FFmpeg: libavcodec/x86/vc1dsp

00001 /*
00002  * VC-1 and WMV3 - DSP functions MMX-optimized
00003  * Copyright (c) 2007 Christophe GISQUET <christophe.gisquet@free.fr>
00004  *
00005  * Permission is hereby granted, free of charge, to any person
00006  * obtaining a copy of this software and associated documentation
00007  * files (the "Software"), to deal in the Software without
00008  * restriction, including without limitation the rights to use,
00009  * copy, modify, merge, publish, distribute, sublicense, and/or sell
00010  * copies of the Software, and to permit persons to whom the
00011  * Software is furnished to do so, subject to the following
00012  * conditions:
00013  *
00014  * The above copyright notice and this permission notice shall be
00015  * included in all copies or substantial portions of the Software.
00016  *
00017  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
00018  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES
00019  * OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
00020  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
00021  * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
00022  * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
00023  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
00024  * OTHER DEALINGS IN THE SOFTWARE.
00025  */
00026 
00027 #include "libavutil/x86_cpu.h"
00028 #include "libavcodec/dsputil.h"
00029 #include "dsputil_mmx.h"
00030 
00032 #define NORMALIZE_MMX(SHIFT)                                    \
00033      "paddw     %%mm7, %%mm3           \n\t" /* +bias-r */      \
00034      "paddw     %%mm7, %%mm4           \n\t" /* +bias-r */      \
00035      "psraw     "SHIFT", %%mm3         \n\t"                    \
00036      "psraw     "SHIFT", %%mm4         \n\t"
00037 
00038 #define TRANSFER_DO_PACK                        \
00039      "packuswb  %%mm4, %%mm3           \n\t"    \
00040      "movq      %%mm3, (%2)            \n\t"
00041 
00042 #define TRANSFER_DONT_PACK                      \
00043      "movq      %%mm3, 0(%2)           \n\t"    \
00044      "movq      %%mm4, 8(%2)           \n\t"
00045 
00047 #define DO_UNPACK(reg)  "punpcklbw %%mm0, " reg "\n\t"
00048 #define DONT_UNPACK(reg)
00049 
00051 #define LOAD_ROUNDER_MMX(ROUND)                 \
00052      "movd      "ROUND", %%mm7         \n\t"    \
00053      "punpcklwd %%mm7, %%mm7           \n\t"    \
00054      "punpckldq %%mm7, %%mm7           \n\t"
00055 
00056 #define SHIFT2_LINE(OFF, R0,R1,R2,R3)           \
00057     "paddw     %%mm"#R2", %%mm"#R1"    \n\t"    \
00058     "movd      (%0,%3), %%mm"#R0"      \n\t"    \
00059     "pmullw    %%mm6, %%mm"#R1"        \n\t"    \
00060     "punpcklbw %%mm0, %%mm"#R0"        \n\t"    \
00061     "movd      (%0,%2), %%mm"#R3"      \n\t"    \
00062     "psubw     %%mm"#R0", %%mm"#R1"    \n\t"    \
00063     "punpcklbw %%mm0, %%mm"#R3"        \n\t"    \
00064     "paddw     %%mm7, %%mm"#R1"        \n\t"    \
00065     "psubw     %%mm"#R3", %%mm"#R1"    \n\t"    \
00066     "psraw     %4, %%mm"#R1"           \n\t"    \
00067     "movq      %%mm"#R1", "#OFF"(%1)   \n\t"    \
00068     "add       %2, %0                  \n\t"
00069 
00070 DECLARE_ALIGNED_16(const uint64_t, ff_pw_9) = 0x0009000900090009ULL;
00071 
00073 static void vc1_put_ver_16b_shift2_mmx(int16_t *dst,
00074                                        const uint8_t *src, x86_reg stride,
00075                                        int rnd, int64_t shift)
00076 {
00077     __asm__ volatile(
00078         "mov       $3, %%"REG_c"           \n\t"
00079         LOAD_ROUNDER_MMX("%5")
00080         "movq      "MANGLE(ff_pw_9)", %%mm6 \n\t"
00081         "1:                                \n\t"
00082         "movd      (%0), %%mm2             \n\t"
00083         "add       %2, %0                  \n\t"
00084         "movd      (%0), %%mm3             \n\t"
00085         "punpcklbw %%mm0, %%mm2            \n\t"
00086         "punpcklbw %%mm0, %%mm3            \n\t"
00087         SHIFT2_LINE(  0, 1, 2, 3, 4)
