AuroraMiddleware/glibc
1
0
Fork 0
mirror of https://sourceware.org/git/glibc.git synced 2024-12-02 01:40:07 +00:00
Code Issues Packages Projects Releases Wiki Activity
f285711506
glibc/sysdeps/x86_64/fpu/multiarch/Makefile

83 lines
2.5 KiB
Makefile
Raw Normal View History

Use rounds{s,d} for x86 rint, ceil, floor
2011-10-17 00:58:17 +00:00
ifeq ($(subdir),math)
libm-sysdep_routines += s_floor-c s_ceil-c s_floorf-c s_ceilf-c \
Add SSE4.1 trunc, truncf (bug 20142). This patch adds SSE4.1 versions of trunc and truncf, using the roundsd / roundss instructions, similar to the versions of ceil, floor, rint and nearbyint functions we already have. In my testing with the glibc benchtests these are about 30% faster than the C versions for double, 20% faster for float. Tested for x86_64. [BZ #20142] * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add s_trunc-c, s_truncf-c, s_trunc-sse4_1 and s_truncf-sse4_1. * sysdeps/x86_64/fpu/multiarch/s_trunc-c.c: New file. * sysdeps/x86_64/fpu/multiarch/s_trunc-sse4_1.S: Likewise. * sysdeps/x86_64/fpu/multiarch/s_trunc.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_truncf-c.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_truncf-sse4_1.S: Likewise. * sysdeps/x86_64/fpu/multiarch/s_truncf.c: Likewise.
2017-09-20 16:54:05 +00:00
s_rint-c s_rintf-c s_nearbyint-c s_nearbyintf-c \
x86_64: roundeven with sse4.1 support This patch adds support for the sse4.1 hardware floating point roundeven. Here is some benchmark results on my systems: =AMD Ryzen 9 3900X 12-Core Processor= * benchmark result before this commit | | roundeven | roundevenf | |------------|--------------|--------------| | duration | 3.75587e+09 | 3.75114e+09 | | iterations | 3.93053e+08 | 4.35402e+08 | | max | 52.592 | 58.71 | | min | 7.98 | 7.22 | | mean | 9.55563 | 8.61535 | * benchmark result after this commit | | roundeven | roundevenf | |------------|---------------|--------------| | duration | 3.73815e+09 | 3.73738e+09 | | iterations | 5.82692e+08 | 5.91498e+08 | | max | 56.468 | 51.642 | | min | 6.27 | 6.156 | | mean | 6.41532 | 6.3185 | =Intel(R) Pentium(R) CPU D1508 @ 2.20GHz= * benchmark result before this commit | | roundeven | roundevenf | |------------|--------------|--------------| | duration | 2.18208e+09 | 2.18258e+09 | | iterations | 2.39932e+08 | 2.46924e+08 | | max | 96.378 | 98.035 | | min | 6.776 | 5.94 | | mean | 9.09456 | 8.83907 | * benchmark result after this commit | | roundeven | roundevenf | |------------|--------------|--------------| | duration | 2.17415e+09 | 2.17005e+09 | | iterations | 3.56193e+08 | 4.09824e+08 | | max | 51.693 | 97.192 | | min | 5.926 | 5.093 | | mean | 6.10385 | 5.29507 | Signed-off-by: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com> Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
2021-05-24 01:43:11 +00:00
s_roundeven-c s_roundevenf-c s_trunc-c s_truncf-c
Optimize accurate 64-bit routines for FMA4 on x86-64
2011-10-25 00:19:17 +00:00
x86-64: Implement libm IFUNC selectors in C * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add s_ceil-sse4_1, s_ceilf-sse4_1, s_floor-sse4_1, s_floorf-sse4_1, s_nearbyint-sse4_1, s_nearbyintf-sse4_1, s_rint-sse4_1 and s_rintf-sse4_1. * sysdeps/x86_64/fpu/multiarch/ifunc-sse4_1.h: New file. * sysdeps/x86_64/fpu/multiarch/s_ceil.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_ceilf.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_floor.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_floorf.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_nearbyint.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_nearbyintf.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_rint.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_rintf.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_ceil.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_ceil-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__ceil): Removed. * sysdeps/x86_64/fpu/multiarch/s_ceilf.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_ceilf-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__ceilf): Removed. * sysdeps/x86_64/fpu/multiarch/s_floor.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_floor-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__floor): Removed. * sysdeps/x86_64/fpu/multiarch/s_floorf.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_floorf-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__floorf): Removed. * sysdeps/x86_64/fpu/multiarch/s_nearbyint.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_nearbyint-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__nearbyint): Removed. * sysdeps/x86_64/fpu/multiarch/s_nearbyintf.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_nearbyintf-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__nearbyintf): Removed. * sysdeps/x86_64/fpu/multiarch/s_rint.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_rint-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__rint): Removed. * sysdeps/x86_64/fpu/multiarch/s_rintf.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_rintf-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__rintf): Removed.
2017-08-04 20:01:59 +00:00
libm-sysdep_routines += s_ceil-sse4_1 s_ceilf-sse4_1 s_floor-sse4_1 \
s_floorf-sse4_1 s_nearbyint-sse4_1 \
x86_64: roundeven with sse4.1 support This patch adds support for the sse4.1 hardware floating point roundeven. Here is some benchmark results on my systems: =AMD Ryzen 9 3900X 12-Core Processor= * benchmark result before this commit | | roundeven | roundevenf | |------------|--------------|--------------| | duration | 3.75587e+09 | 3.75114e+09 | | iterations | 3.93053e+08 | 4.35402e+08 | | max | 52.592 | 58.71 | | min | 7.98 | 7.22 | | mean | 9.55563 | 8.61535 | * benchmark result after this commit | | roundeven | roundevenf | |------------|---------------|--------------| | duration | 3.73815e+09 | 3.73738e+09 | | iterations | 5.82692e+08 | 5.91498e+08 | | max | 56.468 | 51.642 | | min | 6.27 | 6.156 | | mean | 6.41532 | 6.3185 | =Intel(R) Pentium(R) CPU D1508 @ 2.20GHz= * benchmark result before this commit | | roundeven | roundevenf | |------------|--------------|--------------| | duration | 2.18208e+09 | 2.18258e+09 | | iterations | 2.39932e+08 | 2.46924e+08 | | max | 96.378 | 98.035 | | min | 6.776 | 5.94 | | mean | 9.09456 | 8.83907 | * benchmark result after this commit | | roundeven | roundevenf | |------------|--------------|--------------| | duration | 2.17415e+09 | 2.17005e+09 | | iterations | 3.56193e+08 | 4.09824e+08 | | max | 51.693 | 97.192 | | min | 5.926 | 5.093 | | mean | 6.10385 | 5.29507 | Signed-off-by: Shen-Ta Hsieh <ibmibmibm.tw@gmail.com> Reviewed-by: H.J. Lu <hjl.tools@gmail.com>
2021-05-24 01:43:11 +00:00
s_nearbyintf-sse4_1 s_roundeven-sse4_1 \
s_roundevenf-sse4_1 s_rint-sse4_1 s_rintf-sse4_1 \
Add SSE4.1 trunc, truncf (bug 20142). This patch adds SSE4.1 versions of trunc and truncf, using the roundsd / roundss instructions, similar to the versions of ceil, floor, rint and nearbyint functions we already have. In my testing with the glibc benchtests these are about 30% faster than the C versions for double, 20% faster for float. Tested for x86_64. [BZ #20142] * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add s_trunc-c, s_truncf-c, s_trunc-sse4_1 and s_truncf-sse4_1. * sysdeps/x86_64/fpu/multiarch/s_trunc-c.c: New file. * sysdeps/x86_64/fpu/multiarch/s_trunc-sse4_1.S: Likewise. * sysdeps/x86_64/fpu/multiarch/s_trunc.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_truncf-c.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_truncf-sse4_1.S: Likewise. * sysdeps/x86_64/fpu/multiarch/s_truncf.c: Likewise.
2017-09-20 16:54:05 +00:00
s_trunc-sse4_1 s_truncf-sse4_1
x86-64: Implement libm IFUNC selectors in C * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add s_ceil-sse4_1, s_ceilf-sse4_1, s_floor-sse4_1, s_floorf-sse4_1, s_nearbyint-sse4_1, s_nearbyintf-sse4_1, s_rint-sse4_1 and s_rintf-sse4_1. * sysdeps/x86_64/fpu/multiarch/ifunc-sse4_1.h: New file. * sysdeps/x86_64/fpu/multiarch/s_ceil.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_ceilf.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_floor.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_floorf.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_nearbyint.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_nearbyintf.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_rint.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_rintf.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_ceil.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_ceil-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__ceil): Removed. * sysdeps/x86_64/fpu/multiarch/s_ceilf.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_ceilf-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__ceilf): Removed. * sysdeps/x86_64/fpu/multiarch/s_floor.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_floor-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__floor): Removed. * sysdeps/x86_64/fpu/multiarch/s_floorf.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_floorf-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__floorf): Removed. * sysdeps/x86_64/fpu/multiarch/s_nearbyint.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_nearbyint-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__nearbyint): Removed. * sysdeps/x86_64/fpu/multiarch/s_nearbyintf.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_nearbyintf-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__nearbyintf): Removed. * sysdeps/x86_64/fpu/multiarch/s_rint.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_rint-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__rint): Removed. * sysdeps/x86_64/fpu/multiarch/s_rintf.S: Renamed to ... * sysdeps/x86_64/fpu/multiarch/s_rintf-sse4_1.S: This. Don't include <machine/asm.h> nor <init-arch.h>. Include <sysdep.h>. (__rintf): Removed.
2017-08-04 20:01:59 +00:00
x86-64: Add FMA multiarch functions to libm This patch adds multiarch functions optimized with -mfma -mavx2 to libm. e_pow-fma.c is compiled with $(config-cflags-nofma) due to PR 19003. * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add e_exp-fma, e_log-fma, e_pow-fma, s_atan-fma, e_asin-fma, e_atan2-fma, s_sin-fma, s_tan-fma, mplog-fma, mpa-fma, slowexp-fma, slowpow-fma, sincos32-fma, doasin-fma, dosincos-fma, halfulp-fma, mpexp-fma, mpatan2-fma, mpatan-fma, mpsqrt-fma, and mptan-fma. (CFLAGS-doasin-fma.c): New. (CFLAGS-dosincos-fma.c): Likewise. (CFLAGS-e_asin-fma.c): Likewise. (CFLAGS-e_atan2-fma.c): Likewise. (CFLAGS-e_exp-fma.c): Likewise. (CFLAGS-e_log-fma.c): Likewise. (CFLAGS-e_pow-fma.c): Likewise. (CFLAGS-halfulp-fma.c): Likewise. (CFLAGS-mpa-fma.c): Likewise. (CFLAGS-mpatan-fma.c): Likewise. (CFLAGS-mpatan2-fma.c): Likewise. (CFLAGS-mpexp-fma.c): Likewise. (CFLAGS-mplog-fma.c): Likewise. (CFLAGS-mpsqrt-fma.c): Likewise. (CFLAGS-mptan-fma.c): Likewise. (CFLAGS-s_atan-fma.c): Likewise. (CFLAGS-sincos32-fma.c): Likewise. (CFLAGS-slowexp-fma.c): Likewise. (CFLAGS-slowpow-fma.c): Likewise. (CFLAGS-s_sin-fma.c): Likewise. (CFLAGS-s_tan-fma.c): Likewise. * sysdeps/x86_64/fpu/multiarch/doasin-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/dosincos-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_asin-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_atan2-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_exp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_log-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_pow-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/halfulp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/ifunc-avx-fma4.h: Likewise. * sysdeps/x86_64/fpu/multiarch/ifunc-fma4.h: Likewise. * sysdeps/x86_64/fpu/multiarch/mpa-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpatan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpatan2-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpexp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mplog-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpsqrt-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mptan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_atan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sin-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_tan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/sincos32-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/slowexp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/slowpow-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_asin.c: Rewrite. * sysdeps/x86_64/fpu/multiarch/e_atan2.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_exp.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_log.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_pow.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_atan.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sin.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_tan.c: Likewise.
2017-08-07 15:19:59 +00:00
libm-sysdep_routines += e_exp-fma e_log-fma e_pow-fma s_atan-fma \
math: Remove mpa files [BZ #15267] Finally remove all mpa related files, headers, declarations, probes, unused tables and update makefiles. Reviewed-By: Paul Zimmermann <Paul.Zimmermann@inria.fr>
2021-03-10 12:41:47 +00:00
e_asin-fma e_atan2-fma s_sin-fma s_tan-fma
x86-64: Add FMA multiarch functions to libm This patch adds multiarch functions optimized with -mfma -mavx2 to libm. e_pow-fma.c is compiled with $(config-cflags-nofma) due to PR 19003. * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add e_exp-fma, e_log-fma, e_pow-fma, s_atan-fma, e_asin-fma, e_atan2-fma, s_sin-fma, s_tan-fma, mplog-fma, mpa-fma, slowexp-fma, slowpow-fma, sincos32-fma, doasin-fma, dosincos-fma, halfulp-fma, mpexp-fma, mpatan2-fma, mpatan-fma, mpsqrt-fma, and mptan-fma. (CFLAGS-doasin-fma.c): New. (CFLAGS-dosincos-fma.c): Likewise. (CFLAGS-e_asin-fma.c): Likewise. (CFLAGS-e_atan2-fma.c): Likewise. (CFLAGS-e_exp-fma.c): Likewise. (CFLAGS-e_log-fma.c): Likewise. (CFLAGS-e_pow-fma.c): Likewise. (CFLAGS-halfulp-fma.c): Likewise. (CFLAGS-mpa-fma.c): Likewise. (CFLAGS-mpatan-fma.c): Likewise. (CFLAGS-mpatan2-fma.c): Likewise. (CFLAGS-mpexp-fma.c): Likewise. (CFLAGS-mplog-fma.c): Likewise. (CFLAGS-mpsqrt-fma.c): Likewise. (CFLAGS-mptan-fma.c): Likewise. (CFLAGS-s_atan-fma.c): Likewise. (CFLAGS-sincos32-fma.c): Likewise. (CFLAGS-slowexp-fma.c): Likewise. (CFLAGS-slowpow-fma.c): Likewise. (CFLAGS-s_sin-fma.c): Likewise. (CFLAGS-s_tan-fma.c): Likewise. * sysdeps/x86_64/fpu/multiarch/doasin-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/dosincos-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_asin-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_atan2-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_exp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_log-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_pow-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/halfulp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/ifunc-avx-fma4.h: Likewise. * sysdeps/x86_64/fpu/multiarch/ifunc-fma4.h: Likewise. * sysdeps/x86_64/fpu/multiarch/mpa-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpatan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpatan2-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpexp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mplog-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpsqrt-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mptan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_atan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sin-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_tan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/sincos32-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/slowexp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/slowpow-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_asin.c: Rewrite. * sysdeps/x86_64/fpu/multiarch/e_atan2.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_exp.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_log.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_pow.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_atan.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sin.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_tan.c: Likewise.
2017-08-07 15:19:59 +00:00
CFLAGS-e_asin-fma.c = -mfma -mavx2
CFLAGS-e_atan2-fma.c = -mfma -mavx2
CFLAGS-e_exp-fma.c = -mfma -mavx2
CFLAGS-e_log-fma.c = -mfma -mavx2
Add new pow implementation The algorithm is exp(y * log(x)), where log(x) is computed with about 1.3*2^-68 relative error (1.5*2^-68 without fma), returning the result in two doubles, and the exp part uses the same algorithm (and lookup tables) as exp, but takes the input as two doubles and a sign (to handle negative bases with odd integer exponent). The __exp1 internal symbol is no longer necessary. There is separate code path when fma is not available but the worst case error is about 0.54 ULP in both cases. The lookup table and consts for log are 4168 bytes. The .rodata+.text is decreased by 37908 bytes on aarch64. The non-nearest rounding error is less than 1 ULP. Improvements on Cortex-A72 compared to current glibc master: pow thruput: 2.40x in [0.01 11.1]x[0.01 11.1] pow latency: 1.84x in [0.01 11.1]x[0.01 11.1] Tested on aarch64-linux-gnu (defined __FP_FAST_FMA, TOINT_INTRINSICS) and arm-linux-gnueabihf (!defined __FP_FAST_FMA, !TOINT_INTRINSICS) and x86_64-linux-gnu (!defined __FP_FAST_FMA, !TOINT_INTRINSICS) and powerpc64le-linux-gnu (defined __FP_FAST_FMA, !TOINT_INTRINSICS) targets. * NEWS: Mention pow improvements. * math/Makefile (type-double-routines): Add e_pow_log_data. * sysdeps/generic/math_private.h (__exp1): Remove. * sysdeps/i386/fpu/e_pow_log_data.c: New file. * sysdeps/ia64/fpu/e_pow_log_data.c: New file. * sysdeps/ieee754/dbl-64/Makefile (CFLAGS-e_pow.c): Allow fma contraction. * sysdeps/ieee754/dbl-64/e_exp.c (__exp1): Remove. (exp_inline): Remove. (__ieee754_exp): Only single double input is handled. * sysdeps/ieee754/dbl-64/e_pow.c: Rewrite. * sysdeps/ieee754/dbl-64/e_pow_log_data.c: New file. * sysdeps/ieee754/dbl-64/math_config.h (issignaling_inline): Define. (__pow_log_data): Define. * sysdeps/ieee754/dbl-64/upow.h: Remove. * sysdeps/ieee754/dbl-64/upow.tbl: Remove. * sysdeps/m68k/m680x0/fpu/e_pow_log_data.c: New file. * sysdeps/x86_64/fpu/multiarch/Makefile (CFLAGS-e_pow-fma.c): Allow fma contraction. (CFLAGS-e_pow-fma4.c): Likewise.
2018-06-13 16:57:20 +00:00
CFLAGS-e_pow-fma.c = -mfma -mavx2
x86-64: Add FMA multiarch functions to libm This patch adds multiarch functions optimized with -mfma -mavx2 to libm. e_pow-fma.c is compiled with $(config-cflags-nofma) due to PR 19003. * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add e_exp-fma, e_log-fma, e_pow-fma, s_atan-fma, e_asin-fma, e_atan2-fma, s_sin-fma, s_tan-fma, mplog-fma, mpa-fma, slowexp-fma, slowpow-fma, sincos32-fma, doasin-fma, dosincos-fma, halfulp-fma, mpexp-fma, mpatan2-fma, mpatan-fma, mpsqrt-fma, and mptan-fma. (CFLAGS-doasin-fma.c): New. (CFLAGS-dosincos-fma.c): Likewise. (CFLAGS-e_asin-fma.c): Likewise. (CFLAGS-e_atan2-fma.c): Likewise. (CFLAGS-e_exp-fma.c): Likewise. (CFLAGS-e_log-fma.c): Likewise. (CFLAGS-e_pow-fma.c): Likewise. (CFLAGS-halfulp-fma.c): Likewise. (CFLAGS-mpa-fma.c): Likewise. (CFLAGS-mpatan-fma.c): Likewise. (CFLAGS-mpatan2-fma.c): Likewise. (CFLAGS-mpexp-fma.c): Likewise. (CFLAGS-mplog-fma.c): Likewise. (CFLAGS-mpsqrt-fma.c): Likewise. (CFLAGS-mptan-fma.c): Likewise. (CFLAGS-s_atan-fma.c): Likewise. (CFLAGS-sincos32-fma.c): Likewise. (CFLAGS-slowexp-fma.c): Likewise. (CFLAGS-slowpow-fma.c): Likewise. (CFLAGS-s_sin-fma.c): Likewise. (CFLAGS-s_tan-fma.c): Likewise. * sysdeps/x86_64/fpu/multiarch/doasin-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/dosincos-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_asin-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_atan2-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_exp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_log-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_pow-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/halfulp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/ifunc-avx-fma4.h: Likewise. * sysdeps/x86_64/fpu/multiarch/ifunc-fma4.h: Likewise. * sysdeps/x86_64/fpu/multiarch/mpa-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpatan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpatan2-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpexp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mplog-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mpsqrt-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/mptan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_atan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sin-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_tan-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/sincos32-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/slowexp-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/slowpow-fma.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_asin.c: Rewrite. * sysdeps/x86_64/fpu/multiarch/e_atan2.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_exp.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_log.c: Likewise. * sysdeps/x86_64/fpu/multiarch/e_pow.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_atan.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sin.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_tan.c: Likewise.
2017-08-07 15:19:59 +00:00
CFLAGS-s_atan-fma.c = -mfma -mavx2
CFLAGS-s_sin-fma.c = -mfma -mavx2
CFLAGS-s_tan-fma.c = -mfma -mavx2
x86-64: Add sincosf with vector FMA Since the x86-64 assembly version of sincosf is higly optimized with vector instructions, there isn't much room for improvement. However s_sincosf.c written in C with vector math and intrinsics can be optimized by GCC with FMA. On Skylake, bench-sincosf reports performance improvement: Assembly FMA improvement max 104.042 101.008 3% min 9.426 8.586 10% mean 20.6209 18.2238 13% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add s_sincosf-sse2 and s_sincosf-fma. (CFLAGS-s_sincosf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/s_sincosf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/s_sincosf-sse2.S: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sincosf.c: Likewise. * sysdeps/x86_64/fpu/s_sincosf.S: Don't add alias if __sincosf is defined.
2018-01-08 16:04:26 +00:00
libm-sysdep_routines += s_sinf-sse2 s_cosf-sse2 s_sincosf-sse2
x86-64: Add sinf with FMA On Skylake, bench-sinf reports performance improvement: Before After Improvement max 153.996 100.094 54% min 8.546 6.852 25% mean 18.1223 11.802 54% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add s_sinf-sse2 and s_sinf-fma. (CFLAGS-s_sinf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/s_sinf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/s_sinf-sse2.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sinf.c: Likewise.
2017-12-07 18:11:02 +00:00
x86-64: Add powf with FMA For workload-spec2017.wrf, on Skylake, it improves performance by: Before After Improvement reciprocal-throughput 35.4713 27.3842 29% latency 82.4537 66.3175 24% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add e_powf-fma. (CFLAGS-e_powf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/e_powf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/e_powf.c: Likewise.
2017-10-22 15:07:48 +00:00
libm-sysdep_routines += e_exp2f-fma e_expf-fma e_log2f-fma e_logf-fma \
x86-64: Add sincosf with vector FMA Since the x86-64 assembly version of sincosf is higly optimized with vector instructions, there isn't much room for improvement. However s_sincosf.c written in C with vector math and intrinsics can be optimized by GCC with FMA. On Skylake, bench-sincosf reports performance improvement: Assembly FMA improvement max 104.042 101.008 3% min 9.426 8.586 10% mean 20.6209 18.2238 13% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add s_sincosf-sse2 and s_sincosf-fma. (CFLAGS-s_sincosf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/s_sincosf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/s_sincosf-sse2.S: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sincosf.c: Likewise. * sysdeps/x86_64/fpu/s_sincosf.S: Don't add alias if __sincosf is defined.
2018-01-08 16:04:26 +00:00
e_powf-fma s_sinf-fma s_cosf-fma s_sincosf-fma
x86-64: Optimize e_expf with FMA [BZ #21912] FMA optimized e_expf improves performance by more than 50% on Skylake. [BZ #21912] * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add e_expf-fma. * sysdeps/x86_64/fpu/multiarch/e_expf-fma.S: New file. * sysdeps/x86_64/fpu/multiarch/e_expf.c: Likewise. * sysdeps/x86_64/fpu/multiarch/ifunc-fma.h: Likewise.
2017-08-16 15:43:35 +00:00
x86-64: Add exp2f with FMA For workload-spec2017.wrf, on Skylake, it improves performance by: Before After Improvement reciprocal-throughput 13.0291 11.2225 16% latency 44.5154 37.5766 18% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add e_exp2f-fma. (CFLAGS-e_exp2f-fma.c): New. * sysdeps/x86_64/fpu/multiarch/e_exp2f-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/e_exp2f.c: Likewise.
2017-10-22 14:55:49 +00:00
CFLAGS-e_exp2f-fma.c = -mfma -mavx2
x86-64: Replace assembly versions of e_expf with generic e_expf.c This patch replaces x86-64 assembly versions of e_expf with generic e_expf.c. For workload-spec2017.wrf, on Nehalem, it improves performance by: Before After Improvement reciprocal-throughput 36.039 20.7749 73% latency 58.8096 40.8715 43% On Skylake, it improves Before After Improvement reciprocal-throughput 18.4436 11.1693 65% latency 47.5162 37.5411 26% * sysdeps/x86_64/fpu/e_expf.S: Removed. * sysdeps/x86_64/fpu/multiarch/e_expf-fma.S: Likewise. * sysdeps/x86_64/fpu/w_expf.c: Likewise. * sysdeps/x86_64/fpu/libm-test-ulps: Updated for generic e_expf.c. * sysdeps/x86_64/fpu/multiarch/Makefile (CFLAGS-e_expf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/e_expf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/e_expf.c (__redirect_ieee754_expf): Renamed to ... (__redirect_expf): This. (SYMBOL_NAME): Changed to expf. (__ieee754_expf): Renamed to ... (__expf): This. (__GI___expf): This. (__ieee754_expf): Add strong_alias. (__expf_finite): Likewise. (__expf): New. Include <sysdeps/ieee754/flt-32/e_expf.c>.
2017-10-22 14:49:36 +00:00
CFLAGS-e_expf-fma.c = -mfma -mavx2
x86-64: Add log2f with FMA For workload-spec2017.wrf, on Skylake, it improves performance by: Before After Improvement reciprocal-throughput 16.5937 14.0789 17% latency 41.7755 35.3586 18% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add e_log2f-fma. (CFLAGS-e_log2f-fma.c): New. * sysdeps/x86_64/fpu/multiarch/e_log2f-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/e_log2f.c: Likewise.
2017-10-22 15:06:46 +00:00
CFLAGS-e_log2f-fma.c = -mfma -mavx2
x86-64: Add logf with FMA For workload-spec2017.wrf, on Skylake, it improves performance by: Before After Improvement reciprocal-throughput 16.1534 13.8874 16% latency 41.9642 34.3072 22% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add e_logf-fma. (CFLAGS-e_logf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/e_logf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/e_logf.c: Likewise.
2017-10-22 15:03:46 +00:00
CFLAGS-e_logf-fma.c = -mfma -mavx2
x86-64: Add powf with FMA For workload-spec2017.wrf, on Skylake, it improves performance by: Before After Improvement reciprocal-throughput 35.4713 27.3842 29% latency 82.4537 66.3175 24% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add e_powf-fma. (CFLAGS-e_powf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/e_powf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/e_powf.c: Likewise.
2017-10-22 15:07:48 +00:00
CFLAGS-e_powf-fma.c = -mfma -mavx2
x86-64: Add sinf with FMA On Skylake, bench-sinf reports performance improvement: Before After Improvement max 153.996 100.094 54% min 8.546 6.852 25% mean 18.1223 11.802 54% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add s_sinf-sse2 and s_sinf-fma. (CFLAGS-s_sinf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/s_sinf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/s_sinf-sse2.c: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sinf.c: Likewise.
2017-12-07 18:11:02 +00:00
CFLAGS-s_sinf-fma.c = -mfma -mavx2
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>
2017-12-12 23:32:58 +00:00
CFLAGS-s_cosf-fma.c = -mfma -mavx2
x86-64: Add sincosf with vector FMA Since the x86-64 assembly version of sincosf is higly optimized with vector instructions, there isn't much room for improvement. However s_sincosf.c written in C with vector math and intrinsics can be optimized by GCC with FMA. On Skylake, bench-sincosf reports performance improvement: Assembly FMA improvement max 104.042 101.008 3% min 9.426 8.586 10% mean 20.6209 18.2238 13% * sysdeps/x86_64/fpu/multiarch/Makefile (libm-sysdep_routines): Add s_sincosf-sse2 and s_sincosf-fma. (CFLAGS-s_sincosf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/s_sincosf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/s_sincosf-sse2.S: Likewise. * sysdeps/x86_64/fpu/multiarch/s_sincosf.c: Likewise. * sysdeps/x86_64/fpu/s_sincosf.S: Don't add alias if __sincosf is defined.
2018-01-08 16:04:26 +00:00
CFLAGS-s_sincosf-fma.c = -mfma -mavx2
x86-64: Replace assembly versions of e_expf with generic e_expf.c This patch replaces x86-64 assembly versions of e_expf with generic e_expf.c. For workload-spec2017.wrf, on Nehalem, it improves performance by: Before After Improvement reciprocal-throughput 36.039 20.7749 73% latency 58.8096 40.8715 43% On Skylake, it improves Before After Improvement reciprocal-throughput 18.4436 11.1693 65% latency 47.5162 37.5411 26% * sysdeps/x86_64/fpu/e_expf.S: Removed. * sysdeps/x86_64/fpu/multiarch/e_expf-fma.S: Likewise. * sysdeps/x86_64/fpu/w_expf.c: Likewise. * sysdeps/x86_64/fpu/libm-test-ulps: Updated for generic e_expf.c. * sysdeps/x86_64/fpu/multiarch/Makefile (CFLAGS-e_expf-fma.c): New. * sysdeps/x86_64/fpu/multiarch/e_expf-fma.c: New file. * sysdeps/x86_64/fpu/multiarch/e_expf.c (__redirect_ieee754_expf): Renamed to ... (__redirect_expf): This. (SYMBOL_NAME): Changed to expf. (__ieee754_expf): Renamed to ... (__expf): This. (__GI___expf): This. (__ieee754_expf): Add strong_alias. (__expf_finite): Likewise. (__expf): New. Include <sysdeps/ieee754/flt-32/e_expf.c>.
2017-10-22 14:49:36 +00:00
Optimize accurate 64-bit routines for FMA4 on x86-64
2011-10-25 00:19:17 +00:00
libm-sysdep_routines += e_exp-fma4 e_log-fma4 e_pow-fma4 s_atan-fma4 \
math: Remove mpa files [BZ #15267] Finally remove all mpa related files, headers, declarations, probes, unused tables and update makefiles. Reviewed-By: Paul Zimmermann <Paul.Zimmermann@inria.fr>
2021-03-10 12:41:47 +00:00
e_asin-fma4 e_atan2-fma4 s_sin-fma4 s_tan-fma4
Optimize accurate 64-bit routines for FMA4 on x86-64
2011-10-25 00:19:17 +00:00
CFLAGS-e_asin-fma4.c = -mfma4
CFLAGS-e_atan2-fma4.c = -mfma4
CFLAGS-e_exp-fma4.c = -mfma4
CFLAGS-e_log-fma4.c = -mfma4
Add new pow implementation The algorithm is exp(y * log(x)), where log(x) is computed with about 1.3*2^-68 relative error (1.5*2^-68 without fma), returning the result in two doubles, and the exp part uses the same algorithm (and lookup tables) as exp, but takes the input as two doubles and a sign (to handle negative bases with odd integer exponent). The __exp1 internal symbol is no longer necessary. There is separate code path when fma is not available but the worst case error is about 0.54 ULP in both cases. The lookup table and consts for log are 4168 bytes. The .rodata+.text is decreased by 37908 bytes on aarch64. The non-nearest rounding error is less than 1 ULP. Improvements on Cortex-A72 compared to current glibc master: pow thruput: 2.40x in [0.01 11.1]x[0.01 11.1] pow latency: 1.84x in [0.01 11.1]x[0.01 11.1] Tested on aarch64-linux-gnu (defined __FP_FAST_FMA, TOINT_INTRINSICS) and arm-linux-gnueabihf (!defined __FP_FAST_FMA, !TOINT_INTRINSICS) and x86_64-linux-gnu (!defined __FP_FAST_FMA, !TOINT_INTRINSICS) and powerpc64le-linux-gnu (defined __FP_FAST_FMA, !TOINT_INTRINSICS) targets. * NEWS: Mention pow improvements. * math/Makefile (type-double-routines): Add e_pow_log_data. * sysdeps/generic/math_private.h (__exp1): Remove. * sysdeps/i386/fpu/e_pow_log_data.c: New file. * sysdeps/ia64/fpu/e_pow_log_data.c: New file. * sysdeps/ieee754/dbl-64/Makefile (CFLAGS-e_pow.c): Allow fma contraction. * sysdeps/ieee754/dbl-64/e_exp.c (__exp1): Remove. (exp_inline): Remove. (__ieee754_exp): Only single double input is handled. * sysdeps/ieee754/dbl-64/e_pow.c: Rewrite. * sysdeps/ieee754/dbl-64/e_pow_log_data.c: New file. * sysdeps/ieee754/dbl-64/math_config.h (issignaling_inline): Define. (__pow_log_data): Define. * sysdeps/ieee754/dbl-64/upow.h: Remove. * sysdeps/ieee754/dbl-64/upow.tbl: Remove. * sysdeps/m68k/m680x0/fpu/e_pow_log_data.c: New file. * sysdeps/x86_64/fpu/multiarch/Makefile (CFLAGS-e_pow-fma.c): Allow fma contraction. (CFLAGS-e_pow-fma4.c): Likewise.
2018-06-13 16:57:20 +00:00
CFLAGS-e_pow-fma4.c = -mfma4
Optimize accurate 64-bit routines for FMA4 on x86-64
2011-10-25 00:19:17 +00:00
CFLAGS-s_atan-fma4.c = -mfma4
Add AVX optimized versions for some x86-64 math functions
2011-10-26 01:34:55 +00:00
CFLAGS-s_sin-fma4.c = -mfma4
CFLAGS-s_tan-fma4.c = -mfma4
libm-sysdep_routines += e_exp-avx e_log-avx s_atan-avx \
math: Remove mpa files [BZ #15267] Finally remove all mpa related files, headers, declarations, probes, unused tables and update makefiles. Reviewed-By: Paul Zimmermann <Paul.Zimmermann@inria.fr>
2021-03-10 12:41:47 +00:00
e_atan2-avx s_sin-avx s_tan-avx
Add AVX optimized versions for some x86-64 math functions
2011-10-26 01:34:55 +00:00
Use -msse2avx option for x86-64 libm functions
2012-01-28 19:48:46 +00:00
CFLAGS-e_atan2-avx.c = -msse2avx -DSSE2AVX
CFLAGS-e_exp-avx.c = -msse2avx -DSSE2AVX
CFLAGS-e_log-avx.c = -msse2avx -DSSE2AVX
CFLAGS-s_atan-avx.c = -msse2avx -DSSE2AVX
CFLAGS-s_sin-avx.c = -msse2avx -DSSE2AVX
CFLAGS-s_tan-avx.c = -msse2avx -DSSE2AVX
Optimize accurate 64-bit routines for FMA4 on x86-64
2011-10-25 00:19:17 +00:00
endif
Start of series of patches with x86_64 vector math functions. Here is implementation of cos containing SSE, AVX, AVX2 and AVX512 versions according to Vector ABI which had been discussed in <https://groups.google.com/forum/#!topic/x86-64-abi/LmppCfN1rZ4>. Vector math library build and ABI testing enabled by default for x86_64. * sysdeps/x86_64/fpu/Makefile: New file. * sysdeps/x86_64/fpu/Versions: New file. * sysdeps/x86_64/fpu/svml_d_cos_data.S: New file. * sysdeps/x86_64/fpu/svml_d_cos_data.h: New file. * sysdeps/x86_64/fpu/svml_d_cos2_core.S: New file. * sysdeps/x86_64/fpu/svml_d_cos4_core.S: New file. * sysdeps/x86_64/fpu/svml_d_cos4_core_avx.S: New file. * sysdeps/x86_64/fpu/svml_d_cos8_core.S: New file. * sysdeps/x86_64/fpu/svml_d_wrapper_impl.h: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos2_core.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos2_core_sse4.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos4_core.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos4_core_avx2.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos8_core.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos8_core_avx512.S: New file. * sysdeps/x86_64/fpu/multiarch/Makefile (libmvec-sysdep_routines): Added build of SSE, AVX2 and AVX512 IFUNC versions. * sysdeps/x86/fpu/bits/math-vector.h: Added SIMD declaration for cos. * math/bits/mathcalls.h: Added cos declaration with __MATHCALL_VEC. * sysdeps/x86_64/configure.ac: Options for libmvec build. * sysdeps/x86_64/configure: Regenerated. * sysdeps/x86_64/sysdep.h (cfi_offset_rel_rsp): New macro. * sysdeps/unix/sysv/linux/x86_64/libmvec.abilist: New file. * manual/install.texi (Configuring and compiling): Document --disable-mathvec. * INSTALL: Regenerated. * NEWS: Mention addition of libmvec and x86_64 vector cos.
2015-06-09 11:25:49 +00:00
ifeq ($(subdir),mathvec)
x86-64: Add sysdeps/x86_64/fpu/Makeconfig 1. Add sysdeps/x86_64/fpu/Makeconfig to auto-generate libmvec.mk, which contains libmvec ABI test dependencies and CFLAGS, in the build directory. 2. Include libmvec.mk for libmvec ABI test dependencies and CFLAGS. Tested on SSE4, AVX, AVX2 and AVX512 machines. Reviewed-by: Noah Goldstein <goldstein.w.n@gmail.com>
2021-10-12 01:07:09 +00:00
libmvec-multiarch-double-func-list = \
2_core-sse2 \
2_core_sse4 \
4_core-sse \
4_core_avx2 \
8_core-avx2 \
8_core_avx512
libmvec-multiarch-float-func-list = \
f4_core-sse2 \
f4_core_sse4 \
f8_core-sse \
f8_core_avx2 \
f16_core-avx2 \
f16_core_avx512
libmvec-sysdep_routines += \
$(foreach l,$(libmvec-multiarch-double-func-list), \
$(addprefix svml_d_,$(addsuffix $(l),$(libmvec-funcs)))) \
$(foreach l,$(libmvec-multiarch-float-func-list), \
$(addprefix svml_s_,$(addsuffix $(l),$(libmvec-funcs))))
Start of series of patches with x86_64 vector math functions. Here is implementation of cos containing SSE, AVX, AVX2 and AVX512 versions according to Vector ABI which had been discussed in <https://groups.google.com/forum/#!topic/x86-64-abi/LmppCfN1rZ4>. Vector math library build and ABI testing enabled by default for x86_64. * sysdeps/x86_64/fpu/Makefile: New file. * sysdeps/x86_64/fpu/Versions: New file. * sysdeps/x86_64/fpu/svml_d_cos_data.S: New file. * sysdeps/x86_64/fpu/svml_d_cos_data.h: New file. * sysdeps/x86_64/fpu/svml_d_cos2_core.S: New file. * sysdeps/x86_64/fpu/svml_d_cos4_core.S: New file. * sysdeps/x86_64/fpu/svml_d_cos4_core_avx.S: New file. * sysdeps/x86_64/fpu/svml_d_cos8_core.S: New file. * sysdeps/x86_64/fpu/svml_d_wrapper_impl.h: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos2_core.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos2_core_sse4.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos4_core.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos4_core_avx2.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos8_core.S: New file. * sysdeps/x86_64/fpu/multiarch/svml_d_cos8_core_avx512.S: New file. * sysdeps/x86_64/fpu/multiarch/Makefile (libmvec-sysdep_routines): Added build of SSE, AVX2 and AVX512 IFUNC versions. * sysdeps/x86/fpu/bits/math-vector.h: Added SIMD declaration for cos. * math/bits/mathcalls.h: Added cos declaration with __MATHCALL_VEC. * sysdeps/x86_64/configure.ac: Options for libmvec build. * sysdeps/x86_64/configure: Regenerated. * sysdeps/x86_64/sysdep.h (cfi_offset_rel_rsp): New macro. * sysdeps/unix/sysv/linux/x86_64/libmvec.abilist: New file. * manual/install.texi (Configuring and compiling): Document --disable-mathvec. * INSTALL: Regenerated. * NEWS: Mention addition of libmvec and x86_64 vector cos.
2015-06-09 11:25:49 +00:00
endif
Reference in New Issue Copy Permalink