FFmpeg: libavcodec/vp8dsp.c Source File

00001 /*
00002  * Copyright (C) 2010 David Conrad
00003  * Copyright (C) 2010 Ronald S. Bultje
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00027 #include "dsputil.h"
00028 #include "vp8dsp.h"
00029 #include "libavutil/common.h"
00030 
00031 // TODO: Maybe add dequant
00032 static void vp8_luma_dc_wht_c(DCTELEM block[4][4][16], DCTELEM dc[16])
00033 {
00034     int i, t0, t1, t2, t3;
00035 
00036     for (i = 0; i < 4; i++) {
00037         t0 = dc[0*4+i] + dc[3*4+i];
00038         t1 = dc[1*4+i] + dc[2*4+i];
00039         t2 = dc[1*4+i] - dc[2*4+i];
00040         t3 = dc[0*4+i] - dc[3*4+i];
00041 
00042         dc[0*4+i] = t0 + t1;
00043         dc[1*4+i] = t3 + t2;
00044         dc[2*4+i] = t0 - t1;
00045         dc[3*4+i] = t3 - t2;
00046     }
00047 
00048     for (i = 0; i < 4; i++) {
00049         t0 = dc[i*4+0] + dc[i*4+3] + 3; // rounding
00050         t1 = dc[i*4+1] + dc[i*4+2];
00051         t2 = dc[i*4+1] - dc[i*4+2];
00052         t3 = dc[i*4+0] - dc[i*4+3] + 3; // rounding
00053         dc[i*4+0] = 0;
00054         dc[i*4+1] = 0;
00055         dc[i*4+2] = 0;
00056         dc[i*4+3] = 0;
00057 
00058         block[i][0][0] = (t0 + t1) >> 3;
00059         block[i][1][0] = (t3 + t2) >> 3;
00060         block[i][2][0] = (t0 - t1) >> 3;
00061         block[i][3][0] = (t3 - t2) >> 3;
00062     }
00063 }
00064 
00065 static void vp8_luma_dc_wht_dc_c(DCTELEM block[4][4][16], DCTELEM dc[16])
00066 {
00067     int i, val = (dc[0] + 3) >> 3;
00068     dc[0] = 0;
00069 
00070     for (i = 0; i < 4; i++) {
00071         block[i][0][0] = val;
00072         block[i][1][0] = val;
00073         block[i][2][0] = val;
00074         block[i][3][0] = val;
00075     }
00076 }
00077 
00078 #define MUL_20091(a) ((((a)*20091) >> 16) + (a))
00079 #define MUL_35468(a)  (((a)*35468) >> 16)
00080 
00081 static void vp8_idct_add_c(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride)
00082 {
00083     int i, t0, t1, t2, t3;
00084     DCTELEM tmp[16];
00085 
00086     for (i = 0; i < 4; i++) {
00087         t0 = block[0*4+i] + block[2*4+i];
00088         t1 = block[0*4+i] - block[2*4+i];
00089         t2 = MUL_35468(block[1*4+i]) - MUL_20091(block[3*4+i]);
00090         t3 = MUL_20091(block[1*4+i]) + MUL_35468(block[3*4+i]);
00091         block[0*4+i] = 0;
00092         block[1*4+i] = 0;
00093         block[2*4+i] = 0;
00094         block[3*4+i] = 0;
00095 
00096         tmp[i*4+0] = t0 + t3;
00097         tmp[i*4+1] = t1 + t2;
00098         tmp[i*4+2] = t1 - t2;
00099         tmp[i*4+3] = t0 - t3;
00100     }
00101 
00102     for (i = 0; i < 4; i++) {
00103         t0 = tmp[0*4+i] + tmp[2*4+i];
00104         t1 = tmp[0*4+i] - tmp[2*4+i];
