[libc][math] Refactor cbrtf implementation to header-only in src/__support/math folder. #151846

bassiounix · 2025-08-03T04:05:41Z

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 · 2025-08-03T04:06:11Z

[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 👈 (View in Graphite)
[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 · 2025-08-03T04:06:28Z

@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/151846.diff

9 Files Affected:

(modified) libc/shared/math.h (+1)
(added) libc/shared/math/cbrtf.h (+23)
(modified) libc/src/__support/math/CMakeLists.txt (+11)
(added) libc/src/__support/math/cbrtf.h (+161)
(modified) libc/src/math/generic/CMakeLists.txt (+1-5)
(modified) libc/src/math/generic/cbrtf.cpp (+2-145)
(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 (+13-1)

diff --git a/libc/shared/math.h b/libc/shared/math.h
index 3714f380a27dc..ea645f0afedbc 100644
--- a/libc/shared/math.h
+++ b/libc/shared/math.h
@@ -31,6 +31,7 @@
 #include "math/atanhf.h"
 #include "math/atanhf16.h"
 #include "math/cbrt.h"
+#include "math/cbrtf.h"
 #include "math/erff.h"
 #include "math/exp.h"
 #include "math/exp10.h"
diff --git a/libc/shared/math/cbrtf.h b/libc/shared/math/cbrtf.h
new file mode 100644
index 0000000000000..09b86bed3fb7e
--- /dev/null
+++ b/libc/shared/math/cbrtf.h
@@ -0,0 +1,23 @@
+//===-- Shared cbrtf 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 LIBC_SHARED_MATH_CBRTF_H
+#define LIBC_SHARED_MATH_CBRTF_H
+
+#include "shared/libc_common.h"
+#include "src/__support/math/cbrtf.h"
+
+namespace LIBC_NAMESPACE_DECL {
+namespace shared {
+
+using math::cbrtf;
+
+} // namespace shared
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_SHARED_MATH_CBRTF_H
diff --git a/libc/src/__support/math/CMakeLists.txt b/libc/src/__support/math/CMakeLists.txt
index e1076edf1e61c..fe928a8fadd5e 100644
--- a/libc/src/__support/math/CMakeLists.txt
+++ b/libc/src/__support/math/CMakeLists.txt
@@ -346,6 +346,17 @@ add_header_library(
     libc.src.__support.integer_literals
 )
 
