This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [PATCH 4/4] powerpc: Remove f{max,min}{f} assembly implementations
PING.
On 19/12/2016 12:04, Adhemerval Zanella wrote:
> This patch removes the powerpc assembly implementation of fmax/fmin.
> Based on benchtests, the assembly ones shows:
>
> $ ./testrun.sh benchtests/bench-fmax
> "fmax": {
> "": {
> "duration": 5.07586e+09,
> "iterations": 2.01676e+09,
> "max": 1350.39,
> "min": 2.073,
> "mean": 2.51684
> },
> "qNaN": {
> "duration": 5.09315e+09,
> "iterations": 8.4568e+08,
> "max": 2788,
> "min": 5.806,
> "mean": 6.02255
> },
> "sNaN": {
> "duration": 5.09073e+09,
> "iterations": 8.42316e+08,
> "max": 4215.84,
> "min": 5.737,
> "mean": 6.04373
> }
>
> And
>
> $ ./testrun.sh benchtests/bench-fmin
> "fmin": {
> "": {
> "duration": 5.07711e+09,
> "iterations": 2.02982e+09,
> "max": 497.094,
> "min": 2.073,
> "mean": 2.50126
> },
> "qNaN": {
> "duration": 5.09134e+09,
> "iterations": 8.46968e+08,
> "max": 2255.14,
> "min": 5.807,
> "mean": 6.01125
> },
> "sNaN": {
> "duration": 5.09122e+09,
> "iterations": 8.4746e+08,
> "max": 1969.38,
> "min": 5.729,
> "mean": 6.00763
> }
> }
>
> The default implementation (math/s_f{max.min}_template.c) shows slight better
> latency for all cases:
>
> $ ./testrun.sh benchtests/bench-fmax
> "fmax": {
> "": {
> "duration": 5.07044e+09,
> "iterations": 2.38695e+09,
> "max": 2048.58,
> "min": 2.073,
> "mean": 2.12423
> },
> "qNaN": {
> "duration": 5.09004e+09,
> "iterations": 9.45428e+08,
> "max": 3306.93,
> "min": 5.138,
> "mean": 5.38385
> },
> "sNaN": {
> "duration": 5.08458e+09,
> "iterations": 1.15959e+09,
> "max": 972.008,
> "min": 3.321,
> "mean": 4.3848
> }
> }
>
> And:
>
> $ ./testrun.sh benchtests/bench-fmin
> "fmin": {
> "": {
> "duration": 5.06817e+09,
> "iterations": 2.3913e+09,
> "max": 1177.9,
> "min": 2.073,
> "mean": 2.11942
> },
> "qNaN": {
> "duration": 5.08857e+09,
> "iterations": 9.45656e+08,
> "max": 2658.83,
> "min": 5.09,
> "mean": 5.38099
> },
> "sNaN": {
> "duration": 5.08093e+09,
> "iterations": 1.16725e+09,
> "max": 1030.74,
> "min": 3.323,
> "mean": 4.3529
> }
> }
>
> Both were run with GCC 5.4 (ubuntu 16 default installation) using default
> compiler flags on POWER8E 3.4GHz (powerpc64le-linux-gnu).
> ---
> ChangeLog | 10 +++++
> sysdeps/powerpc/fpu/s_fmax.S | 77 ----------------------------------
> sysdeps/powerpc/fpu/s_fmaxf.S | 1 -
> sysdeps/powerpc/fpu/s_fmin.S | 77 ----------------------------------
> sysdeps/powerpc/fpu/s_fminf.S | 1 -
> sysdeps/powerpc/powerpc32/fpu/s_fmax.S | 5 ---
> sysdeps/powerpc/powerpc32/fpu/s_fmin.S | 5 ---
> sysdeps/powerpc/powerpc64/fpu/s_fmax.S | 5 ---
> sysdeps/powerpc/powerpc64/fpu/s_fmin.S | 5 ---
> 9 files changed, 10 insertions(+), 176 deletions(-)
> delete mode 100644 sysdeps/powerpc/fpu/s_fmax.S
> delete mode 100644 sysdeps/powerpc/fpu/s_fmaxf.S
> delete mode 100644 sysdeps/powerpc/fpu/s_fmin.S
> delete mode 100644 sysdeps/powerpc/fpu/s_fminf.S
> delete mode 100644 sysdeps/powerpc/powerpc32/fpu/s_fmax.S
> delete mode 100644 sysdeps/powerpc/powerpc32/fpu/s_fmin.S
> delete mode 100644 sysdeps/powerpc/powerpc64/fpu/s_fmax.S
> delete mode 100644 sysdeps/powerpc/powerpc64/fpu/s_fmin.S
>
> diff --git a/sysdeps/powerpc/fpu/s_fmax.S b/sysdeps/powerpc/fpu/s_fmax.S
> deleted file mode 100644
> index e6405c0..0000000
> --- a/sysdeps/powerpc/fpu/s_fmax.S
> +++ /dev/null
> @@ -1,77 +0,0 @@
> -/* Floating-point maximum. PowerPC version.
> - Copyright (C) 1997-2016 Free Software Foundation, Inc.
> - This file is part of the GNU C Library.
> -
> - The GNU C Library is free software; you can redistribute it and/or
> - modify it under the terms of the GNU Lesser General Public
> - License as published by the Free Software Foundation; either
> - version 2.1 of the License, or (at your option) any later version.
> -
> - The GNU C Library is distributed in the hope that it will be useful,
> - but WITHOUT ANY WARRANTY; without even the implied warranty of
> - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> - Lesser General Public License for more details.
> -
> - You should have received a copy of the GNU Lesser General Public
> - License along with the GNU C Library; if not, see
> - <http://www.gnu.org/licenses/>. */
> -
> -#include <sysdep.h>
> -
> -ENTRY(__fmax)
> -/* double [f1] fmax (double [f1] x, double [f2] y); */
> - fcmpu cr0,fp1,fp2
> - blt cr0,0f /* if x < y, neither x nor y can be NaN... */
> - bnulr+ cr0
> -/* x and y are unordered, so one of x or y must be a NaN... */
> - fcmpu cr1,fp2,fp2
> - bun cr1,1f
> -/* x is a NaN; y is not. Test if x is signaling. */
> -#ifdef __powerpc64__
> - stfd fp1,-8(r1)
> - lwz r3,-8+HIWORD(r1)
> -#else
> - stwu r1,-16(r1)
> - cfi_adjust_cfa_offset (16)
> - stfd fp1,8(r1)
> - lwz r3,8+HIWORD(r1)
> - addi r1,r1,16
> - cfi_adjust_cfa_offset (-16)
> -#endif
> - andis. r3,r3,8
> - bne cr0,0f
> - b 2f
> -1: /* y is a NaN; x may or may not be. */
> - fcmpu cr1,fp1,fp1
> - bun cr1,2f
> -/* y is a NaN; x is not. Test if y is signaling. */
> -#ifdef __powerpc64__
> - stfd fp2,-8(r1)
> - lwz r3,-8+HIWORD(r1)
> -#else
> - stwu r1,-16(r1)
> - cfi_adjust_cfa_offset (16)
> - stfd fp2,8(r1)
> - lwz r3,8+HIWORD(r1)
> - addi r1,r1,16
> - cfi_adjust_cfa_offset (-16)
> -#endif
> - andis. r3,r3,8
> - bnelr cr0
> -2: /* x and y are NaNs, or one is a signaling NaN. */
> - fadd fp1,fp1,fp2
> - blr
> -0: fmr fp1,fp2
> - blr
> -END(__fmax)
> -
> -weak_alias (__fmax,fmax)
> -
> -/* It turns out that it's safe to use this code even for single-precision. */
> -strong_alias(__fmax,__fmaxf)
> -weak_alias (__fmax,fmaxf)
> -
> -#ifdef NO_LONG_DOUBLE
> -weak_alias (__fmax,__fmaxl)
> -weak_alias (__fmax,fmaxl)
> -#endif
> diff --git a/sysdeps/powerpc/fpu/s_fmaxf.S b/sysdeps/powerpc/fpu/s_fmaxf.S
> deleted file mode 100644
> index 3c2d62b..0000000
> --- a/sysdeps/powerpc/fpu/s_fmaxf.S
> +++ /dev/null
> @@ -1 +0,0 @@
> -/* __fmaxf is in s_fmax.c */
> diff --git a/sysdeps/powerpc/fpu/s_fmin.S b/sysdeps/powerpc/fpu/s_fmin.S
> deleted file mode 100644
> index 9ae77fe..0000000
> --- a/sysdeps/powerpc/fpu/s_fmin.S
> +++ /dev/null
> @@ -1,77 +0,0 @@
> -/* Floating-point minimum. PowerPC version.
> - Copyright (C) 1997-2016 Free Software Foundation, Inc.
> - This file is part of the GNU C Library.
> -
> - The GNU C Library is free software; you can redistribute it and/or
> - modify it under the terms of the GNU Lesser General Public
> - License as published by the Free Software Foundation; either
> - version 2.1 of the License, or (at your option) any later version.
> -
> - The GNU C Library is distributed in the hope that it will be useful,
> - but WITHOUT ANY WARRANTY; without even the implied warranty of
> - MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
> - Lesser General Public License for more details.