00088         SHIFT2_LINE( 24, 2, 3, 4, 1)
00089         SHIFT2_LINE( 48, 3, 4, 1, 2)
00090         SHIFT2_LINE( 72, 4, 1, 2, 3)
00091         SHIFT2_LINE( 96, 1, 2, 3, 4)
00092         SHIFT2_LINE(120, 2, 3, 4, 1)
00093         SHIFT2_LINE(144, 3, 4, 1, 2)
00094         SHIFT2_LINE(168, 4, 1, 2, 3)
00095         "sub       %6, %0                  \n\t"
00096         "add       $8, %1                  \n\t"
00097         "dec       %%"REG_c"               \n\t"
00098         "jnz 1b                            \n\t"
00099         : "+r"(src), "+r"(dst)
00100         : "r"(stride), "r"(-2*stride),
00101           "m"(shift), "m"(rnd), "r"(9*stride-4)
00102         : "%"REG_c, "memory"
00103     );
00104 }
00105 
00110 static void vc1_put_hor_16b_shift2_mmx(uint8_t *dst, x86_reg stride,
00111                                        const int16_t *src, int rnd)
00112 {
00113     int h = 8;
00114 
00115     src -= 1;
00116     rnd -= (-1+9+9-1)*1024; /* Add -1024 bias */
00117     __asm__ volatile(
00118         LOAD_ROUNDER_MMX("%4")
00119         "movq      "MANGLE(ff_pw_128)", %%mm6\n\t"
00120         "movq      "MANGLE(ff_pw_9)", %%mm5 \n\t"
00121         "1:                                \n\t"
00122         "movq      2*0+0(%1), %%mm1        \n\t"
00123         "movq      2*0+8(%1), %%mm2        \n\t"
00124         "movq      2*1+0(%1), %%mm3        \n\t"
00125         "movq      2*1+8(%1), %%mm4        \n\t"
00126         "paddw     2*3+0(%1), %%mm1        \n\t"
00127         "paddw     2*3+8(%1), %%mm2        \n\t"
00128         "paddw     2*2+0(%1), %%mm3        \n\t"
00129         "paddw     2*2+8(%1), %%mm4        \n\t"
00130         "pmullw    %%mm5, %%mm3            \n\t"
00131         "pmullw    %%mm5, %%mm4            \n\t"
00132         "psubw     %%mm1, %%mm3            \n\t"
00133         "psubw     %%mm2, %%mm4            \n\t"
00134         NORMALIZE_MMX("$7")
00135         /* Remove bias */
00136         "paddw     %%mm6, %%mm3            \n\t"
00137         "paddw     %%mm6, %%mm4            \n\t"
00138         TRANSFER_DO_PACK
00139         "add       $24, %1                 \n\t"
00140         "add       %3, %2                  \n\t"
00141         "decl      %0                      \n\t"
00142         "jnz 1b                            \n\t"
00143         : "+r"(h), "+r" (src),  "+r" (dst)
00144         : "r"(stride), "m"(rnd)
00145         : "memory"
00146     );
00147 }
00148 
00149 
00154 static void vc1_put_shift2_mmx(uint8_t *dst, const uint8_t *src,
00155                                x86_reg stride, int rnd, x86_reg offset)
00156 {
00157     rnd = 8-rnd;
00158     __asm__ volatile(
00159         "mov       $8, %%"REG_c"           \n\t"
00160         LOAD_ROUNDER_MMX("%5")
00161         "movq      "MANGLE(ff_pw_9)", %%mm6\n\t"
00162         "1:                                \n\t"
00163         "movd      0(%0   ), %%mm3         \n\t"
00164         "movd      4(%0   ), %%mm4         \n\t"
00165         "movd      0(%0,%2), %%mm1         \n\t"
00166         "movd      4(%0,%2), %%mm2         \n\t"
00167         "add       %2, %0                  \n\t"
00168         "punpcklbw %%mm0, %%mm3            \n\t"
00169         "punpcklbw %%mm0, %%mm4            \n\t"
00170         "punpcklbw %%mm0, %%mm1            \n\t"
00171         "punpcklbw %%mm0, %%mm2            \n\t"
00172         "paddw     %%mm1, %%mm3            \n\t"
00173         "paddw     %%mm2, %%mm4            \n\t"
00174         "movd      0(%0,%3), %%mm1         \n\t"
00175         "movd      4(%0,%3), %%mm2         \n\t"
00176         "pmullw    %%mm6, %%mm3            \n\t" /* 0,9,9,0*/
00177         "pmullw    %%mm6, %%mm4            \n\t" /* 0,9,9,0*/
00178         "punpcklbw %%mm0, %%mm1            \n\t"
00179         "punpcklbw %%mm0, %%mm2            \n\t"
00180         "psubw     %%mm1, %%mm3            \n\t" /*-1,9,9,0*/
00181         "psubw     %%mm2, %%mm4            \n\t" /*-1,9,9,0*/
00182         "movd      0(%0,%2), %%mm1         \n\t"