+add_header_library(
+  cbrtf
+  HDRS
+    cbrtf.h
+  DEPENDS
+    libc.src.__support.FPUtil.fenv_impl
+    libc.src.__support.FPUtil.fp_bits
+    libc.src.__support.FPUtil.multiply_add
+    libc.src.__support.macros.optimization
+)
+
 add_header_library(
   erff
   HDRS
diff --git a/libc/src/__support/math/cbrtf.h b/libc/src/__support/math/cbrtf.h
new file mode 100644
index 0000000000000..f82892bbbe61b
--- /dev/null
+++ b/libc/src/__support/math/cbrtf.h
@@ -0,0 +1,161 @@
+//===-- Implementation header for cbrtf -------------------------*- 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 LIBC_SRC___SUPPORT_MATH_CBRTF_H
+#define LIBC_SRC___SUPPORT_MATH_CBRTF_H
+
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/__support/FPUtil/multiply_add.h"
+#include "src/__support/macros/config.h"
+#include "src/__support/macros/optimization.h" // LIBC_UNLIKELY
+
+namespace LIBC_NAMESPACE_DECL {
+
+namespace math {
+
+LIBC_INLINE static constexpr float cbrtf(float x) {
+  // Look up table for 2^(i/3) for i = 0, 1, 2.
+  constexpr double CBRT2[3] = {1.0, 0x1.428a2f98d728bp0, 0x1.965fea53d6e3dp0};
+
+  // Degree-7 polynomials approximation of ((1 + x)^(1/3) - 1)/x for 0 <= x <= 1
+  // generated by Sollya with:
+  // > for i from 0 to 15 do {
+  //     P = fpminimax(((1 + x)^(1/3) - 1)/x, 6, [|D...|], [i/16, (i + 1)/16]);
+  //     print("{", coeff(P, 0), ",", coeff(P, 1), ",", coeff(P, 2), ",",
+  //           coeff(P, 3), ",", coeff(P, 4), ",", coeff(P, 5), ",",
+  //           coeff(P, 6), "},");
+  // };
+  // Then (1 + x)^(1/3) ~ 1 + x * P(x).
+  constexpr double COEFFS[16][7] = {
+      {0x1.55555555554ebp-2, -0x1.c71c71c678c0cp-4, 0x1.f9add2776de81p-5,
+       -0x1.511e10aa964a7p-5, 0x1.ee44165937fa2p-6, -0x1.7c5c9e059345dp-6,
+       0x1.047f75e0aff14p-6},
+      {0x1.5555554d1149ap-2, -0x1.c71c676fcb5bp-4, 0x1.f9ab127dc57ebp-5,
+       -0x1.50ea8fd1d4c15p-5, 0x1.e9d68f28ced43p-6, -0x1.60e0e1e661311p-6,
+       0x1.716eca1d6e3bcp-7},
+      {0x1.5555546377d45p-2, -0x1.c71bc1c6d49d2p-4, 0x1.f9924cc0ed24dp-5,
+       -0x1.4fea3beb53b3bp-5, 0x1.de028a9a07b1bp-6, -0x1.3b090d2233524p-6,
+       0x1.0aeca34893785p-7},
+      {0x1.55554dce9f649p-2, -0x1.c7188b34b98f8p-4, 0x1.f93e1af34af49p-5,
+       -0x1.4d9a06be75c63p-5, 0x1.cb943f4f68992p-6, -0x1.139a685a5e3c4p-6,
+       0x1.88410674c6a5dp-8},
+      {0x1.5555347d211c3p-2, -0x1.c70f2a4b1a5fap-4, 0x1.f88420e8602c3p-5,
+       -0x1.49becfa4ed3ep-5, 0x1.b475cd9013162p-6, -0x1.dcfee1dd2f8efp-7,
+       0x1.249bb51a1c498p-8},
+      {0x1.5554f01b33dbap-2, -0x1.c6facb929dbf1p-4, 0x1.f73fb7861252ep-5,
+       -0x1.4459a4a0071fap-5, 0x1.9a8df2b504fc2p-6, -0x1.9a7ce3006d06ep-7,
+       0x1.ba9230918fa2ep-9},
+      {0x1.55545c695db5fp-2, -0x1.c6d6089f20275p-4, 0x1.f556e0ea80efp-5,
+       -0x1.3d91372d083f4p-5, 0x1.7f66cff331f4p-6, -0x1.606a562491737p-7,
+       0x1.52e3e17c71069p-9},
+      {0x1.55534a879232ap-2, -0x1.c69b836998b84p-4, 0x1.f2bb26dac0e4cp-5,
+       -0x1.359eed43716d7p-5, 0x1.64218cd824fbcp-6, -0x1.2e703e2e091e8p-7,
+       0x1.0677d9af6aad4p-9},
+      {0x1.5551836bb5494p-2, -0x1.c64658c15353bp-4, 0x1.ef68517451a6ep-5,
+       -0x1.2cc20a980dceep-5, 0x1.49843e0fad93ap-6, -0x1.03c59ccb68e54p-7,
+       0x1.9ad325dc7adcbp-10},
+      {0x1.554ecacb0d035p-2, -0x1.c5d2664026ffcp-4, 0x1.eb624796ba809p-5,
+       -0x1.233803d19a535p-5, 0x1.300decb1c3c28p-6, -0x1.befe18031ec3dp-8,
+       0x1.449f5ee175c69p-10},
+      {0x1.554ae1f5ae815p-2, -0x1.c53c6b14ff6b2p-4, 0x1.e6b2d5127bb5bp-5,
+       -0x1.19387336788a3p-5, 0x1.180955a6ab255p-6, -0x1.81696703ba369p-8,
+       0x1.02cb36389bd79p-10},
+      {0x1.55458a59f356ep-2, -0x1.c4820dd631ae9p-4, 0x1.e167af818bd15p-5,
+       -0x1.0ef35f6f72e52p-5, 0x1.019c33b65e4ebp-6, -0x1.4d25bdd52d3a5p-8,
+       0x1.a008ae91f5936p-11},
+      {0x1.553e878eafee1p-2, -0x1.c3a1d0b2a3db2p-4, 0x1.db90d8ed9f89bp-5,
