diff options
Diffstat (limited to 'src/libffmpeg/libavcodec/i386/dsputil_mmx.c')
| -rw-r--r-- | src/libffmpeg/libavcodec/i386/dsputil_mmx.c | 1104 |
1 files changed, 228 insertions, 876 deletions
diff --git a/src/libffmpeg/libavcodec/i386/dsputil_mmx.c b/src/libffmpeg/libavcodec/i386/dsputil_mmx.c index 2c71850ee..b8eaa5fbd 100644 --- a/src/libffmpeg/libavcodec/i386/dsputil_mmx.c +++ b/src/libffmpeg/libavcodec/i386/dsputil_mmx.c @@ -1,25 +1,24 @@ /* * MMX optimized DSP utils - * Copyright (c) 2000, 2001 Gerard Lantau. + * Copyright (c) 2000, 2001 Fabrice Bellard. * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License as published by - * the Free Software Foundation; either version 2 of the License, or - * (at your option) any later version. + * 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. * - * This program is distributed in the hope that it will be useful, + * This library is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of - * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the - * GNU General Public License for more details. + * 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 General Public License - * along with this program; if not, write to the Free Software - * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + * 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 * * MMX optimization by Nick Kurshev <nickols_k@mail.ru> */ -#include "xine-utils/xineutils.h" #include "../dsputil.h" #include "../simple_idct.h" @@ -45,38 +44,124 @@ 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); - /* external functions, from idct_mmx.c */ void ff_mmx_idct(DCTELEM *block); void ff_mmxext_idct(DCTELEM *block); /* pixel operations */ -static const unsigned long long int mm_wone __attribute__ ((aligned(8))) = 0x0001000100010001LL; -static const unsigned long long int mm_wtwo __attribute__ ((aligned(8))) = 0x0002000200020002LL; -//static const unsigned short mm_wone[4] __attribute__ ((aligned(8))) = { 0x1, 0x1, 0x1, 0x1 }; -//static const unsigned short mm_wtwo[4] __attribute__ ((aligned(8))) = { 0x2, 0x2, 0x2, 0x2 }; +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; #define JUMPALIGN() __asm __volatile (".balign 8"::) #define MOVQ_ZERO(regd) __asm __volatile ("pxor %%" #regd ", %%" #regd ::) +#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" ::) + #ifndef PIC -#define MOVQ_WONE(regd) __asm __volatile ("movq %0, %%" #regd " \n\t" ::"m"(mm_wone)) +#define MOVQ_BONE(regd) __asm __volatile ("movq %0, %%" #regd " \n\t" ::"m"(mm_bone)) #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 -#define MOVQ_WONE(regd) \ +#define MOVQ_BONE(regd) \ __asm __volatile ( \ - "pcmpeqd %%" #regd ", %%" #regd " \n\t" \ - "psrlw $15, %%" #regd ::) + "pcmpeqd %%" #regd ", %%" #regd " \n\t" \ + "psrlw $15, %%" #regd " \n\t" \ + "packuswb %%" #regd ", %%" #regd " \n\t" ::) #define MOVQ_WTWO(regd) \ __asm __volatile ( \ - "pcmpeqd %%" #regd ", %%" #regd " \n\t" \ - "psrlw $15, %%" #regd " \n\t" \ - "psllw $1, %%" #regd ::) + "pcmpeqd %%" #regd ", %%" #regd " \n\t" \ + "psrlw $15, %%" #regd " \n\t" \ + "psllw $1, %%" #regd " \n\t"::) + #endif +// using regr as temporary and for the output result +// first argument is unmodifed and second is trashed +// regfe is supposed to contain 0xfefefefefefefefe +#define PAVGB_MMX_NO_RND(rega, regb, regr, regfe) \ + "movq " #rega ", " #regr " \n\t"\ + "pand " #regb ", " #regr " \n\t"\ + "pxor " #rega ", " #regb " \n\t"\ + "pand " #regfe "," #regb " \n\t"\ + "psrlq $1, " #regb " \n\t"\ + "paddb " #regb ", " #regr " \n\t" + +#define PAVGB_MMX(rega, regb, regr, regfe) \ + "movq " #rega ", " #regr " \n\t"\ + "por " #regb ", " #regr " \n\t"\ + "pxor " #rega ", " #regb " \n\t"\ + "pand " #regfe "," #regb " \n\t"\ + "psrlq $1, " #regb " \n\t"\ + "psubb " #regb ", " #regr " \n\t" + +// mm6 is supposed to contain 0xfefefefefefefefe +#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"\ + "pand %%mm6, " #regb " \n\t"\ + "pand %%mm6, " #regd " \n\t"\ + "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"\ + "pand %%mm6, " #regb " \n\t"\ + "pand %%mm6, " #regd " \n\t"\ + "psrlq $1, " #regd " \n\t"\ + "psrlq $1, " #regb " \n\t"\ + "psubb " #regb ", " #regr " \n\t"\ + "psubb " #regd ", " #regp " \n\t" + +/***********************************/ +/* MMX no rounding */ +#define DEF(x, y) x ## _no_rnd_ ## y ##_mmx +#define SET_RND MOVQ_WONE +#define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX_NO_RND(a, b, c, d, e, f) +#define PAVGB(a, b, c, e) PAVGB_MMX_NO_RND(a, b, c, e) + +#include "dsputil_mmx_rnd.h" + +#undef DEF +#undef SET_RND +#undef PAVGBP +#undef PAVGB +/***********************************/ +/* MMX rounding */ + +#define DEF(x, y) x ## _ ## y ##_mmx +#define SET_RND MOVQ_WTWO +#define PAVGBP(a, b, c, d, e, f) PAVGBP_MMX(a, b, c, d, e, f) +#define PAVGB(a, b, c, e) PAVGB_MMX(a, b, c, e) + +#include "dsputil_mmx_rnd.h" + +#undef DEF +#undef SET_RND +#undef PAVGBP +#undef PAVGB + /***********************************/ /* 3Dnow specific */ @@ -92,7 +177,7 @@ static const unsigned long long int mm_wtwo __attribute__ ((aligned(8))) = 0x000 /***********************************/ /* MMX2 specific */ -#define DEF(x) x ## _sse +#define DEF(x) x ## _mmx2 /* Introduced only in MMX2 set */ #define PAVGB "pavgb" @@ -107,34 +192,59 @@ static const unsigned long long int mm_wtwo __attribute__ ((aligned(8))) = 0x000 static void get_pixels_mmx(DCTELEM *block, const UINT8 *pixels, int line_size) { - DCTELEM *p; - const UINT8 *pix; - int i; + 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" + ); +} - /* read the pixels */ - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - for(i=0;i<4;i++) { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %2, %%mm1\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "movq %%mm0, %0\n\t" - "movq %%mm2, 8%0\n\t" - "movq %%mm1, 16%0\n\t" - "movq %%mm3, 24%0\n\t" - :"=m"(*p) - :"m"(*pix), "m"(*(pix+line_size)) - :"memory"); - pix += line_size*2; - p += 16; - } +static void diff_pixels_mmx(DCTELEM *block, const UINT8 *s1, const UINT8 *s2, int stride) +{ + asm volatile( + "pxor %%mm7, %%mm7 \n\t" + "movl $-128, %%eax \n\t" + ".balign 16 \n\t" + "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" + ); } static void put_pixels_clamped_mmx(const DCTELEM *block, UINT8 *pixels, int line_size) @@ -203,12 +313,12 @@ static void add_pixels_clamped_mmx(const DCTELEM *block, UINT8 *pixels, int line pix = pixels; MOVQ_ZERO(mm7); i = 4; - while (i) { + do { __asm __volatile( - "movq %2, %%mm0\n\t" - "movq 8%2, %%mm1\n\t" - "movq 16%2, %%mm2\n\t" - "movq 24%2, %%mm3\n\t" + "movq (%2), %%mm0\n\t" + "movq 8(%2), %%mm1\n\t" + "movq 16(%2), %%mm2\n\t" + "movq 24(%2), %%mm3\n\t" "movq %0, %%mm4\n\t" "movq %1, %%mm6\n\t" "movq %%mm4, %%mm5\n\t" @@ -226,809 +336,42 @@ static void add_pixels_clamped_mmx(const