[PATCH 00/13] math: Use CORE-MATH sin, cos, tan, and sincos

Adhemerval Zanella adhemerval.zanella@linaro.org
Tue Nov 17 20:46:38 GMT 2026


This series replaces the IBM binary64 sin, cos, tan, and sincos with
the CORE-MATH implementations, which are correctly rounded in all
rounding modes.

  - Patches 1-2 make roundeven_finite an architecture hook and add an
    inline LoongArch version, used by the argument reductions.

  - Patches 3-4 add benchtests workloads for each code path (fast,
    moderate, slow) of sincos and tan, as sin and cos already have.

  - Patch 5 moves the IBM cos to s_cos.c, so sin and cos can be
    replaced separately.

  - Patches 6-9 import sin, cos, and tan.  Patch 7 makes j0, j1, y0,
    and y1 compute their trigonometric terms in round-to-nearest, as
    they relied on the IBM behavior (otherwise j1 and y1 exceed the
    9 ulp threshold in the directed rounding modes).

  - Patches 10-12 share the tables and routines: sin, cos, and their
    ifunc variants use a single copy, sincos is built on top of the sin
    and cos code (with bitwise identical results), and the tan tables
    move to the same file.

  - Patch 13 removes the x86_64 AVX variants, which no longer gain
    anything over the SSE2 ones.

Reciprocal throughput improvement in benchtests, on x86_64-v3 (Ryzen 9
5900X, gcc 16.1.1), aarch64 (Neoverse N1), power10 (POWER10), and
loongarch64 (Loongson-3C5000L-LL), the last three with gcc 16.2.1:

                      x86_64-v3      aarch64      power10  loongarch64
  sin    fast              +28%         -13%        +171%          -5%
         moderate          +38%         +23%        +249%         +19%
         slow             11.2x         6.9x         9.7x         6.3x

  cos    fast              +40%          +4%        +200%         +10%
         moderate          +46%         +37%        +256%         +22%
         slow             11.5x         6.9x         9.9x         6.4x

  tan    fast              -42%         -48%         +45%         -38%
         moderate          -38%         -42%         +28%         -32%
         slow              5.3x         3.2x         4.1x         3.0x

  sincos fast               -1%         -29%        +130%          -4%
         moderate          +18%          -3%        +163%         +18%
         slow              8.6x         5.1x         7.5x         5.2x

The main regressions are tan on the fast and moderate paths, except on
power10, and sincos on the aarch64 fast path.  Without FMA (x86-64
baseline and v2), the software fma makes sin and cos about 4x slower
and tan about 27x slower on these paths.

Checked on x86_64-linux-gnu and i686-linux-gnu.

Adhemerval Zanella (13):
  math: Make roundeven_finite an architecture hook
  loongarch: Add an inline roundeven_finite
  benchtests: Add path-oriented workloads to sincos
  benchtests: Add a moderate workload to tan
  math: Move the IBM cos implementation to s_cos.c
  math: Use sin from CORE-MATH
  math: Compute the j0, j1, y0, and y1 trigonometric terms in
    round-to-nearest
  math: Use cos from CORE-MATH
  math: Use tan from CORE-MATH
  math: Consolidate the sin and cos tables and routines
  math: Use sincos from CORE-MATH
  math: Move the tan tables to s_sincos_data.c
  x86_64: Remove the AVX variants of sin, cos, tan, and sincos