+       -0x1.0490e20f1ae91p-5, 0x1.d9a5d1fc42fe3p-7, -0x1.20bf8227c2abfp-8,
+       0x1.50f8174cdb6e9p-11},
+      {0x1.5535a0dedf1b1p-2, -0x1.c29afb8bd01a1p-4, 0x1.d53f6371c1e27p-5,
+       -0x1.f463209b433e2p-6, 0x1.b35222a17e44p-7, -0x1.f5efbf505e133p-9,
+       0x1.12e0e94e8586dp-11},
+      {0x1.552aa25e57bfdp-2, -0x1.c16d811e4acadp-4, 0x1.ce8489b47aa51p-5,
+       -0x1.dfde7ff758ea8p-6, 0x1.901f43aac38c8p-7, -0x1.b581d07df5ad5p-9,
+       0x1.c3726535f1fc6p-12},
+      {0x1.551d5d9b204d3p-2, -0x1.c019e328f8db1p-4, 0x1.c7710f44fc3cep-5,
+       -0x1.cbbbe25ea8ba4p-6, 0x1.6fe270088623dp-7, -0x1.7e6fc79733761p-9,
+       0x1.75077abf18d84p-12},
+  };
+
+  using FloatBits = typename fputil::FPBits<float>;
+  using DoubleBits = typename fputil::FPBits<double>;
+
+  FloatBits x_bits(x);
+
+  uint32_t x_abs = x_bits.uintval() & 0x7fff'ffff;
+  uint32_t sign_bit = (x_bits.uintval() >> 31) << DoubleBits::EXP_LEN;
+
+  if (LIBC_UNLIKELY(x == 0.0f || x_abs >= 0x7f80'0000)) {
+    // x is 0, Inf, or NaN.
+    // Make sure it works for FTZ/DAZ modes.
+    return x + x;
+  }
+
+  double xd = static_cast<double>(x);
+  DoubleBits xd_bits(xd);
+
+  // When using biased exponent of x in double precision,
+  //   x_e = real_exponent_of_x + 1023
+  // Then:
+  //   x_e / 3 = real_exponent_of_x / 3 + 1023/3
+  //           = real_exponent_of_x / 3 + 341
+  // So to make it the correct biased exponent of x^(1/3), we add
+  //   1023 - 341 = 682
+  // to the quotient x_e / 3.
+  unsigned x_e = static_cast<unsigned>(xd_bits.get_biased_exponent());
+  unsigned out_e = (x_e / 3 + 682) | sign_bit;
+  unsigned shift_e = x_e % 3;
+
+  // Set x_m = 2^(x_e % 3) * (1.mantissa)
+  uint64_t x_m = xd_bits.get_mantissa();
+  // Use the leading 4 bits for look up table
+  unsigned idx = static_cast<unsigned>(x_m >> (DoubleBits::FRACTION_LEN - 4));
+
+  x_m |= static_cast<uint64_t>(DoubleBits::EXP_BIAS)
+         << DoubleBits::FRACTION_LEN;
+
+  double x_reduced = DoubleBits(x_m).get_val();
+  double dx = x_reduced - 1.0;
+
+  double dx_sq = dx * dx;
+  double c0 = fputil::multiply_add(dx, COEFFS[idx][0], 1.0);
+  double c1 = fputil::multiply_add(dx, COEFFS[idx][2], COEFFS[idx][1]);
+  double c2 = fputil::multiply_add(dx, COEFFS[idx][4], COEFFS[idx][3]);
+  double c3 = fputil::multiply_add(dx, COEFFS[idx][6], COEFFS[idx][5]);
+
+  double dx_4 = dx_sq * dx_sq;
+  double p0 = fputil::multiply_add(dx_sq, c1, c0);
+  double p1 = fputil::multiply_add(dx_sq, c3, c2);
+
+  double r = fputil::multiply_add(dx_4, p1, p0) * CBRT2[shift_e];
+
+  uint64_t r_m = DoubleBits(r).get_mantissa();
+  // Check if the output is exact.  To be exact, the smallest 1-bit of the
+  // output has to be at least 2^-7 or higher.  So we check the lowest 44 bits
+  // to see if they are within 2^(-52 + 3) errors from all zeros, then the
+  // result cube root is exact.
+  if (LIBC_UNLIKELY(((r_m + 8) & 0xfffffffffff) <= 16)) {
+    if ((r_m & 0xfffffffffff) <= 8)
+      r_m &= 0xffff'ffff'ffff'ffe0;
+    else
+      r_m = (r_m & 0xffff'ffff'ffff'ffe0) + 0x20;
+    fputil::clear_except_if_required(FE_INEXACT);
+  }
+  // Adjust exponent and sign.
+  uint64_t r_bits =
+      r_m | (static_cast<uint64_t>(out_e) << DoubleBits::FRACTION_LEN);
+
+  return static_cast<float>(DoubleBits(r_bits).get_val());
+}
+
+} // namespace math
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_SRC___SUPPORT_MATH_CBRTF_H
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index a86619576cfc6..44c17cf5b0770 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -4739,11 +4739,7 @@ add_entrypoint_object(
   HDRS
     ../cbrtf.h
   DEPENDS
-    libc.hdr.fenv_macros
-    libc.src.__support.FPUtil.fenv_impl
-    libc.src.__support.FPUtil.fp_bits
-    libc.src.__support.FPUtil.multiply_add
-    libc.src.__support.macros.optimization
+    libc.src.__support.math.cbrtf
 )
 
