2001-07-22 14:18:56 +00:00
|
|
|
/*
|
|
|
|
* MMX optimized DSP utils
|
2002-05-25 22:45:33 +00:00
|
|
|
* Copyright (c) 2000, 2001 Fabrice Bellard.
|
2001-07-22 14:18:56 +00:00
|
|
|
*
|
2002-05-25 22:45:33 +00:00
|
|
|
* This 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 of the License, or (at your option) any later version.
|
2001-07-22 14:18:56 +00:00
|
|
|
*
|
2002-05-25 22:45:33 +00:00
|
|
|
* This library is distributed in the hope that it will be useful,
|
2001-07-22 14:18:56 +00:00
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of
|
2002-05-25 22:45:33 +00:00
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
|
|
|
* Lesser General Public License for more details.
|
2001-07-22 14:18:56 +00:00
|
|
|
*
|
2002-05-25 22:45:33 +00:00
|
|
|
* You should have received a copy of the GNU Lesser General Public
|
|
|
|
* License along with this library; if not, write to the Free Software
|
|
|
|
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
|
2001-07-22 14:18:56 +00:00
|
|
|
*
|
|
|
|
* MMX optimization by Nick Kurshev <nickols_k@mail.ru>
|
|
|
|
*/
|
|
|
|
|
|
|
|
#include "../dsputil.h"
|
|
|
|
|
2001-07-24 20:38:55 +00:00
|
|
|
int mm_flags; /* multimedia extension flags */
|
2002-11-11 09:40:17 +00:00
|
|
|
/* FIXME use them in static form */
|
2002-03-27 21:25:22 +00:00
|
|
|
int pix_abs16x16_mmx(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs16x16_x2_mmx(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs16x16_y2_mmx(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs16x16_xy2_mmx(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
|
|
|
|
int pix_abs16x16_mmx2(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs16x16_x2_mmx2(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs16x16_y2_mmx2(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs16x16_xy2_mmx2(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
|
|
|
|
int pix_abs8x8_mmx(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs8x8_x2_mmx(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs8x8_y2_mmx(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs8x8_xy2_mmx(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
|
|
|
|
int pix_abs8x8_mmx2(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs8x8_x2_mmx2(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs8x8_y2_mmx2(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int pix_abs8x8_xy2_mmx2(UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
|
2002-12-27 23:51:46 +00:00
|
|
|
int sad16x16_mmx(void *s, UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int sad8x8_mmx(void *s, UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int sad16x16_mmx2(void *s, UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
int sad8x8_mmx2(void *s, UINT8 *blk1, UINT8 *blk2, int lx);
|
|
|
|
|
2001-07-22 14:18:56 +00:00
|
|
|
/* pixel operations */
|
2002-05-17 13:01:01 +00:00
|
|
|
static const uint64_t mm_bone __attribute__ ((aligned(8))) = 0x0101010101010101ULL;
|
|
|
|
static const uint64_t mm_wone __attribute__ ((aligned(8))) = 0x0001000100010001ULL;
|
|
|
|
static const uint64_t mm_wtwo __attribute__ ((aligned(8))) = 0x0002000200020002ULL;
|
2001-07-22 14:18:56 +00:00
|
|
|
|
2002-02-18 09:40:05 +00:00
|
|
|
#define JUMPALIGN() __asm __volatile (".balign 8"::)
|
|
|
|
#define MOVQ_ZERO(regd) __asm __volatile ("pxor %%" #regd ", %%" #regd ::)
|
|
|
|
|
2002-05-30 15:14:56 +00:00
|
|
|
#define MOVQ_WONE(regd) \
|
|
|
|
__asm __volatile ( \
|
|
|
|
"pcmpeqd %%" #regd ", %%" #regd " \n\t" \
|
|
|
|
"psrlw $15, %%" #regd ::)
|
|
|
|
|
|
|
|
#define MOVQ_BFE(regd) \
|
|
|
|
__asm __volatile ( \
|
|
|
|
"pcmpeqd %%" #regd ", %%" #regd " \n\t"\
|
|
|
|
"paddb %%" #regd ", %%" #regd " \n\t" ::)
|
|
|
|
|
2002-02-18 09:40:05 +00:00
|
|
|
#ifndef PIC
|
2002-05-30 15:14:56 +00:00
|
|
|
#define MOVQ_BONE(regd) __asm __volatile ("movq %0, %%" #regd " \n\t" ::"m"(mm_bone))
|
2002-02-18 09:40:05 +00:00
|
|
|
#define MOVQ_WTWO(regd) __asm __volatile ("movq %0, %%" #regd " \n\t" ::"m"(mm_wtwo))
|
|
|
|
#else
|
|
|
|
// for shared library it's better to use this way for accessing constants
|
|
|
|
// pcmpeqd -> -1
|
2002-05-30 15:14:56 +00:00
|
|
|
#define MOVQ_BONE(regd) \
|
2002-02-18 09:40:05 +00:00
|
|
|
__asm __volatile ( \
|
2002-05-30 15:14:56 +00:00
