yading@10: /*
yading@10:  * Copyright (C) 2010 David Conrad
yading@10:  * Copyright (C) 2010 Ronald S. Bultje
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
yading@10:  * Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public
yading@10:  * License along with FFmpeg; if not, write to the Free Software
yading@10:  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: /**
yading@10:  * @file
yading@10:  * VP8 compatible video decoder
yading@10:  */
yading@10: 
yading@10: #include "dsputil.h"
yading@10: #include "vp8dsp.h"
yading@10: #include "libavutil/common.h"
yading@10: 
yading@10: // TODO: Maybe add dequant
yading@10: static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
yading@10: {
yading@10:     int i, t0, t1, t2, t3;
yading@10: 
yading@10:     for (i = 0; i < 4; i++) {
yading@10:         t0 = dc[0*4+i] + dc[3*4+i];
yading@10:         t1 = dc[1*4+i] + dc[2*4+i];
yading@10:         t2 = dc[1*4+i] - dc[2*4+i];
yading@10:         t3 = dc[0*4+i] - dc[3*4+i];
yading@10: 
yading@10:         dc[0*4+i] = t0 + t1;
yading@10:         dc[1*4+i] = t3 + t2;
yading@10:         dc[2*4+i] = t0 - t1;
yading@10:         dc[3*4+i] = t3 - t2;
yading@10:     }
yading@10: 
yading@10:     for (i = 0; i < 4; i++) {
yading@10:         t0 = dc[i*4+0] + dc[i*4+3] + 3; // rounding
yading@10:         t1 = dc[i*4+1] + dc[i*4+2];
yading@10:         t2 = dc[i*4+1] - dc[i*4+2];
yading@10:         t3 = dc[i*4+0] - dc[i*4+3] + 3; // rounding
yading@10:         dc[i*4+0] = 0;
yading@10:         dc[i*4+1] = 0;
yading@10:         dc[i*4+2] = 0;
yading@10:         dc[i*4+3] = 0;
yading@10: 
yading@10:         block[i][0][0] = (t0 + t1) >> 3;
yading@10:         block[i][1][0] = (t3 + t2) >> 3;
yading@10:         block[i][2][0] = (t0 - t1) >> 3;
yading@10:         block[i][3][0] = (t3 - t2) >> 3;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vp8_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16])
yading@10: {
yading@10:     int i, val = (dc[0] + 3) >> 3;
yading@10:     dc[0] = 0;
yading@10: 
yading@10:     for (i = 0; i < 4; i++) {
yading@10:         block[i][0][0] = val;
yading@10:         block[i][1][0] = val;
yading@10:         block[i][2][0] = val;
yading@10:         block[i][3][0] = val;
yading@10:     }
yading@10: }
yading@10: 
yading@10: #define MUL_20091(a) ((((a)*20091) >> 16) + (a))
yading@10: #define MUL_35468(a)  (((a)*35468) >> 16)
yading@10: 
yading@10: static void vp8_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
yading@10: {
yading@10:     int i, t0, t1, t2, t3;
yading@10:     int16_t tmp[16];
yading@10: 
yading@10:     for (i = 0; i < 4; i++) {
yading@10:         t0 = block[0*4+i] + block[2*4+i];
yading@10:         t1 = block[0*4+i] - block[2*4+i];
yading@10:         t2 = MUL_35468(block[1*4+i]) - MUL_20091(block[3*4+i]);
yading@10:         t3 = MUL_20091(block[1*4+i]) + MUL_35468(block[3*4+i]);
yading@10:         block[0*4+i] = 0;
yading@10:         block[1*4+i] = 0;
yading@10:         block[2*4+i] = 0;
yading@10:         block[3*4+i] = 0;
yading@10: 
yading@10:         tmp[i*4+0] = t0 + t3;
yading@10:         tmp[i*4+1] = t1 + t2;