> -
> - You should have received a copy of the GNU Lesser General Public
> - License along with the GNU C Library; if not, see
> - <http://www.gnu.org/licenses/>. */
> -
> -#include <sysdep.h>
> -
> -ENTRY(__fmin)
> -/* double [f1] fmin (double [f1] x, double [f2] y); */
> - fcmpu cr0,fp1,fp2
> - bgt cr0,0f /* if x > y, neither x nor y can be NaN... */
> - bnulr+ cr0
> -/* x and y are unordered, so one of x or y must be a NaN... */
> - fcmpu cr1,fp2,fp2
> - bun cr1,1f
> -/* x is a NaN; y is not. Test if x is signaling. */
> -#ifdef __powerpc64__
> - stfd fp1,-8(r1)
> - lwz r3,-8+HIWORD(r1)
> -#else
> - stwu r1,-16(r1)
> - cfi_adjust_cfa_offset (16)
> - stfd fp1,8(r1)
> - lwz r3,8+HIWORD(r1)
> - addi r1,r1,16
> - cfi_adjust_cfa_offset (-16)
> -#endif
> - andis. r3,r3,8
> - bne cr0,0f
> - b 2f
> -1: /* y is a NaN; x may or may not be. */
> - fcmpu cr1,fp1,fp1
> - bun cr1,2f
> -/* y is a NaN; x is not. Test if y is signaling. */
> -#ifdef __powerpc64__
> - stfd fp2,-8(r1)
> - lwz r3,-8+HIWORD(r1)
> -#else
> - stwu r1,-16(r1)
> - cfi_adjust_cfa_offset (16)
> - stfd fp2,8(r1)
> - lwz r3,8+HIWORD(r1)
> - addi r1,r1,16
> - cfi_adjust_cfa_offset (-16)
> -#endif
> - andis. r3,r3,8
> - bnelr cr0
> -2: /* x and y are NaNs, or one is a signaling NaN. */
> - fadd fp1,fp1,fp2
> - blr
> -0: fmr fp1,fp2
> - blr
> -END(__fmin)
> -
> -weak_alias (__fmin,fmin)
> -
> -/* It turns out that it's safe to use this code even for single-precision. */
> -strong_alias(__fmin,__fminf)
> -weak_alias (__fmin,fminf)
> -
> -#ifdef NO_LONG_DOUBLE
> -weak_alias (__fmin,__fminl)
> -weak_alias (__fmin,fminl)
> -#endif
> diff --git a/sysdeps/powerpc/fpu/s_fminf.S b/sysdeps/powerpc/fpu/s_fminf.S
> deleted file mode 100644
> index 10ab7fe..0000000
> --- a/sysdeps/powerpc/fpu/s_fminf.S
> +++ /dev/null
> @@ -1 +0,0 @@
> -/* __fminf is in s_fmin.c */
> diff --git a/sysdeps/powerpc/powerpc32/fpu/s_fmax.S b/sysdeps/powerpc/powerpc32/fpu/s_fmax.S
> deleted file mode 100644
> index 6973576..0000000
> --- a/sysdeps/powerpc/powerpc32/fpu/s_fmax.S
> +++ /dev/null
> @@ -1,5 +0,0 @@
> -#include <math_ldbl_opt.h>
> -#include <sysdeps/powerpc/fpu/s_fmax.S>
> -#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
> -compat_symbol (libm, __fmax, fmaxl, GLIBC_2_1)
> -#endif
> diff --git a/sysdeps/powerpc/powerpc32/fpu/s_fmin.S b/sysdeps/powerpc/powerpc32/fpu/s_fmin.S
> deleted file mode 100644
> index 6d4a0a9..0000000
> --- a/sysdeps/powerpc/powerpc32/fpu/s_fmin.S
> +++ /dev/null
> @@ -1,5 +0,0 @@
> -#include <math_ldbl_opt.h>
> -#include <sysdeps/powerpc/fpu/s_fmin.S>
> -#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
> -compat_symbol (libm, __fmin, fminl, GLIBC_2_1)
> -#endif
> diff --git a/sysdeps/powerpc/powerpc64/fpu/s_fmax.S b/sysdeps/powerpc/powerpc64/fpu/s_fmax.S
> deleted file mode 100644
> index 6973576..0000000
> --- a/sysdeps/powerpc/powerpc64/fpu/s_fmax.S
> +++ /dev/null
> @@ -1,5 +0,0 @@
> -#include <math_ldbl_opt.h>
> -#include <sysdeps/powerpc/fpu/s_fmax.S>
> -#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
> -compat_symbol (libm, __fmax, fmaxl, GLIBC_2_1)
> -#endif
> diff --git a/sysdeps/powerpc/powerpc64/fpu/s_fmin.S b/sysdeps/powerpc/powerpc64/fpu/s_fmin.S
> deleted file mode 100644
> index 6d4a0a9..0000000
> --- a/sysdeps/powerpc/powerpc64/fpu/s_fmin.S
> +++ /dev/null
> @@ -1,5 +0,0 @@
> -#include <math_ldbl_opt.h>
> -#include <sysdeps/powerpc/fpu/s_fmin.S>
> -#if LONG_DOUBLE_COMPAT(libm, GLIBC_2_1)
> -compat_symbol (libm, __fmin, fminl, GLIBC_2_1)
> -#endif
>