 add_entrypoint_object(
diff --git a/libc/src/math/generic/cbrtf.cpp b/libc/src/math/generic/cbrtf.cpp
index 71b23c4a8c742..0bd8f71365300 100644
--- a/libc/src/math/generic/cbrtf.cpp
+++ b/libc/src/math/generic/cbrtf.cpp
@@ -7,153 +7,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/math/cbrtf.h"
-#include "hdr/fenv_macros.h"
-#include "src/__support/FPUtil/FEnvImpl.h"
-#include "src/__support/FPUtil/FPBits.h"
-#include "src/__support/FPUtil/multiply_add.h"
-#include "src/__support/common.h"
-#include "src/__support/macros/config.h"
-#include "src/__support/macros/optimization.h" // LIBC_UNLIKELY
+#include "src/__support/math/cbrtf.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-namespace {
-
-// Look up table for 2^(i/3) for i = 0, 1, 2.
-constexpr double CBRT2[3] = {1.0, 0x1.428a2f98d728bp0, 0x1.965fea53d6e3dp0};
-
-// Degree-7 polynomials approximation of ((1 + x)^(1/3) - 1)/x for 0 <= x <= 1
-// generated by Sollya with:
-// > for i from 0 to 15 do {
-//     P = fpminimax(((1 + x)^(1/3) - 1)/x, 6, [|D...|], [i/16, (i + 1)/16]);
-//     print("{", coeff(P, 0), ",", coeff(P, 1), ",", coeff(P, 2), ",",
-//           coeff(P, 3), ",", coeff(P, 4), ",", coeff(P, 5), ",",
-//           coeff(P, 6), "},");
-// };
-// Then (1 + x)^(1/3) ~ 1 + x * P(x).
-constexpr double COEFFS[16][7] = {
-    {0x1.55555555554ebp-2, -0x1.c71c71c678c0cp-4, 0x1.f9add2776de81p-5,
-     -0x1.511e10aa964a7p-5, 0x1.ee44165937fa2p-6, -0x1.7c5c9e059345dp-6,
-     0x1.047f75e0aff14p-6},
-    {0x1.5555554d1149ap-2, -0x1.c71c676fcb5bp-4, 0x1.f9ab127dc57ebp-5,
-     -0x1.50ea8fd1d4c15p-5, 0x1.e9d68f28ced43p-6, -0x1.60e0e1e661311p-6,
-     0x1.716eca1d6e3bcp-7},
-    {0x1.5555546377d45p-2, -0x1.c71bc1c6d49d2p-4, 0x1.f9924cc0ed24dp-5,
-     -0x1.4fea3beb53b3bp-5, 0x1.de028a9a07b1bp-6, -0x1.3b090d2233524p-6,
-     0x1.0aeca34893785p-7},
-    {0x1.55554dce9f649p-2, -0x1.c7188b34b98f8p-4, 0x1.f93e1af34af49p-5,
-     -0x1.4d9a06be75c63p-5, 0x1.cb943f4f68992p-6, -0x1.139a685a5e3c4p-6,
-     0x1.88410674c6a5dp-8},
-    {0x1.5555347d211c3p-2, -0x1.c70f2a4b1a5fap-4, 0x1.f88420e8602c3p-5,
-     -0x1.49becfa4ed3ep-5, 0x1.b475cd9013162p-6, -0x1.dcfee1dd2f8efp-7,
-     0x1.249bb51a1c498p-8},
-    {0x1.5554f01b33dbap-2, -0x1.c6facb929dbf1p-4, 0x1.f73fb7861252ep-5,
-     -0x1.4459a4a0071fap-5, 0x1.9a8df2b504fc2p-6, -0x1.9a7ce3006d06ep-7,
-     0x1.ba9230918fa2ep-9},
-    {0x1.55545c695db5fp-2, -0x1.c6d6089f20275p-4, 0x1.f556e0ea80efp-5,
-     -0x1.3d91372d083f4p-5, 0x1.7f66cff331f4p-6, -0x1.606a562491737p-7,
-     0x1.52e3e17c71069p-9},
-    {0x1.55534a879232ap-2, -0x1.c69b836998b84p-4, 0x1.f2bb26dac0e4cp-5,
-     -0x1.359eed43716d7p-5, 0x1.64218cd824fbcp-6, -0x1.2e703e2e091e8p-7,
-     0x1.0677d9af6aad4p-9},
-    {0x1.5551836bb5494p-2, -0x1.c64658c15353bp-4, 0x1.ef68517451a6ep-5,