|
|
|
"pcmpeqd %%" #regd ", %%" #regd " \n\t" \
|
|
|
|
"psrlw $15, %%" #regd " \n\t" \
|
|
|
|
"packuswb %%" #regd ", %%" #regd " \n\t" ::)
|
2002-02-18 09:40:05 +00:00
|
|
|
|
|
|
|
#define MOVQ_WTWO(regd) \
|
|
|
|
__asm __volatile ( \
|
2002-05-30 15:14:56 +00:00
|
|
|
"pcmpeqd %%" #regd ", %%" #regd " \n\t" \
|
|
|
|
"psrlw $15, %%" #regd " \n\t" \
|
|
|
|
"psllw $1, %%" #regd " \n\t"::)
|
2002-05-17 13:01:01 +00:00
|
|
|
|
2002-02-18 09:40:05 +00:00
|
|
|
#endif
|
|
|
|
|
2002-05-30 15:14:56 +00:00
|
|
|
// using regr as temporary and for the output result
|
2002-05-29 14:29:48 +00:00
|
|
|
// first argument is unmodifed and second is trashed
|
2002-06-03 17:35:35 +00:00
|
|
|
// regfe is supposed to contain 0xfefefefefefefefe
|
|
|
|
#define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \
|
2002-05-29 17:16:22 +00:00
|
|
|
"movq " #rega ", " #regr " \n\t"\
|
|
|
|
"pand " #regb ", " #regr " \n\t"\
|
2002-05-29 14:29:48 +00:00
|
|
|
"pxor " #rega ", " #regb " \n\t"\
|
2002-06-03 17:35:35 +00:00
|
|
|
"pand " #regfe "," #regb " \n\t"\
|
2002-05-29 14:29:48 +00:00
|
|
|
"psrlq $1, " #regb " \n\t"\
|
2002-05-29 17:16:22 +00:00
|
|
|
"paddb " #regb ", " #regr " \n\t"
|
2002-05-29 14:29:48 +00:00
|
|
|
|
2002-06-03 17:35:35 +00:00
|
|
|
#define PAVGB_MMX(rega, regb, regr, regfe) \
|
2002-05-29 17:16:22 +00:00
|
|
|
"movq " #rega ", " #regr " \n\t"\
|
|
|
|
"por " #regb ", " #regr " \n\t"\
|
2002-05-29 14:29:48 +00:00
|
|
|
"pxor " #rega ", " #regb " \n\t"\
|
2002-06-03 17:35:35 +00:00
|
|
|
"pand " #regfe "," #regb " \n\t"\
|
2002-05-29 14:29:48 +00:00
|
|
|
"psrlq $1, " #regb " \n\t"\
|
2002-05-29 17:16:22 +00:00
|
|
|
"psubb " #regb ", " #regr " \n\t"
|
2002-05-29 14:29:48 +00:00
|
|
|
|
2002-06-03 17:35:35 +00:00
|
|
|
// mm6 is supposed to contain 0xfefefefefefefefe
|
2002-05-29 19:57:21 +00:00
|
|
|
#define PAVGBP_MMX_NO_RND(rega, regb, regr, regc, regd, regp) \
|
|
|
|
"movq " #rega ", " #regr " \n\t"\
|
|
|
|
"movq " #regc ", " #regp " \n\t"\
|
|
|
|
"pand " #regb ", " #regr " \n\t"\
|
|
|
|
"pand " #regd ", " #regp " \n\t"\
|
|
|
|
"pxor " #rega ", " #regb " \n\t"\
|
|
|
|
"pxor " #regc ", " #regd " \n\t"\
|
2002-05-30 15:14:56 +00:00
|
|
|
"pand %%mm6, " #regb " \n\t"\
|
|
|
|
"pand %%mm6, " #regd " \n\t"\
|
2002-05-29 19:57:21 +00:00
|
|
|
"psrlq $1, " #regb " \n\t"\
|
|
|
|
"psrlq $1, " #regd " \n\t"\
|
|
|
|
"paddb " #regb ", " #regr " \n\t"\
|
|
|
|
"paddb " #regd ", " #regp " \n\t"
|
|
|
|
|
|
|
|
#define PAVGBP_MMX(rega, regb, regr, regc, regd, regp) \
|
|
|
|
"movq " #rega ", " #regr " \n\t"\
|
|
|
|
"movq " #regc ", " #regp " \n\t"\
|
|
|
|
"por " #regb ", " #regr " \n\t"\
|
|
|
|
"por " #regd ", " #regp " \n\t"\
|
|
|
|
"pxor " #rega ", " #regb " \n\t"\
|
|
|
|
"pxor " #regc ", " #regd " \n\t"\
|
2002-05-30 15:14:56 +00:00
|
|
|
"pand %%mm6, " #regb " \n\t"\
|
|
|
|
"pand %%mm6, " #regd " \n\t"\
|
2002-05-29 19:57:21 +00:00
|
|
|
"psrlq $1, " #regd " \n\t"\
|
|
|
|
"psrlq $1, " #regb " \n\t"\
|
|
|
|
"psubb " #regb ", " #regr " \n\t"\
|
|
|
|
"psubb " #regd ", " #regp " \n\t"
|
|
|
|
|
2002-05-29 17:16:22 +00:00
|
|
|
/***********************************/
|
|
|
|
/* MMX no rounding */
|
|
|
|
#define DEF(x, y) x ## _no_rnd_ ## y ##_mmx
|
2002-05-30 15:14:56 +00:00
|
|
|
#define SET_RND MOVQ_WONE
|
2002-05-29 19:57:21 +00:00
|
|
|
#define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX_NO_RND(a, b, c, d, e, f)
|
2002-06-03 17:35:35 +00:00
|
|
|
#define PAVGB(a, b, c, e) PAVGB_MMX_NO_RND(a, b, c, e)
|
2002-05-30 15:14:56 +00:00
|
|
|
|
2002-05-29 17:16:22 +00:00
|
|
|
#include "dsputil_mmx_rnd.h"
|
|
|
|
|
|
|
|
#undef DEF
|
2002-05-30 15:14:56 +00:00
|
|
|
#undef SET_RND
|
2002-05-29 19:57:21 +00:00
|
|
|
#undef PAVGBP
|
2002-06-03 17:35:35 +00:00
|
|
|
#undef PAVGB
|
2002-05-29 17:16:22 +00:00
|
|
|
/***********************************/
|
|
|
|
/* MMX rounding */
|
|
|
|
|
|
|
|
#define DEF(x, y) x ## _ ## y ##_mmx
|
2002-05-30 15:14:56 +00:00
|
|
|
#define SET_RND MOVQ_WTWO
|
2002-05-29 19:57:21 +00:00
|
|
|
#define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX(a, b, c, d, e, f)
|
2002-06-03 17:35:35 +00:00
|
|
|
#define PAVGB(a, b, c, e) PAVGB_MMX(a, b, c, e)
|
2002-05-30 15:14:56 +00:00
|
|
|
|
2002-05-29 17:16:22 +00:00
|
|
|
#include "dsputil_mmx_rnd.h"