yading@10:         tmp[i*4+2] = t1 - t2;
yading@10:         tmp[i*4+3] = t0 - t3;
yading@10:     }
yading@10: 
yading@10:     for (i = 0; i < 4; i++) {
yading@10:         t0 = tmp[0*4+i] + tmp[2*4+i];
yading@10:         t1 = tmp[0*4+i] - tmp[2*4+i];
yading@10:         t2 = MUL_35468(tmp[1*4+i]) - MUL_20091(tmp[3*4+i]);
yading@10:         t3 = MUL_20091(tmp[1*4+i]) + MUL_35468(tmp[3*4+i]);
yading@10: 
yading@10:         dst[0] = av_clip_uint8(dst[0] + ((t0 + t3 + 4) >> 3));
yading@10:         dst[1] = av_clip_uint8(dst[1] + ((t1 + t2 + 4) >> 3));
yading@10:         dst[2] = av_clip_uint8(dst[2] + ((t1 - t2 + 4) >> 3));
yading@10:         dst[3] = av_clip_uint8(dst[3] + ((t0 - t3 + 4) >> 3));
yading@10:         dst += stride;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vp8_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
yading@10: {
yading@10:     int i, dc = (block[0] + 4) >> 3;
yading@10:     block[0] = 0;
yading@10: 
yading@10:     for (i = 0; i < 4; i++) {
yading@10:         dst[0] = av_clip_uint8(dst[0] + dc);
yading@10:         dst[1] = av_clip_uint8(dst[1] + dc);
yading@10:         dst[2] = av_clip_uint8(dst[2] + dc);
yading@10:         dst[3] = av_clip_uint8(dst[3] + dc);
yading@10:         dst += stride;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static void vp8_idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)
yading@10: {
yading@10:     vp8_idct_dc_add_c(dst+stride*0+0, block[0], stride);
yading@10:     vp8_idct_dc_add_c(dst+stride*0+4, block[1], stride);
yading@10:     vp8_idct_dc_add_c(dst+stride*4+0, block[2], stride);
yading@10:     vp8_idct_dc_add_c(dst+stride*4+4, block[3], stride);
yading@10: }
yading@10: 
yading@10: static void vp8_idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)
yading@10: {
yading@10:     vp8_idct_dc_add_c(dst+ 0, block[0], stride);
yading@10:     vp8_idct_dc_add_c(dst+ 4, block[1], stride);
yading@10:     vp8_idct_dc_add_c(dst+ 8, block[2], stride);
yading@10:     vp8_idct_dc_add_c(dst+12, block[3], stride);
yading@10: }
yading@10: 
yading@10: // because I like only having two parameters to pass functions...
yading@10: #define LOAD_PIXELS\
yading@10:     int av_unused p3 = p[-4*stride];\
yading@10:     int av_unused p2 = p[-3*stride];\
yading@10:     int av_unused p1 = p[-2*stride];\
yading@10:     int av_unused p0 = p[-1*stride];\
yading@10:     int av_unused q0 = p[ 0*stride];\
yading@10:     int av_unused q1 = p[ 1*stride];\
yading@10:     int av_unused q2 = p[ 2*stride];\
yading@10:     int av_unused q3 = p[ 3*stride];
yading@10: 
yading@10: #define clip_int8(n) (cm[n+0x80]-0x80)
yading@10: 
yading@10: static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4tap)
yading@10: {
yading@10:     LOAD_PIXELS
yading@10:     int a, f1, f2;
yading@10:     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
yading@10: 
yading@10:     a = 3*(q0 - p0);
yading@10: 
yading@10:     if (is4tap)
yading@10:         a += clip_int8(p1 - q1);
yading@10: 
yading@10:     a = clip_int8(a);
yading@10: 
yading@10:     // We deviate from the spec here with c(a+3) >> 3
yading@10:     // since that's what libvpx does.