-     -0x1.2cc20a980dceep-5, 0x1.49843e0fad93ap-6, -0x1.03c59ccb68e54p-7,
-     0x1.9ad325dc7adcbp-10},
-    {0x1.554ecacb0d035p-2, -0x1.c5d2664026ffcp-4, 0x1.eb624796ba809p-5,
-     -0x1.233803d19a535p-5, 0x1.300decb1c3c28p-6, -0x1.befe18031ec3dp-8,
-     0x1.449f5ee175c69p-10},
-    {0x1.554ae1f5ae815p-2, -0x1.c53c6b14ff6b2p-4, 0x1.e6b2d5127bb5bp-5,
-     -0x1.19387336788a3p-5, 0x1.180955a6ab255p-6, -0x1.81696703ba369p-8,
-     0x1.02cb36389bd79p-10},
-    {0x1.55458a59f356ep-2, -0x1.c4820dd631ae9p-4, 0x1.e167af818bd15p-5,
-     -0x1.0ef35f6f72e52p-5, 0x1.019c33b65e4ebp-6, -0x1.4d25bdd52d3a5p-8,
-     0x1.a008ae91f5936p-11},
-    {0x1.553e878eafee1p-2, -0x1.c3a1d0b2a3db2p-4, 0x1.db90d8ed9f89bp-5,
-     -0x1.0490e20f1ae91p-5, 0x1.d9a5d1fc42fe3p-7, -0x1.20bf8227c2abfp-8,
-     0x1.50f8174cdb6e9p-11},
-    {0x1.5535a0dedf1b1p-2, -0x1.c29afb8bd01a1p-4, 0x1.d53f6371c1e27p-5,
-     -0x1.f463209b433e2p-6, 0x1.b35222a17e44p-7, -0x1.f5efbf505e133p-9,
-     0x1.12e0e94e8586dp-11},
-    {0x1.552aa25e57bfdp-2, -0x1.c16d811e4acadp-4, 0x1.ce8489b47aa51p-5,
-     -0x1.dfde7ff758ea8p-6, 0x1.901f43aac38c8p-7, -0x1.b581d07df5ad5p-9,
-     0x1.c3726535f1fc6p-12},
-    {0x1.551d5d9b204d3p-2, -0x1.c019e328f8db1p-4, 0x1.c7710f44fc3cep-5,
-     -0x1.cbbbe25ea8ba4p-6, 0x1.6fe270088623dp-7, -0x1.7e6fc79733761p-9,
-     0x1.75077abf18d84p-12},
-};
-
-} // anonymous namespace
-
-LLVM_LIBC_FUNCTION(float, cbrtf, (float x)) {
-  using FloatBits = typename fputil::FPBits<float>;
-  using DoubleBits = typename fputil::FPBits<double>;
-
-  FloatBits x_bits(x);
-
-  uint32_t x_abs = x_bits.uintval() & 0x7fff'ffff;
-  uint32_t sign_bit = (x_bits.uintval() >> 31) << DoubleBits::EXP_LEN;
-
-  if (LIBC_UNLIKELY(x == 0.0f || x_abs >= 0x7f80'0000)) {
-    // x is 0, Inf, or NaN.
-    // Make sure it works for FTZ/DAZ modes.
-    return x + x;
-  }
-
-  double xd = static_cast<double>(x);
-  DoubleBits xd_bits(xd);
-
-  // When using biased exponent of x in double precision,
-  //   x_e = real_exponent_of_x + 1023
-  // Then:
-  //   x_e / 3 = real_exponent_of_x / 3 + 1023/3
-  //           = real_exponent_of_x / 3 + 341
-  // So to make it the correct biased exponent of x^(1/3), we add
-  //   1023 - 341 = 682
-  // to the quotient x_e / 3.
-  unsigned x_e = static_cast<unsigned>(xd_bits.get_biased_exponent());
-  unsigned out_e = (x_e / 3 + 682) | sign_bit;
-  unsigned shift_e = x_e % 3;
-
-  // Set x_m = 2^(x_e % 3) * (1.mantissa)
-  uint64_t x_m = xd_bits.get_mantissa();
-  // Use the leading 4 bits for look up table
-  unsigned idx = static_cast<unsigned>(x_m >> (DoubleBits::FRACTION_LEN - 4));
-
-  x_m |= static_cast<uint64_t>(DoubleBits::EXP_BIAS)
-         << DoubleBits::FRACTION_LEN;
-
