• Main Page
  • Related Pages
  • Modules
  • Data Structures
  • Files
  • File List
  • Globals

libavcodec/x86/motion_est_mmx.c

Go to the documentation of this file.
00001 /*
00002  * MMX optimized motion estimation
00003  * Copyright (c) 2001 Fabrice Bellard
00004  * Copyright (c) 2002-2004 Michael Niedermayer
00005  *
00006  * mostly by Michael Niedermayer <michaelni@gmx.at>
00007  *
00008  * This file is part of FFmpeg.
00009  *
00010  * FFmpeg is free software; you can redistribute it and/or
00011  * modify it under the terms of the GNU Lesser General Public
00012  * License as published by the Free Software Foundation; either
00013  * version 2.1 of the License, or (at your option) any later version.
00014  *
00015  * FFmpeg is distributed in the hope that it will be useful,
00016  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00017  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00018  * Lesser General Public License for more details.
00019  *
00020  * You should have received a copy of the GNU Lesser General Public
00021  * License along with FFmpeg; if not, write to the Free Software
00022  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00023  */
00024 
00025 #include "libavutil/x86_cpu.h"
00026 #include "libavcodec/dsputil.h"
00027 #include "dsputil_mmx.h"
00028 
00029 DECLARE_ASM_CONST(8, uint64_t, round_tab)[3]={
00030 0x0000000000000000ULL,
00031 0x0001000100010001ULL,
00032 0x0002000200020002ULL,
00033 };
00034 
00035 DECLARE_ASM_CONST(8, uint64_t, bone)= 0x0101010101010101LL;
00036 
00037 static inline void sad8_1_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
00038 {
00039     x86_reg len= -(stride*h);
00040     __asm__ volatile(
00041         ASMALIGN(4)
00042         "1:                             \n\t"
00043         "movq (%1, %%"REG_a"), %%mm0    \n\t"
00044         "movq (%2, %%"REG_a"), %%mm2    \n\t"
00045         "movq (%2, %%"REG_a"), %%mm4    \n\t"
00046         "add %3, %%"REG_a"              \n\t"
00047         "psubusb %%mm0, %%mm2           \n\t"
00048         "psubusb %%mm4, %%mm0           \n\t"
00049         "movq (%1, %%"REG_a"), %%mm1    \n\t"
00050         "movq (%2, %%"REG_a"), %%mm3    \n\t"
00051         "movq (%2, %%"REG_a"), %%mm5    \n\t"
00052         "psubusb %%mm1, %%mm3           \n\t"
00053         "psubusb %%mm5, %%mm1           \n\t"
00054         "por %%mm2, %%mm0               \n\t"
00055         "por %%mm1, %%mm3               \n\t"
00056         "movq %%mm0, %%mm1              \n\t"
00057         "movq %%mm3, %%mm2              \n\t"
00058         "punpcklbw %%mm7, %%mm0         \n\t"
00059         "punpckhbw %%mm7, %%mm1         \n\t"
00060         "punpcklbw %%mm7, %%mm3         \n\t"
00061         "punpckhbw %%mm7, %%mm2         \n\t"
00062         "paddw %%mm1, %%mm0             \n\t"
00063         "paddw %%mm3, %%mm2             \n\t"
00064         "paddw %%mm2, %%mm0             \n\t"
00065         "paddw %%mm0, %%mm6             \n\t"
00066         "add %3, %%"REG_a"              \n\t"
00067         " js 1b                         \n\t"
00068         : "+a" (len)
00069         : "r" (blk1 - len), "r" (blk2 - len), "r" ((x86_reg)stride)
00070     );
00071 }
00072 
00073 static inline void sad8_1_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
00074 {
00075     __asm__ volatile(
00076         ASMALIGN(4)
00077         "1:                             \n\t"
00078         "movq (%1), %%mm0               \n\t"
00079         "movq (%1, %3), %%mm1           \n\t"
00080         "psadbw (%2), %%mm0             \n\t"
00081         "psadbw (%2, %3), %%mm1         \n\t"
00082         "paddw %%mm0, %%mm6             \n\t"
00083         "paddw %%mm1, %%mm6             \n\t"
00084         "lea (%1,%3,2), %1              \n\t"