DCTELEM *block, UINT8 *pixels, int line "movq %%mm0, %0\n\t" "movq %%mm2, %1\n\t" :"+m"(*pix), "+m"(*(pix+line_size)) - :"m"(*p) + :"r"(p) :"memory"); pix += line_size*2; p += 16; - i--; - }; + } while (--i); } static void put_pixels_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h) { - int hh; - UINT8 *p; - const UINT8 *pix; - - p = block; - pix = pixels; // 2s -#if 0 - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"=m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; - p += line_size; - } while (--h); -#else - // this optimized code is not very usefull - // the above loop is definitely faster - // at least on Celeron 500MHz - hh = h & 3; - while (hh) { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"=m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; - p += line_size; - hh--; - } - hh=h>>2; - while (hh) { - __asm __volatile( - "movq (%1), %%mm0 \n\t" - "movq (%1, %2), %%mm1 \n\t" - "movq (%1, %2, 2), %%mm2 \n\t" - "movq (%1, %3), %%mm3 \n\t" - "movq %%mm0, (%0) \n\t" - "movq %%mm1, (%0, %2) \n\t" - "movq %%mm2, (%0, %2, 2) \n\t" - "movq %%mm3, (%0, %3) \n\t" - ::"r"(p), "r"(pix), "r"(line_size), "r"(line_size*3) - :"memory"); - pix += line_size*4; - p += line_size*4; - hh--; - } -#endif -} - -static void put_pixels_x2_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WONE(mm4); - JUMPALIGN(); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq 1%1, %%mm1\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "paddusw %%mm4, %%mm0\n\t" - "paddusw %%mm4, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"=m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; p += line_size; - } while (--h); -} - -static void put_pixels_y2_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WONE(mm4); - JUMPALIGN(); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %2, %%mm1\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "paddusw %%mm4, %%mm0\n\t" - "paddusw %%mm4, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"=m"(*p) - :"m"(*pix), - "m"(*(pix+line_size)) - :"memory"); - pix += line_size; - p += line_size; - } while (--h); -} - -static void put_pixels_xy2_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; // 1s - MOVQ_ZERO(mm7); - MOVQ_WTWO(mm6); - JUMPALIGN(); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %2, %%mm1\n\t" - "movq 1%1, %%mm4\n\t" - "movq 1%2, %%mm5\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "movq %%mm4, %%mm1\n\t" - "movq %%mm5, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpcklbw %%mm7, %%mm5\n\t" - "punpckhbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm5, %%mm4\n\t" - "paddusw %%mm3, %%mm1\n\t" - "paddusw %%mm6, %%mm4\n\t" - "paddusw %%mm6, %%mm1\n\t" - "paddusw %%mm4, %%mm0\n\t" - "paddusw %%mm1, %%mm2\n\t" - "psrlw $2, %%mm0\n\t" - "psrlw $2, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"=m"(*p) - :"m"(*pix), - "m"(*(pix+line_size)) - :"memory"); - pix += line_size; - p += line_size; - } while(--h); -} - -static void put_no_rnd_pixels_x2_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq 1%1, %%mm1\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"=m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; - p += line_size; - } while (--h); -} - -static void put_no_rnd_pixels_y2_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - JUMPALIGN(); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %2, %%mm1\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"=m"(*p) - :"m"(*pix), - "m"(*(pix+line_size)) - :"memory"); - pix += line_size; - p += line_size; - } while(--h); -} - -static void put_no_rnd_pixels_xy2_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WONE(mm6); - JUMPALIGN(); - do { __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %2, %%mm1\n\t" - "movq 1%1, %%mm4\n\t" - "movq 1%2, %%mm5\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "movq %%mm4, %%mm1\n\t" - "movq %%mm5, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpcklbw %%mm7, %%mm5\n\t" - "punpckhbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm5, %%mm4\n\t" - "paddusw %%mm3, %%mm1\n\t" - "paddusw %%mm6, %%mm4\n\t" - "paddusw %%mm6, %%mm1\n\t" - "paddusw %%mm4, %%mm0\n\t" - "paddusw %%mm1, %%mm2\n\t" - "psrlw $2, %%mm0\n\t" - "psrlw $2, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"=m"(*p) - :"m"(*pix), - "m"(*(pix+line_size)) - :"memory"); - pix += line_size; - p += line_size; - } while(--h); -} - -static void avg_pixels_mmx(UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WONE(mm6); - JUMPALIGN(); - do { - __asm __volatile( - "movq %0, %%mm0\n\t" - "movq %1, %%mm1\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "paddusw %%mm6, %%mm0\n\t" - "paddusw %%mm6, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"+m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; - p += line_size; - } - while (--h); -} - -static void avg_pixels_x2_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WONE(mm6); - JUMPALIGN(); - do { - __asm __volatile( - "movq %1, %%mm1\n\t" - "movq %0, %%mm0\n\t" - "movq 1%1, %%mm4\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "movq %%mm4, %%mm5\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpckhbw %%mm7, %%mm5\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "paddusw %%mm4, %%mm1\n\t" - "paddusw %%mm5, %%mm3\n\t" - "paddusw %%mm6, %%mm1\n\t" - "paddusw %%mm6, %%mm3\n\t" - "psrlw $1, %%mm1\n\t" - "psrlw $1, %%mm3\n\t" - "paddusw %%mm6, %%mm0\n\t" - "paddusw %%mm6, %%mm2\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"+m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; - p += line_size; - } while (--h); -} - -static void avg_pixels_y2_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WONE(mm6); - JUMPALIGN(); - do { - __asm __volatile( - "movq %1, %%mm1\n\t" - "movq %0, %%mm0\n\t" - "movq %2, %%mm4\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "movq %%mm4, %%mm5\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpckhbw %%mm7, %%mm5\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "paddusw %%mm4, %%mm1\n\t" - "paddusw %%mm5, %%mm3\n\t" - "paddusw %%mm6, %%mm1\n\t" - "paddusw %%mm6, %%mm3\n\t" - "psrlw $1, %%mm1\n\t" - "psrlw $1, %%mm3\n\t" - "paddusw %%mm6, %%mm0\n\t" - "paddusw %%mm6, %%mm2\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"+m"(*p) - :"m"(*pix), "m"(*(pix+line_size)) - :"memory"); - pix += line_size; - p += line_size ; - } while(--h); -} - -static void avg_pixels_xy2_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - // this doesn't seem to be used offten - so - // the inside usage of mm_wone is not optimized - MOVQ_WTWO(mm6); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %2, %%mm1\n\t" - "movq 1%1, %%mm4\n\t" - "movq 1%2, %%mm5\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "movq %%mm4, %%mm1\n\t" - "movq %%mm5, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpcklbw %%mm7, %%mm5\n\t" - "punpckhbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm5, %%mm4\n\t" - "paddusw %%mm3, %%mm1\n\t" - "paddusw %%mm6, %%mm4\n\t" - "paddusw %%mm6, %%mm1\n\t" - "paddusw %%mm4, %%mm0\n\t" - "paddusw %%mm1, %%mm2\n\t" - "movq %3, %%mm5\n\t" - "psrlw $2, %%mm0\n\t" - "movq %0, %%mm1\n\t" - "psrlw $2, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "paddusw %%mm5, %%mm0\n\t" - "paddusw %%mm5, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"+m"(*p) - :"m"(*pix), - "m"(*(pix+line_size)), "m"(mm_wone) - :"memory"); - pix += line_size; - p += line_size ; - } while(--h); -} - -static void avg_no_rnd_pixels_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %0, %%mm1\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"+m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; - p += line_size ; - } while (--h); -} - -static void avg_no_rnd_pixels_x2_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq 1%1, %%mm1\n\t" - "movq %0, %%mm4\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "movq %%mm4, %%mm5\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpckhbw %%mm7, %%mm5\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "paddusw %%mm4, %%mm0\n\t" - "paddusw %%mm5, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"+m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; - p += line_size; - } while (--h); -} - -static void avg_no_rnd_pixels_y2_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %2, %%mm1\n\t" - "movq %0, %%mm4\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "movq %%mm4, %%mm5\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpckhbw %%mm7, %%mm5\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "paddusw %%mm4, %%mm0\n\t" - "paddusw %%mm5, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"+m"(*p) - :"m"(*pix), "m"(*(pix+line_size)) - :"memory"); - pix += line_size; - p += line_size ; - } while(--h); -} - -static void avg_no_rnd_pixels_xy2_mmx( UINT8 *block, const UINT8 *pixels, int line_size, int h) -{ - UINT8 *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WONE(mm6); - JUMPALIGN(); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %2, %%mm1\n\t" - "movq 1%1, %%mm4\n\t" - "movq 1%2, %%mm5\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "movq %%mm4, %%mm1\n\t" - "movq %%mm5, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpcklbw %%mm7, %%mm5\n\t" - "punpckhbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm5, %%mm4\n\t" - "paddusw %%mm3, %%mm1\n\t" - "paddusw %%mm6, %%mm4\n\t" - "paddusw %%mm6, %%mm1\n\t" - "paddusw %%mm4, %%mm0\n\t" - "paddusw %%mm1, %%mm2\n\t" - "movq %0, %%mm1\n\t" - "psrlw $2, %%mm0\n\t" - "movq %%mm1, %%mm3\n\t" - "psrlw $2, %%mm2\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "psrlw $1, %%mm0\n\t" - "psrlw $1, %%mm2\n\t" - "packuswb %%mm2, %%mm0\n\t" - "movq %%mm0, %0\n\t" - :"+m"(*p) - :"m"(*pix), - "m"(*(pix+line_size)) - :"memory"); - pix += line_size; - p += line_size; - } while(--h); -} - -static void sub_pixels_mmx( DCTELEM *block, const UINT8 *pixels, int line_size, int h) -{ - DCTELEM *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - do { - __asm __volatile( - "movq %0, %%mm0\n\t" - "movq %1, %%mm2\n\t" - "movq 8%0, %%mm1\n\t" - "movq %%mm2, %%mm3\n\t" - "punpcklbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "psubsw %%mm2, %%mm0\n\t" - "psubsw %%mm3, %%mm1\n\t" - "movq %%mm0, %0\n\t" - "movq %%mm1, 