 Compare July 29, 2025 17:57
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf_implementation_to_header-only_in_src___support_math_folder branch 2 times, most recently from 26d6828 to fa8739a Compare July 29, 2025 18:18
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf16_implementation_to_header-only_in_src___support_math_folder branch from 0f24379 to 4cc63c7 Compare July 29, 2025 18:18
Base automatically changed from users/bassiounix/spr/07-28-_libc_math_refactor_atanf_implementation_to_header-only_in_src___support_math_folder to main July 30, 2025 04:24
@bassiounix bassiounix force-pushed the users/bassiounix/spr/07-28-_libc_math_refactor_atanf16_implementation_to_header-only_in_src___support_math_folder branch from 4cc63c7 to 8ca5b0c Compare July 30, 2025 04:26
Copy link
Contributor Author

bassiounix commented Jul 30, 2025

Merge activity

  • Jul 30, 8:29 PM UTC: A user started a stack merge that includes this pull request via Graphite.
  • Jul 30, 8:31 PM UTC: @bassiounix merged this pull request with Graphite.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bazel "Peripheral" support tier build system: utils/bazel libc
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants