Mercurial > libavcodec.hg
diff i386/motion_est_mmx.c @ 294:944632089814 libavcodec
4MV motion estimation (not finished yet)
SAD functions rewritten (8x8 support & MMX2 optimizations)
HQ inter/intra decission
msmpeg4 encoding bugfix (MV where too long)
| author | michaelni |
|---|---|
| date | Wed, 27 Mar 2002 21:25:22 +0000 |
| parents | 3049d6d452a3 |
| children | 54d86f074a4b |
line wrap: on
line diff
--- a/i386/motion_est_mmx.c Mon Mar 25 15:36:16 2002 +0000 +++ b/i386/motion_est_mmx.c Wed Mar 27 21:25:22 2002 +0000 @@ -16,229 +16,347 @@ * along with this program; if not, write to the Free Software * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. * + * mostly by Michael Niedermayer <michaelni@gmx.at> */ #include "../dsputil.h" -#include "mmx.h" -static const unsigned long long int mm_wone __attribute__ ((aligned(8))) = 0x0001000100010001; -static const unsigned long long int mm_wtwo __attribute__ ((aligned(8))) = 0x0002000200020002; +static const __attribute__ ((aligned(8))) UINT64 round_tab[3]={ +0x0000000000000000, +0x0001000100010001, +0x0002000200020002, +}; -/* mm7 is accumulator, mm6 is zero */ -static inline void sad_add(const UINT8 *p1, const UINT8 *p2) +static inline void sad8_mmx(UINT8 *blk1, UINT8 *blk2, int stride, int h) { - movq_m2r(*p1, mm0); - movq_m2r(*p2, mm1); - movq_r2r(mm0, mm2); - psubusb_r2r(mm1, mm0); - psubusb_r2r(mm2, mm1); - por_r2r(mm1, mm0); /* mm0 is absolute value */ - - movq_r2r(mm0, mm1); - punpcklbw_r2r(mm6, mm0); - punpckhbw_r2r(mm6, mm1); - paddusw_r2r(mm0, mm7); - paddusw_r2r(mm1, mm7); + int len= -(stride<<h); + asm volatile( + ".balign 16 \n\t" + "1: \n\t" + "movq (%1, %%eax), %%mm0 \n\t" + "movq (%2, %%eax), %%mm2 \n\t" + "movq (%2, %%eax), %%mm4 \n\t" + "addl %3, %%eax \n\t" + "psubusb %%mm0, %%mm2 \n\t" + "psubusb %%mm4, %%mm0 \n\t" + "movq (%1, %%eax), %%mm1 \n\t" + "movq (%2, %%eax), %%mm3 \n\t" + "movq (%2, %%eax), %%mm5 \n\t" + "psubusb %%mm1, %%mm3 \n\t" + "psubusb %%mm5, %%mm1 \n\t" + "por %%mm2, %%mm0 \n\t" + "por %%mm1, %%mm3 \n\t" + "movq %%mm0, %%mm1 \n\t" + "movq %%mm3, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "punpcklbw %%mm7, %%mm3 \n\t" + "punpckhbw %%mm7, %%mm2 \n\t" + "paddw %%mm1, %%mm0 \n\t" + "paddw %%mm3, %%mm2 \n\t" + "paddw %%mm2, %%mm0 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "addl %3, %%eax \n\t" + " js 1b \n\t" + : "+a" (len) + : "r" (blk1 - len), "r" (blk2 - len), "r" (stride) + ); } -/* convert mm7 to value */ -static inline int sad_end(void) +static inline void sad8_mmx2(UINT8 *blk1, UINT8 *blk2, int stride, int h) { - int res; - - movq_r2r(mm7, mm0); - psrlq_i2r(32, mm7); - paddusw_r2r(mm0, mm7); - - movq_r2r(mm7, mm0); - psrlq_i2r(16, mm7); - paddusw_r2r(mm0, mm7); - __asm __volatile ("movd %%mm7, %0" : "=a" (res)); - return res & 0xffff; + int len= -(stride<<h); + asm volatile( + ".balign 16 \n\t" + "1: \n\t" + "movq (%1, %%eax), %%mm0 \n\t" + "movq (%2, %%eax), %%mm2 \n\t" + "psadbw %%mm2, %%mm0 \n\t" + "addl %3, %%eax \n\t" + "movq (%1, %%eax), %%mm1 \n\t" + "movq (%2, %%eax), %%mm3 \n\t" + "psadbw %%mm1, %%mm3 \n\t" + "paddw %%mm3, %%mm0 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "addl %3, %%eax \n\t" + " js 1b \n\t" + : "+a" (len) + : "r" (blk1 - len), "r" (blk2 - len), "r" (stride) + ); } -int pix_abs16x16_mmx(UINT8 *blk1, UINT8 *blk2, int lx, int h) +static inline void sad8_2_mmx2(UINT8 *blk1a, UINT8 *blk1b, UINT8 *blk2, int stride, int h) { - const UINT8 *p1, *p2; - - h >>= 1; - p1 = blk1; - p2 = blk2; - pxor_r2r(mm7, mm7); /* mm7 is accumulator */ - pxor_r2r(mm6, mm6); /* mm7 is zero constant */ - do { - sad_add(p1, p2); - sad_add(p1 + 8, p2 + 8); - p1 += lx; - p2 += lx; - sad_add(p1, p2); - sad_add(p1 + 8, p2 + 8); - p1 += lx; - p2 += lx; - } while (--h); - return sad_end(); + int len= -(stride<<h); + asm volatile( + ".balign 16 \n\t" + "1: \n\t" + "movq (%1, %%eax), %%mm0 \n\t" + "movq (%2, %%eax), %%mm2 \n\t" + "pavgb %%mm2, %%mm0 \n\t" + "movq (%3, %%eax), %%mm2 \n\t" + "psadbw %%mm2, %%mm0 \n\t" + "addl %4, %%eax \n\t" + "movq (%1, %%eax), %%mm1 \n\t" + "movq (%2, %%eax), %%mm3 \n\t" + "pavgb %%mm1, %%mm3 \n\t" + "movq (%3, %%eax), %%mm1 \n\t" + "psadbw %%mm1, %%mm3 \n\t" + "paddw %%mm3, %%mm0 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "addl %4, %%eax \n\t" + " js 1b \n\t" + : "+a" (len) + : "r" (blk1a - len), "r" (blk1b -len), "r" (blk2 - len), "r" (stride) + ); } -/* please test it ! */ -static inline void sad_add_sse(const UINT8 *p1, const UINT8 *p2) -{ - movq_m2r(*(p1 + 0), mm0); - movq_m2r(*(p1 + 8), mm1); - psadbw_m2r(*(p2 + 0), mm0); - psadbw_m2r(*(p2 + 8), mm1); - paddusw_r2r(mm0, mm7); - paddusw_r2r(mm1, mm7); +static inline void sad8_4_mmx2(UINT8 *blk1, UINT8 *blk2, int stride, int h) +{ //FIXME reuse src + int len= -(stride<<h); + asm volatile( + ".balign 16 \n\t" + "1: \n\t" + "movq (%1, %%eax), %%mm0 \n\t" + "movq (%2, %%eax), %%mm2 \n\t" + "movq 1(%1, %%eax), %%mm1 \n\t" + "movq 1(%2, %%eax), %%mm3 \n\t" + "pavgb %%mm2, %%mm0 \n\t" + "pavgb %%mm1, %%mm3 \n\t" + "pavgb %%mm3, %%mm0 \n\t" + "movq (%3, %%eax), %%mm2 \n\t" + "psadbw %%mm2, %%mm0 \n\t" + "addl %4, %%eax \n\t" + "movq (%1, %%eax), %%mm1 \n\t" + "movq (%2, %%eax), %%mm3 \n\t" + "movq 1(%1, %%eax), %%mm2 \n\t" + "movq 1(%2, %%eax), %%mm4 \n\t" + "pavgb %%mm3, %%mm1 \n\t" + "pavgb %%mm4, %%mm2 \n\t" + "pavgb %%mm1, %%mm2 \n\t" + "movq (%3, %%eax), %%mm1 \n\t" + "psadbw %%mm1, %%mm2 \n\t" + "paddw %%mm2, %%mm0 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "addl %4, %%eax \n\t" + " js 1b \n\t" + : "+a" (len) + : "r" (blk1 - len), "r" (blk1 - len + stride), "r" (blk2 - len), "r" (stride) + ); } -int pix_abs16x16_sse(UINT8 *blk1, UINT8 *blk2, int lx, int h) +static inline void sad8_2_mmx(UINT8 *blk1a, UINT8 *blk1b, UINT8 *blk2, int