RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity
ffmpeg/libavcodec/x86/dsputilenc_mmx.c

1062 lines
30 KiB
C
Raw Normal View History

split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
/*
* MMX optimized DSP utils
cosmetics: Remove pointless period after copyright statement non-sentences. Originally committed as revision 16684 to svn://svn.ffmpeg.org/ffmpeg/trunk
2009-01-19 15:46:40 +00:00
* Copyright (c) 2000, 2001 Fabrice Bellard
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
* Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
*
Move misplaced file author information where it belongs
2013-04-10 17:26:49 +00:00
* MMX optimization by Nick Kurshev <nickols_k@mail.ru>
*
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
* This file is part of FFmpeg.
*
* FFmpeg 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.
*
* FFmpeg 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 FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
Add av_cold attributes to arch-specific init functions
2013-02-01 09:31:59 +00:00
#include "libavutil/attributes.h"
Move mm_support() from libavcodec to libavutil, make it a public function and rename it to av_get_cpu_flags(). Originally committed as revision 25076 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-09-08 15:07:14 +00:00
#include "libavutil/cpu.h"
x86: rename libavutil/x86_cpu.h to libavutil/x86/asm.h This puts x86-specific things in the x86/ subdirectory where they belong. Signed-off-by: Mans Rullgard <mans@mansr.com>
2012-08-08 12:51:52 +00:00
#include "libavutil/x86/asm.h"
x86: Replace checks for CPU extensions and flags by convenience macros This separates code relying on inline from that relying on external assembly and fixes instances where the coalesced check was incorrect.
2012-08-29 17:01:05 +00:00
#include "libavutil/x86/cpu.h"
dsputil: Move fdct function declarations to dct.h
2013-02-07 20:58:48 +00:00
#include "libavcodec/dct.h"
Use full path for #includes from another directory. Originally committed as revision 13098 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-09 11:56:36 +00:00
#include "libavcodec/dsputil.h"
#include "libavcodec/mpegvideo.h"
moves mid_pred() into mathops.h (with arch specific code split by directory) Originally committed as revision 16681 to svn://svn.ffmpeg.org/ffmpeg/trunk
2009-01-18 22:57:40 +00:00
#include "libavcodec/mathops.h"
x86: dsputil: Rename dsputil_mmx.h --> dsputil_x86.h The header is not (anymore) MMX-specific.
2013-05-07 23:52:57 +00:00
#include "dsputil_x86.h"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
Drop DCTELEM typedef It does not help as an abstraction and adds dsputil dependencies. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
2013-01-20 00:02:29 +00:00
void ff_get_pixels_mmx(int16_t *block, const uint8_t *pixels, int line_size);
void ff_get_pixels_sse2(int16_t *block, const uint8_t *pixels, int line_size);
void ff_diff_pixels_mmx(int16_t *block, const uint8_t *s1, const uint8_t *s2, int stride);
dsputilenc: x86: Convert pixel inline asm to yasm Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
2013-01-16 07:41:39 +00:00
int ff_pix_sum16_mmx(uint8_t * pix, int line_size);
int ff_pix_norm1_mmx(uint8_t *pix, int line_size);
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
#if HAVE_INLINE_ASM
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
static int sse8_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h) {
int tmp;
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile (
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"movl %4,%%ecx\n"
"shr $1,%%ecx\n"
"pxor %%mm0,%%mm0\n" /* mm0 = 0 */
"pxor %%mm7,%%mm7\n" /* mm7 holds the sum */
"1:\n"
"movq (%0),%%mm1\n" /* mm1 = pix1[0][0-7] */
"movq (%1),%%mm2\n" /* mm2 = pix2[0][0-7] */
"movq (%0,%3),%%mm3\n" /* mm3 = pix1[1][0-7] */
"movq (%1,%3),%%mm4\n" /* mm4 = pix2[1][0-7] */
/* todo: mm1-mm2, mm3-mm4 */
/* algo: subtract mm1 from mm2 with saturation and vice versa */
/* OR the results to get absolute difference */
"movq %%mm1,%%mm5\n"
"movq %%mm3,%%mm6\n"
"psubusb %%mm2,%%mm1\n"
"psubusb %%mm4,%%mm3\n"
"psubusb %%mm5,%%mm2\n"
"psubusb %%mm6,%%mm4\n"
"por %%mm1,%%mm2\n"
"por %%mm3,%%mm4\n"
/* now convert to 16-bit vectors so we can square them */
"movq %%mm2,%%mm1\n"
"movq %%mm4,%%mm3\n"
"punpckhbw %%mm0,%%mm2\n"
"punpckhbw %%mm0,%%mm4\n"
"punpcklbw %%mm0,%%mm1\n" /* mm1 now spread over (mm1,mm2) */
"punpcklbw %%mm0,%%mm3\n" /* mm4 now spread over (mm3,mm4) */
"pmaddwd %%mm2,%%mm2\n"
"pmaddwd %%mm4,%%mm4\n"
"pmaddwd %%mm1,%%mm1\n"
"pmaddwd %%mm3,%%mm3\n"
"lea (%0,%3,2), %0\n" /* pix1 += 2*line_size */
"lea (%1,%3,2), %1\n" /* pix2 += 2*line_size */
"paddd %%mm2,%%mm1\n"
"paddd %%mm4,%%mm3\n"
"paddd %%mm1,%%mm7\n"
"paddd %%mm3,%%mm7\n"
"decl %%ecx\n"
"jnz 1b\n"
"movq %%mm7,%%mm1\n"
"psrlq $32, %%mm7\n" /* shift hi dword to lo */
"paddd %%mm7,%%mm1\n"
"movd %%mm1,%2\n"
: "+r" (pix1), "+r" (pix2), "=r"(tmp)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r" ((x86_reg)line_size) , "m" (h)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
: "%ecx");
return tmp;
}
static int sse16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h) {
int tmp;
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile (
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"movl %4,%%ecx\n"
"pxor %%mm0,%%mm0\n" /* mm0 = 0 */
"pxor %%mm7,%%mm7\n" /* mm7 holds the sum */
"1:\n"
"movq (%0),%%mm1\n" /* mm1 = pix1[0-7] */
"movq (%1),%%mm2\n" /* mm2 = pix2[0-7] */
"movq 8(%0),%%mm3\n" /* mm3 = pix1[8-15] */
"movq 8(%1),%%mm4\n" /* mm4 = pix2[8-15] */
/* todo: mm1-mm2, mm3-mm4 */
/* algo: subtract mm1 from mm2 with saturation and vice versa */
/* OR the results to get absolute difference */
"movq %%mm1,%%mm5\n"
"movq %%mm3,%%mm6\n"
"psubusb %%mm2,%%mm1\n"
"psubusb %%mm4,%%mm3\n"
"psubusb %%mm5,%%mm2\n"
"psubusb %%mm6,%%mm4\n"
"por %%mm1,%%mm2\n"
"por %%mm3,%%mm4\n"
/* now convert to 16-bit vectors so we can square them */
"movq %%mm2,%%mm1\n"
"movq %%mm4,%%mm3\n"
"punpckhbw %%mm0,%%mm2\n"
"punpckhbw %%mm0,%%mm4\n"
"punpcklbw %%mm0,%%mm1\n" /* mm1 now spread over (mm1,mm2) */
"punpcklbw %%mm0,%%mm3\n" /* mm4 now spread over (mm3,mm4) */
"pmaddwd %%mm2,%%mm2\n"
"pmaddwd %%mm4,%%mm4\n"
"pmaddwd %%mm1,%%mm1\n"
"pmaddwd %%mm3,%%mm3\n"
"add %3,%0\n"
"add %3,%1\n"
"paddd %%mm2,%%mm1\n"
"paddd %%mm4,%%mm3\n"
"paddd %%mm1,%%mm7\n"
"paddd %%mm3,%%mm7\n"
"decl %%ecx\n"
"jnz 1b\n"
"movq %%mm7,%%mm1\n"
"psrlq $32, %%mm7\n" /* shift hi dword to lo */
"paddd %%mm7,%%mm1\n"
"movd %%mm1,%2\n"
: "+r" (pix1), "+r" (pix2), "=r"(tmp)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r" ((x86_reg)line_size) , "m" (h)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
: "%ecx");
return tmp;
}
static int hf_noise8_mmx(uint8_t * pix1, int line_size, int h) {
int tmp;
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile (
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"movl %3,%%ecx\n"
"pxor %%mm7,%%mm7\n"
"pxor %%mm6,%%mm6\n"
"movq (%0),%%mm0\n"
"movq %%mm0, %%mm1\n"
"psllq $8, %%mm0\n"
"psrlq $8, %%mm1\n"
"psrlq $8, %%mm0\n"
"movq %%mm0, %%mm2\n"
"movq %%mm1, %%mm3\n"
"punpcklbw %%mm7,%%mm0\n"
"punpcklbw %%mm7,%%mm1\n"
"punpckhbw %%mm7,%%mm2\n"
"punpckhbw %%mm7,%%mm3\n"
"psubw %%mm1, %%mm0\n"
"psubw %%mm3, %%mm2\n"
"add %2,%0\n"
"movq (%0),%%mm4\n"
"movq %%mm4, %%mm1\n"
"psllq $8, %%mm4\n"
"psrlq $8, %%mm1\n"
"psrlq $8, %%mm4\n"
"movq %%mm4, %%mm5\n"
"movq %%mm1, %%mm3\n"
"punpcklbw %%mm7,%%mm4\n"
"punpcklbw %%mm7,%%mm1\n"
"punpckhbw %%mm7,%%mm5\n"
"punpckhbw %%mm7,%%mm3\n"
"psubw %%mm1, %%mm4\n"
"psubw %%mm3, %%mm5\n"
"psubw %%mm4, %%mm0\n"
"psubw %%mm5, %%mm2\n"
"pxor %%mm3, %%mm3\n"
"pxor %%mm1, %%mm1\n"
"pcmpgtw %%mm0, %%mm3\n\t"
"pcmpgtw %%mm2, %%mm1\n\t"
"pxor %%mm3, %%mm0\n"
"pxor %%mm1, %%mm2\n"
"psubw %%mm3, %%mm0\n"
"psubw %%mm1, %%mm2\n"
"paddw %%mm0, %%mm2\n"
"paddw %%mm2, %%mm6\n"
"add %2,%0\n"
"1:\n"
"movq (%0),%%mm0\n"
"movq %%mm0, %%mm1\n"
"psllq $8, %%mm0\n"
"psrlq $8, %%mm1\n"
"psrlq $8, %%mm0\n"
"movq %%mm0, %%mm2\n"
"movq %%mm1, %%mm3\n"
"punpcklbw %%mm7,%%mm0\n"
"punpcklbw %%mm7,%%mm1\n"
"punpckhbw %%mm7,%%mm2\n"
"punpckhbw %%mm7,%%mm3\n"
"psubw %%mm1, %%mm0\n"
"psubw %%mm3, %%mm2\n"
"psubw %%mm0, %%mm4\n"
"psubw %%mm2, %%mm5\n"
"pxor %%mm3, %%mm3\n"
"pxor %%mm1, %%mm1\n"
"pcmpgtw %%mm4, %%mm3\n\t"
"pcmpgtw %%mm5, %%mm1\n\t"
"pxor %%mm3, %%mm4\n"
"pxor %%mm1, %%mm5\n"
"psubw %%mm3, %%mm4\n"
"psubw %%mm1, %%mm5\n"
"paddw %%mm4, %%mm5\n"
"paddw %%mm5, %%mm6\n"
"add %2,%0\n"
"movq (%0),%%mm4\n"
"movq %%mm4, %%mm1\n"
"psllq $8, %%mm4\n"
"psrlq $8, %%mm1\n"
"psrlq $8, %%mm4\n"
"movq %%mm4, %%mm5\n"
"movq %%mm1, %%mm3\n"
"punpcklbw %%mm7,%%mm4\n"
"punpcklbw %%mm7,%%mm1\n"
"punpckhbw %%mm7,%%mm5\n"
"punpckhbw %%mm7,%%mm3\n"
"psubw %%mm1, %%mm4\n"
"psubw %%mm3, %%mm5\n"
"psubw %%mm4, %%mm0\n"
"psubw %%mm5, %%mm2\n"
"pxor %%mm3, %%mm3\n"
"pxor %%mm1, %%mm1\n"
"pcmpgtw %%mm0, %%mm3\n\t"
"pcmpgtw %%mm2, %%mm1\n\t"
"pxor %%mm3, %%mm0\n"
"pxor %%mm1, %%mm2\n"
"psubw %%mm3, %%mm0\n"
"psubw %%mm1, %%mm2\n"
"paddw %%mm0, %%mm2\n"
"paddw %%mm2, %%mm6\n"
"add %2,%0\n"
"subl $2, %%ecx\n"
" jnz 1b\n"
"movq %%mm6, %%mm0\n"
"punpcklwd %%mm7,%%mm0\n"
"punpckhwd %%mm7,%%mm6\n"
"paddd %%mm0, %%mm6\n"
"movq %%mm6,%%mm0\n"
"psrlq $32, %%mm6\n"
"paddd %%mm6,%%mm0\n"
"movd %%mm0,%1\n"
: "+r" (pix1), "=r"(tmp)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r" ((x86_reg)line_size) , "g" (h-2)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
: "%ecx");
return tmp;
}
static int hf_noise16_mmx(uint8_t * pix1, int line_size, int h) {
int tmp;
uint8_t * pix= pix1;
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile (
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"movl %3,%%ecx\n"
"pxor %%mm7,%%mm7\n"
"pxor %%mm6,%%mm6\n"
"movq (%0),%%mm0\n"
"movq 1(%0),%%mm1\n"
"movq %%mm0, %%mm2\n"
"movq %%mm1, %%mm3\n"
"punpcklbw %%mm7,%%mm0\n"
"punpcklbw %%mm7,%%mm1\n"
"punpckhbw %%mm7,%%mm2\n"
"punpckhbw %%mm7,%%mm3\n"
"psubw %%mm1, %%mm0\n"
"psubw %%mm3, %%mm2\n"
"add %2,%0\n"
"movq (%0),%%mm4\n"
"movq 1(%0),%%mm1\n"
"movq %%mm4, %%mm5\n"
"movq %%mm1, %%mm3\n"
"punpcklbw %%mm7,%%mm4\n"
"punpcklbw %%mm7,%%mm1\n"
"punpckhbw %%mm7,%%mm5\n"
"punpckhbw %%mm7,%%mm3\n"
"psubw %%mm1, %%mm4\n"
"psubw %%mm3, %%mm5\n"
"psubw %%mm4, %%mm0\n"
"psubw %%mm5, %%mm2\n"
"pxor %%mm3, %%mm3\n"
"pxor %%mm1, %%mm1\n"
"pcmpgtw %%mm0, %%mm3\n\t"
"pcmpgtw %%mm2, %%mm1\n\t"
"pxor %%mm3, %%mm0\n"
"pxor %%mm1, %%mm2\n"
"psubw %%mm3, %%mm0\n"
"psubw %%mm1, %%mm2\n"
"paddw %%mm0, %%mm2\n"
"paddw %%mm2, %%mm6\n"
"add %2,%0\n"
"1:\n"
"movq (%0),%%mm0\n"
"movq 1(%0),%%mm1\n"
"movq %%mm0, %%mm2\n"
"movq %%mm1, %%mm3\n"
"punpcklbw %%mm7,%%mm0\n"
"punpcklbw %%mm7,%%mm1\n"
"punpckhbw %%mm7,%%mm2\n"
"punpckhbw %%mm7,%%mm3\n"
"psubw %%mm1, %%mm0\n"
"psubw %%mm3, %%mm2\n"
"psubw %%mm0, %%mm4\n"
"psubw %%mm2, %%mm5\n"
"pxor %%mm3, %%mm3\n"
"pxor %%mm1, %%mm1\n"
"pcmpgtw %%mm4, %%mm3\n\t"
"pcmpgtw %%mm5, %%mm1\n\t"
"pxor %%mm3, %%mm4\n"
"pxor %%mm1, %%mm5\n"
"psubw %%mm3, %%mm4\n"
"psubw %%mm1, %%mm5\n"
"paddw %%mm4, %%mm5\n"
"paddw %%mm5, %%mm6\n"
"add %2,%0\n"
"movq (%0),%%mm4\n"
"movq 1(%0),%%mm1\n"
"movq %%mm4, %%mm5\n"
"movq %%mm1, %%mm3\n"
"punpcklbw %%mm7,%%mm4\n"
"punpcklbw %%mm7,%%mm1\n"
"punpckhbw %%mm7,%%mm5\n"
"punpckhbw %%mm7,%%mm3\n"
"psubw %%mm1, %%mm4\n"
"psubw %%mm3, %%mm5\n"
"psubw %%mm4, %%mm0\n"
"psubw %%mm5, %%mm2\n"
"pxor %%mm3, %%mm3\n"
"pxor %%mm1, %%mm1\n"
"pcmpgtw %%mm0, %%mm3\n\t"
"pcmpgtw %%mm2, %%mm1\n\t"
"pxor %%mm3, %%mm0\n"
"pxor %%mm1, %%mm2\n"
"psubw %%mm3, %%mm0\n"
"psubw %%mm1, %%mm2\n"
"paddw %%mm0, %%mm2\n"
"paddw %%mm2, %%mm6\n"
"add %2,%0\n"
"subl $2, %%ecx\n"
" jnz 1b\n"
"movq %%mm6, %%mm0\n"
"punpcklwd %%mm7,%%mm0\n"
"punpckhwd %%mm7,%%mm6\n"
"paddd %%mm0, %%mm6\n"
"movq %%mm6,%%mm0\n"
"psrlq $32, %%mm6\n"
"paddd %%mm6,%%mm0\n"
"movd %%mm0,%1\n"
: "+r" (pix1), "=r"(tmp)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r" ((x86_reg)line_size) , "g" (h-2)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
: "%ecx");
return tmp + hf_noise8_mmx(pix+8, line_size, h);
}
static int nsse16_mmx(void *p, uint8_t * pix1, uint8_t * pix2, int line_size, int h) {
MpegEncContext *c = p;
int score1, score2;
if(c) score1 = c->dsp.sse[0](c, pix1, pix2, line_size, h);
else score1 = sse16_mmx(c, pix1, pix2, line_size, h);
score2= hf_noise16_mmx(pix1, line_size, h) - hf_noise16_mmx(pix2, line_size, h);
if(c) return score1 + FFABS(score2)*c->avctx->nsse_weight;
else return score1 + FFABS(score2)*8;
}
static int nsse8_mmx(void *p, uint8_t * pix1, uint8_t * pix2, int line_size, int h) {
MpegEncContext *c = p;
int score1= sse8_mmx(c, pix1, pix2, line_size, h);
int score2= hf_noise8_mmx(pix1, line_size, h) - hf_noise8_mmx(pix2, line_size, h);
if(c) return score1 + FFABS(score2)*c->avctx->nsse_weight;
else return score1 + FFABS(score2)*8;
}
static int vsad_intra16_mmx(void *v, uint8_t * pix, uint8_t * dummy, int line_size, int h) {
int tmp;
dsputil_mmx: switch to av_assert2() Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-12-10 13:32:23 +00:00
av_assert2( (((int)pix) & 7) == 0);
av_assert2((line_size &7) ==0);
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#define SUM(in0, in1, out0, out1) \
"movq (%0), %%mm2\n"\
"movq 8(%0), %%mm3\n"\
"add %2,%0\n"\
"movq %%mm2, " #out0 "\n"\
"movq %%mm3, " #out1 "\n"\
"psubusb " #in0 ", %%mm2\n"\
"psubusb " #in1 ", %%mm3\n"\
"psubusb " #out0 ", " #in0 "\n"\
"psubusb " #out1 ", " #in1 "\n"\
"por %%mm2, " #in0 "\n"\
"por %%mm3, " #in1 "\n"\
"movq " #in0 ", %%mm2\n"\
"movq " #in1 ", %%mm3\n"\
"punpcklbw %%mm7, " #in0 "\n"\
"punpcklbw %%mm7, " #in1 "\n"\
"punpckhbw %%mm7, %%mm2\n"\
"punpckhbw %%mm7, %%mm3\n"\
"paddw " #in1 ", " #in0 "\n"\
"paddw %%mm3, %%mm2\n"\
"paddw %%mm2, " #in0 "\n"\
"paddw " #in0 ", %%mm6\n"
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile (
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"movl %3,%%ecx\n"
"pxor %%mm6,%%mm6\n"
"pxor %%mm7,%%mm7\n"
"movq (%0),%%mm0\n"
"movq 8(%0),%%mm1\n"
"add %2,%0\n"
Simplify vsad_intra16_mmx() Originally committed as revision 13188 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:31:10 +00:00
"jmp 2f\n"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"1:\n"
SUM(%%mm4, %%mm5, %%mm0, %%mm1)
Simplify vsad_intra16_mmx() Originally committed as revision 13188 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:31:10 +00:00
"2:\n"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