00183         "movd      4(%0,%2), %%mm2         \n\t"
00184         "punpcklbw %%mm0, %%mm1            \n\t"
00185         "punpcklbw %%mm0, %%mm2            \n\t"
00186         "psubw     %%mm1, %%mm3            \n\t" /*-1,9,9,-1*/
00187         "psubw     %%mm2, %%mm4            \n\t" /*-1,9,9,-1*/
00188         NORMALIZE_MMX("$4")
00189         "packuswb  %%mm4, %%mm3            \n\t"
00190         "movq      %%mm3, (%1)             \n\t"
00191         "add       %6, %0                  \n\t"
00192         "add       %4, %1                  \n\t"
00193         "dec       %%"REG_c"               \n\t"
00194         "jnz 1b                            \n\t"
00195         : "+r"(src),  "+r"(dst)
00196         : "r"(offset), "r"(-2*offset), "g"(stride), "m"(rnd),
00197           "g"(stride-offset)
00198         : "%"REG_c, "memory"
00199     );
00200 }
00201 
00206 DECLARE_ASM_CONST(16, uint64_t, ff_pw_53) = 0x0035003500350035ULL;
00207 DECLARE_ASM_CONST(16, uint64_t, ff_pw_18) = 0x0012001200120012ULL;
00208 
00219 #define MSPEL_FILTER13_CORE(UNPACK, MOVQ, A1, A2, A3, A4)       \
00220      MOVQ "*0+"A1", %%mm1       \n\t"                           \
00221      MOVQ "*4+"A1", %%mm2       \n\t"                           \
00222      UNPACK("%%mm1")                                            \
00223      UNPACK("%%mm2")                                            \
00224      "pmullw    "MANGLE(ff_pw_3)", %%mm1\n\t"                   \
00225      "pmullw    "MANGLE(ff_pw_3)", %%mm2\n\t"                   \
00226      MOVQ "*0+"A2", %%mm3       \n\t"                           \
00227      MOVQ "*4+"A2", %%mm4       \n\t"                           \
00228      UNPACK("%%mm3")                                            \
00229      UNPACK("%%mm4")                                            \
00230      "pmullw    %%mm6, %%mm3    \n\t" /* *18 */                 \
00231      "pmullw    %%mm6, %%mm4    \n\t" /* *18 */                 \
00232      "psubw     %%mm1, %%mm3    \n\t" /* 18,-3 */               \
00233      "psubw     %%mm2, %%mm4    \n\t" /* 18,-3 */               \
00234      MOVQ "*0+"A4", %%mm1       \n\t"                           \
00235      MOVQ "*4+"A4", %%mm2       \n\t"                           \
00236      UNPACK("%%mm1")                                            \
00237      UNPACK("%%mm2")                                            \
00238      "psllw     $2, %%mm1       \n\t" /* 4* */                  \
00239      "psllw     $2, %%mm2       \n\t" /* 4* */                  \
00240      "psubw     %%mm1, %%mm3    \n\t" /* -4,18,-3 */            \
00241      "psubw     %%mm2, %%mm4    \n\t" /* -4,18,-3 */            \
00242      MOVQ "*0+"A3", %%mm1       \n\t"                           \
00243      MOVQ "*4+"A3", %%mm2       \n\t"                           \
00244      UNPACK("%%mm1")                                            \
00245      UNPACK("%%mm2")                                            \
00246      "pmullw    %%mm5, %%mm1    \n\t" /* *53 */                 \
00247      "pmullw    %%mm5, %%mm2    \n\t" /* *53 */                 \
00248      "paddw     %%mm1, %%mm3    \n\t" /* 4,53,18,-3 */          \
00249      "paddw     %%mm2, %%mm4    \n\t" /* 4,53,18,-3 */
00250 
00259 #define MSPEL_FILTER13_VER_16B(NAME, A1, A2, A3, A4)                    \
00260 static void                                                             \
00261 vc1_put_ver_16b_ ## NAME ## _mmx(int16_t *dst, const uint8_t *src,      \
00262                                  x86_reg src_stride,                   \
00263                                  int rnd, int64_t shift)                \
00264 {                                                                       \
00265     int h = 8;                                                          \
00266     src -= src_stride;                                                  \
00267     __asm__ volatile(                                                       \
00268         LOAD_ROUNDER_MMX("%5")                                          \