00085         "lea (%2,%3,2), %2              \n\t"
00086         "sub $2, %0                     \n\t"
00087         " jg 1b                         \n\t"
00088         : "+r" (h), "+r" (blk1), "+r" (blk2)
00089         : "r" ((x86_reg)stride)
00090     );
00091 }
00092 
00093 static int sad16_sse2(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)
00094 {
00095     int ret;
00096     __asm__ volatile(
00097         "pxor %%xmm6, %%xmm6            \n\t"
00098         ASMALIGN(4)
00099         "1:                             \n\t"
00100         "movdqu (%1), %%xmm0            \n\t"
00101         "movdqu (%1, %3), %%xmm1        \n\t"
00102         "psadbw (%2), %%xmm0            \n\t"
00103         "psadbw (%2, %3), %%xmm1        \n\t"
00104         "paddw %%xmm0, %%xmm6           \n\t"
00105         "paddw %%xmm1, %%xmm6           \n\t"
00106         "lea (%1,%3,2), %1              \n\t"
00107         "lea (%2,%3,2), %2              \n\t"
00108         "sub $2, %0                     \n\t"
00109         " jg 1b                         \n\t"
00110         : "+r" (h), "+r" (blk1), "+r" (blk2)
00111         : "r" ((x86_reg)stride)
00112     );
00113     __asm__ volatile(
00114         "movhlps %%xmm6, %%xmm0         \n\t"
00115         "paddw   %%xmm0, %%xmm6         \n\t"
00116         "movd    %%xmm6, %0             \n\t"
00117         : "=r"(ret)
00118     );
00119     return ret;
00120 }
00121 
00122 static inline void sad8_x2a_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
00123 {
00124     __asm__ volatile(
00125         ASMALIGN(4)
00126         "1:                             \n\t"
00127         "movq (%1), %%mm0               \n\t"
00128         "movq (%1, %3), %%mm1           \n\t"
00129         "pavgb 1(%1), %%mm0             \n\t"
00130         "pavgb 1(%1, %3), %%mm1         \n\t"
00131         "psadbw (%2), %%mm0             \n\t"
00132         "psadbw (%2, %3), %%mm1         \n\t"
00133         "paddw %%mm0, %%mm6             \n\t"
00134         "paddw %%mm1, %%mm6             \n\t"
00135         "lea (%1,%3,2), %1              \n\t"
00136         "lea (%2,%3,2), %2              \n\t"
00137         "sub $2, %0                     \n\t"
00138         " jg 1b                         \n\t"
00139         : "+r" (h), "+r" (blk1), "+r" (blk2)
00140         : "r" ((x86_reg)stride)
00141     );
00142 }
00143 
00144 static inline void sad8_y2a_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
00145 {
00146     __asm__ volatile(
00147         "movq (%1), %%mm0               \n\t"
00148         "add %3, %1                     \n\t"
00149         ASMALIGN(4)
00150         "1:                             \n\t"
00151         "movq (%1), %%mm1               \n\t"
00152         "movq (%1, %3), %%mm2           \n\t"
00153         "pavgb %%mm1, %%mm0             \n\t"
00154         "pavgb %%mm2, %%mm1             \n\t"
00155         "psadbw (%2), %%mm0             \n\t"
00156         "psadbw (%2, %3), %%mm1         \n\t"
00157         "paddw %%mm0, %%mm6             \n\t"
00158         "paddw %%mm1, %%mm6             \n\t"
00159         "movq %%mm2, %%mm0              \n\t"
00160         "lea (%1,%3,2), %1              \n\t"
00161         "lea (%2,%3,2), %2              \n\t"
00162         "sub $2, %0                     \n\t"
00163         " jg 1b                         \n\t"
00164         : "+r" (h), "+r" (blk1), "+r" (blk2)
00165         : "r" ((x86_reg)stride)
00166     );
00167 }
00168 
00169 static inline void sad8_4_mmx2(uint8_t *blk1, uint8_t *blk2, int stride, int h)
00170 {
00171     __asm__ volatile(
00172         "movq "MANGLE(bone)", %%mm5     \n\t"
00173         "movq (%1), %%mm0               \n\t"
00174         "pavgb 1(%1), %%mm0             \n\t"
00175         "add %3, %1                     \n\t"
00176         ASMALIGN(4)