|
|
|
|
|
|
|
|
#undef DEF
|
2002-05-30 15:14:56 +00:00
|
|
|
#undef SET_RND
|
2002-05-29 19:57:21 +00:00
|
|
|
#undef PAVGBP
|
2002-06-03 17:35:35 +00:00
|
|
|
#undef PAVGB
|
2002-05-17 13:01:01 +00:00
|
|
|
|
2001-07-22 14:18:56 +00:00
|
|
|
/***********************************/
|
|
|
|
/* 3Dnow specific */
|
|
|
|
|
|
|
|
#define DEF(x) x ## _3dnow
|
|
|
|
/* for Athlons PAVGUSB is prefered */
|
|
|
|
#define PAVGB "pavgusb"
|
|
|
|
|
|
|
|
#include "dsputil_mmx_avg.h"
|
|
|
|
|
|
|
|
#undef DEF
|
|
|
|
#undef PAVGB
|
|
|
|
|
|
|
|
/***********************************/
|
|
|
|
/* MMX2 specific */
|
|
|
|
|
2002-05-17 01:04:14 +00:00
|
|
|
#define DEF(x) x ## _mmx2
|
2001-07-22 14:18:56 +00:00
|
|
|
|
|
|
|
/* Introduced only in MMX2 set */
|
|
|
|
#define PAVGB "pavgb"
|
|
|
|
|
|
|
|
#include "dsputil_mmx_avg.h"
|
|
|
|
|
|
|
|
#undef DEF
|
|
|
|
#undef PAVGB
|
|
|
|
|
|
|
|
/***********************************/
|
|
|
|
/* standard MMX */
|
|
|
|
|
|
|
|
static void get_pixels_mmx(DCTELEM *block, const UINT8 *pixels, int line_size)
|
|
|
|
{
|
2002-05-17 01:04:14 +00:00
|
|
|
asm volatile(
|
|
|
|
"movl $-128, %%eax \n\t"
|
|
|
|
"pxor %%mm7, %%mm7 \n\t"
|
|
|
|
".balign 16 \n\t"
|
|
|
|
"1: \n\t"
|
|
|
|
"movq (%0), %%mm0 \n\t"
|
|
|
|
"movq (%0, %2), %%mm2 \n\t"
|
|
|
|
"movq %%mm0, %%mm1 \n\t"
|
|
|
|
"movq %%mm2, %%mm3 \n\t"
|
|
|
|
"punpcklbw %%mm7, %%mm0 \n\t"
|
|
|
|
"punpckhbw %%mm7, %%mm1 \n\t"
|
|
|
|
"punpcklbw %%mm7, %%mm2 \n\t"
|
|
|
|
"punpckhbw %%mm7, %%mm3 \n\t"
|
|
|
|
"movq %%mm0, (%1, %%eax)\n\t"
|
|
|
|
"movq %%mm1, 8(%1, %%eax)\n\t"
|
|
|
|
"movq %%mm2, 16(%1, %%eax)\n\t"
|
|
|
|
"movq %%mm3, 24(%1, %%eax)\n\t"
|
|
|
|
"addl %3, %0 \n\t"
|
|
|
|
"addl $32, %%eax \n\t"
|
|
|
|
"js 1b \n\t"
|
|
|
|
: "+r" (pixels)
|
|
|
|
: "r" (block+64), "r" (line_size), "r" (line_size*2)
|
|
|
|
: "%eax"
|
|
|
|
);
|
2001-07-22 14:18:56 +00:00
|
|
|
}
|
|
|
|
|
2002-12-27 23:51:46 +00:00
|
|
|
static inline void diff_pixels_mmx(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride)
|
2002-04-17 04:32:12 +00:00
|
|
|
{
|
|
|
|
asm volatile(
|
2002-05-17 01:04:14 +00:00
|
|
|
"pxor %%mm7, %%mm7 \n\t"
|
2002-04-17 04:32:12 +00:00
|
|
|
"movl $-128, %%eax \n\t"
|
2002-05-17 01:04:14 +00:00
|
|
|
".balign 16 \n\t"
|
2002-04-17 04:32:12 +00:00
|
|
|
"1: \n\t"
|
|
|
|
"movq (%0), %%mm0 \n\t"
|
|
|
|
"movq (%1), %%mm2 \n\t"
|
|
|
|
"movq %%mm0, %%mm1 \n\t"
|
|
|
|
"movq %%mm2, %%mm3 \n\t"
|
|
|
|
"punpcklbw %%mm7, %%mm0 \n\t"
|
|
|
|
"punpckhbw %%mm7, %%mm1 \n\t"
|
|
|
|
"punpcklbw %%mm7, %%mm2 \n\t"
|
|
|
|
"punpckhbw %%mm7, %%mm3 \n\t"
|
|
|
|
"psubw %%mm2, %%mm0 \n\t"
|
|
|
|
"psubw %%mm3, %%mm1 \n\t"
|
|
|
|
"movq %%mm0, (%2, %%eax)\n\t"
|
|
|
|
"movq %%mm1, 8(%2, %%eax)\n\t"
|
|
|
|
"addl %3, %0 \n\t"
|
|
|
|
"addl %3, %1 \n\t"
|
|
|
|
"addl $16, %%eax \n\t"
|
|
|
|
"jnz 1b \n\t"
|
|
|
|
: "+r" (s1), "+r" (s2)
|
|
|
|
: "r" (block+64), "r" (stride)
|
|
|
|
: "%eax"
|
|
|
|
);
|
|
|
|
}
|
|
|
|
|
2002-11-11 09:40:17 +00:00
|
|
|
void put_pixels_clamped_mmx(const DCTELEM *block, UINT8 *pixels, int line_size)
|
2001-07-22 14:18:56 +00:00
|
|
|
{
|
|
|
|
const DCTELEM *p;
|
|
|
|
UINT8 *pix;
|
|
|
|
|
|
|
|
/* read the pixels */
|
|
|
|
p = block;
|
|
|
|
pix = pixels;
|
2002-02-18 09:40:05 +00:00
|
|
|
/* unrolled loop */
|
2001-07-22 14:18:56 +00:00
|
|
|
__asm __volatile(
|
2001-11-10 14:36:12 +00:00
|
|
|
"movq %3, %%mm0\n\t"
|
|
|
|
"movq 8%3, %%mm1\n\t"
|
|
|
|
"movq 16%3, %%mm2\n\t"
|
|
|
|
"movq 24%3, %%mm3\n\t"
|
|
|
|
"movq 32%3, %%mm4\n\t"
|
|
|
|
"movq 40%3, %%mm5\n\t"
|
|
|
|
"movq 48%3, %%mm6\n\t"
|
|
|
|
"movq 56%3, %%mm7\n\t"
|
2001-07-22 14:18:56 +00:00
|
|
|
"packuswb %%mm1, %%mm0\n\t"
|
|
|
|
"packuswb %%mm3, %%mm2\n\t"
|
|
|
|
"packuswb %%mm5, %%mm4\n\t"
|
|
|
|
"packuswb %%mm7, %%mm6\n\t"
|
2001-11-10 14:36:12 +00:00
|
|
|
"movq %%mm0, (%0)\n\t"
|
|
|
|
"movq %%mm2, (%0, %1)\n\t"
|
|
|
|
"movq %%mm4, (%0, %1, 2)\n\t"
|
|
|
|
"movq %%mm6, (%0, %2)\n\t"
|
|
|
|
::"r" (pix), "r" (line_size), "r" (line_size*3), "m"(*p)
|
2001-07-22 14:18:56 +00:00
|
|
|
:"memory");
|
|
|
|
pix += line_size*4;
|
|
|
|
p += 32;
|
2002-02-18 09:40:05 +00:00
|
|
|
|
|
|
|
// if here would be an exact copy of the code above
|
|
|
|
// compiler would generate some very strange code
|
|
|
|
// thus using "r"
|
|
|
|
__asm __volatile(
|
|
|
|
"movq (%3), %%mm0\n\t"
|
|
|
|
"movq 8(%3), %%mm1\n\t"
|
|
|
|
"movq 16(%3), %%mm2\n\t"
|
|
|
|
"movq 24(%3), %%mm3\n\t"
|
|
|
|
"movq 32(%3), %%mm4\n\t"
|
|
|
|
"movq 40(%3), %%mm5\n\t"
|
|
|
|
"movq 48(%3), %%mm6\n\t"
|
|
|
|
"movq 56(%3), %%mm7\n\t"
|
|
|
|
"packuswb %%mm1, %%mm0\n\t"
|
|
|
|
"packuswb %%mm3, %%mm2\n\t"
|
|
|
|
"packuswb %%mm5, %%mm4\n\t"
|
|
|
|
"packuswb %%mm7, %%mm6\n\t"
|
|
|
|
"movq %%mm0, (%0)\n\t"
|
|
|
|
"movq %%mm2, (%0, %1)\n\t"
|
|
|
|
"movq %%mm4, (%0, %1, 2)\n\t"
|
|
|
|
"movq %%mm6, (%0, %2)\n\t"
|
|
|
|
::"r" (pix), "r" (line_size), "r" (line_size*3), "r"(p)
|
|
|
|
:"memory");
|
2001-07-22 14:18:56 +00:00
|
|
|
}
|
|
|
|
|
2002-11-11 09:40:17 +00:00
|
|
|
void add_pixels_clamped_mmx(const DCTELEM *block, UINT8 *pixels, int line_size)
|
2001-07-22 14:18:56 +00:00
|
|
|
{
|
|
|
|
const DCTELEM *p;
|
|
|
|
UINT8 *pix;
|
|
|
|
int i;
|
|
|
|
|
|
|
|
/* read the pixels */
|
|
|
|
p = block;
|
|
|
|
pix = pixels;
|
2002-02-18 09:40:05 +00:00
|
|
|
MOVQ_ZERO(mm7);
|
|
|
|
i = 4;
|
2002-04-24 12:18:59 +00:00
|
|
|
do {
|
2001-07-22 14:18:56 +00:00
|
|
|
__asm __volatile(
|
2002-04-24 12:18:59 +00:00
|
|
|
"movq (%2), %%mm0\n\t"
|
|
|
|
"movq 8(%2), %%mm1\n\t"
|
|
|
|
"movq 16(%2), %%mm2\n\t"
|
|
|
|
"movq 24(%2), %%mm3\n\t"
|
2001-07-22 14:18:56 +00:00
|
|
|
"movq %0, %%mm4\n\t"
|
|
|
|
"movq %1, %%mm6\n\t"
|
|
|
|
"movq %%mm4, %%mm5\n\t"
|
|
|
|
"punpcklbw %%mm7, %%mm4\n\t"
|
|
|
|
"punpckhbw %%mm7, %%mm5\n\t"
|
|
|
|
"paddsw %%mm4, %%mm0\n\t"
|
|
|
|
"paddsw %%mm5, %%mm1\n\t"
|
|
|
|
"movq %%mm6, %%mm5\n\t"
|
|
|
|
"punpcklbw %%mm7, %%mm6\n\t"
|
|
|
|
"punpckhbw %%mm7, %%mm5\n\t"
|
|
|
|
"paddsw %%mm6, %%mm2\n\t"
|
|
|
|
"paddsw %%mm5, %%mm3\n\t"
|
|
|
|
"packuswb %%mm1, %%mm0\n\t"
|
|
|
|
"packuswb %%mm3, %%mm2\n\t"
|
|
|
|
"movq %%mm0, %0\n\t"
|
|
|
|
"movq %%mm2, %1\n\t"
|
2001-11-10 14:36:12 +00:00
|
|
|
:"+m"(*pix), "+m"(*(pix+line_size))
|
2002-04-24 12:18:59 +00:00
|
|
|
:"r"(p)
|
2001-07-22 14:18:56 +00:00
|
|
|
:"memory");
|
|
|
|
pix += line_size*2;
|
|
|
|
p += 16;
|
2002-04-24 12:18:59 +00:00
|
|
|
} while (--i);
|
2001-07-22 14:18:56 +00:00
|
|
|
}
|
|
|
|
|
2002-09-11 12:39:53 +00:00
|
|
|
static void put_pixels8_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h)
|
2001-07-22 14:18:56 +00:00
|
|
|
{
|
2002-06-03 17:35:35 +00:00
|
|
|
__asm __volatile(
|
2002-05-23 15:48:30 +00:00
|
|
|
"lea (%3, %3), %%eax \n\t"
|
2002-05-23 16:19:07 +00:00
|
|
|
".balign 8 \n\t"
|
2002-05-23 15:48:30 +00:00
|
|
|
"1: \n\t"
|
|
|
|
"movq (%1), %%mm0 \n\t"
|
|
|
|
"movq (%1, %3), %%mm1 \n\t"
|
|
|
|
"movq %%mm0, (%2) \n\t"
|
|
|
|
"movq %%mm1, (%2, %3) \n\t"
|
|
|
|
"addl %%eax, %1 \n\t"
|
|
|
|
"addl %%eax, %2 \n\t"
|
|
|
|
"movq (%1), %%mm0 \n\t"
|
|
|
|
"movq (%1, %3), %%mm1 \n\t"
|
|
|
|
"movq %%mm0, (%2) \n\t"
|
|
|
|
"movq %%mm1, (%2, %3) \n\t"
|
|
|
|
"addl %%eax, %1 \n\t"
|
|
|
|
"addl %%eax, %2 \n\t"
|
|
|
|
"subl $4, %0 \n\t"
|
|
|
|
"jnz 1b \n\t"
|
|
|
|
: "+g"(h), "+r" (pixels), "+r" (block)
|
|
|
|
: "r"(line_size)
|
|
|
|
: "%eax", "memory"
|
|
|
|
);
|
2001-07-22 14:18:56 +00:00
|
|
|
}
|
|
|
|
|
2002-09-11 12:39:53 +00:00
|
|
|
static void put_pixels16_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h)
|
|
|
|
{
|
|
|
|
__asm __volatile(
|
|
|
|
"lea (%3, %3), %%eax \n\t"
|
|
|
|
".balign 8 \n\t"
|
|
|
|
"1: \n\t"
|
|
|
|
"movq (%1), %%mm0 \n\t"
|
|
|
|
"movq 8(%1), %%mm4 \n\t"
|
|
|
|
"movq (%1, %3), %%mm1 \n\t"
|
|
|
|
"movq 8(%1, %3), %%mm5 \n\t"
|
|
|
|
"movq %%mm0, (%2) \n\t"
|
|
|
|
"movq %%mm4, 8(%2) \n\t"
|
|
|
|
"movq %%mm1, (%2, %3) \n\t"
|
|
|
|
"movq %%mm5, 8(%2, %3) \n\t"
|
|
|
|
"addl %%eax, %1 \n\t"
|
|
|
|
"addl %%eax, %2 \n\t"
|
|
|
|
"movq (%1), %%mm0 \n\t"
|
|
|
|
"movq 8(%1), %%mm4 \n\t"
|
|
|
|
"movq (%1, %3), %%mm1 \n\t"
|
|
|
|
"movq 8(%1, %3), %%mm5 \n\t"
|
|
|
|
"movq %%mm0, (%2) \n\t"
|
|
|
|
"movq %%mm4, 8(%2) \n\t"
|
|
|
|
"movq %%mm1, (%2, %3) \n\t"
|
|
|
|
"movq %%mm5, 8(%2, %3) \n\t"
|
|
|
|
"addl %%eax, %1 \n\t"
|
|
|
|
"addl %%eax, %2 \n\t"
|
|
|
|
"subl $4, %0 \n\t"
|
|
|
|
"jnz 1b \n\t"
|
|
|
|
: "+g"(h), "+r" (pixels), "+r" (block)
|
|
|
|
: "r"(line_size)
|
|
|
|
: "%eax", "memory"
|
|
|
|
);
|
|
|
|
}
|
|
|
|
|
2002-03-28 13:41:04 +00:00
|
|
|
static void clear_blocks_mmx(DCTELEM *blocks)
|
|
|
|
{
|
2002-06-03 17:35:35 +00:00
|
|
|
__asm __volatile(
|
2002-03-28 13:41:04 +00:00
|
|
|
"pxor %%mm7, %%mm7 \n\t"
|
|
|
|
"movl $-128*6, %%eax \n\t"
|
|
|
|
"1: \n\t"
|
|
|
|
"movq %%mm7, (%0, %%eax) \n\t"
|
|
|
|
"movq %%mm7, 8(%0, %%eax) \n\t"
|
|
|
|
"movq %%mm7, 16(%0, %%eax) \n\t"
|
|
|
|
"movq %%mm7, 24(%0, %%eax) \n\t"
|
|
|
|
"addl $32, %%eax \n\t"
|
|
|
|
" js 1b \n\t"
|
|
|
|
: : "r" (((int)blocks)+128*6)
|
|
|
|
: "%eax"
|
|
|
|
);
|
|
|
|
}
|
|
|
|
|
2002-09-20 10:03:01 +00:00
|
|
|
static int pix_sum16_mmx(UINT8 * pix, int line_size){
|
|
|
|
const int h=16;
|
|
|
|
int sum;
|
|
|
|
int index= -line_size*h;
|
|
|
|
|
|
|
|
__asm __volatile(
|
|
|
|
"pxor %%mm7, %%mm7 \n\t"
|
|
|
|
"pxor %%mm6, %%mm6 \n\t"
|
|
|
|
"1: \n\t"
|
|
|
|
"movq (%2, %1), %%mm0 \n\t"
|
|
|
|
"movq (%2, %1), %%mm1 \n\t"
|
|
|
|
"movq 8(%2, %1), %%mm2 \n\t"
|
|
|
|
"movq 8(%2, %1), %%mm3 \n\t"
|
|
|
|
"punpcklbw %%mm7, %%mm0 \n\t"
|
|
|
|
"punpckhbw %%mm7, %%mm1 \n\t"
|
|
|
|
"punpcklbw %%mm7, %%mm2 \n\t"
|
|
|
|
"punpckhbw %%mm7, %%mm3 \n\t"
|
|
|
|
"paddw %%mm0, %%mm1 \n\t"
|
|
|
|
"paddw %%mm2, %%mm3 \n\t"
|
|
|
|
"paddw %%mm1, %%mm3 \n\t"
|
|
|
|
"paddw %%mm3, %%mm6 \n\t"
|
|
|
|
"addl %3, %1 \n\t"
|
|
|
|
" js 1b \n\t"
|
|
|
|
"movq %%mm6, %%mm5 \n\t"
|
|
|
|
"psrlq $32, %%mm6 \n\t"
|
|
|
|
"paddw %%mm5, %%mm6 \n\t"
|
|
|
|
"movq %%mm6, %%mm5 \n\t"
|
|
|
|
"psrlq $16, %%mm6 \n\t"
|
|
|
|
"paddw %%mm5, %%mm6 \n\t"
|
|
|
|
"movd %%mm6, %0 \n\t"
|
|
|
|
"andl $0xFFFF, %0 \n\t"
|
|
|
|
: "=&r" (sum), "+r" (index)
|
|
|
|
: "r" (pix - index), "r" (line_size)
|
|
|
|
);
|
|
|
|
|
|
|
|
return sum;
|
|
|
|
}
|
|
|
|
|
2002-11-14 19:20:04 +00:00
|
|
|
static void add_bytes_mmx(uint8_t *dst, uint8_t *src, int w){
|
|
|
|
int i=0;
|
|
|
|
asm volatile(
|
|
|
|
"1: \n\t"
|
|
|
|
"movq (%1, %0), %%mm0 \n\t"
|
|
|
|
"movq (%2, %0), %%mm1 \n\t"
|
|
|
|
"paddb %%mm0, %%mm1 \n\t"
|
|
|
|
"movq %%mm1, (%2, %0) \n\t"
|
|
|
|
"movq 8(%1, %0), %%mm0 \n\t"
|
|
|
|
"movq 8(%2, %0), %%mm1 \n\t"
|
|
|
|
"paddb %%mm0, %%mm1 \n\t"
|
|
|
|
"movq %%mm1, 8(%2, %0) \n\t"
|
|
|
|
"addl $16, %0 \n\t"
|
|
|
|
"cmpl %3, %0 \n\t"
|
|
|
|
" jb 1b \n\t"
|
|
|
|
: "+r" (i)
|
|
|
|
: "r"(src), "r"(dst), "r"(w-15)
|
|
|
|
);
|
|
|
|
for(; i<w; i++)
|
|
|
|
dst[i+0] += src[i+0];
|
|
|
|
}
|
|
|
|
|
|
|
|
static void diff_bytes_mmx(uint8_t *dst, uint8_t *src1, uint8_t *src2, int w){
|
|
|
|
int i=0;
|
|
|
|
asm volatile(
|
|
|
|
"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"
|
|
|
|
"addl $16, %0 \n\t"
|
|
|
|
"cmpl %4, %0 \n\t"
|
|
|
|
" jb 1b \n\t"
|
|
|
|
: "+r" (i)
|
|
|
|
: "r"(src1), "r"(src2), "r"(dst), "r"(w-15)
|
|
|
|
);
|
|
|
|
for(; i<w; i++)
|
|
|
|
dst[i+0] = src1[i+0]-src2[i+0];
|
|
|
|
}
|
2002-12-27 23:51:46 +00:00
|
|
|
#define LBUTTERFLY(a,b)\
|
|
|
|
"paddw " #b ", " #a " \n\t"\
|
|
|
|
"paddw " #b ", " #b " \n\t"\
|
|
|
|
"psubw " #a ", " #b " \n\t"
|
|
|
|
|
|
|
|
#define HADAMARD48\
|
|
|
|
LBUTTERFLY(%%mm0, %%mm1)\
|
|
|
|
LBUTTERFLY(%%mm2, %%mm3)\
|
|
|
|
LBUTTERFLY(%%mm4, %%mm5)\
|
|
|
|
LBUTTERFLY(%%mm6, %%mm7)\
|
|
|
|
\
|
|
|
|
LBUTTERFLY(%%mm0, %%mm2)\
|
|
|
|
LBUTTERFLY(%%mm1, %%mm3)\
|
|
|
|
LBUTTERFLY(%%mm4, %%mm6)\
|
|
|
|
LBUTTERFLY(%%mm5, %%mm7)\
|
|
|
|
\
|
|
|
|
LBUTTERFLY(%%mm0, %%mm4)\
|
|
|
|
LBUTTERFLY(%%mm1, %%mm5)\
|
|
|
|
LBUTTERFLY(%%mm2, %%mm6)\
|
|
|
|
LBUTTERFLY(%%mm3, %%mm7)
|
|
|
|
|
|
|
|
#define MMABS(a,z)\
|
|
|
|
"pxor " #z ", " #z " \n\t"\
|
|
|
|
"pcmpgtw " #a ", " #z " \n\t"\
|
|
|
|
"pxor " #z ", " #a " \n\t"\
|
|
|
|
"psubw " #z ", " #a " \n\t"
|
|
|
|
|
|
|
|
#define MMABS_SUM(a,z, sum)\
|
|
|
|
"pxor " #z ", " #z " \n\t"\
|
|
|
|
"pcmpgtw " #a ", " #z " \n\t"\
|
|
|
|
"pxor " #z ", " #a " \n\t"\
|
|
|
|
"psubw " #z ", " #a " \n\t"\
|
|
|
|
"paddusw " #a ", " #sum " \n\t"
|
|
|
|
|
|
|
|
|
|
|
|
#define SBUTTERFLY(a,b,t,n)\
|
|
|
|
"movq " #a ", " #t " \n\t" /* abcd */\
|
|
|
|
"punpckl" #n " " #b ", " #a " \n\t" /* aebf */\
|
|
|
|
"punpckh" #n " " #b ", " #t " \n\t" /* cgdh */\
|
|
|
|
|
|
|
|
#define TRANSPOSE4(a,b,c,d,t)\
|
|
|
|
SBUTTERFLY(a,b,t,wd) /* a=aebf t=cgdh */\
|
|
|
|
SBUTTERFLY(c,d,b,wd) /* c=imjn b=kolp */\
|
|
|
|
SBUTTERFLY(a,c,d,dq) /* a=aeim d=bfjn */\
|
|
|
|
SBUTTERFLY(t,b,c,dq) /* t=cgko c=dhlp */
|
|
|
|
|
|
|
|
#define LOAD4(o, a, b, c, d)\