yading@10:     f1 = FFMIN(a+4, 127) >> 3;
yading@10:     f2 = FFMIN(a+3, 127) >> 3;
yading@10: 
yading@10:     // Despite what the spec says, we do need to clamp here to
yading@10:     // be bitexact with libvpx.
yading@10:     p[-1*stride] = cm[p0 + f2];
yading@10:     p[ 0*stride] = cm[q0 - f1];
yading@10: 
yading@10:     // only used for _inner on blocks without high edge variance
yading@10:     if (!is4tap) {
yading@10:         a = (f1+1)>>1;
yading@10:         p[-2*stride] = cm[p1 + a];
yading@10:         p[ 1*stride] = cm[q1 - a];
yading@10:     }
yading@10: }
yading@10: 
yading@10: static av_always_inline int simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
yading@10: {
yading@10:     LOAD_PIXELS
yading@10:     return 2*FFABS(p0-q0) + (FFABS(p1-q1) >> 1) <= flim;
yading@10: }
yading@10: 
yading@10: /**
yading@10:  * E - limit at the macroblock edge
yading@10:  * I - limit for interior difference
yading@10:  */
yading@10: static av_always_inline int normal_limit(uint8_t *p, ptrdiff_t stride, int E, int I)
yading@10: {
yading@10:     LOAD_PIXELS
yading@10:     return simple_limit(p, stride, E)
yading@10:         && FFABS(p3-p2) <= I && FFABS(p2-p1) <= I && FFABS(p1-p0) <= I
yading@10:         && FFABS(q3-q2) <= I && FFABS(q2-q1) <= I && FFABS(q1-q0) <= I;
yading@10: }
yading@10: 
yading@10: // high edge variance
yading@10: static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh)
yading@10: {
yading@10:     LOAD_PIXELS
yading@10:     return FFABS(p1-p0) > thresh || FFABS(q1-q0) > thresh;
yading@10: }
yading@10: 
yading@10: static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
yading@10: {
yading@10:     int a0, a1, a2, w;
yading@10:     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
yading@10: 
yading@10:     LOAD_PIXELS
yading@10: 
yading@10:     w = clip_int8(p1-q1);
yading@10:     w = clip_int8(w + 3*(q0-p0));
yading@10: 
yading@10:     a0 = (27*w + 63) >> 7;
yading@10:     a1 = (18*w + 63) >> 7;
yading@10:     a2 = ( 9*w + 63) >> 7;
yading@10: 
yading@10:     p[-3*stride] = cm[p2 + a2];
yading@10:     p[-2*stride] = cm[p1 + a1];
yading@10:     p[-1*stride] = cm[p0 + a0];
yading@10:     p[ 0*stride] = cm[q0 - a0];
yading@10:     p[ 1*stride] = cm[q1 - a1];
yading@10:     p[ 2*stride] = cm[q2 - a2];
yading@10: }
yading@10: 
yading@10: #define LOOP_FILTER(dir, size, stridea, strideb, maybe_inline) \
yading@10: static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, ptrdiff_t stride,\
yading@10:                                      int flim_E, int flim_I, int hev_thresh)\
yading@10: {\
yading@10:     int i;\
yading@10: \
yading@10:     for (i = 0; i < size; i++)\
yading@10:         if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
yading@10:             if (hev(dst+i*stridea, strideb, hev_thresh))\
yading@10:                 filter_common(dst+i*stridea, strideb, 1);\
yading@10:             else\
yading@10:                 filter_mbedge(dst+i*stridea, strideb);\
yading@10:         }\
yading@10: }\
yading@10: \
yading@10: static maybe_inline void vp8_ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, ptrdiff_t stride,\
yading@10:                                       int flim_E, int flim_I, int hev_thresh)\
yading@10: {\
yading@10:     int i;\
yading@10: \
yading@10:     for (i = 0; i < size; i++)\
yading@10:         if (normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
yading@10:             int hv = hev(dst+i*stridea, strideb, hev_thresh);\
yading@10:             if (hv) \
