2014-01-24 13:55:16 +03:00
|
|
|
/*
|
2014-07-01 15:38:57 +03:00
|
|
|
* This file is part of FFmpeg.
|
2014-01-24 13:55:16 +03:00
|
|
|
*
|
2014-07-01 15:38:57 +03:00
|
|
|
* FFmpeg is free software; you can redistribute it and/or
|
2014-01-24 13:55:16 +03:00
|
|
|
* 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.
|
|
|
|
*
|
2014-07-01 15:38:57 +03:00
|
|
|
* FFmpeg is distributed in the hope that it will be useful,
|
2014-01-24 13:55:16 +03:00
|
|
|
* 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
|
2014-07-01 15:38:57 +03:00
|
|
|
* License along with FFmpeg; if not, write to the Free Software
|
2014-01-24 13:55:16 +03:00
|
|
|
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "config.h"
|
|
|
|
#include "libavutil/attributes.h"
|
|
|
|
#include "libavutil/cpu.h"
|
|
|
|
#include "libavutil/x86/cpu.h"
|
|
|
|
#include "libavcodec/avcodec.h"
|
|
|
|
#include "libavcodec/idctdsp.h"
|
|
|
|
#include "idctdsp.h"
|
2014-02-18 15:20:01 +03:00
|
|
|
#include "simple_idct.h"
|
2014-01-24 13:55:16 +03:00
|
|
|
|
|
|
|
/* Input permutation for the simple_idct_mmx */
|
|
|
|
static const uint8_t simple_mmx_permutation[64] = {
|
|
|
|
0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D,
|
|
|
|
0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D,
|
|
|
|
0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D,
|
|
|
|
0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F,
|
|
|
|
0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F,
|
|
|
|
0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D,
|
|
|
|
0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F,
|
|
|
|
0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
|
|
|
|
};
|
|
|
|
|
2014-08-08 13:10:05 +03:00
|
|
|
static const uint8_t idct_sse2_row_perm[8] = { 0, 4, 1, 5, 2, 6, 3, 7 };
|
|
|
|
|
2014-01-24 13:55:16 +03:00
|
|
|
av_cold int ff_init_scantable_permutation_x86(uint8_t *idct_permutation,
|
2014-02-18 13:30:55 +03:00
|
|
|
enum idct_permutation_type perm_type)
|
2014-01-24 13:55:16 +03:00
|
|
|
{
|
|
|
|
int i;
|
|
|
|
|
2014-02-18 13:30:55 +03:00
|
|
|
switch (perm_type) {
|
|
|
|
case FF_IDCT_PERM_SIMPLE:
|
2014-01-24 13:55:16 +03:00
|
|
|
for (i = 0; i < 64; i++)
|
|
|
|
idct_permutation[i] = simple_mmx_permutation[i];
|
|
|
|
return 1;
|
2014-08-08 13:10:05 +03:00
|
|
|
case FF_IDCT_PERM_SSE2:
|
|
|
|
for (i = 0; i < 64; i++)
|
|
|
|
idct_permutation[i] = (i & 0x38) | idct_sse2_row_perm[i & 7];
|
|
|
|
return 1;
|
2014-01-24 13:55:16 +03:00
|
|
|
}
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
av_cold void ff_idctdsp_init_x86(IDCTDSPContext *c, AVCodecContext *avctx,
|
|
|
|
unsigned high_bit_depth)
|
|
|
|
{
|
|
|
|
int cpu_flags = av_get_cpu_flags();
|
|
|
|
|
2014-07-01 15:38:57 +03:00
|
|
|
if (EXTERNAL_MMX(cpu_flags)) {
|
|
|
|
c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_mmx;
|
2014-09-25 01:53:07 +03:00
|
|
|
c->put_pixels_clamped = ff_put_pixels_clamped_mmx;
|
|
|
|
c->add_pixels_clamped = ff_add_pixels_clamped_mmx;
|
2017-04-04 18:42:58 +02:00
|
|
|
|
|
|
|
if (INLINE_MMX(cpu_flags)) {
|
|
|
|
if (!high_bit_depth &&
|
|
|
|
avctx->lowres == 0 &&
|
|
|
|
(avctx->idct_algo == FF_IDCT_AUTO ||
|
|
|
|
avctx->idct_algo == FF_IDCT_SIMPLEAUTO ||
|
|
|
|
avctx->idct_algo == FF_IDCT_SIMPLEMMX)) {
|
|
|
|
c->idct_put = ff_simple_idct_put_mmx;
|
|
|
|
c->idct_add = ff_simple_idct_add_mmx;
|
|
|
|
c->idct = ff_simple_idct_mmx;
|
|
|
|
c->perm_type = FF_IDCT_PERM_SIMPLE;
|
|
|
|
}
|
|
|
|
}
|
2014-07-01 15:38:57 +03:00
|
|
|
}
|
2017-04-04 18:42:58 +02:00
|
|
|
|
2014-07-01 15:38:57 +03:00
|
|
|
if (EXTERNAL_SSE2(cpu_flags)) {
|
|
|
|
c->put_signed_pixels_clamped = ff_put_signed_pixels_clamped_sse2;
|
2014-09-25 01:53:07 +03:00
|
|
|
c->put_pixels_clamped = ff_put_pixels_clamped_sse2;
|
|
|
|
c->add_pixels_clamped = ff_add_pixels_clamped_sse2;
|
2017-04-04 18:42:58 +02:00
|
|
|
|
|
|
|
if (INLINE_SSE2(cpu_flags)) {
|
|
|
|
if (!high_bit_depth &&
|
|
|
|
avctx->lowres == 0 &&
|
|
|
|
(avctx->idct_algo == FF_IDCT_AUTO ||
|
|
|
|
avctx->idct_algo == FF_IDCT_SIMPLEAUTO ||
|
|
|
|
avctx->idct_algo == FF_IDCT_SIMPLEMMX)) {
|
|
|
|
c->idct_put = ff_simple_idct_put_sse2;
|
|
|
|
c->idct_add = ff_simple_idct_add_sse2;
|
|
|
|
c->perm_type = FF_IDCT_PERM_SIMPLE;
|
|
|
|
}
|
|
|
|
}
|
2014-07-01 15:38:57 +03:00
|
|
|
}
|
2015-10-12 19:37:47 +02:00
|
|
|
|
x86: simple_idct: 12bits versions
On 12 frames of a 444p 12 bits DNxHR sequence, _put function:
C: 78902 decicycles in idct, 262071 runs, 73 skips
avx: 32478 decicycles in idct, 262045 runs, 99 skips
Difference between the 2:
stddev: 0.39 PSNR:104.47 MAXDIFF: 2
This is unavoidable and due to the scale factors used in the x86
version, which cannot match the C ones.