 NEWS                                          |    3 +-
 SHARED-FILES                                  |    8 +
 benchtests/sincos-inputs                      | 8133 ++++++-----------
 benchtests/tan-inputs                         | 1002 ++
 math/Makefile                                 |    5 +-
 math/auto-libm-test-in                        |  117 +
 math/auto-libm-test-out-cos                   | 1286 +++
 math/auto-libm-test-out-sin                   | 1587 ++++
 math/auto-libm-test-out-sincos                | 1555 ++++
 math/auto-libm-test-out-tan                   | 1654 ++++
 sysdeps/generic/math-roundeven-finite.h       |   45 +
 sysdeps/generic/math_uint128.h                |   14 +
 sysdeps/i386/Makefile                         |    4 +
 sysdeps/i386/fpu/libm-test-ulps               |    5 +
 sysdeps/ieee754/dbl-64/e_j0.c                 |   17 +-
 sysdeps/ieee754/dbl-64/e_j1.c                 |   15 +-
 sysdeps/ieee754/dbl-64/e_rem_pio2.c           |   76 -
 sysdeps/ieee754/dbl-64/libm-test-ulps         |   48 +
 sysdeps/ieee754/dbl-64/math_config.h          |   27 +-
 sysdeps/ieee754/dbl-64/s_cos.c                |  223 +-
 sysdeps/ieee754/dbl-64/s_sin.c                |  494 +-
 sysdeps/ieee754/dbl-64/s_sin_common.h         |  285 +
 sysdeps/ieee754/dbl-64/s_sincos.c             |  367 +-
 sysdeps/ieee754/dbl-64/s_sincos_data.c        | 1694 ++++
 sysdeps/ieee754/dbl-64/s_sincos_data.h        |   72 +
 sysdeps/ieee754/dbl-64/s_tan.c                | 1804 +++-
 sysdeps/ieee754/dbl-64/sincostab.c            |  913 --
 sysdeps/ieee754/dbl-64/usncs.h                |   47 -
 sysdeps/ieee754/dbl-64/utan.h                 |   96 -
 sysdeps/ieee754/dbl-64/utan.tbl               | 1524 ---
 sysdeps/ieee754/flt-32/math_config.h          |   52 +-
 sysdeps/loongarch/fpu/math-roundeven-finite.h |   46 +
 sysdeps/m68k/m680x0/fpu/e_rem_pio2.c          |    1 -
 .../fpu/{sincostab.c => s_sincos_data.c}      |    0
 sysdeps/powerpc/fpu/math-roundeven-finite.h   |   60 +
 sysdeps/powerpc/fpu/math_private.h            |    6 -
 sysdeps/x86_64/fpu/multiarch/Makefile         |   10 +-
 sysdeps/x86_64/fpu/multiarch/s_cos-fma.c      |    4 +
 sysdeps/x86_64/fpu/multiarch/s_cos-fma4.c     |    4 +
 sysdeps/x86_64/fpu/multiarch/s_cos.c          |   33 +
 sysdeps/x86_64/fpu/multiarch/s_sin-avx.c      |    5 -
 sysdeps/x86_64/fpu/multiarch/s_sin-fma.c      |    7 -
 sysdeps/x86_64/fpu/multiarch/s_sin-fma4.c     |    7 -
 sysdeps/x86_64/fpu/multiarch/s_sin.c          |   13 +-
 sysdeps/x86_64/fpu/multiarch/s_sincos-avx.c   |    3 -
 sysdeps/x86_64/fpu/multiarch/s_tan-avx.c      |    6 -
 sysdeps/x86_64/fpu/multiarch/s_tan-fma.c      |    4 -
 sysdeps/x86_64/fpu/multiarch/s_tan-fma4.c     |    4 -
 sysdeps/x86_64/fpu/multiarch/s_tan.c          |    2 +-
 49 files changed, 14706 insertions(+), 8681 deletions(-)
 create mode 100644 sysdeps/generic/math-roundeven-finite.h
 delete mode 100644 sysdeps/ieee754/dbl-64/e_rem_pio2.c
 create mode 100644 sysdeps/ieee754/dbl-64/s_sin_common.h
 create mode 100644 sysdeps/ieee754/dbl-64/s_sincos_data.c
 create mode 100644 sysdeps/ieee754/dbl-64/s_sincos_data.h
 delete mode 100644 sysdeps/ieee754/dbl-64/sincostab.c
 delete mode 100644 sysdeps/ieee754/dbl-64/usncs.h
 delete mode 100644 sysdeps/ieee754/dbl-64/utan.h
 delete mode 100644 sysdeps/ieee754/dbl-64/utan.tbl
 create mode 100644 sysdeps/loongarch/fpu/math-roundeven-finite.h
 delete mode 100644 sysdeps/m68k/m680x0/fpu/e_rem_pio2.c
 rename sysdeps/m68k/m680x0/fpu/{sincostab.c => s_sincos_data.c} (100%)
 create mode 100644 sysdeps/powerpc/fpu/math-roundeven-finite.h
 create mode 100644 sysdeps/x86_64/fpu/multiarch/s_cos-fma.c
 create mode 100644 sysdeps/x86_64/fpu/multiarch/s_cos-fma4.c
 create mode 100644 sysdeps/x86_64/fpu/multiarch/s_cos.c
 delete mode 100644 sysdeps/x86_64/fpu/multiarch/s_sin-avx.c
 delete mode 100644 sysdeps/x86_64/fpu/multiarch/s_sincos-avx.c
 delete mode 100644 sysdeps/x86_64/fpu/multiarch/s_tan-avx.c

-- 
2.53.0



More information about the Libc-alpha mailing list