|
|
|
|
"movq "#o"(%1), " #a " \n\t"\
|
|
|
|
"movq "#o"+16(%1), " #b " \n\t"\
|
|
|
|
"movq "#o"+32(%1), " #c " \n\t"\
|
|
|
|
"movq "#o"+48(%1), " #d " \n\t"
|
|
|
|
|
|
|
|
#define STORE4(o, a, b, c, d)\
|
|
|
|
"movq "#a", "#o"(%1) \n\t"\
|
|
|
|
"movq "#b", "#o"+16(%1) \n\t"\
|
|
|
|
"movq "#c", "#o"+32(%1) \n\t"\
|
|
|
|
"movq "#d", "#o"+48(%1) \n\t"\
|
|
|
|
|
|
|
|
static int hadamard8_diff_mmx(void *s, uint8_t *src1, uint8_t *src2, int stride){
|
|
|
|
uint64_t temp[16] __align8;
|
|
|
|
int sum=0;
|
|
|
|
|
|
|
|
diff_pixels_mmx((DCTELEM*)temp, src1, src2, stride);
|
2002-11-14 19:20:04 +00:00
|
|
|
|
2002-12-27 23:51:46 +00:00
|
|
|
asm volatile(
|
|
|
|
LOAD4(0 , %%mm0, %%mm1, %%mm2, %%mm3)
|
|
|
|
LOAD4(64, %%mm4, %%mm5, %%mm6, %%mm7)
|
|
|
|
|
|
|
|
HADAMARD48
|
|
|
|
|
|
|
|
"movq %%mm7, 112(%1) \n\t"
|
|
|
|
|
|
|
|
TRANSPOSE4(%%mm0, %%mm1, %%mm2, %%mm3, %%mm7)
|
|
|
|
STORE4(0 , %%mm0, %%mm3, %%mm7, %%mm2)
|
|
|
|
|
|
|
|
"movq 112(%1), %%mm7 \n\t"
|
|
|
|
TRANSPOSE4(%%mm4, %%mm5, %%mm6, %%mm7, %%mm0)
|
|
|
|
STORE4(64, %%mm4, %%mm7, %%mm0, %%mm6)
|
|
|
|
|
|
|
|
LOAD4(8 , %%mm0, %%mm1, %%mm2, %%mm3)
|
|
|
|
LOAD4(72, %%mm4, %%mm5, %%mm6, %%mm7)
|
|
|
|
|
|
|
|
HADAMARD48
|
|
|
|
|
|
|
|
"movq %%mm7, 120(%1) \n\t"
|
|
|
|
|
|
|
|
TRANSPOSE4(%%mm0, %%mm1, %%mm2, %%mm3, %%mm7)
|
|
|
|
STORE4(8 , %%mm0, %%mm3, %%mm7, %%mm2)
|
|
|
|
|
|
|
|
"movq 120(%1), %%mm7 \n\t"
|
|
|
|
TRANSPOSE4(%%mm4, %%mm5, %%mm6, %%mm7, %%mm0)
|
|
|
|
"movq %%mm7, %%mm5 \n\t"//FIXME remove
|
|
|
|
"movq %%mm6, %%mm7 \n\t"
|
|
|
|
"movq %%mm0, %%mm6 \n\t"
|
|
|
|
// STORE4(72, %%mm4, %%mm7, %%mm0, %%mm6) //FIXME remove
|
|
|
|
|
|
|
|
LOAD4(64, %%mm0, %%mm1, %%mm2, %%mm3)
|
|
|
|
// LOAD4(72, %%mm4, %%mm5, %%mm6, %%mm7)
|
|
|
|
|
|
|
|
HADAMARD48
|
|
|
|
"movq %%mm7, 64(%1) \n\t"
|
|
|
|
MMABS(%%mm0, %%mm7)
|
|
|
|
MMABS_SUM(%%mm1, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm2, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm3, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm4, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm5, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm6, %%mm7, %%mm0)
|
|
|
|
"movq 64(%1), %%mm1 \n\t"
|
|
|
|
MMABS_SUM(%%mm1, %%mm7, %%mm0)
|
|
|
|
"movq %%mm0, 64(%1) \n\t"
|
|
|
|
|
|
|
|
LOAD4(0 , %%mm0, %%mm1, %%mm2, %%mm3)
|
|
|
|
LOAD4(8 , %%mm4, %%mm5, %%mm6, %%mm7)
|
|
|
|
|
|
|
|
HADAMARD48
|
|
|
|
"movq %%mm7, (%1) \n\t"
|
|
|
|
MMABS(%%mm0, %%mm7)
|
|
|
|
MMABS_SUM(%%mm1, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm2, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm3, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm4, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm5, %%mm7, %%mm0)
|
|
|
|
MMABS_SUM(%%mm6, %%mm7, %%mm0)
|
|
|
|
"movq (%1), %%mm1 \n\t"
|
|
|
|
MMABS_SUM(%%mm1, %%mm7, %%mm0)
|
|
|
|
"movq 64(%1), %%mm1 \n\t"
|
|
|
|
MMABS_SUM(%%mm1, %%mm7, %%mm0)
|
|
|
|
|
|
|
|
"movq %%mm0, %%mm1 \n\t"
|
|
|
|
"psrlq $32, %%mm0 \n\t"
|
|
|
|
"paddusw %%mm1, %%mm0 \n\t"
|
|
|
|
"movq %%mm0, %%mm1 \n\t"
|
|
|
|
"psrlq $16, %%mm0 \n\t"
|
|
|
|
"paddusw %%mm1, %%mm0 \n\t"
|
|
|
|
"movd %%mm0, %0 \n\t"
|
|
|
|
|
|
|
|
: "=r" (sum)
|
|
|
|
: "r"(temp)
|
|
|
|
);
|
|
|
|
return sum&0xFFFF;
|
|
|
|
}
|
|
|
|
|
|
|
|
WARPER88_1616(hadamard8_diff_mmx, hadamard8_diff16_mmx)
|
2002-11-14 19:20:04 +00:00
|
|
|
|
2002-05-18 22:59:12 +00:00
|
|
|
#if 0
|
2002-02-18 09:40:05 +00:00
|
|
|
static void just_return() { return; }
|
2002-05-18 22:59:12 +00:00
|
|
|
#endif
|
2002-02-18 09:40:05 +00:00
|
|
|
|
2002-11-11 09:40:17 +00:00
|
|
|
void dsputil_init_mmx(DSPContext* c, unsigned mask)
|
2001-07-22 14:18:56 +00:00
|
|
|
{
|
|
|
|
mm_flags = mm_support();
|
2002-09-01 07:26:43 +00:00
|
|
|
#if 0
|
|
|
|
fprintf(stderr, "libavcodec: CPU flags:");
|
2001-07-22 14:18:56 +00:00
|
|
|
if (mm_flags & MM_MMX)
|
2002-09-01 07:26:43 +00:00
|
|
|
fprintf(stderr, " mmx");
|
2001-07-22 14:18:56 +00:00
|
|
|
if (mm_flags & MM_MMXEXT)
|
2002-09-01 07:26:43 +00:00
|
|
|
fprintf(stderr, " mmxext");
|
2001-07-22 14:18:56 +00:00
|
|
|
if (mm_flags & MM_3DNOW)
|
2002-09-01 07:26:43 +00:00
|
|
|
fprintf(stderr, " 3dnow");
|
2001-07-22 14:18:56 +00:00
|
|
|
if (mm_flags & MM_SSE)
|
2002-09-01 07:26:43 +00:00
|
|
|
fprintf(stderr, " sse");
|