-  double x_reduced = DoubleBits(x_m).get_val();
-  double dx = x_reduced - 1.0;
-
-  double dx_sq = dx * dx;
-  double c0 = fputil::multiply_add(dx, COEFFS[idx][0], 1.0);
-  double c1 = fputil::multiply_add(dx, COEFFS[idx][2], COEFFS[idx][1]);
-  double c2 = fputil::multiply_add(dx, COEFFS[idx][4], COEFFS[idx][3]);
-  double c3 = fputil::multiply_add(dx, COEFFS[idx][6], COEFFS[idx][5]);
-
-  double dx_4 = dx_sq * dx_sq;
-  double p0 = fputil::multiply_add(dx_sq, c1, c0);
-  double p1 = fputil::multiply_add(dx_sq, c3, c2);
-
-  double r = fputil::multiply_add(dx_4, p1, p0) * CBRT2[shift_e];
-
-  uint64_t r_m = DoubleBits(r).get_mantissa();
-  // Check if the output is exact.  To be exact, the smallest 1-bit of the
-  // output has to be at least 2^-7 or higher.  So we check the lowest 44 bits
-  // to see if they are within 2^(-52 + 3) errors from all zeros, then the
-  // result cube root is exact.
-  if (LIBC_UNLIKELY(((r_m + 8) & 0xfffffffffff) <= 16)) {
-    if ((r_m & 0xfffffffffff) <= 8)
-      r_m &= 0xffff'ffff'ffff'ffe0;
-    else
-      r_m = (r_m & 0xffff'ffff'ffff'ffe0) + 0x20;
-    fputil::clear_except_if_required(FE_INEXACT);
-  }
-  // Adjust exponent and sign.
-  uint64_t r_bits =
-      r_m | (static_cast<uint64_t>(out_e) << DoubleBits::FRACTION_LEN);
-
-  return static_cast<float>(DoubleBits(r_bits).get_val());
-}
+LLVM_LIBC_FUNCTION(float, cbrtf, (float x)) { return math::cbrtf(x); }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/test/shared/CMakeLists.txt b/libc/test/shared/CMakeLists.txt
index f5ea510180366..9685aeaee7d8d 100644
--- a/libc/test/shared/CMakeLists.txt
+++ b/libc/test/shared/CMakeLists.txt
@@ -27,6 +27,7 @@ add_fp_unittest(
     libc.src.__support.math.atanhf
     libc.src.__support.math.atanhf16
     libc.src.__support.math.cbrt
+    libc.src.__support.math.cbrtf
     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 3d64e5e68882b..5e57c49ed23d9 100644
--- a/libc/test/shared/shared_math_test.cpp
+++ b/libc/test/shared/shared_math_test.cpp
@@ -49,6 +49,7 @@ TEST(LlvmLibcSharedMathTest, AllFloat) {
   EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::atan2f(0.0f, 0.0f));
   EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::atanf(0.0f));
   EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::atanhf(0.0f));
+  EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::cbrtf(0.0f));
   EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::erff(0.0f));
   EXPECT_FP_EQ(0x1p+0f, LIBC_NAMESPACE::shared::exp10f(0.0f));
   EXPECT_FP_EQ(0x1p+0f, LIBC_NAMESPACE::shared::expf(0.0f));
diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index b46f334512979..d30c725524d17 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -2412,6 +2412,18 @@ libc_support_library(
     ],
 )
 