SUM(%%mm0, %%mm1, %%mm4, %%mm5)
"subl $2, %%ecx\n"
"jnz 1b\n"
"movq %%mm6,%%mm0\n"
"psrlq $32, %%mm6\n"
"paddw %%mm6,%%mm0\n"
"movq %%mm0,%%mm6\n"
"psrlq $16, %%mm0\n"
"paddw %%mm6,%%mm0\n"
"movd %%mm0,%1\n"
: "+r" (pix), "=r"(tmp)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r" ((x86_reg)line_size) , "m" (h)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
: "%ecx");
return tmp & 0xFFFF;
}
#undef SUM
x86: mmx2 ---> mmxext in function names
2012-07-08 17:56:10 +00:00
static int vsad_intra16_mmxext(void *v, uint8_t *pix, uint8_t *dummy,
int line_size, int h)
{
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
int tmp;
dsputil_mmx: switch to av_assert2() Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-12-10 13:32:23 +00:00
av_assert2( (((int)pix) & 7) == 0);
av_assert2((line_size &7) ==0);
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#define SUM(in0, in1, out0, out1) \
"movq (%0), " #out0 "\n"\
"movq 8(%0), " #out1 "\n"\
"add %2,%0\n"\
"psadbw " #out0 ", " #in0 "\n"\
"psadbw " #out1 ", " #in1 "\n"\
"paddw " #in1 ", " #in0 "\n"\
"paddw " #in0 ", %%mm6\n"
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile (
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"movl %3,%%ecx\n"
"pxor %%mm6,%%mm6\n"
"pxor %%mm7,%%mm7\n"
"movq (%0),%%mm0\n"
"movq 8(%0),%%mm1\n"
"add %2,%0\n"
Simplify vsad_intra16_mmx2() Originally committed as revision 13189 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:33:01 +00:00
"jmp 2f\n"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"1:\n"
SUM(%%mm4, %%mm5, %%mm0, %%mm1)
Simplify vsad_intra16_mmx2() Originally committed as revision 13189 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:33:01 +00:00
"2:\n"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
SUM(%%mm0, %%mm1, %%mm4, %%mm5)
"subl $2, %%ecx\n"
"jnz 1b\n"
"movd %%mm6,%1\n"
: "+r" (pix), "=r"(tmp)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r" ((x86_reg)line_size) , "m" (h)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
: "%ecx");
return tmp;
}
#undef SUM
static int vsad16_mmx(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h) {
int tmp;
dsputil_mmx: switch to av_assert2() Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-12-10 13:32:23 +00:00
av_assert2( (((int)pix1) & 7) == 0);
av_assert2( (((int)pix2) & 7) == 0);
av_assert2((line_size &7) ==0);
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#define SUM(in0, in1, out0, out1) \
"movq (%0),%%mm2\n"\
"movq (%1)," #out0 "\n"\
"movq 8(%0),%%mm3\n"\
"movq 8(%1)," #out1 "\n"\
"add %3,%0\n"\
"add %3,%1\n"\
"psubb " #out0 ", %%mm2\n"\
"psubb " #out1 ", %%mm3\n"\
"pxor %%mm7, %%mm2\n"\
"pxor %%mm7, %%mm3\n"\
"movq %%mm2, " #out0 "\n"\
"movq %%mm3, " #out1 "\n"\
"psubusb " #in0 ", %%mm2\n"\
"psubusb " #in1 ", %%mm3\n"\
"psubusb " #out0 ", " #in0 "\n"\
"psubusb " #out1 ", " #in1 "\n"\
"por %%mm2, " #in0 "\n"\
"por %%mm3, " #in1 "\n"\
"movq " #in0 ", %%mm2\n"\
"movq " #in1 ", %%mm3\n"\
"punpcklbw %%mm7, " #in0 "\n"\
"punpcklbw %%mm7, " #in1 "\n"\
"punpckhbw %%mm7, %%mm2\n"\
"punpckhbw %%mm7, %%mm3\n"\
"paddw " #in1 ", " #in0 "\n"\
"paddw %%mm3, %%mm2\n"\
"paddw %%mm2, " #in0 "\n"\
"paddw " #in0 ", %%mm6\n"
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile (
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"movl %4,%%ecx\n"
"pxor %%mm6,%%mm6\n"
"pcmpeqw %%mm7,%%mm7\n"
"psllw $15, %%mm7\n"
"packsswb %%mm7, %%mm7\n"
"movq (%0),%%mm0\n"
"movq (%1),%%mm2\n"
"movq 8(%0),%%mm1\n"
"movq 8(%1),%%mm3\n"
"add %3,%0\n"
"add %3,%1\n"
"psubb %%mm2, %%mm0\n"
"psubb %%mm3, %%mm1\n"
"pxor %%mm7, %%mm0\n"
"pxor %%mm7, %%mm1\n"
Simplify vsad16_mmx(). Originally committed as revision 13191 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:35:14 +00:00
"jmp 2f\n"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"1:\n"
SUM(%%mm4, %%mm5, %%mm0, %%mm1)
Simplify vsad16_mmx(). Originally committed as revision 13191 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:35:14 +00:00
"2:\n"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
SUM(%%mm0, %%mm1, %%mm4, %%mm5)
"subl $2, %%ecx\n"
"jnz 1b\n"
"movq %%mm6,%%mm0\n"
"psrlq $32, %%mm6\n"
"paddw %%mm6,%%mm0\n"
"movq %%mm0,%%mm6\n"
"psrlq $16, %%mm0\n"
"paddw %%mm6,%%mm0\n"
"movd %%mm0,%2\n"
: "+r" (pix1), "+r" (pix2), "=r"(tmp)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r" ((x86_reg)line_size) , "m" (h)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
: "%ecx");
return tmp & 0x7FFF;
}
#undef SUM
x86: mmx2 ---> mmxext in function names
2012-07-08 17:56:10 +00:00
static int vsad16_mmxext(void *v, uint8_t *pix1, uint8_t *pix2,
int line_size, int h)
{
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
int tmp;
dsputil_mmx: switch to av_assert2() Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-12-10 13:32:23 +00:00
av_assert2( (((int)pix1) & 7) == 0);
av_assert2( (((int)pix2) & 7) == 0);
av_assert2((line_size &7) ==0);
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#define SUM(in0, in1, out0, out1) \
"movq (%0)," #out0 "\n"\
"movq (%1),%%mm2\n"\
"movq 8(%0)," #out1 "\n"\
"movq 8(%1),%%mm3\n"\
"add %3,%0\n"\
"add %3,%1\n"\
"psubb %%mm2, " #out0 "\n"\
"psubb %%mm3, " #out1 "\n"\
"pxor %%mm7, " #out0 "\n"\
"pxor %%mm7, " #out1 "\n"\
"psadbw " #out0 ", " #in0 "\n"\
"psadbw " #out1 ", " #in1 "\n"\
"paddw " #in1 ", " #in0 "\n"\
"paddw " #in0 ", %%mm6\n"
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile (
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"movl %4,%%ecx\n"
"pxor %%mm6,%%mm6\n"
"pcmpeqw %%mm7,%%mm7\n"
"psllw $15, %%mm7\n"
"packsswb %%mm7, %%mm7\n"
"movq (%0),%%mm0\n"
"movq (%1),%%mm2\n"
"movq 8(%0),%%mm1\n"
"movq 8(%1),%%mm3\n"
"add %3,%0\n"
"add %3,%1\n"
"psubb %%mm2, %%mm0\n"
"psubb %%mm3, %%mm1\n"
"pxor %%mm7, %%mm0\n"
"pxor %%mm7, %%mm1\n"
Simplify vsad16_mmx2(). Originally committed as revision 13193 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:36:44 +00:00
"jmp 2f\n"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"1:\n"
SUM(%%mm4, %%mm5, %%mm0, %%mm1)
Simplify vsad16_mmx2(). Originally committed as revision 13193 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-17 14:36:44 +00:00
"2:\n"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
SUM(%%mm0, %%mm1, %%mm4, %%mm5)
"subl $2, %%ecx\n"
"jnz 1b\n"
"movd %%mm6,%2\n"
: "+r" (pix1), "+r" (pix2), "=r"(tmp)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r" ((x86_reg)line_size) , "m" (h)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
: "%ecx");
return tmp;
}
#undef SUM
dsputil: mark source of diff_bytes as const. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-03-15 21:16:25 +00:00
static void diff_bytes_mmx(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w){
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
x86_reg i=0;
dsp: fix diff_bytes_mmx() with small width Fixes Ticket1068 Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-03-17 19:45:45 +00:00
if(w>=16)
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile(
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"1: \n\t"
"movq (%2, %0), %%mm0 \n\t"
"movq (%1, %0), %%mm1 \n\t"
"psubb %%mm0, %%mm1 \n\t"
"movq %%mm1, (%3, %0) \n\t"
"movq 8(%2, %0), %%mm0 \n\t"
"movq 8(%1, %0), %%mm1 \n\t"
"psubb %%mm0, %%mm1 \n\t"
"movq %%mm1, 8(%3, %0) \n\t"
"add $16, %0 \n\t"
"cmp %4, %0 \n\t"
" jb 1b \n\t"
: "+r" (i)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r"(src1), "r"(src2), "r"(dst), "r"((x86_reg)w-15)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
);
for(; i<w; i++)
dst[i+0] = src1[i+0]-src2[i+0];
}
x86: mmx2 ---> mmxext in function names
2012-07-08 17:56:10 +00:00
static void sub_hfyu_median_prediction_mmxext(uint8_t *dst, const uint8_t *src1,
const uint8_t *src2, int w,
int *left, int *left_top)
{
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
x86_reg i=0;
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
uint8_t l, lt;
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile(
dsputil_mmx: fix reading prior of the src array in sub_hfyu_median_prediction() This should fix the utvideoenc valgrind failure Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-09-27 13:52:21 +00:00
"movq (%1, %0), %%mm0 \n\t" // LT
"psllq $8, %%mm0 \n\t"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"1: \n\t"
"movq (%1, %0), %%mm1 \n\t" // T
"movq -1(%2, %0), %%mm2 \n\t" // L
"movq (%2, %0), %%mm3 \n\t" // X
"movq %%mm2, %%mm4 \n\t" // L
"psubb %%mm0, %%mm2 \n\t"
"paddb %%mm1, %%mm2 \n\t" // L + T - LT
"movq %%mm4, %%mm5 \n\t" // L
"pmaxub %%mm1, %%mm4 \n\t" // max(T, L)
"pminub %%mm5, %%mm1 \n\t" // min(T, L)
"pminub %%mm2, %%mm4 \n\t"
"pmaxub %%mm1, %%mm4 \n\t"
"psubb %%mm4, %%mm3 \n\t" // dst - pred
"movq %%mm3, (%3, %0) \n\t"
"add $8, %0 \n\t"
dsputil_mmx: fix reading prior of the src array in sub_hfyu_median_prediction() This should fix the utvideoenc valgrind failure Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
2012-09-27 13:52:21 +00:00
"movq -1(%1, %0), %%mm0 \n\t" // LT
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"cmp %4, %0 \n\t"
" jb 1b \n\t"
: "+r" (i)
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
: "r"(src1), "r"(src2), "r"(dst), "r"((x86_reg)w)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
);
l= *left;
lt= *left_top;
dst[0]= src2[0] - mid_pred(l, src1[0], (l + src1[0] - lt)&0xFF);
*left_top= src1[w-1];
*left = src2[w-1];
}
#define MMABS_MMX(a,z)\
"pxor " #z ", " #z " \n\t"\
"pcmpgtw " #a ", " #z " \n\t"\
"pxor " #z ", " #a " \n\t"\
"psubw " #z ", " #a " \n\t"
x86: MMX2 ---> MMXEXT in macro names
2012-07-08 17:28:57 +00:00
#define MMABS_MMXEXT(a, z) \
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"pxor " #z ", " #z " \n\t"\