00105         t2 = MUL_35468(tmp[1*4+i]) - MUL_20091(tmp[3*4+i]);
00106         t3 = MUL_20091(tmp[1*4+i]) + MUL_35468(tmp[3*4+i]);
00107 
00108         dst[0] = av_clip_uint8(dst[0] + ((t0 + t3 + 4) >> 3));
00109         dst[1] = av_clip_uint8(dst[1] + ((t1 + t2 + 4) >> 3));
00110         dst[2] = av_clip_uint8(dst[2] + ((t1 - t2 + 4) >> 3));
00111         dst[3] = av_clip_uint8(dst[3] + ((t0 - t3 + 4) >> 3));
00112         dst += stride;
00113     }
00114 }
00115 
00116 static void vp8_idct_dc_add_c(uint8_t *dst, DCTELEM block[16], ptrdiff_t stride)
00117 {
00118     int i, dc = (block[0] + 4) >> 3;
00119     block[0] = 0;
00120 
00121     for (i = 0; i < 4; i++) {
00122         dst[0] = av_clip_uint8(dst[0] + dc);
00123         dst[1] = av_clip_uint8(dst[1] + dc);
00124         dst[2] = av_clip_uint8(dst[2] + dc);
00125         dst[3] = av_clip_uint8(dst[3] + dc);
00126         dst += stride;
00127     }
00128 }
00129 
00130 static void vp8_idct_dc_add4uv_c(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride)
00131 {
00132     vp8_idct_dc_add_c(dst+stride*0+0, block[0], stride);
00133     vp8_idct_dc_add_c(dst+stride*0+4, block[1], stride);
00134     vp8_idct_dc_add_c(dst+stride*4+0, block[2], stride);
00135     vp8_idct_dc_add_c(dst+stride*4+4, block[3], stride);
00136 }
00137 
00138 static void vp8_idct_dc_add4y_c(uint8_t *dst, DCTELEM block[4][16], ptrdiff_t stride)
00139 {
00140     vp8_idct_dc_add_c(dst+ 0, block[0], stride);
00141     vp8_idct_dc_add_c(dst+ 4, block[1], stride);
00142     vp8_idct_dc_add_c(dst+ 8, block[2], stride);
00143     vp8_idct_dc_add_c(dst+12, block[3], stride);
00144 }
00145 
00146 // because I like only having two parameters to pass functions...
00147 #define LOAD_PIXELS\
00148     int av_unused p3 = p[-4*stride];\
00149     int av_unused p2 = p[-3*stride];\
00150     int av_unused p1 = p[-2*stride];\
00151     int av_unused p0 = p[-1*stride];\
00152     int av_unused q0 = p[ 0*stride];\
00153     int av_unused q1 = p[ 1*stride];\
00154     int av_unused q2 = p[ 2*stride];\
00155     int av_unused q3 = p[ 3*stride];
00156 
00157 #define clip_int8(n) (cm[n+0x80]-0x80)
00158 
00159 static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4tap)
00160 {
00161     LOAD_PIXELS
00162     int a, f1, f2;
00163     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
00164 
00165     a = 3*(q0 - p0);
00166 
00167     if (is4tap)
00168         a += clip_int8(p1 - q1);
00169 
00170     a = clip_int8(a);
00171 
00172     // We deviate from the spec here with c(a+3) >> 3
00173     // since that's what libvpx does.
00174     f1 = FFMIN(a+4, 127) >> 3;
00175     f2 = FFMIN(a+3, 127) >> 3;
00176 
00177     // Despite what the spec says, we do need to clamp here to
00178     // be bitexact with libvpx.
00179     p[-1*stride] = cm[p0 + f2];
00180     p[ 0*stride] = cm[q0 - f1];
00181 
00182     // only used for _inner on blocks without high edge variance
00183     if (!is4tap) {
00184         a = (f1+1)>>1;
00185         p[-2*stride] = cm[p1 + a];
00186         p[ 1*stride] = cm[q1 - a];
00187     }
00188 }
00189 
00190 static av_always_inline int simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
00191 {
00192     LOAD_PIXELS
00193     return 2*FFABS(p0-q0) + (FFABS(p1-q1) >> 1) <= flim;
00194 }
00195 
00200 static av_always_inline int normal_limit(uint8_t *p, ptrdiff_t stride, int E, int I)
00201 {
00202     LOAD_PIXELS
00203     return simple_limit(p, stride, E)
00204         && FFABS(p3-p2) <= I && FFABS(p2-p1) <= I && FFABS(p1-p0) <= I
00205         && FFABS(q3-q2) <= I && FFABS(q2-q1) <= I && FFABS(q1-q0) <= I;
00206 }
00207 
00208 // high edge variance
00209 static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh)
00210 {
00211     LOAD_PIXELS
00212     return FFABS(p1-p0) > thresh || FFABS(q1-q0) > thresh;
00213 }
00214 
00215 static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
00216 {
00217     int a0, a1, a2, w;
00218     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
00219 
00220     LOAD_PIXELS
00221 
00222     w = clip_int8(p1-q1);
00223     w = clip_int8(w + 3*(q0-p0));
00224 
00225     a0 = (27*w + 63) >> 7;
00226     a1 = (18*w + 63) >> 7;
00227     a2 = ( 9*w + 63) >> 7;
00228 
00229     p[-3*stride] = cm[p2 + a2];
00230     p[-2*stride] = cm[p1 + a1];
00231     p[-1*stride] = cm[p0 + a0];
00232     p[ 0*stride] = cm[q0 - a0];
00233     p[ 1*stride] = cm[q1 - a1];
00234     p[ 2*stride] = cm[q2 - a2];
00235 }
00236 
00237 #define LOOP_FILTER(dir, size, stridea, strideb, maybe_inline) \
00238 static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, ptrdiff_t stride,\
00239                                      int flim_E, int flim_I, int hev_thresh)\
00240 {\
00241     int i;\
00242 \
00243     for (i = 0; i < size; i++)\
00244         if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
00245             if (hev(dst+i*stridea, strideb, hev_thresh))\
00246                 filter_common(dst+i*stridea, strideb, 1);\
00247             else\
00248                 filter_mbedge(dst+i*stridea, strideb);\
00249         }\
00250 }\
00251 \
00252 static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, ptrdiff_t stride,\
00253                                       int flim_E, int flim_I, int hev_thresh)\
00254 {\
00255     int i;\
00256 \
00257     for (i = 0; i < size; i++)\
00258         if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
00259             int hv = hev(dst+i*stridea, strideb, hev_thresh);\
00260             if (hv) \
00261                 filter_common(dst+i*stridea, strideb, 1);\
00262             else \
00263                 filter_common(dst+i*stridea, strideb, 0);\
00264         }\
00265 }
00266 
00267 LOOP_FILTER(v, 16, 1, stride,)
00268 LOOP_FILTER(h, 16, stride, 1,)
00269 
00270 #define UV_LOOP_FILTER(dir, stridea, strideb) \
00271 LOOP_FILTER(dir, 8, stridea, strideb, av_always_inline) \
00272 static void vp8_ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
00273                                       int fE, int fI, int hev_thresh)\
00274 {\
00275   vp8_ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh);\
00276   vp8_ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh);\
00277 }\
00278 static void vp8_ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
00279                                       int fE, int fI, int hev_thresh)\
00280 {\
00281   vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh);\
00282   vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh);\
00283 }
00284 
00285 UV_LOOP_FILTER(v, 1, stride)
00286 UV_LOOP_FILTER(h, stride, 1)
00287 
00288 static void vp8_v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)
00289 {
00290     int i;
00291 
00292     for (i = 0; i < 16; i++)
00293         if (simple_limit(dst+i, stride, flim))
00294             filter_common(dst+i, stride, 1);
00295 }
00296 
00297 static void vp8_h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)
00298 {
00299     int i;
00300 
00301     for (i = 0; i < 16; i++)
00302         if (simple_limit(dst+i*stride, 1, flim))
00303             filter_common(dst+i*stride, 1, 1);
00304 }
00305 
00306 static const uint8_t subpel_filters[7][6] = {
00307     { 0,   6, 123,  12,   1,   0 },
00308     { 2,  11, 108,  36,   8,   1 },
00309     { 0,   9,  93,  50,   6,   0 },
00310     { 3,  16,  77,  77,  16,   3 },
00311     { 0,   6,  50,  93,   9,   0 },
00312     { 1,   8,  36, 108,  11,   2 },
00313     { 0,   1,  12, 123,   6,   0 },
00314 };
00315 
00316 #define PUT_PIXELS(WIDTH) \
00317 static void put_vp8_pixels ## WIDTH ##_c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int x, int y) { \
00318     int i; \
00319     for (i = 0; i < h; i++, dst+= dststride, src+= srcstride) { \
00320         memcpy(dst, src, WIDTH); \
00321     } \
00322 }
00323 
00324 PUT_PIXELS(16)
00325 PUT_PIXELS(8)
00326 PUT_PIXELS(4)
00327 
00328 #define FILTER_6TAP(src, F, stride) \
00329     cm[(F[2]*src[x+0*stride] - F[1]*src[x-1*stride] + F[0]*src[x-2*stride] + \
00330         F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + F[5]*src[x+3*stride] + 64) >> 7]
00331 
00332 #define FILTER_4TAP(src, F, stride) \
00333     cm[(F[2]*src[x+0*stride] - F[1]*src[x-1*stride] + \
00334         F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + 64) >> 7]
00335 
00336 #define VP8_EPEL_H(SIZE, TAPS) \
00337 static void put_vp8_epel ## SIZE ## _h ## TAPS ## _c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my) \
00338 { \
00339     const uint8_t *filter = subpel_filters[mx-1]; \
00340     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
00341     int x, y; \
00342 \
00343     for (y = 0; y < h; y++) { \
00344         for (x = 0; x < SIZE; x++) \
00345             dst[x] = FILTER_ ## TAPS ## TAP(src, filter, 1); \
00346         dst += dststride; \
00347         src += srcstride; \
00348     } \
00349 }
00350 #define VP8_EPEL_V(SIZE, TAPS) \
00351 static void put_vp8_epel ## SIZE ## _v ## TAPS ## _c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my) \
00352 { \
00353     const uint8_t *filter = subpel_filters[my-1]; \
00354     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
00355     int x, y; \
00356 \
00357     for (y = 0; y < h; y++) { \
00358         for (x = 0; x < SIZE; x++) \
00359             dst[x] = FILTER_ ## TAPS ## TAP(src, filter, srcstride); \
00360         dst += dststride; \
00361         src += srcstride; \
00362     } \
00363 }
00364 #define VP8_EPEL_HV(SIZE, HTAPS, VTAPS) \