+libc_support_library(
+    name = "__support_math_cbrtf",
+    hdrs = ["src/__support/math/cbrtf.h"],
+    deps = [
+        ":__support_fputil_double_double",
+        ":__support_fputil_polyeval",
+        ":__support_fputil_fenv_impl",
+        ":__support_fputil_dyadic_float",
+        ":__support_integer_literals",
+    ],
+)
+
 libc_support_library(
     name = "__support_math_erff",
     hdrs = ["src/__support/math/erff.h"],
@@ -3088,7 +3100,7 @@ libc_math_function(
 libc_math_function(
     name = "cbrtf",
     additional_deps = [
-        ":__support_fputil_polyeval",
+        ":__support_math_cbrtf",
     ],
 )

bassiounix · 2025-08-08T14:38:08Z

Merge activity

Aug 8, 2:38 PM UTC: A user started a stack merge that includes this pull request via Graphite.
Aug 8, 2:39 PM UTC: Graphite rebased this pull request as part of a merge.
Aug 8, 2:43 PM UTC: Graphite rebased this pull request as part of a merge.
Aug 8, 2:46 PM UTC: Graphite rebased this pull request as part of a merge.
Aug 8, 2:58 PM UTC: Graphite rebased this pull request as part of a merge.
Aug 8, 3:07 PM UTC: Graphite rebased this pull request as part of a merge.
Aug 8, 3:10 PM UTC: Graphite rebased this pull request as part of a merge.
Aug 8, 3:13 PM UTC: Graphite rebased this pull request as part of a merge.
Aug 8, 3:23 PM UTC: Graphite rebased this pull request as part of a merge.
Aug 8, 3:28 PM UTC: @bassiounix merged this pull request with Graphite.