2001-07-22 14:18:56 +00:00
|
|
|
if (mm_flags & MM_SSE2)
|
2002-09-01 07:26:43 +00:00
|
|
|
fprintf(stderr, " sse2");
|
|
|
|
fprintf(stderr, "\n");
|
2001-07-22 14:18:56 +00:00
|
|
|
#endif
|
|
|
|
|
|
|
|
if (mm_flags & MM_MMX) {
|
2002-11-11 09:40:17 +00:00
|
|
|
c->get_pixels = get_pixels_mmx;
|
|
|
|
c->diff_pixels = diff_pixels_mmx;
|
|
|
|
c->put_pixels_clamped = put_pixels_clamped_mmx;
|
|
|
|
c->add_pixels_clamped = add_pixels_clamped_mmx;
|
|
|
|
c->clear_blocks = clear_blocks_mmx;
|
|
|
|
c->pix_sum = pix_sum16_mmx;
|
|
|
|
|
|
|
|
c->pix_abs16x16 = pix_abs16x16_mmx;
|
|
|
|
c->pix_abs16x16_x2 = pix_abs16x16_x2_mmx;
|
|
|
|
c->pix_abs16x16_y2 = pix_abs16x16_y2_mmx;
|
|
|
|
c->pix_abs16x16_xy2 = pix_abs16x16_xy2_mmx;
|
|
|
|
c->pix_abs8x8 = pix_abs8x8_mmx;
|
|
|
|
c->pix_abs8x8_x2 = pix_abs8x8_x2_mmx;
|
|
|
|
c->pix_abs8x8_y2 = pix_abs8x8_y2_mmx;
|
|
|
|
c->pix_abs8x8_xy2 = pix_abs8x8_xy2_mmx;
|
|
|
|
|
|
|
|
c->put_pixels_tab[0][0] = put_pixels16_mmx;
|
|
|
|
c->put_pixels_tab[0][1] = put_pixels16_x2_mmx;
|
|
|
|
c->put_pixels_tab[0][2] = put_pixels16_y2_mmx;
|
|
|
|
c->put_pixels_tab[0][3] = put_pixels16_xy2_mmx;
|
|
|
|
|
|
|
|
c->put_no_rnd_pixels_tab[0][0] = put_pixels16_mmx;
|
|
|
|
c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmx;
|
|
|
|
c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_mmx;
|
|
|
|
c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy2_mmx;
|
|
|
|
|
|
|
|
c->avg_pixels_tab[0][0] = avg_pixels16_mmx;
|
|
|
|
c->avg_pixels_tab[0][1] = avg_pixels16_x2_mmx;
|
|
|
|
c->avg_pixels_tab[0][2] = avg_pixels16_y2_mmx;
|
|
|
|
c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmx;
|
|
|
|
|
|
|
|
c->avg_no_rnd_pixels_tab[0][0] = avg_no_rnd_pixels16_mmx;
|
|
|
|
c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x2_mmx;
|
|
|
|
c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y2_mmx;
|
|
|
|
c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy2_mmx;
|
|
|
|
|
|
|
|
c->put_pixels_tab[1][0] = put_pixels8_mmx;
|
|
|
|
c->put_pixels_tab[1][1] = put_pixels8_x2_mmx;
|
|
|
|
c->put_pixels_tab[1][2] = put_pixels8_y2_mmx;
|
|
|
|
c->put_pixels_tab[1][3] = put_pixels8_xy2_mmx;
|
|
|
|
|
|
|
|
c->put_no_rnd_pixels_tab[1][0] = put_pixels8_mmx;
|
|
|
|
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_mmx;
|
|
|
|
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_mmx;
|
|
|
|
c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy2_mmx;
|
|
|
|
|
|
|
|
c->avg_pixels_tab[1][0] = avg_pixels8_mmx;
|
|
|
|
c->avg_pixels_tab[1][1] = avg_pixels8_x2_mmx;
|
|
|
|
c->avg_pixels_tab[1][2] = avg_pixels8_y2_mmx;
|
|
|
|
c->avg_pixels_tab[1][3] = avg_pixels8_xy2_mmx;
|
|
|
|
|
|
|
|
c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_mmx;
|
|
|
|
c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x2_mmx;
|
|
|
|
c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y2_mmx;
|
|
|
|
c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy2_mmx;
|
2002-11-14 19:20:04 +00:00
|
|
|
|
|
|
|
c->add_bytes= add_bytes_mmx;
|
|
|
|
c->diff_bytes= diff_bytes_mmx;
|
2002-12-27 23:51:46 +00:00
|
|
|
|
|
|
|
c->hadamard8_diff[0]= hadamard8_diff16_mmx;
|
|
|
|
c->hadamard8_diff[1]= hadamard8_diff_mmx;
|
|
|
|
|
|
|
|
c->sad[0]= sad16x16_mmx;
|
|
|
|
c->sad[1]= sad8x8_mmx;
|
|
|
|
|
2001-07-22 14:18:56 +00:00
|
|
|
if (mm_flags & MM_MMXEXT) {
|
2002-11-11 09:40:17 +00:00
|
|
|
c->pix_abs16x16 = pix_abs16x16_mmx2;
|
|
|
|
c->pix_abs16x16_x2 = pix_abs16x16_x2_mmx2;
|
|
|
|
c->pix_abs16x16_y2 = pix_abs16x16_y2_mmx2;
|
|
|
|
c->pix_abs16x16_xy2 = pix_abs16x16_xy2_mmx2;
|
|
|
|
|
|
|
|
c->pix_abs8x8 = pix_abs8x8_mmx2;
|
|
|
|
c->pix_abs8x8_x2 = pix_abs8x8_x2_mmx2;
|
|
|
|
c->pix_abs8x8_y2 = pix_abs8x8_y2_mmx2;
|
|
|
|
c->pix_abs8x8_xy2 = pix_abs8x8_xy2_mmx2;
|
|
|
|
|
2002-12-27 23:51:46 +00:00
|
|
|
c->sad[0]= sad16x16_mmx2;
|
|
|
|
c->sad[1]= sad8x8_mmx2;
|
|
|
|
|
2002-11-11 09:40:17 +00:00
|
|
|
c->put_pixels_tab[0][1] = put_pixels16_x2_mmx2;
|
|
|
|
c->put_pixels_tab[0][2] = put_pixels16_y2_mmx2;
|
|
|
|
c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmx2;
|
|
|
|
c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_mmx2;
|
|
|
|
|
|
|
|
c->avg_pixels_tab[0][0] = avg_pixels16_mmx2;
|
|
|
|
c->avg_pixels_tab[0][1] = avg_pixels16_x2_mmx2;
|
|
|
|
c->avg_pixels_tab[0][2] = avg_pixels16_y2_mmx2;