"psubw " #a ", " #z " \n\t"\
"pmaxsw " #z ", " #a " \n\t"
#define MMABS_SSSE3(a,z)\
"pabsw " #a ", " #a " \n\t"
#define MMABS_SUM(a,z, sum)\
MMABS(a,z)\
"paddusw " #a ", " #sum " \n\t"
/* FIXME: HSUM_* saturates at 64k, while an 8x8 hadamard or dct block can get up to
* about 100k on extreme inputs. But that's very unlikely to occur in natural video,
* and it's even more unlikely to not have any alternative mvs/modes with lower cost. */
#define HSUM_MMX(a, t, dst)\
"movq "#a", "#t" \n\t"\
"psrlq $32, "#a" \n\t"\
"paddusw "#t", "#a" \n\t"\
"movq "#a", "#t" \n\t"\
"psrlq $16, "#a" \n\t"\
"paddusw "#t", "#a" \n\t"\
"movd "#a", "#dst" \n\t"\
x86: MMX2 ---> MMXEXT in macro names
2012-07-08 17:28:57 +00:00
#define HSUM_MMXEXT(a, t, dst) \
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"pshufw $0x0E, "#a", "#t" \n\t"\
"paddusw "#t", "#a" \n\t"\
"pshufw $0x01, "#a", "#t" \n\t"\
"paddusw "#t", "#a" \n\t"\
"movd "#a", "#dst" \n\t"\
#define HSUM_SSE2(a, t, dst)\
"movhlps "#a", "#t" \n\t"\
"paddusw "#t", "#a" \n\t"\
"pshuflw $0x0E, "#a", "#t" \n\t"\
"paddusw "#t", "#a" \n\t"\
"pshuflw $0x01, "#a", "#t" \n\t"\
"paddusw "#t", "#a" \n\t"\
"movd "#a", "#dst" \n\t"\
#define DCT_SAD4(m,mm,o)\
"mov"#m" "#o"+ 0(%1), "#mm"2 \n\t"\
"mov"#m" "#o"+16(%1), "#mm"3 \n\t"\
"mov"#m" "#o"+32(%1), "#mm"4 \n\t"\
"mov"#m" "#o"+48(%1), "#mm"5 \n\t"\
MMABS_SUM(mm##2, mm##6, mm##0)\
MMABS_SUM(mm##3, mm##7, mm##1)\
MMABS_SUM(mm##4, mm##6, mm##0)\
MMABS_SUM(mm##5, mm##7, mm##1)\
#define DCT_SAD_MMX\
"pxor %%mm0, %%mm0 \n\t"\
"pxor %%mm1, %%mm1 \n\t"\
DCT_SAD4(q, %%mm, 0)\
DCT_SAD4(q, %%mm, 8)\
DCT_SAD4(q, %%mm, 64)\
DCT_SAD4(q, %%mm, 72)\
"paddusw %%mm1, %%mm0 \n\t"\
HSUM(%%mm0, %%mm1, %0)
#define DCT_SAD_SSE2\
"pxor %%xmm0, %%xmm0 \n\t"\
"pxor %%xmm1, %%xmm1 \n\t"\
DCT_SAD4(dqa, %%xmm, 0)\
DCT_SAD4(dqa, %%xmm, 64)\
"paddusw %%xmm1, %%xmm0 \n\t"\
HSUM(%%xmm0, %%xmm1, %0)
#define DCT_SAD_FUNC(cpu) \
Drop DCTELEM typedef It does not help as an abstraction and adds dsputil dependencies. Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
2013-01-20 00:02:29 +00:00
static int sum_abs_dctelem_##cpu(int16_t *block){\
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
int sum;\
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile(\
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
DCT_SAD\
:"=r"(sum)\
:"r"(block)\
);\
return sum&0xFFFF;\
}
#define DCT_SAD DCT_SAD_MMX
#define HSUM(a,t,dst) HSUM_MMX(a,t,dst)
#define MMABS(a,z) MMABS_MMX(a,z)
DCT_SAD_FUNC(mmx)
#undef MMABS
#undef HSUM
x86: MMX2 ---> MMXEXT in macro names
2012-07-08 17:28:57 +00:00
#define HSUM(a,t,dst) HSUM_MMXEXT(a,t,dst)
#define MMABS(a,z) MMABS_MMXEXT(a,z)
x86: mmx2 ---> mmxext in function names
2012-07-08 17:56:10 +00:00
DCT_SAD_FUNC(mmxext)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#undef HSUM
#undef DCT_SAD
#define DCT_SAD DCT_SAD_SSE2
#define HSUM(a,t,dst) HSUM_SSE2(a,t,dst)
DCT_SAD_FUNC(sse2)
#undef MMABS
x86: Split inline and external assembly #ifdefs
2012-08-28 12:53:33 +00:00
#if HAVE_SSSE3_INLINE
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#define MMABS(a,z) MMABS_SSSE3(a,z)
DCT_SAD_FUNC(ssse3)
#undef MMABS
#endif
#undef HSUM
#undef DCT_SAD
static int ssd_int8_vs_int16_mmx(const int8_t *pix1, const int16_t *pix2, int size){
int sum;
Do not misuse long as the size of a register in x86. typedef x86_reg as the appropriate size and use it instead. Originally committed as revision 13081 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-05-08 21:11:24 +00:00
x86_reg i=size;
Convert asm keyword into __asm__. Neither the asm() nor the __asm__() keyword is part of the C99 standard, but while GCC accepts the former in C89 syntax, it is not accepted in C99 unless GNU extensions are turned on (with -fasm). The latter form is accepted in any syntax as an extension (without requiring further command-line options). Sun Studio C99 compiler also does not accept asm() while accepting __asm__(), albeit reporting warnings that it's not valid C99 syntax. Originally committed as revision 15627 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-10-16 13:34:09 +00:00
__asm__ volatile(
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
"pxor %%mm4, %%mm4 \n"
"1: \n"
"sub $8, %0 \n"
"movq (%2,%0), %%mm2 \n"
"movq (%3,%0,2), %%mm0 \n"
"movq 8(%3,%0,2), %%mm1 \n"
"punpckhbw %%mm2, %%mm3 \n"
"punpcklbw %%mm2, %%mm2 \n"
"psraw $8, %%mm3 \n"
"psraw $8, %%mm2 \n"
"psubw %%mm3, %%mm1 \n"
"psubw %%mm2, %%mm0 \n"
"pmaddwd %%mm1, %%mm1 \n"
"pmaddwd %%mm0, %%mm0 \n"
"paddd %%mm1, %%mm4 \n"
"paddd %%mm0, %%mm4 \n"
"jg 1b \n"
"movq %%mm4, %%mm3 \n"
"psrlq $32, %%mm3 \n"
"paddd %%mm3, %%mm4 \n"
"movd %%mm4, %1 \n"
:"+r"(i), "=r"(sum)
:"r"(pix1), "r"(pix2)
);
return sum;
}
#define PHADDD(a, t)\
"movq "#a", "#t" \n\t"\