00365 static void put_vp8_epel ## SIZE ## _h ## HTAPS ## v ## VTAPS ## _c(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, ptrdiff_t srcstride, int h, int mx, int my) \
00366 { \
00367     const uint8_t *filter = subpel_filters[mx-1]; \
00368     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
00369     int x, y; \
00370     uint8_t tmp_array[(2*SIZE+VTAPS-1)*SIZE]; \
00371     uint8_t *tmp = tmp_array; \
00372     src -= (2-(VTAPS==4))*srcstride; \
00373 \
00374     for (y = 0; y < h+VTAPS-1; y++) { \
00375         for (x = 0; x < SIZE; x++) \
00376             tmp[x] = FILTER_ ## HTAPS ## TAP(src, filter, 1); \
00377         tmp += SIZE; \
00378         src += srcstride; \
00379     } \
00380 \
00381     tmp = tmp_array + (2-(VTAPS==4))*SIZE; \
00382     filter = subpel_filters[my-1]; \
00383 \
00384     for (y = 0; y < h; y++) { \
00385         for (x = 0; x < SIZE; x++) \
00386             dst[x] = FILTER_ ## VTAPS ## TAP(tmp, filter, SIZE); \
00387         dst += dststride; \
00388         tmp += SIZE; \
00389     } \
00390 }
00391 
00392 VP8_EPEL_H(16, 4)
00393 VP8_EPEL_H(8,  4)
00394 VP8_EPEL_H(4,  4)
00395 VP8_EPEL_H(16, 6)
00396 VP8_EPEL_H(8,  6)
00397 VP8_EPEL_H(4,  6)
00398 VP8_EPEL_V(16, 4)
00399 VP8_EPEL_V(8,  4)
00400 VP8_EPEL_V(4,  4)
00401 VP8_EPEL_V(16, 6)
00402 VP8_EPEL_V(8,  6)
00403 VP8_EPEL_V(4,  6)
00404 VP8_EPEL_HV(16, 4, 4)
00405 VP8_EPEL_HV(8,  4, 4)
00406 VP8_EPEL_HV(4,  4, 4)
00407 VP8_EPEL_HV(16, 4, 6)
00408 VP8_EPEL_HV(8,  4, 6)
00409 VP8_EPEL_HV(4,  4, 6)
00410 VP8_EPEL_HV(16, 6, 4)
00411 VP8_EPEL_HV(8,  6, 4)
00412 VP8_EPEL_HV(4,  6, 4)
00413 VP8_EPEL_HV(16, 6, 6)
00414 VP8_EPEL_HV(8,  6, 6)
00415 VP8_EPEL_HV(4,  6, 6)
00416 
00417 #define VP8_BILINEAR(SIZE) \
00418 static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, ptrdiff_t stride, uint8_t *src, ptrdiff_t s2, int h, int mx, int my) \
00419 { \
00420     int a = 8-mx, b = mx; \
00421     int x, y; \
00422 \
00423     for (y = 0; y < h; y++) { \
00424         for (x = 0; x < SIZE; x++) \
00425             dst[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
00426         dst += stride; \
00427         src += stride; \
00428     } \
00429 } \
00430 static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, ptrdiff_t stride, uint8_t *src, ptrdiff_t s2, int h, int mx, int my) \
00431 { \
00432     int c = 8-my, d = my; \
00433     int x, y; \
00434 \
00435     for (y = 0; y < h; y++) { \
00436         for (x = 0; x < SIZE; x++) \
00437             dst[x] = (c*src[x] + d*src[x+stride] + 4) >> 3; \
00438         dst += stride; \
00439         src += stride; \
00440     } \
00441 } \
00442 \
00443 static void put_vp8_bilinear ## SIZE ## _hv_c(uint8_t *dst, ptrdiff_t stride, uint8_t *src, ptrdiff_t s2, int h, int mx, int my) \
00444 { \
00445     int a = 8-mx, b = mx; \
00446     int c = 8-my, d = my; \
00447     int x, y; \
00448     uint8_t tmp_array[(2*SIZE+1)*SIZE]; \
00449     uint8_t *tmp = tmp_array; \
00450 \
00451     for (y = 0; y < h+1; y++) { \
00452         for (x = 0; x < SIZE; x++) \
00453             tmp[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
00454         tmp += SIZE; \
00455         src += stride; \
00456     } \
00457 \
00458     tmp = tmp_array; \
00459 \
00460     for (y = 0; y < h; y++) { \
00461         for (x = 0; x < SIZE; x++) \
00462             dst[x] = (c*tmp[x] + d*tmp[x+SIZE] + 4) >> 3; \