00269         "movq      "MANGLE(ff_pw_53)", %%mm5\n\t"                       \
00270         "movq      "MANGLE(ff_pw_18)", %%mm6\n\t"                       \
00271         ASMALIGN(3)                                                     \
00272         "1:                        \n\t"                                \
00273         MSPEL_FILTER13_CORE(DO_UNPACK, "movd  1", A1, A2, A3, A4)       \
00274         NORMALIZE_MMX("%6")                                             \
00275         TRANSFER_DONT_PACK                                              \
00276         /* Last 3 (in fact 4) bytes on the line */                      \
00277         "movd      8+"A1", %%mm1   \n\t"                                \
00278         DO_UNPACK("%%mm1")                                              \
00279         "movq      %%mm1, %%mm3    \n\t"                                \
00280         "paddw     %%mm1, %%mm1    \n\t"                                \
00281         "paddw     %%mm3, %%mm1    \n\t" /* 3* */                       \
00282         "movd      8+"A2", %%mm3   \n\t"                                \
00283         DO_UNPACK("%%mm3")                                              \
00284         "pmullw    %%mm6, %%mm3    \n\t" /* *18 */                      \
00285         "psubw     %%mm1, %%mm3    \n\t" /*18,-3 */                     \
00286         "movd      8+"A3", %%mm1   \n\t"                                \
00287         DO_UNPACK("%%mm1")                                              \
00288         "pmullw    %%mm5, %%mm1    \n\t" /* *53 */                      \
00289         "paddw     %%mm1, %%mm3    \n\t" /*53,18,-3 */                  \
00290         "movd      8+"A4", %%mm1   \n\t"                                \
00291         DO_UNPACK("%%mm1")                                              \
00292         "psllw     $2, %%mm1       \n\t" /* 4* */                       \
00293         "psubw     %%mm1, %%mm3    \n\t"                                \
00294         "paddw     %%mm7, %%mm3    \n\t"                                \
00295         "psraw     %6, %%mm3       \n\t"                                \
00296         "movq      %%mm3, 16(%2)   \n\t"                                \
00297         "add       %3, %1          \n\t"                                \
00298         "add       $24, %2         \n\t"                                \
00299         "decl      %0              \n\t"                                \
00300         "jnz 1b                    \n\t"                                \
00301         : "+r"(h), "+r" (src),  "+r" (dst)                              \
00302         : "r"(src_stride), "r"(3*src_stride),                           \
00303           "m"(rnd), "m"(shift)                                          \
00304         : "memory"                                                      \
00305     );                                                                  \
00306 }
00307 
00315 #define MSPEL_FILTER13_HOR_16B(NAME, A1, A2, A3, A4)                    \
00316 static void                                                             \
00317 vc1_put_hor_16b_ ## NAME ## _mmx(uint8_t *dst, x86_reg stride,         \
00318                                  const int16_t *src, int rnd)           \
00319 {                                                                       \
00320     int h = 8;                                                          \
00321     src -= 1;                                                           \
00322     rnd -= (-4+58+13-3)*256; /* Add -256 bias */                        \
00323     __asm__ volatile(                                                       \
00324         LOAD_ROUNDER_MMX("%4")                                          \
00325         "movq      "MANGLE(ff_pw_18)", %%mm6   \n\t"                    \
00326         "movq      "MANGLE(ff_pw_53)", %%mm5   \n\t"                    \
00327         ASMALIGN(3)                                                     \