"psrlq $32, "#a" \n\t"\
"paddd "#t", "#a" \n\t"
/*
pmulhw: dst[0-15]=(src[0-15]*dst[0-15])[16-31]
pmulhrw: dst[0-15]=(src[0-15]*dst[0-15] + 0x8000)[16-31]
pmulhrsw: dst[0-15]=(src[0-15]*dst[0-15] + 0x4000)[15-30]
*/
#define PMULHRW(x, y, s, o)\
"pmulhw " #s ", "#x " \n\t"\
"pmulhw " #s ", "#y " \n\t"\
"paddw " #o ", "#x " \n\t"\
"paddw " #o ", "#y " \n\t"\
"psraw $1, "#x " \n\t"\
"psraw $1, "#y " \n\t"
#define DEF(x) x ## _mmx
#define SET_RND MOVQ_WONE
#define SCALE_OFFSET 1
x86: avcodec: Drop silly "_mmx" suffix from dsputil template names
2012-08-11 20:35:27 +00:00
#include "dsputil_qns_template.c"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#undef DEF
#undef SET_RND
#undef SCALE_OFFSET
#undef PMULHRW
#define DEF(x) x ## _3dnow
#define SET_RND(x)
#define SCALE_OFFSET 0
#define PMULHRW(x, y, s, o)\
"pmulhrw " #s ", "#x " \n\t"\
"pmulhrw " #s ", "#y " \n\t"
x86: avcodec: Drop silly "_mmx" suffix from dsputil template names
2012-08-11 20:35:27 +00:00
#include "dsputil_qns_template.c"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#undef DEF
#undef SET_RND
#undef SCALE_OFFSET
#undef PMULHRW
x86: Split inline and external assembly #ifdefs
2012-08-28 12:53:33 +00:00
#if HAVE_SSSE3_INLINE
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#undef PHADDD
#define DEF(x) x ## _ssse3
#define SET_RND(x)
#define SCALE_OFFSET -1
#define PHADDD(a, t)\
"pshufw $0x0E, "#a", "#t" \n\t"\
"paddd "#t", "#a" \n\t" /* faster than phaddd on core2 */
#define PMULHRW(x, y, s, o)\
"pmulhrsw " #s ", "#x " \n\t"\
"pmulhrsw " #s ", "#y " \n\t"
x86: avcodec: Drop silly "_mmx" suffix from dsputil template names
2012-08-11 20:35:27 +00:00
#include "dsputil_qns_template.c"
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#undef DEF
#undef SET_RND
#undef SCALE_OFFSET
#undef PMULHRW
#undef PHADDD
x86: Split inline and external assembly #ifdefs
2012-08-28 12:53:33 +00:00
#endif /* HAVE_SSSE3_INLINE */
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
#endif /* HAVE_INLINE_ASM */
int ff_sse16_sse2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h);
#define hadamard_func(cpu) \
int ff_hadamard8_diff_##cpu (void *s, uint8_t *src1, uint8_t *src2, \
int stride, int h); \
int ff_hadamard8_diff16_##cpu(void *s, uint8_t *src1, uint8_t *src2, \
int stride, int h);
hadamard_func(mmx)
x86: mmx2 ---> mmxext in asm constructs
2012-07-09 22:04:18 +00:00
hadamard_func(mmxext)
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
hadamard_func(sse2)
hadamard_func(ssse3)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
Add av_cold attributes to arch-specific init functions
2013-02-01 09:31:59 +00:00
av_cold void ff_dsputilenc_init_mmx(DSPContext *c, AVCodecContext *avctx)
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
{
Consistently use "cpu_flags" as variable/parameter name for CPU flags
2013-07-17 18:19:24 +00:00
int cpu_flags = av_get_cpu_flags();
Remove global mm_flags variable Originally committed as revision 24909 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-08-24 17:47:05 +00:00
dsputilenc: x86: Convert pixel inline asm to yasm Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
2013-01-16 07:41:39 +00:00
#if HAVE_YASM
x86: Move some conditional code around to avoid unused variable warnings
2013-04-22 02:57:48 +00:00
int bit_depth = avctx->bits_per_raw_sample;
Consistently use "cpu_flags" as variable/parameter name for CPU flags
2013-07-17 18:19:24 +00:00
if (EXTERNAL_MMX(cpu_flags)) {
dsputilenc: x86: Convert pixel inline asm to yasm Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
2013-01-16 07:41:39 +00:00
if (bit_depth <= 8)
c->get_pixels = ff_get_pixels_mmx;
c->diff_pixels = ff_diff_pixels_mmx;
c->pix_sum = ff_pix_sum16_mmx;
c->pix_norm1 = ff_pix_norm1_mmx;
}
Consistently use "cpu_flags" as variable/parameter name for CPU flags
2013-07-17 18:19:24 +00:00
if (EXTERNAL_SSE2(cpu_flags))
dsputilenc: x86: Convert pixel inline asm to yasm Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
2013-01-16 07:41:39 +00:00
if (bit_depth <= 8)
c->get_pixels = ff_get_pixels_sse2;
#endif /* HAVE_YASM */
#if HAVE_INLINE_ASM
x86: avcodec: Use convenience macros to check for CPU flags
2013-08-20 12:46:58 +00:00
if (INLINE_MMX(cpu_flags)) {
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
const int dct_algo = avctx->dct_algo;
jfdctint: add 10-bit version Signed-off-by: Mans Rullgard <mans@mansr.com>
2011-07-20 19:01:56 +00:00
if (avctx->bits_per_raw_sample <= 8 &&
(dct_algo==FF_DCT_AUTO || dct_algo==FF_DCT_MMX)) {
Consistently use "cpu_flags" as variable/parameter name for CPU flags
2013-07-17 18:19:24 +00:00
if (cpu_flags & AV_CPU_FLAG_SSE2) {
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
c->fdct = ff_fdct_sse2;
Consistently use "cpu_flags" as variable/parameter name for CPU flags
2013-07-17 18:19:24 +00:00