|
|
|
|
c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmx2;
|
|
|
|
|
|
|
|
c->put_pixels_tab[1][1] = put_pixels8_x2_mmx2;
|
|
|
|
c->put_pixels_tab[1][2] = put_pixels8_y2_mmx2;
|
|
|
|
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_mmx2;
|
|
|
|
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_mmx2;
|
|
|
|
|
|
|
|
c->avg_pixels_tab[1][0] = avg_pixels8_mmx2;
|
|
|
|
c->avg_pixels_tab[1][1] = avg_pixels8_x2_mmx2;
|
|
|
|
c->avg_pixels_tab[1][2] = avg_pixels8_y2_mmx2;
|
|
|
|
c->avg_pixels_tab[1][3] = avg_pixels8_xy2_mmx2;
|
2001-07-22 14:18:56 +00:00
|
|
|
} else if (mm_flags & MM_3DNOW) {
|
2002-11-11 09:40:17 +00:00
|
|
|
c->put_pixels_tab[0][1] = put_pixels16_x2_3dnow;
|
|
|
|
c->put_pixels_tab[0][2] = put_pixels16_y2_3dnow;
|
|
|
|
c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_3dnow;
|
|
|
|
c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_3dnow;
|
|
|
|
|
|
|
|
c->avg_pixels_tab[0][0] = avg_pixels16_3dnow;
|
|
|
|
c->avg_pixels_tab[0][1] = avg_pixels16_x2_3dnow;
|
|
|
|
c->avg_pixels_tab[0][2] = avg_pixels16_y2_3dnow;
|
|
|
|
c->avg_pixels_tab[0][3] = avg_pixels16_xy2_3dnow;
|
|
|
|
|
|
|
|
c->put_pixels_tab[1][1] = put_pixels8_x2_3dnow;
|
|
|
|
c->put_pixels_tab[1][2] = put_pixels8_y2_3dnow;
|
|
|
|
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_3dnow;
|
|
|
|
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_3dnow;
|
|
|
|
|
|
|
|
c->avg_pixels_tab[1][0] = avg_pixels8_3dnow;
|
|
|
|
c->avg_pixels_tab[1][1] = avg_pixels8_x2_3dnow;
|
|
|
|
c->avg_pixels_tab[1][2] = avg_pixels8_y2_3dnow;
|
|
|
|
c->avg_pixels_tab[1][3] = avg_pixels8_xy2_3dnow;
|
2001-07-22 14:18:56 +00:00
|
|
|
}
|
|
|
|
}
|
2002-02-18 09:40:05 +00:00
|
|
|
|
|
|
|
#if 0
|
|
|
|
// for speed testing
|
|
|
|
get_pixels = just_return;
|
|
|
|
put_pixels_clamped = just_return;
|
|
|
|
add_pixels_clamped = just_return;
|
|
|
|
|
|
|
|
pix_abs16x16 = just_return;
|
|
|
|
pix_abs16x16_x2 = just_return;
|
|
|
|
pix_abs16x16_y2 = just_return;
|
|
|
|
pix_abs16x16_xy2 = just_return;
|
|
|
|
|
|
|
|
put_pixels_tab[0] = just_return;
|
|
|
|
put_pixels_tab[1] = just_return;
|
|
|
|
put_pixels_tab[2] = just_return;
|
|
|
|
put_pixels_tab[3] = just_return;
|
|
|
|
|
|
|
|
put_no_rnd_pixels_tab[0] = just_return;
|
|
|
|
put_no_rnd_pixels_tab[1] = just_return;
|
|
|
|
put_no_rnd_pixels_tab[2] = just_return;
|
|
|
|
put_no_rnd_pixels_tab[3] = just_return;
|
|
|
|
|
|
|
|
avg_pixels_tab[0] = just_return;
|
|
|
|
avg_pixels_tab[1] = just_return;
|
|
|
|
avg_pixels_tab[2] = just_return;
|
|
|
|
avg_pixels_tab[3] = just_return;
|
|
|
|
|
|
|
|
avg_no_rnd_pixels_tab[0] = just_return;
|
|
|
|
avg_no_rnd_pixels_tab[1] = just_return;
|
|
|
|
avg_no_rnd_pixels_tab[2] = just_return;
|
|
|
|
avg_no_rnd_pixels_tab[3] = just_return;
|
|
|
|
|
|
|
|
//av_fdct = just_return;
|
|
|
|
//ff_idct = just_return;
|
|
|
|
#endif
|
2001-07-22 14:18:56 +00:00
|
|
|
}
|
2002-05-20 16:25:09 +00:00
|
|
|
|
|
|
|
/* remove any non bit exact operation (testing purpose). NOTE that
|
|
|
|
this function should be kept as small as possible because it is
|
|
|
|
always difficult to test automatically non bit exact cases. */
|
2002-11-11 09:40:17 +00:00
|
|
|
void dsputil_set_bit_exact_mmx(DSPContext* c, unsigned mask)
|
2002-05-20 16:25:09 +00:00
|
|
|
{
|
|
|
|
if (mm_flags & MM_MMX) {
|
2002-09-11 12:39:53 +00:00
|
|
|
/* MMX2 & 3DNOW */
|
2002-11-11 09:40:17 +00:00
|
|
|
c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_mmx;
|
|
|
|
c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_mmx;
|
|
|
|
c->avg_pixels_tab[0][3] = avg_pixels16_xy2_mmx;
|
|
|
|
c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_mmx;
|
|
|
|
c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_mmx;
|
|
|
|
c->avg_pixels_tab[1][3] = avg_pixels8_xy2_mmx;
|
2002-07-21 00:48:50 +00:00
|
|
|
|
2002-09-11 12:39:53 +00:00
|
|
|
if (mm_flags & MM_MMXEXT) {
|
2002-11-11 09:40:17 +00:00
|
|
|
c->pix_abs16x16_x2 = pix_abs16x16_x2_mmx;
|
|
|
|
c->pix_abs16x16_y2 = pix_abs16x16_y2_mmx;
|
|
|
|
c->pix_abs16x16_xy2 = pix_abs16x16_xy2_mmx;
|
|
|
|
c->pix_abs8x8_x2 = pix_abs8x8_x2_mmx;
|
|
|
|
c->pix_abs8x8_y2 = pix_abs8x8_y2_mmx;
|
|
|
|
c->pix_abs8x8_xy2= pix_abs8x8_xy2_mmx;
|
2002-05-20 16:25:09 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|