00328         "1:                        \n\t"                                \
00329         MSPEL_FILTER13_CORE(DONT_UNPACK, "movq 2", A1, A2, A3, A4)      \
00330         NORMALIZE_MMX("$7")                                             \
00331         /* Remove bias */                                               \
00332         "paddw     "MANGLE(ff_pw_128)", %%mm3  \n\t"                    \
00333         "paddw     "MANGLE(ff_pw_128)", %%mm4  \n\t"                    \
00334         TRANSFER_DO_PACK                                                \
00335         "add       $24, %1         \n\t"                                \
00336         "add       %3, %2          \n\t"                                \
00337         "decl      %0              \n\t"                                \
00338         "jnz 1b                    \n\t"                                \
00339         : "+r"(h), "+r" (src),  "+r" (dst)                              \
00340         : "r"(stride), "m"(rnd)                                         \
00341         : "memory"                                                      \
00342     );                                                                  \
00343 }
00344 
00353 #define MSPEL_FILTER13_8B(NAME, A1, A2, A3, A4)                         \
00354 static void                                                             \
00355 vc1_put_## NAME ## _mmx(uint8_t *dst, const uint8_t *src,               \
00356                         x86_reg stride, int rnd, x86_reg offset)      \
00357 {                                                                       \
00358     int h = 8;                                                          \
00359     src -= offset;                                                      \
00360     rnd = 32-rnd;                                                       \
00361     __asm__ volatile (                                                      \
00362         LOAD_ROUNDER_MMX("%6")                                          \
00363         "movq      "MANGLE(ff_pw_53)", %%mm5       \n\t"                \
00364         "movq      "MANGLE(ff_pw_18)", %%mm6       \n\t"                \
00365         ASMALIGN(3)                                                     \
00366         "1:                        \n\t"                                \
00367         MSPEL_FILTER13_CORE(DO_UNPACK, "movd   1", A1, A2, A3, A4)      \
00368         NORMALIZE_MMX("$6")                                             \
00369         TRANSFER_DO_PACK                                                \
00370         "add       %5, %1          \n\t"                                \
00371         "add       %5, %2          \n\t"                                \
00372         "decl      %0              \n\t"                                \
00373         "jnz 1b                    \n\t"                                \
00374         : "+r"(h), "+r" (src),  "+r" (dst)                              \
00375         : "r"(offset), "r"(3*offset), "g"(stride), "m"(rnd)             \
00376         : "memory"                                                      \
00377     );                                                                  \
00378 }
00379 
00381 MSPEL_FILTER13_8B     (shift1, "0(%1,%4  )", "0(%1,%3,2)", "0(%1,%3  )", "0(%1     )")
00382 MSPEL_FILTER13_VER_16B(shift1, "0(%1,%4  )", "0(%1,%3,2)", "0(%1,%3  )", "0(%1     )")
00383 MSPEL_FILTER13_HOR_16B(shift1, "2*3(%1)", "2*2(%1)", "2*1(%1)", "2*0(%1)")
00384 
00386 MSPEL_FILTER13_8B     (shift3, "0(%1     )", "0(%1,%3  )", "0(%1,%3,2)", "0(%1,%4  )")
00387 MSPEL_FILTER13_VER_16B(shift3, "0(%1     )", "0(%1,%3  )", "0(%1,%3,2)", "0(%1,%4  )")
00388 MSPEL_FILTER13_HOR_16B(shift3, "2*0(%1)", "2*1(%1)", "2*2(%1)", "2*3(%1)")
00389 
00390 typedef void (*vc1_mspel_mc_filter_ver_16bits)(int16_t *dst, const uint8_t *src, x86_reg src_stride, int rnd, int64_t shift);
00391 typedef void (*vc1_mspel_mc_filter_hor_16bits)(uint8_t *dst, x86_reg dst_stride, const int16_t *src, int rnd);