…pport/math folder.

bassiounix added bazel "Peripheral" support tier build system: utils/bazel libc labels Aug 3, 2025 — with Graphite App

bassiounix requested a review from lntue August 3, 2025 04:06

bassiounix mentioned this pull request Aug 3, 2025

[libc][math] Refactor cbrt implementation to header-only in src/__support/math folder. #151837

Merged

bassiounix marked this pull request as ready for review August 3, 2025 04:06

bassiounix requested review from rupprecht, keith and aaronmondal as code owners August 3, 2025 04:06

bassiounix mentioned this pull request Aug 2, 2025

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

Open

bassiounix removed request for keith, rupprecht and aaronmondal August 3, 2025 04:07

bassiounix force-pushed the users/bassiounix/spr/08-02-_libc_math_refactor_cbrt_implementation_to_header-only_in_src___support_math_folder branch from 150b50e to b200607 Compare August 3, 2025 17:54

Base automatically changed from users/bassiounix/spr/08-02-_libc_math_refactor_cbrt_implementation_to_header-only_in_src___support_math_folder to main August 3, 2025 18:02

bassiounix force-pushed the users/bassiounix/spr/08-03-_libc_math_refactor_cbrtf_implementation_to_header-only_in_src___support_math_folder branch 2 times, most recently from 2539a02 to 4aa6063 Compare August 3, 2025 19:25

bassiounix mentioned this pull request Aug 3, 2025

[libc][math] Refactor cos implementation to header-only in src/__support/math folder. #151883

Merged

bassiounix force-pushed the users/bassiounix/spr/08-03-_libc_math_refactor_cbrtf_implementation_to_header-only_in_src___support_math_folder branch from 4aa6063 to ae312dc Compare August 5, 2025 03:31

bassiounix mentioned this pull request Aug 5, 2025

[libc][math] Refactor cosf implementation to header-only in src/__support/math folder. #152069

Merged

bassiounix force-pushed the users/bassiounix/spr/08-03-_libc_math_refactor_cbrtf_implementation_to_header-only_in_src___support_math_folder branch from ae312dc to 1b752ff Compare August 5, 2025 03:48

lntue approved these changes Aug 5, 2025

View reviewed changes

bassiounix force-pushed the users/bassiounix/spr/08-03-_libc_math_refactor_cbrtf_implementation_to_header-only_in_src___support_math_folder branch 7 times, most recently from 847602c to fd28f6c Compare August 8, 2025 15:13

[libc][math] Refactor cbrtf implementation to header-only in src/__su…

50a8445

…pport/math folder.

bassiounix force-pushed the users/bassiounix/spr/08-03-_libc_math_refactor_cbrtf_implementation_to_header-only_in_src___support_math_folder branch from fd28f6c to 50a8445 Compare August 8, 2025 15:22

bassiounix merged commit 66734f4 into main Aug 8, 2025
19 checks passed

bassiounix deleted the users/bassiounix/spr/08-03-_libc_math_refactor_cbrtf_implementation_to_header-only_in_src___support_math_folder branch August 8, 2025 15:28

bassiounix mentioned this pull request Aug 9, 2025

[libc][math] Refactor cosf16 implementation to header-only in src/__support/math folder. #152871

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

[libc][math] Refactor cbrtf implementation to header-only in src/__support/math folder. #151846

[libc][math] Refactor cbrtf implementation to header-only in src/__support/math folder. #151846

Uh oh!

bassiounix commented Aug 3, 2025 •

edited

Loading

Uh oh!

bassiounix commented Aug 3, 2025 •

edited

Loading

Uh oh!

llvmbot commented Aug 3, 2025

Uh oh!

bassiounix commented Aug 8, 2025 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

[libc][math] Refactor cbrtf implementation to header-only in src/__support/math folder. #151846

[libc][math] Refactor cbrtf implementation to header-only in src/__support/math folder. #151846

Uh oh!

Conversation

bassiounix commented Aug 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

bassiounix commented Aug 3, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

llvmbot commented Aug 3, 2025

Uh oh!

bassiounix commented Aug 8, 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 Aug 3, 2025 •

edited

Loading

bassiounix commented Aug 3, 2025 •

edited

Loading

bassiounix commented Aug 8, 2025 •

edited

Loading