x86-64: Add cosf with FMA
commitac817e083b37a5c25d05cde8bde302d7a93ffc5e
authorH.J. Lu <hjl.tools@gmail.com>
Tue, 12 Dec 2017 23:32:58 +0000 (12 15:32 -0800)
committerH.J. Lu <hjl.tools@gmail.com>
Tue, 12 Dec 2017 23:32:58 +0000 (12 15:32 -0800)
tree779f6b5284d8548ff24bf96db06e347b011ea46c
parenteb4285768bb0608e2c3381d0a5214cae83b1f282
x86-64: Add cosf with FMA

On Skylake, bench-cosf reports performance improvement:

            Before        After         Improvement
max        135.362       94.552            43%
min        8.532         7.688             11%
mean       17.1446       11.8128           45%

* sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines):
Add s_cosf-sse2 and s_cosf-fma.
(CFLAGS-s_cosf-fma.c): New.
* sysdeps/x86_64/fpu/multiarch/s_cosf-fma.c: New file.
* sysdeps/x86_64/fpu/multiarch/s_cosf-sse2.c: Likewise.
* sysdeps/x86_64/fpu/multiarch/s_cosf.c: Likewise.

Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
ChangeLog
NEWS
sysdeps/x86_64/fpu/multiarch/Makefile
sysdeps/x86_64/fpu/multiarch/s_cosf-fma.c [new file with mode: 0644]
sysdeps/x86_64/fpu/multiarch/s_cosf-sse2.c [new file with mode: 0644]
sysdeps/x86_64/fpu/multiarch/s_cosf.c [new file with mode: 0644]