00177         "1:                             \n\t"
00178         "movq (%1), %%mm1               \n\t"
00179         "movq (%1,%3), %%mm2            \n\t"
00180         "pavgb 1(%1), %%mm1             \n\t"
00181         "pavgb 1(%1,%3), %%mm2          \n\t"
00182         "psubusb %%mm5, %%mm1           \n\t"
00183         "pavgb %%mm1, %%mm0             \n\t"
00184         "pavgb %%mm2, %%mm1             \n\t"
00185         "psadbw (%2), %%mm0             \n\t"
00186         "psadbw (%2,%3), %%mm1          \n\t"
00187         "paddw %%mm0, %%mm6             \n\t"
00188         "paddw %%mm1, %%mm6             \n\t"
00189         "movq %%mm2, %%mm0              \n\t"
00190         "lea (%1,%3,2), %1              \n\t"
00191         "lea (%2,%3,2), %2              \n\t"
00192         "sub $2, %0                     \n\t"
00193         " jg 1b                         \n\t"
00194         : "+r" (h), "+r" (blk1), "+r" (blk2)
00195         : "r" ((x86_reg)stride)
00196     );
00197 }
00198 
00199 static inline void sad8_2_mmx(uint8_t *blk1a, uint8_t *blk1b, uint8_t *blk2, int stride, int h)
00200 {
00201     x86_reg len= -(stride*h);
00202     __asm__ volatile(
00203         ASMALIGN(4)
00204         "1:                             \n\t"
00205         "movq (%1, %%"REG_a"), %%mm0    \n\t"
00206         "movq (%2, %%"REG_a"), %%mm1    \n\t"
00207         "movq (%1, %%"REG_a"), %%mm2    \n\t"
00208         "movq (%2, %%"REG_a"), %%mm3    \n\t"
00209         "punpcklbw %%mm7, %%mm0         \n\t"
00210         "punpcklbw %%mm7, %%mm1         \n\t"
00211         "punpckhbw %%mm7, %%mm2         \n\t"
00212         "punpckhbw %%mm7, %%mm3         \n\t"
00213         "paddw %%mm0, %%mm1             \n\t"
00214         "paddw %%mm2, %%mm3             \n\t"
00215         "movq (%3, %%"REG_a"), %%mm4    \n\t"
00216         "movq (%3, %%"REG_a"), %%mm2    \n\t"
00217         "paddw %%mm5, %%mm1             \n\t"
00218         "paddw %%mm5, %%mm3             \n\t"
00219         "psrlw $1, %%mm1                \n\t"
00220         "psrlw $1, %%mm3                \n\t"
00221         "packuswb %%mm3, %%mm1          \n\t"
00222         "psubusb %%mm1, %%mm4           \n\t"
00223         "psubusb %%mm2, %%mm1           \n\t"
00224         "por %%mm4, %%mm1               \n\t"
00225         "movq %%mm1, %%mm0              \n\t"
00226         "punpcklbw %%mm7, %%mm0         \n\t"
00227         "punpckhbw %%mm7, %%mm1         \n\t"
00228         "paddw %%mm1, %%mm0             \n\t"
00229         "paddw %%mm0, %%mm6             \n\t"
00230         "add %4, %%"REG_a"              \n\t"
00231         " js 1b                         \n\t"
00232         : "+a" (len)
00233         : "r" (blk1a - len), "r" (blk1b -len), "r" (blk2 - len), "r" ((x86_reg)stride)
00234     );
00235 }
00236 
00237 static inline void sad8_4_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
00238 {
00239     x86_reg len= -(stride*h);
00240     __asm__ volatile(
00241         "movq (%1, %%"REG_a"), %%mm0    \n\t"
00242         "movq 1(%1, %%"REG_a"), %%mm2   \n\t"
00243         "movq %%mm0, %%mm1              \n\t"
00244         "movq %%mm2, %%mm3              \n\t"
00245         "punpcklbw %%mm7, %%mm0         \n\t"
00246         "punpckhbw %%mm7, %%mm1         \n\t"
00247         "punpcklbw %%mm7, %%mm2         \n\t"
00248         "punpckhbw %%mm7, %%mm3         \n\t"
00249         "paddw %%mm2, %%mm0             \n\t"
00250         "paddw %%mm3, %%mm1             \n\t"
00251         ASMALIGN(4)
00252         "1:                             \n\t"
00253         "movq (%2, %%"REG_a"), %%mm2    \n\t"
00254         "movq 1(%2, %%"REG_a"), %%mm4   \n\t"
00255         "movq %%mm2, %%mm3              \n\t"
00256         "movq %%mm4, %%mm5              \n\t"