} else if (cpu_flags & AV_CPU_FLAG_MMXEXT) {
x86: mmx2 ---> mmxext in function names
2012-07-08 17:56:10 +00:00
c->fdct = ff_fdct_mmxext;
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
}else{
c->fdct = ff_fdct_mmx;
}
}
c->diff_bytes= diff_bytes_mmx;
c->sum_abs_dctelem= sum_abs_dctelem_mmx;
dsputilenc_mmx: split assignment of ff_sse16_sse2 to SSE2 section.
2012-06-14 22:05:10 +00:00
c->sse[0] = sse16_mmx;
c->sse[1] = sse8_mmx;
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
c->vsad[4]= vsad_intra16_mmx;
c->nsse[0] = nsse16_mmx;
c->nsse[1] = nsse8_mmx;
if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
c->vsad[0] = vsad16_mmx;
}
if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
c->try_8x8basis= try_8x8basis_mmx;
}
c->add_8x8basis= add_8x8basis_mmx;
c->ssd_int8_vs_int16 = ssd_int8_vs_int16_mmx;
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
}
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
if (INLINE_MMXEXT(cpu_flags)) {
c->sum_abs_dctelem = sum_abs_dctelem_mmxext;
c->vsad[4] = vsad_intra16_mmxext;
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
if (!(avctx->flags & CODEC_FLAG_BITEXACT)){
c->vsad[0] = vsad16_mmxext;
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
}
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
c->sub_hfyu_median_prediction = sub_hfyu_median_prediction_mmxext;
}
if (INLINE_SSE2(cpu_flags)) {
c->sum_abs_dctelem= sum_abs_dctelem_sse2;
}
Remove FF_MM_SSE2/3 flags for CPUs where this is generally not faster than regular MMX code. Examples of this are the Core1 CPU. Instead, set a new flag, FF_MM_SSE2/3SLOW, which can be checked for particular SSE2/3 functions that have been checked specifically on such CPUs and are actually faster than their MMX counterparts. In addition, use this flag to enable particular VP8 and LPC SSE2 functions that are faster than their MMX counterparts. Based on a patch by Loren Merritt <lorenm AT u washington edu>. Originally committed as revision 24340 to svn://svn.ffmpeg.org/ffmpeg/trunk
2010-07-19 22:38:23 +00:00
x86: Split inline and external assembly #ifdefs
2012-08-28 12:53:33 +00:00
#if HAVE_SSSE3_INLINE
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
if (INLINE_SSSE3(cpu_flags)) {
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
c->try_8x8basis = try_8x8basis_ssse3;
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
}
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
c->add_8x8basis = add_8x8basis_ssse3;
c->sum_abs_dctelem = sum_abs_dctelem_ssse3;
}
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
#endif
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
if (INLINE_AMD3DNOW(cpu_flags)) {
if (!(avctx->flags & CODEC_FLAG_BITEXACT)) {
c->try_8x8basis = try_8x8basis_3dnow;
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
}
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
c->add_8x8basis = add_8x8basis_3dnow;
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
}
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
#endif /* HAVE_INLINE_ASM */
Consistently use "cpu_flags" as variable/parameter name for CPU flags
2013-07-17 18:19:24 +00:00
if (EXTERNAL_MMX(cpu_flags)) {
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
c->hadamard8_diff[0] = ff_hadamard8_diff16_mmx;
c->hadamard8_diff[1] = ff_hadamard8_diff_mmx;
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
}
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
if (EXTERNAL_MMXEXT(cpu_flags)) {
c->hadamard8_diff[0] = ff_hadamard8_diff16_mmxext;
c->hadamard8_diff[1] = ff_hadamard8_diff_mmxext;
}
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
if (EXTERNAL_SSE2(cpu_flags)) {
c->sse[0] = ff_sse16_sse2;
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
#if HAVE_ALIGNED_STACK
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
c->hadamard8_diff[0] = ff_hadamard8_diff16_sse2;
c->hadamard8_diff[1] = ff_hadamard8_diff_sse2;
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
#endif
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
}
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
x86: avcodec: Consistently structure CPU extension initialization
2013-08-20 13:32:00 +00:00
if (EXTERNAL_SSSE3(cpu_flags) && HAVE_ALIGNED_STACK) {
c->hadamard8_diff[0] = ff_hadamard8_diff16_ssse3;
c->hadamard8_diff[1] = ff_hadamard8_diff_ssse3;
x86/dsputilenc: bury inline asm under HAVE_INLINE_ASM.
2012-07-26 03:30:19 +00:00
}
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
dsputil: Add ff_ prefix to the dsputil*_init* functions Signed-off-by: Martin Storsjö <martin@martin.st>
2012-02-15 10:06:44 +00:00
ff_dsputil_init_pix_mmx(c, avctx);
split encoding part of dsputil_mmx into its own file Originally committed as revision 12223 to svn://svn.ffmpeg.org/ffmpeg/trunk
2008-02-25 23:14:22 +00:00
}