00463         dst += stride; \
00464         tmp += SIZE; \
00465     } \
00466 }
00467 
00468 VP8_BILINEAR(16)
00469 VP8_BILINEAR(8)
00470 VP8_BILINEAR(4)
00471 
00472 #define VP8_MC_FUNC(IDX, SIZE) \
00473     dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
00474     dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c; \
00475     dsp->put_vp8_epel_pixels_tab[IDX][0][2] = put_vp8_epel ## SIZE ## _h6_c; \
00476     dsp->put_vp8_epel_pixels_tab[IDX][1][0] = put_vp8_epel ## SIZE ## _v4_c; \
00477     dsp->put_vp8_epel_pixels_tab[IDX][1][1] = put_vp8_epel ## SIZE ## _h4v4_c; \
00478     dsp->put_vp8_epel_pixels_tab[IDX][1][2] = put_vp8_epel ## SIZE ## _h6v4_c; \
00479     dsp->put_vp8_epel_pixels_tab[IDX][2][0] = put_vp8_epel ## SIZE ## _v6_c; \
00480     dsp->put_vp8_epel_pixels_tab[IDX][2][1] = put_vp8_epel ## SIZE ## _h4v6_c; \
00481     dsp->put_vp8_epel_pixels_tab[IDX][2][2] = put_vp8_epel ## SIZE ## _h6v6_c
00482 
00483 #define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \
00484     dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
00485     dsp->put_vp8_bilinear_pixels_tab[IDX][0][1] = put_vp8_bilinear ## SIZE ## _h_c; \
00486     dsp->put_vp8_bilinear_pixels_tab[IDX][0][2] = put_vp8_bilinear ## SIZE ## _h_c; \
00487     dsp->put_vp8_bilinear_pixels_tab[IDX][1][0] = put_vp8_bilinear ## SIZE ## _v_c; \
00488     dsp->put_vp8_bilinear_pixels_tab[IDX][1][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
00489     dsp->put_vp8_bilinear_pixels_tab[IDX][1][2] = put_vp8_bilinear ## SIZE ## _hv_c; \
00490     dsp->put_vp8_bilinear_pixels_tab[IDX][2][0] = put_vp8_bilinear ## SIZE ## _v_c; \
00491     dsp->put_vp8_bilinear_pixels_tab[IDX][2][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
00492     dsp->put_vp8_bilinear_pixels_tab[IDX][2][2] = put_vp8_bilinear ## SIZE ## _hv_c
00493 
00494 av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
00495 {
00496     dsp->vp8_luma_dc_wht    = vp8_luma_dc_wht_c;
00497     dsp->vp8_luma_dc_wht_dc = vp8_luma_dc_wht_dc_c;
00498     dsp->vp8_idct_add       = vp8_idct_add_c;
00499     dsp->vp8_idct_dc_add    = vp8_idct_dc_add_c;
00500     dsp->vp8_idct_dc_add4y  = vp8_idct_dc_add4y_c;
00501     dsp->vp8_idct_dc_add4uv = vp8_idct_dc_add4uv_c;
00502 
00503     dsp->vp8_v_loop_filter16y = vp8_v_loop_filter16_c;
00504     dsp->vp8_h_loop_filter16y = vp8_h_loop_filter16_c;
00505     dsp->vp8_v_loop_filter8uv = vp8_v_loop_filter8uv_c;
00506     dsp->vp8_h_loop_filter8uv = vp8_h_loop_filter8uv_c;
00507 
00508     dsp->vp8_v_loop_filter16y_inner = vp8_v_loop_filter16_inner_c;
00509     dsp->vp8_h_loop_filter16y_inner = vp8_h_loop_filter16_inner_c;
00510     dsp->vp8_v_loop_filter8uv_inner = vp8_v_loop_filter8uv_inner_c;
00511     dsp->vp8_h_loop_filter8uv_inner = vp8_h_loop_filter8uv_inner_c;
00512 
00513     dsp->vp8_v_loop_filter_simple = vp8_v_loop_filter_simple_c;
00514     dsp->vp8_h_loop_filter_simple = vp8_h_loop_filter_simple_c;
00515 
00516     VP8_MC_FUNC(0, 16);
00517     VP8_MC_FUNC(1, 8);
00518     VP8_MC_FUNC(2, 4);
00519 
00520     VP8_BILINEAR_MC_FUNC(0, 16);
00521     VP8_BILINEAR_MC_FUNC(1, 8);
00522     VP8_BILINEAR_MC_FUNC(2, 4);
00523 
00524     if (HAVE_MMX)
00525         ff_vp8dsp_init_x86(dsp);
00526     if (HAVE_ALTIVEC)
00527         ff_vp8dsp_init_altivec(dsp);
00528     if (ARCH_ARM)
00529         ff_vp8dsp_init_arm(dsp);
00530 }