00257         "punpcklbw %%mm7, %%mm2         \n\t"
00258         "punpckhbw %%mm7, %%mm3         \n\t"
00259         "punpcklbw %%mm7, %%mm4         \n\t"
00260         "punpckhbw %%mm7, %%mm5         \n\t"
00261         "paddw %%mm4, %%mm2             \n\t"
00262         "paddw %%mm5, %%mm3             \n\t"
00263         "movq 16+"MANGLE(round_tab)", %%mm5 \n\t"
00264         "paddw %%mm2, %%mm0             \n\t"
00265         "paddw %%mm3, %%mm1             \n\t"
00266         "paddw %%mm5, %%mm0             \n\t"
00267         "paddw %%mm5, %%mm1             \n\t"
00268         "movq (%3, %%"REG_a"), %%mm4    \n\t"
00269         "movq (%3, %%"REG_a"), %%mm5    \n\t"
00270         "psrlw $2, %%mm0                \n\t"
00271         "psrlw $2, %%mm1                \n\t"
00272         "packuswb %%mm1, %%mm0          \n\t"
00273         "psubusb %%mm0, %%mm4           \n\t"
00274         "psubusb %%mm5, %%mm0           \n\t"
00275         "por %%mm4, %%mm0               \n\t"
00276         "movq %%mm0, %%mm4              \n\t"
00277         "punpcklbw %%mm7, %%mm0         \n\t"
00278         "punpckhbw %%mm7, %%mm4         \n\t"
00279         "paddw %%mm0, %%mm6             \n\t"
00280         "paddw %%mm4, %%mm6             \n\t"
00281         "movq  %%mm2, %%mm0             \n\t"
00282         "movq  %%mm3, %%mm1             \n\t"
00283         "add %4, %%"REG_a"              \n\t"
00284         " js 1b                         \n\t"
00285         : "+a" (len)
00286         : "r" (blk1 - len), "r" (blk1 -len + stride), "r" (blk2 - len), "r" ((x86_reg)stride)
00287     );
00288 }
00289 
00290 static inline int sum_mmx(void)
00291 {
00292     int ret;
00293     __asm__ volatile(
00294         "movq %%mm6, %%mm0              \n\t"
00295         "psrlq $32, %%mm6               \n\t"
00296         "paddw %%mm0, %%mm6             \n\t"
00297         "movq %%mm6, %%mm0              \n\t"
00298         "psrlq $16, %%mm6               \n\t"
00299         "paddw %%mm0, %%mm6             \n\t"
00300         "movd %%mm6, %0                 \n\t"
00301         : "=r" (ret)
00302     );
00303     return ret&0xFFFF;
00304 }
00305 
00306 static inline int sum_mmx2(void)
00307 {
00308     int ret;
00309     __asm__ volatile(
00310         "movd %%mm6, %0                 \n\t"
00311         : "=r" (ret)
00312     );
00313     return ret;
00314 }
00315 
00316 static inline void sad8_x2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
00317 {
00318     sad8_2_mmx(blk1, blk1+1, blk2, stride, h);
00319 }
00320 static inline void sad8_y2a_mmx(uint8_t *blk1, uint8_t *blk2, int stride, int h)
00321 {
00322     sad8_2_mmx(blk1, blk1+stride, blk2, stride, h);
00323 }
00324 
00325 
00326 #define PIX_SAD(suf)\
00327 static int sad8_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
00328 {\
00329     assert(h==8);\
00330     __asm__ volatile("pxor %%mm7, %%mm7     \n\t"\
00331                  "pxor %%mm6, %%mm6     \n\t":);\
00332 \
00333     sad8_1_ ## suf(blk1, blk2, stride, 8);\
00334 \
00335     return sum_ ## suf();\
00336 }\
00337 static int sad8_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
00338 {\
00339     assert(h==8);\
00340     __asm__ volatile("pxor %%mm7, %%mm7     \n\t"\
00341                  "pxor %%mm6, %%mm6     \n\t"\
00342                  "movq %0, %%mm5        \n\t"\
00343                  :: "m"(round_tab[1]) \
00344                  );\
00345 \
00346     sad8_x2a_ ## suf(blk1, blk2, stride, 8);\
00347 \
00348     return sum_ ## suf();\
00349 }\
00350 \
00351 static int sad8_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
00352 {\
00353     assert(h==8);\
00354     __asm__ volatile("pxor %%mm7, %%mm7     \n\t"\
00355                  "pxor %%mm6, %%mm6     \n\t"\
00356                  "movq %0, %%mm5        \n\t"\
