2013-06-18 21:30:42 +00:00
|
|
|
/*
|
|
|
|
* linear least squares model
|
|
|
|
*
|
|
|
|
* Copyright (c) 2013 Loren Merritt
|
|
|
|
*
|
2013-06-30 09:30:07 +00:00
|
|
|
* This file is part of FFmpeg.
|
2013-06-18 21:30:42 +00:00
|
|
|
*
|
2013-06-30 09:30:07 +00:00
|
|
|
* FFmpeg is free software; you can redistribute it and/or
|
2013-06-18 21:30:42 +00:00
|
|
|
* 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.
|
|
|
|
*
|
2013-06-30 09:30:07 +00:00
|
|
|
* FFmpeg is distributed in the hope that it will be useful,
|
2013-06-18 21:30:42 +00:00
|
|
|
* 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
|
2013-06-30 09:30:07 +00:00
|
|
|
* License along with FFmpeg; if not, write to the Free Software
|
2013-06-18 21:30:42 +00:00
|
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
|
|
*/
|
|
|
|
|
2014-08-09 21:03:26 +00:00
|
|
|
#include "libavutil/lls.h"
|
2013-06-18 21:30:42 +00:00
|
|
|
#include "libavutil/x86/cpu.h"
|
|
|
|
|
2014-09-28 13:44:38 +00:00
|
|
|
void ff_update_lls_sse2(LLSModel *m, const double *var);
|
|
|
|
void ff_update_lls_avx(LLSModel *m, const double *var);
|
lavu/x86/lls: add fma3 optimizations for update_lls
This improves accuracy (very slightly) and speed for processors having
fma3.
Sample benchmark (fate flac-16-lpc-cholesky, Haswell):
old:
5993610 decicycles in ff_lpc_calc_coefs, 64 runs, 0 skips
5951528 decicycles in ff_lpc_calc_coefs, 128 runs, 0 skips
new:
5252410 decicycles in ff_lpc_calc_coefs, 64 runs, 0 skips
5232869 decicycles in ff_lpc_calc_coefs, 128 runs, 0 skips
Tested with FATE and --disable-fma3, also examined contents of
lavu/lls-test.
Reviewed-by: James Almer <jamrial@gmail.com>
Reviewed-by: Henrik Gramner <henrik@gramner.com>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
2016-01-13 22:59:26 +00:00
|
|
|
void ff_update_lls_fma3(LLSModel *m, const double *var);
|
2014-09-28 13:44:38 +00:00
|
|
|
double ff_evaluate_lls_sse2(LLSModel *m, const double *var, int order);
|
2013-06-18 21:30:42 +00:00
|
|
|
|
2014-08-09 21:03:26 +00:00
|
|
|
av_cold void ff_init_lls_x86(LLSModel *m)
|
2013-06-18 21:30:42 +00:00
|
|
|
{
|
|
|
|
int cpu_flags = av_get_cpu_flags();
|
|
|
|
if (EXTERNAL_SSE2(cpu_flags)) {
|
|
|
|
m->update_lls = ff_update_lls_sse2;
|
2013-07-01 00:27:47 +00:00
|
|
|
if (m->indep_count >= 4)
|
2013-06-18 21:30:43 +00:00
|
|
|
m->evaluate_lls = ff_evaluate_lls_sse2;
|
2013-06-18 21:30:42 +00:00
|
|
|
}
|
2015-05-31 17:20:29 +00:00
|
|
|
if (EXTERNAL_AVX_FAST(cpu_flags)) {
|
2013-06-18 21:30:42 +00:00
|
|
|
m->update_lls = ff_update_lls_avx;
|
|
|
|
}
|
2016-02-07 03:52:06 +00:00
|
|
|
if (EXTERNAL_FMA3_FAST(cpu_flags)) {
|
lavu/x86/lls: add fma3 optimizations for update_lls
This improves accuracy (very slightly) and speed for processors having
fma3.
Sample benchmark (fate flac-16-lpc-cholesky, Haswell):
old:
5993610 decicycles in ff_lpc_calc_coefs, 64 runs, 0 skips
5951528 decicycles in ff_lpc_calc_coefs, 128 runs, 0 skips
new:
5252410 decicycles in ff_lpc_calc_coefs, 64 runs, 0 skips
5232869 decicycles in ff_lpc_calc_coefs, 128 runs, 0 skips
Tested with FATE and --disable-fma3, also examined contents of
lavu/lls-test.
Reviewed-by: James Almer <jamrial@gmail.com>
Reviewed-by: Henrik Gramner <henrik@gramner.com>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
2016-01-13 22:59:26 +00:00
|
|
|
m->update_lls = ff_update_lls_fma3;
|
|
|
|
}
|
2013-06-18 21:30:42 +00:00
|
|
|
}
|