In addition, the trick of adding an initial bias to the input of a
pass can overflow, as the input coefficients are already 15bits,
which is the maximum this function can handle.
Overall, however, the omse on 12 bits samples goes from 0.16916 to
0.16883. Reducing rowshift by 1 improves to 0.0908, but causes
overflows.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-12 19:37:49 +02:00
|
|
|
if (ARCH_X86_64 && avctx->lowres == 0) {
|
|
|
|
if (avctx->bits_per_raw_sample == 10 &&
|
2017-04-05 13:49:14 +02:00
|
|
|
(avctx->idct_algo == FF_IDCT_AUTO ||
|
|
|
|
avctx->idct_algo == FF_IDCT_SIMPLEAUTO ||
|
|
|
|
avctx->idct_algo == FF_IDCT_SIMPLE)) {
|
|
|
|
if (EXTERNAL_SSE2(cpu_flags)) {
|
|
|
|
c->idct_put = ff_simple_idct10_put_sse2;
|
|
|
|
c->idct_add = NULL;
|
|
|
|
c->idct = ff_simple_idct10_sse2;
|
|
|
|
c->perm_type = FF_IDCT_PERM_TRANSPOSE;
|
2015-10-12 19:37:47 +02:00
|
|
|
|
2017-04-05 13:49:14 +02:00
|
|
|
}
|
|
|
|
if (EXTERNAL_AVX(cpu_flags)) {
|
|
|
|
c->idct_put = ff_simple_idct10_put_avx;
|
|
|
|
c->idct_add = NULL;
|
|
|
|
c->idct = ff_simple_idct10_avx;
|
|
|
|
c->perm_type = FF_IDCT_PERM_TRANSPOSE;
|
|
|
|
}
|
x86: simple_idct: 12bits versions
On 12 frames of a 444p 12 bits DNxHR sequence, _put function:
C: 78902 decicycles in idct, 262071 runs, 73 skips
avx: 32478 decicycles in idct, 262045 runs, 99 skips
Difference between the 2:
stddev: 0.39 PSNR:104.47 MAXDIFF: 2
This is unavoidable and due to the scale factors used in the x86
version, which cannot match the C ones.
In addition, the trick of adding an initial bias to the input of a
pass can overflow, as the input coefficients are already 15bits,
which is the maximum this function can handle.
Overall, however, the omse on 12 bits samples goes from 0.16916 to
0.16883. Reducing rowshift by 1 improves to 0.0908, but causes
overflows.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
2015-10-12 19:37:49 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
if (avctx->bits_per_raw_sample == 12 &&
|
|
|
|
(avctx->idct_algo == FF_IDCT_AUTO ||
|
|
|
|
avctx->idct_algo == FF_IDCT_SIMPLEMMX)) {
|
|
|
|
if (EXTERNAL_SSE2(cpu_flags)) {
|
|
|
|
c->idct_put = ff_simple_idct12_put_sse2;
|
|
|
|
c->idct_add = NULL;
|
|
|
|
c->idct = ff_simple_idct12_sse2;
|
|
|
|
c->perm_type = FF_IDCT_PERM_TRANSPOSE;
|
|
|
|
}
|
|
|
|
if (EXTERNAL_AVX(cpu_flags)) {
|
|
|
|
c->idct_put = ff_simple_idct12_put_avx;
|
|
|
|
c->idct_add = NULL;
|
|
|
|
c->idct = ff_simple_idct12_avx;
|
|
|
|
c->perm_type = FF_IDCT_PERM_TRANSPOSE;
|
|
|
|
}
|
|
|
|
}
|
2015-10-12 19:37:47 +02:00
|
|
|
}
|
2014-01-24 13:55:16 +03:00
|
|
|
}
|