00357                  :: "m"(round_tab[1]) \
00358                  );\
00359 \
00360     sad8_y2a_ ## suf(blk1, blk2, stride, 8);\
00361 \
00362     return sum_ ## suf();\
00363 }\
00364 \
00365 static int sad8_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
00366 {\
00367     assert(h==8);\
00368     __asm__ volatile("pxor %%mm7, %%mm7     \n\t"\
00369                  "pxor %%mm6, %%mm6     \n\t"\
00370                  ::);\
00371 \
00372     sad8_4_ ## suf(blk1, blk2, stride, 8);\
00373 \
00374     return sum_ ## suf();\
00375 }\
00376 \
00377 static int sad16_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
00378 {\
00379     __asm__ volatile("pxor %%mm7, %%mm7     \n\t"\
00380                  "pxor %%mm6, %%mm6     \n\t":);\
00381 \
00382     sad8_1_ ## suf(blk1  , blk2  , stride, h);\
00383     sad8_1_ ## suf(blk1+8, blk2+8, stride, h);\
00384 \
00385     return sum_ ## suf();\
00386 }\
00387 static int sad16_x2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
00388 {\
00389     __asm__ volatile("pxor %%mm7, %%mm7     \n\t"\
00390                  "pxor %%mm6, %%mm6     \n\t"\
00391                  "movq %0, %%mm5        \n\t"\
00392                  :: "m"(round_tab[1]) \
00393                  );\
00394 \
00395     sad8_x2a_ ## suf(blk1  , blk2  , stride, h);\
00396     sad8_x2a_ ## suf(blk1+8, blk2+8, stride, h);\
00397 \
00398     return sum_ ## suf();\
00399 }\
00400 static int sad16_y2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
00401 {\
00402     __asm__ volatile("pxor %%mm7, %%mm7     \n\t"\
00403                  "pxor %%mm6, %%mm6     \n\t"\
00404                  "movq %0, %%mm5        \n\t"\
00405                  :: "m"(round_tab[1]) \
00406                  );\
00407 \
00408     sad8_y2a_ ## suf(blk1  , blk2  , stride, h);\
00409     sad8_y2a_ ## suf(blk1+8, blk2+8, stride, h);\
00410 \
00411     return sum_ ## suf();\
00412 }\
00413 static int sad16_xy2_ ## suf(void *v, uint8_t *blk2, uint8_t *blk1, int stride, int h)\
00414 {\
00415     __asm__ volatile("pxor %%mm7, %%mm7     \n\t"\
00416                  "pxor %%mm6, %%mm6     \n\t"\
00417                  ::);\
00418 \
00419     sad8_4_ ## suf(blk1  , blk2  , stride, h);\
00420     sad8_4_ ## suf(blk1+8, blk2+8, stride, h);\
00421 \
00422     return sum_ ## suf();\
00423 }\
00424 
00425 PIX_SAD(mmx)
00426 PIX_SAD(mmx2)
00427 
00428 void dsputil_init_pix_mmx(DSPContext* c, AVCodecContext *avctx)
00429 {
00430     if (mm_flags & FF_MM_MMX) {
00431         c->pix_abs[0][0] = sad16_mmx;
00432         c->pix_abs[0][1] = sad16_x2_mmx;
00433         c->pix_abs[0][2] = sad16_y2_mmx;
00434         c->pix_abs[0][3] = sad16_xy2_mmx;
00435         c->pix_abs[1][0] = sad8_mmx;
00436         c->pix_abs[1][1] = sad8_x2_mmx;
00437         c->pix_abs[1][2] = sad8_y2_mmx;
00438         c->pix_abs[1][3] = sad8_xy2_mmx;
00439 
00440         c->sad[0]= sad16_mmx;
00441         c->sad[1]= sad8_mmx;
00442     }
00443     if (mm_flags & FF_MM_MMX2) {
00444         c->pix_abs[0][0] = sad16_mmx2;
00445         c->pix_abs[1][0] = sad8_mmx2;
00446 
00447         c->sad[0]= sad16_mmx2;
00448         c->sad[1]= sad8_mmx2;
00449 
00450         if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
00451             c->pix_abs[0][1] = sad16_x2_mmx2;
00452             c->pix_abs[0][2] = sad16_y2_mmx2;
00453             c->pix_abs[0][3] = sad16_xy2_mmx2;
00454             c->pix_abs[1][1] = sad8_x2_mmx2;
00455             c->pix_abs[1][2] = sad8_y2_mmx2;
00456             c->pix_abs[1][3] = sad8_xy2_mmx2;
00457         }
00458     }
00459     if ((mm_flags & FF_MM_SSE2) && !(mm_flags & FF_MM_3DNOW) && avctx->codec_id != CODEC_ID_SNOW) {
00460         c->sad[0]= sad16_sse2;
00461     }
00462 }

Generated on Fri Sep 16 2011 17:17:46 for FFmpeg by  doxygen 1.7.1