stride, int h) { - const UINT8 *p1, *p2; - - h >>= 1; - p1 = blk1; - p2 = blk2; - pxor_r2r(mm7, mm7); /* mm7 is accumulator */ - do { - sad_add_sse(p1, p2); - p1 += lx; - p2 += lx; - sad_add_sse(p1, p2); - p1 += lx; - p2 += lx; - } while (--h); - return sad_end(); + int len= -(stride<<h); + asm volatile( + ".balign 16 \n\t" + "1: \n\t" + "movq (%1, %%eax), %%mm0 \n\t" + "movq (%2, %%eax), %%mm1 \n\t" + "movq (%1, %%eax), %%mm2 \n\t" + "movq (%2, %%eax), %%mm3 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm1 \n\t" + "punpckhbw %%mm7, %%mm2 \n\t" + "punpckhbw %%mm7, %%mm3 \n\t" + "paddw %%mm0, %%mm1 \n\t" + "paddw %%mm2, %%mm3 \n\t" + "movq (%3, %%eax), %%mm4 \n\t" + "movq (%3, %%eax), %%mm2 \n\t" + "paddw %%mm5, %%mm1 \n\t" + "paddw %%mm5, %%mm3 \n\t" + "psrlw $1, %%mm1 \n\t" + "psrlw $1, %%mm3 \n\t" + "packuswb %%mm3, %%mm1 \n\t" + "psubusb %%mm1, %%mm4 \n\t" + "psubusb %%mm2, %%mm1 \n\t" + "por %%mm4, %%mm1 \n\t" + "movq %%mm1, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "paddw %%mm1, %%mm0 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "addl %4, %%eax \n\t" + " js 1b \n\t" + : "+a" (len) + : "r" (blk1a - len), "r" (blk1b -len), "r" (blk2 - len), "r" (stride) + ); } -#define DUMP(reg) { mmx_t tmp; movq_r2m(reg, tmp); printf(#reg "=%016Lx\n", tmp.uq); } - -/* mm7 is accumulator, mm6 is zero */ -static inline void sad_add_x2(const UINT8 *p1, const UINT8 *p2, const UINT8 *p3) +static inline void sad8_4_mmx(UINT8 *blk1, UINT8 *blk2, int stride, int h) { - movq_m2r(*(p2 + 0), mm0); - movq_m2r(*(p3 + 0), mm1); - movq_r2r(mm0, mm2); - movq_r2r(mm1, mm3); - punpcklbw_r2r(mm6, mm0); /* extract 4 bytes low */ - punpcklbw_r2r(mm6, mm1); - punpckhbw_r2r(mm6, mm2); /* high */ - punpckhbw_r2r(mm6, mm3); - paddusw_r2r(mm1, mm0); - paddusw_r2r(mm3, mm2); - movq_m2r(*(p1 + 0), mm1); /* mm1 : other value */ - paddusw_r2r(mm5, mm0); /* + 1 */ - paddusw_r2r(mm5, mm2); /* + 1 */ - psrlw_i2r(1, mm0); - psrlw_i2r(1, mm2); - packuswb_r2r(mm2, mm0); /* average is in mm0 */ - - movq_r2r(mm1, mm2); - psubusb_r2r(mm0, mm1); - psubusb_r2r(mm2, mm0); - por_r2r(mm1, mm0); /* mm0 is absolute value */ - - movq_r2r(mm0, mm1); - punpcklbw_r2r(mm6, mm0); - punpckhbw_r2r(mm6, mm1); - paddusw_r2r(mm0, mm7); - paddusw_r2r(mm1, mm7); + int len= -(stride<<h); + asm volatile( + ".balign 16 \n\t" + "1: \n\t" + "movq (%1, %%eax), %%mm0 \n\t" + "movq (%2, %%eax), %%mm1 \n\t" + "movq %%mm0, %%mm4 \n\t" + "movq %%mm1, %%mm2 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm1 \n\t" + "punpckhbw %%mm7, %%mm4 \n\t" + "punpckhbw %%mm7, %%mm2 \n\t" + "paddw %%mm1, %%mm0 \n\t" + "paddw %%mm2, %%mm4 \n\t" + "movq 1(%1, %%eax), %%mm2 \n\t" + "movq 1(%2, %%eax), %%mm3 \n\t" + "movq %%mm2, %%mm1 \n\t" + "punpcklbw %%mm7, %%mm2 \n\t" + "punpckhbw %%mm7, %%mm1 \n\t" + "paddw %%mm0, %%mm2 \n\t" + "paddw %%mm4, %%mm1 \n\t" + "movq %%mm3, %%mm4 \n\t" + "punpcklbw %%mm7, %%mm3 \n\t" + "punpckhbw %%mm7, %%mm4 \n\t" + "paddw %%mm3, %%mm2 \n\t" + "paddw %%mm4, %%mm1 \n\t" + "movq (%3, %%eax), %%mm3 \n\t" + "movq (%3, %%eax), %%mm4 \n\t" + "paddw %%mm5, %%mm2 \n\t" + "paddw %%mm5, %%mm1 \n\t" + "psrlw $2, %%mm2 \n\t" + "psrlw $2, %%mm1 \n\t" + "packuswb %%mm1, %%mm2 \n\t" + "psubusb %%mm2, %%mm3 \n\t" + "psubusb %%mm4, %%mm2 \n\t" + "por %%mm3, %%mm2 \n\t" + "movq %%mm2, %%mm0 \n\t" + "punpcklbw %%mm7, %%mm0 \n\t" + "punpckhbw %%mm7, %%mm2 \n\t" + "paddw %%mm2, %%mm0 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "addl %4, %%eax \n\t" + " js 1b \n\t" + : "+a" (len) + : "r" (blk1 - len), "r" (blk1 -len + stride), "r" (blk2 - len), "r" (stride) + ); } -int pix_abs16x16_x2_mmx(UINT8 *blk1, UINT8 *blk2, int lx, int h) +static inline int sum_mmx() { - const UINT8 *p1, *p2; + int ret; + asm volatile( + "movq %%mm6, %%mm0 \n\t" + "psrlq $32, %%mm6 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "movq %%mm6, %%mm0 \n\t" + "psrlq $16, %%mm6 \n\t" + "paddw %%mm0, %%mm6 \n\t" + "movd %%mm6, %0 \n\t" + : "=r" (ret) + ); + return ret&0xFFFF; +} - p1 = blk1; - p2 = blk2; - pxor_r2r(mm7, mm7); /* mm7 is accumulator */ - pxor_r2r(mm6, mm6); /* mm7 is zero constant */ - movq_m2r(mm_wone, mm5); /* one constant */ - do { - sad_add_x2(p1, p2, p2 + 1); - sad_add_x2(p1 + 8, p2 + 8, p2 + 9); - p1 += lx; - p2 += lx; - } while (--h); - return sad_end(); +static inline int sum_mmx2() +{ + int ret; + asm volatile( + "movd %%mm6, %0 \n\t" + : "=r" (ret) + ); + return ret; } -int pix_abs16x16_y2_mmx(UINT8 *blk1, UINT8 *blk2, int lx, int h) -{ - const UINT8 *p1, *p2; - - p1 = blk1; - p2 = blk2; - pxor_r2r(mm7, mm7); /* mm7 is accumulator */ - pxor_r2r(mm6, mm6); /* mm7 is zero constant */ - movq_m2r(mm_wone, mm5); /* one constant */ - do { - sad_add_x2(p1, p2, p2 + lx); - sad_add_x2(p1 + 8, p2 + 8, p2 + 8 + lx); - p1 += lx; - p2 += lx; - } while (--h); - return sad_end(); -} - -/* mm7 is accumulator, mm6 is zero */ -static inline void sad_add_xy2(const UINT8 *p1, const UINT8 *p2, const UINT8 *p3) -{ - movq_m2r(*(p2 + 0), mm0); - movq_m2r(*(p3 + 0), mm1); - movq_r2r(mm0, mm2); - movq_r2r(mm1, mm3); - punpcklbw_r2r(mm6, mm0); /* extract 4 bytes low */ - punpcklbw_r2r(mm6, mm1); - punpckhbw_r2r(mm6, mm2); /* high */ - punpckhbw_r2r(mm6, mm3); - paddusw_r2r(mm1, mm0); - paddusw_r2r(mm3, mm2); +#define PIX_SAD(suf)\ +int pix_abs8x8_ ## suf(UINT8 *blk2, UINT8 *blk1, int stride)\ +{\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t":);\ +\ + sad8_ ## suf(blk1, blk2, stride, 3);\ +\ + return sum_ ## suf();\ +}\ +\ +int pix_abs8x8_x2_ ## suf(UINT8 *blk2, UINT8 *blk1, int stride)\ +{\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ + :: "m"(round_tab[1]) \ + );\ +\ + sad8_2_ ## suf(blk1, blk2+1, blk2, stride, 3);\ +\ + return sum_ ## suf();\ +}\ +\ +int pix_abs8x8_y2_ ## suf(UINT8 *blk2, UINT8 *blk1, int stride)\ +{\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ + :: "m"(round_tab[1]) \ + );\ +\ + sad8_2_ ## suf(blk1, blk1+stride, blk2, stride, 3);\ +\ + return sum_ ## suf();\ +}\ +\ +int pix_abs8x8_xy2_ ## suf(UINT8 *blk2, UINT8 *blk1, int stride)\ +{\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ + :: "m"(round_tab[2]) \ + );\ +\ + sad8_4_ ## suf(blk1, blk2, stride, 3);\ +\ + return sum_ ## suf();\ +}\ +\ +int pix_abs16x16_ ## suf(UINT8 *blk2, UINT8 *blk1, int stride)\ +{\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t":);\ +\ + sad8_ ## suf(blk1 , blk2 , stride, 4);\ + sad8_ ## suf(blk1+8, blk2+8, stride, 4);\ +\ + return sum_ ## suf();\ +}\ +int pix_abs16x16_x2_ ## suf(UINT8 *blk2, UINT8 *blk1, int stride)\ +{\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ + :: "m"(round_tab[1]) \ + );\ +\ + sad8_2_ ## suf(blk1 , blk1+1, blk2 , stride, 4);\ + sad8_2_ ## suf(blk1+8, blk1+9, blk2+8, stride, 4);\ +\ + return sum_ ## suf();\ +}\ +int pix_abs16x16_y2_ ## suf(UINT8 *blk2, UINT8 *blk1, int stride)\ +{\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ + :: "m"(round_tab[1]) \ + );\ +\ + sad8_2_ ## suf(blk1 , blk1+stride, blk2 , stride, 4);\ + sad8_2_ ## suf(blk1+8, blk1+stride+8,blk2+8, stride, 4);\ +\ + return sum_ ## suf();\ +}\ +int pix_abs16x16_xy2_ ## suf(UINT8 *blk2, UINT8 *blk1, int stride)\ +{\ + asm volatile("pxor %%mm7, %%mm7 \n\t"\ + "pxor %%mm6, %%mm6 \n\t"\ + "movq %0, %%mm5 \n\t"\ + :: "m"(round_tab[2]) \ + );\ +\ + sad8_4_ ## suf(blk1 , blk2 , stride, 4);\ + sad8_4_ ## suf(blk1+8, blk2+8, stride, 4);\ +\ + return sum_ ## suf();\ +}\ - movq_m2r(*(p2 + 1), mm1); - movq_m2r(*(p3 + 1), mm3); - movq_r2r(mm1, mm4); - punpcklbw_r2r(mm6, mm1); /* low */ - punpckhbw_r2r(mm6, mm4); /* high */ - paddusw_r2r(mm1, mm0); - paddusw_r2r(mm4, mm2); - movq_r2r(mm3, mm4); - punpcklbw_r2r(mm6, mm3); /* low */ - punpckhbw_r2r(mm6, mm4); /* high */ - paddusw_r2r(mm3, mm0); - paddusw_r2r(mm4, mm2); - - movq_m2r(*(p1 + 0), mm1); /* mm1 : other value */ - paddusw_r2r(mm5, mm0); /* + 2 */ - paddusw_r2r(mm5, mm2); /* + 2 */ - psrlw_i2r(2, mm0); - psrlw_i2r(2, mm2); - packuswb_r2r(mm2, mm0); /* average is in mm0 */ - - movq_r2r(mm1, mm2); - psubusb_r2r(mm0, mm1); - psubusb_r2r(mm2, mm0); - por_r2r(mm1, mm0); /* mm0 is absolute value */ - - movq_r2r(mm0, mm1); - punpcklbw_r2r(mm6, mm0); - punpckhbw_r2r(mm6, mm1); - paddusw_r2r(mm0, mm7); - paddusw_r2r(mm1, mm7); -} - -int pix_abs16x16_xy2_mmx(UINT8 *blk1, UINT8 *blk2, int lx, int h) -{ - const UINT8 *p1, *p2, *p3; - - p1 = blk1; - p2 = blk2; - p3 = blk2 + lx; - pxor_r2r(mm7, mm7); /* mm7 is accumulator */ - pxor_r2r(mm6, mm6); /* mm7 is zero constant */ - movq_m2r(mm_wtwo, mm5); /* one constant */ - do { - sad_add_xy2(p1, p2, p2 + lx); - sad_add_xy2(p1 + 8, p2 + 8, p2 + 8 + lx); - p1 += lx; - p2 += lx; - } while (--h); - return sad_end(); -} +PIX_SAD(mmx) +PIX_SAD(mmx2)