00392 typedef void (*vc1_mspel_mc_filter_8bits)(uint8_t *dst, const uint8_t *src, x86_reg stride, int rnd, x86_reg offset);
00393 
00405 static void vc1_mspel_mc(uint8_t *dst, const uint8_t *src, int stride,
00406                          int hmode, int vmode, int rnd)
00407 {
00408     static const vc1_mspel_mc_filter_ver_16bits vc1_put_shift_ver_16bits[] =
00409          { NULL, vc1_put_ver_16b_shift1_mmx, vc1_put_ver_16b_shift2_mmx, vc1_put_ver_16b_shift3_mmx };
00410     static const vc1_mspel_mc_filter_hor_16bits vc1_put_shift_hor_16bits[] =
00411          { NULL, vc1_put_hor_16b_shift1_mmx, vc1_put_hor_16b_shift2_mmx, vc1_put_hor_16b_shift3_mmx };
00412     static const vc1_mspel_mc_filter_8bits vc1_put_shift_8bits[] =
00413          { NULL, vc1_put_shift1_mmx, vc1_put_shift2_mmx, vc1_put_shift3_mmx };
00414 
00415     __asm__ volatile(
00416         "pxor %%mm0, %%mm0         \n\t"
00417         ::: "memory"
00418     );
00419 
00420     if (vmode) { /* Vertical filter to apply */
00421         if (hmode) { /* Horizontal filter to apply, output to tmp */
00422             static const int shift_value[] = { 0, 5, 1, 5 };
00423             int              shift = (shift_value[hmode]+shift_value[vmode])>>1;
00424             int              r;
00425             DECLARE_ALIGNED_16(int16_t, tmp[12*8]);
00426 
00427             r = (1<<(shift-1)) + rnd-1;
00428             vc1_put_shift_ver_16bits[vmode](tmp, src-1, stride, r, shift);
00429 
00430             vc1_put_shift_hor_16bits[hmode](dst, stride, tmp+1, 64-rnd);
00431             return;
00432         }
00433         else { /* No horizontal filter, output 8 lines to dst */
00434             vc1_put_shift_8bits[vmode](dst, src, stride, 1-rnd, stride);
00435             return;
00436         }
00437     }
00438 
00439     /* Horizontal mode with no vertical mode */
00440     vc1_put_shift_8bits[hmode](dst, src, stride, rnd, 1);
00441 }
00442 
00443 void ff_put_vc1_mspel_mc00_mmx(uint8_t *dst, const uint8_t *src, int stride, int rnd);
00444 
00446 #define DECLARE_FUNCTION(a, b)                                          \
00447 static void put_vc1_mspel_mc ## a ## b ## _mmx(uint8_t *dst, const uint8_t *src, int stride, int rnd) { \
00448      vc1_mspel_mc(dst, src, stride, a, b, rnd);                         \
00449 }
00450 
00451 DECLARE_FUNCTION(0, 1)
00452 DECLARE_FUNCTION(0, 2)
00453 DECLARE_FUNCTION(0, 3)
00454 
00455 DECLARE_FUNCTION(1, 0)
00456 DECLARE_FUNCTION(1, 1)
00457 DECLARE_FUNCTION(1, 2)
00458 DECLARE_FUNCTION(1, 3)
00459 
00460 DECLARE_FUNCTION(2, 0)
00461 DECLARE_FUNCTION(2, 1)
00462 DECLARE_FUNCTION(2, 2)
00463 DECLARE_FUNCTION(2, 3)
00464 
00465 DECLARE_FUNCTION(3, 0)
00466 DECLARE_FUNCTION(3, 1)
00467 DECLARE_FUNCTION(3, 2)
00468 DECLARE_FUNCTION(3, 3)
00469 
00470 void ff_vc1dsp_init_mmx(DSPContext* dsp, AVCodecContext *avctx) {
00471     dsp->put_vc1_mspel_pixels_tab[ 0] = ff_put_vc1_mspel_mc00_mmx;
00472     dsp->put_vc1_mspel_pixels_tab[ 4] = put_vc1_mspel_mc01_mmx;
00473     dsp->put_vc1_mspel_pixels_tab[ 8] = put_vc1_mspel_mc02_mmx;
00474     dsp->put_vc1_mspel_pixels_tab[12] = put_vc1_mspel_mc03_mmx;
00475 
00476     dsp->put_vc1_mspel_pixels_tab[ 1] = put_vc1_mspel_mc10_mmx;
00477     dsp->put_vc1_mspel_pixels_tab[ 5] = put_vc1_mspel_mc11_mmx;
00478     dsp->put_vc1_mspel_pixels_tab[ 9] = put_vc1_mspel_mc12_mmx;
00479     dsp->put_vc1_mspel_pixels_tab[13] = put_vc1_mspel_mc13_mmx;
00480 
00481     dsp->put_vc1_mspel_pixels_tab[ 2] = put_vc1_mspel_mc20_mmx;
00482     dsp->put_vc1_mspel_pixels_tab[ 6] = put_vc1_mspel_mc21_mmx;
00483     dsp->put_vc1_mspel_pixels_tab[10] = put_vc1_mspel_mc22_mmx;
00484     dsp->put_vc1_mspel_pixels_tab[14] = put_vc1_mspel_mc23_mmx;
00485 
00486     dsp->put_vc1_mspel_pixels_tab[ 3] = put_vc1_mspel_mc30_mmx;
00487     dsp->put_vc1_mspel_pixels_tab[ 7] = put_vc1_mspel_mc31_mmx;
00488     dsp->put_vc1_mspel_pixels_tab[11] = put_vc1_mspel_mc32_mmx;
00489     dsp->put_vc1_mspel_pixels_tab[15] = put_vc1_mspel_mc33_mmx;
00490 }
libavcodec/x86/vc1dsp_mmx.c