8%0\n\t" - :"+m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; - p += 8; - } while (--h); -} - -static void sub_pixels_x2_mmx( DCTELEM *block, const UINT8 *pixels, int line_size, int h) -{ - DCTELEM *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WONE(mm6); - JUMPALIGN(); - do { - __asm __volatile( - "movq %0, %%mm0\n\t" - "movq %1, %%mm2\n\t" - "movq 8%0, %%mm1\n\t" - "movq 1%1, %%mm4\n\t" - "movq %%mm2, %%mm3\n\t" - "movq %%mm4, %%mm5\n\t" - "punpcklbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpckhbw %%mm7, %%mm5\n\t" - "paddusw %%mm4, %%mm2\n\t" - "paddusw %%mm5, %%mm3\n\t" - "paddusw %%mm6, %%mm2\n\t" - "paddusw %%mm6, %%mm3\n\t" - "psrlw $1, %%mm2\n\t" - "psrlw $1, %%mm3\n\t" - "psubsw %%mm2, %%mm0\n\t" - "psubsw %%mm3, %%mm1\n\t" - "movq %%mm0, %0\n\t" - "movq %%mm1, 8%0\n\t" - :"+m"(*p) - :"m"(*pix) - :"memory"); - pix += line_size; - p += 8; - } while (--h); -} - -static void sub_pixels_y2_mmx( DCTELEM *block, const UINT8 *pixels, int line_size, int h) -{ - DCTELEM *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WONE(mm6); - do { - __asm __volatile( - "movq %0, %%mm0\n\t" - "movq %1, %%mm2\n\t" - "movq 8%0, %%mm1\n\t" - "movq %2, %%mm4\n\t" - "movq %%mm2, %%mm3\n\t" - "movq %%mm4, %%mm5\n\t" - "punpcklbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpckhbw %%mm7, %%mm5\n\t" - "paddusw %%mm4, %%mm2\n\t" - "paddusw %%mm5, %%mm3\n\t" - "paddusw %%mm6, %%mm2\n\t" - "paddusw %%mm6, %%mm3\n\t" - "psrlw $1, %%mm2\n\t" - "psrlw $1, %%mm3\n\t" - "psubsw %%mm2, %%mm0\n\t" - "psubsw %%mm3, %%mm1\n\t" - "movq %%mm0, %0\n\t" - "movq %%mm1, 8%0\n\t" - :"+m"(*p) - :"m"(*pix), "m"(*(pix+line_size)) - :"memory"); - pix += line_size; - p += 8; - } while (--h); -} - -static void sub_pixels_xy2_mmx( DCTELEM *block, const UINT8 *pixels, int line_size, int h) -{ - DCTELEM *p; - const UINT8 *pix; - p = block; - pix = pixels; - MOVQ_ZERO(mm7); - MOVQ_WTWO(mm6); - JUMPALIGN(); - do { - __asm __volatile( - "movq %1, %%mm0\n\t" - "movq %2, %%mm1\n\t" - "movq 1%1, %%mm4\n\t" - "movq 1%2, %%mm5\n\t" - "movq %%mm0, %%mm2\n\t" - "movq %%mm1, %%mm3\n\t" - "punpcklbw %%mm7, %%mm0\n\t" - "punpcklbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm2\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm1, %%mm0\n\t" - "paddusw %%mm3, %%mm2\n\t" - "movq %%mm4, %%mm1\n\t" - "movq %%mm5, %%mm3\n\t" - "punpcklbw %%mm7, %%mm4\n\t" - "punpcklbw %%mm7, %%mm5\n\t" - "punpckhbw %%mm7, %%mm1\n\t" - "punpckhbw %%mm7, %%mm3\n\t" - "paddusw %%mm5, %%mm4\n\t" - "paddusw %%mm3, %%mm1\n\t" - "paddusw %%mm6, %%mm4\n\t" - "paddusw %%mm6, %%mm1\n\t" - "paddusw %%mm4, %%mm0\n\t" - "paddusw %%mm1, %%mm2\n\t" - "movq %0, %%mm1\n\t" - "movq 8%0, %%mm3\n\t" - "psrlw $2, %%mm0\n\t" - "psrlw $2, %%mm2\n\t" - "psubsw %%mm0, %%mm1\n\t" - "psubsw %%mm2, %%mm3\n\t" - "movq %%mm1, %0\n\t" - "movq %%mm3, 8%0\n\t" - :"+m"(*p) - :"m"(*pix), - "m"(*(pix+line_size)) - :"memory"); - pix += line_size; - p += 8 ; - } while(--h); + "lea (%3, %3), %%eax \n\t" + ".balign 8 \n\t" + "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" + ); } static void clear_blocks_mmx(DCTELEM *blocks) { - asm volatile( + __asm __volatile( "pxor %%mm7, %%mm7 \n\t" "movl $-128*6, %%eax \n\t" "1: \n\t" @@ -1043,7 +386,9 @@ static void clear_blocks_mmx(DCTELEM *blocks) ); } +#if 0 static void just_return() { return; } +#endif void dsputil_init_mmx(void) { @@ -1065,10 +410,11 @@ void dsputil_init_mmx(void) if (mm_flags & MM_MMX) { get_pixels = get_pixels_mmx; + diff_pixels = diff_pixels_mmx; put_pixels_clamped = put_pixels_clamped_mmx; add_pixels_clamped = add_pixels_clamped_mmx; clear_blocks= clear_blocks_mmx; - + pix_abs16x16 = pix_abs16x16_mmx; pix_abs16x16_x2 = pix_abs16x16_x2_mmx; pix_abs16x16_y2 = pix_abs16x16_y2_mmx; @@ -1088,7 +434,7 @@ void dsputil_init_mmx(void) put_no_rnd_pixels_tab[1] = put_no_rnd_pixels_x2_mmx; put_no_rnd_pixels_tab[2] = put_no_rnd_pixels_y2_mmx; put_no_rnd_pixels_tab[3] = put_no_rnd_pixels_xy2_mmx; - + avg_pixels_tab[0] = avg_pixels_mmx; avg_pixels_tab[1] = avg_pixels_x2_mmx; avg_pixels_tab[2] = avg_pixels_y2_mmx; @@ -1098,44 +444,37 @@ void dsputil_init_mmx(void) avg_no_rnd_pixels_tab[1] = avg_no_rnd_pixels_x2_mmx; avg_no_rnd_pixels_tab[2] = avg_no_rnd_pixels_y2_mmx; avg_no_rnd_pixels_tab[3] = avg_no_rnd_pixels_xy2_mmx; - - sub_pixels_tab[0] = sub_pixels_mmx; - sub_pixels_tab[1] = sub_pixels_x2_mmx; - sub_pixels_tab[2] = sub_pixels_y2_mmx; - sub_pixels_tab[3] = sub_pixels_xy2_mmx; if (mm_flags & MM_MMXEXT) { pix_abs16x16 = pix_abs16x16_mmx2; pix_abs16x16_x2 = pix_abs16x16_x2_mmx2; pix_abs16x16_y2 = pix_abs16x16_y2_mmx2; pix_abs16x16_xy2= pix_abs16x16_xy2_mmx2; - + pix_abs8x8 = pix_abs8x8_mmx2; pix_abs8x8_x2 = pix_abs8x8_x2_mmx2; pix_abs8x8_y2 = pix_abs8x8_y2_mmx2; pix_abs8x8_xy2= pix_abs8x8_xy2_mmx2; - - put_pixels_tab[1] = put_pixels_x2_sse; - put_pixels_tab[2] = put_pixels_y2_sse; - - avg_pixels_tab[0] = avg_pixels_sse; - avg_pixels_tab[1] = avg_pixels_x2_sse; - avg_pixels_tab[2] = avg_pixels_y2_sse; - avg_pixels_tab[3] = avg_pixels_xy2_sse; - - sub_pixels_tab[1] = sub_pixels_x2_sse; - sub_pixels_tab[2] = sub_pixels_y2_sse; + + put_pixels_tab[1] = put_pixels_x2_mmx2; + put_pixels_tab[2] = put_pixels_y2_mmx2; + put_no_rnd_pixels_tab[1] = put_no_rnd_pixels_x2_mmx2; + put_no_rnd_pixels_tab[2] = put_no_rnd_pixels_y2_mmx2; + + avg_pixels_tab[0] = avg_pixels_mmx2; + avg_pixels_tab[1] = avg_pixels_x2_mmx2; + avg_pixels_tab[2] = avg_pixels_y2_mmx2; + avg_pixels_tab[3] = avg_pixels_xy2_mmx2; } else if (mm_flags & MM_3DNOW) { put_pixels_tab[1] = put_pixels_x2_3dnow; put_pixels_tab[2] = put_pixels_y2_3dnow; - + put_no_rnd_pixels_tab[1] = put_no_rnd_pixels_x2_3dnow; + put_no_rnd_pixels_tab[2] = put_no_rnd_pixels_y2_3dnow; + avg_pixels_tab[0] = avg_pixels_3dnow; avg_pixels_tab[1] = avg_pixels_x2_3dnow; avg_pixels_tab[2] = avg_pixels_y2_3dnow; avg_pixels_tab[3] = avg_pixels_xy2_3dnow; - - sub_pixels_tab[1] = sub_pixels_x2_3dnow; - sub_pixels_tab[2] = sub_pixels_y2_3dnow; } /* idct */ @@ -1181,12 +520,25 @@ void dsputil_init_mmx(void) avg_no_rnd_pixels_tab[2] = just_return; avg_no_rnd_pixels_tab[3] = just_return; - sub_pixels_tab[0] = just_return; - sub_pixels_tab[1] = just_return; - sub_pixels_tab[2] = just_return; - sub_pixels_tab[3] = just_return; - //av_fdct = just_return; //ff_idct = just_return; #endif } + +/* 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. */ +void dsputil_set_bit_exact_mmx(void) +{ + if (mm_flags & MM_MMX) { + if (mm_flags & MM_MMXEXT) { + put_no_rnd_pixels_tab[1] = put_no_rnd_pixels_x2_mmx; + put_no_rnd_pixels_tab[2] = put_no_rnd_pixels_y2_mmx; + avg_pixels_tab[3] = avg_pixels_xy2_mmx; + } else if (mm_flags & MM_3DNOW) { + put_no_rnd_pixels_tab[1] = put_no_rnd_pixels_x2_mmx; + put_no_rnd_pixels_tab[2] = put_no_rnd_pixels_y2_mmx; + avg_pixels_tab[3] = avg_pixels_xy2_mmx; + } + } +} |