yading@10:                 filter_common(dst+i*stridea, strideb, 1);\
yading@10:             else \
yading@10:                 filter_common(dst+i*stridea, strideb, 0);\
yading@10:         }\
yading@10: }
yading@10: 
yading@10: LOOP_FILTER(v, 16, 1, stride,)
yading@10: LOOP_FILTER(h, 16, stride, 1,)
yading@10: 
yading@10: #define UV_LOOP_FILTER(dir, stridea, strideb) \
yading@10: LOOP_FILTER(dir, 8, stridea, strideb, av_always_inline) \
yading@10: static void vp8_ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
yading@10:                                       int fE, int fI, int hev_thresh)\
yading@10: {\
yading@10:   vp8_ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh);\
yading@10:   vp8_ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh);\
yading@10: }\
yading@10: static void vp8_ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
yading@10:                                       int fE, int fI, int hev_thresh)\
yading@10: {\
yading@10:   vp8_ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh);\
yading@10:   vp8_ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh);\
yading@10: }
yading@10: 
yading@10: UV_LOOP_FILTER(v, 1, stride)
yading@10: UV_LOOP_FILTER(h, stride, 1)
yading@10: 
yading@10: static void vp8_v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for (i = 0; i < 16; i++)
yading@10:         if (simple_limit(dst+i, stride, flim))
yading@10:             filter_common(dst+i, stride, 1);
yading@10: }
yading@10: 
yading@10: static void vp8_h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)
yading@10: {
yading@10:     int i;
yading@10: 
yading@10:     for (i = 0; i < 16; i++)
yading@10:         if (simple_limit(dst+i*stride, 1, flim))
yading@10:             filter_common(dst+i*stride, 1, 1);
yading@10: }
yading@10: 
yading@10: static const uint8_t subpel_filters[7][6] = {
yading@10:     { 0,   6, 123,  12,   1,   0 },
yading@10:     { 2,  11, 108,  36,   8,   1 },
yading@10:     { 0,   9,  93,  50,   6,   0 },
yading@10:     { 3,  16,  77,  77,  16,   3 },
yading@10:     { 0,   6,  50,  93,   9,   0 },
yading@10:     { 1,   8,  36, 108,  11,   2 },
yading@10:     { 0,   1,  12, 123,   6,   0 },
yading@10: };
yading@10: 
yading@10: #define PUT_PIXELS(WIDTH) \
yading@10: 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) { \
yading@10:     int i; \
yading@10:     for (i = 0; i < h; i++, dst+= dststride, src+= srcstride) { \
yading@10:         memcpy(dst, src, WIDTH); \
yading@10:     } \
yading@10: }
yading@10: 
yading@10: PUT_PIXELS(16)
yading@10: PUT_PIXELS(8)
yading@10: PUT_PIXELS(4)
yading@10: 
yading@10: #define FILTER_6TAP(src, F, stride) \
yading@10:     cm[(F[2]*src[x+0*stride] - F[1]*src[x-1*stride] + F[0]*src[x-2*stride] + \
yading@10:         F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + F[5]*src[x+3*stride] + 64) >> 7]
yading@10: 
yading@10: #define FILTER_4TAP(src, F, stride) \
yading@10:     cm[(F[2]*src[x+0*stride] - F[1]*src[x-1*stride] + \
yading@10:         F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + 64) >> 7]
yading@10: 
yading@10: #define VP8_EPEL_H(SIZE, TAPS) \
yading@10: 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) \
yading@10: { \
yading@10:     const uint8_t *filter = subpel_filters[mx-1]; \
yading@10:     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
yading@10:     int x, y; \
yading@10: \
yading@10:     for (y = 0; y < h; y++) { \
yading@10:         for (x = 0; x < SIZE; x++) \
yading@10:             dst[x] = FILTER_ ## TAPS ## TAP(src, filter, 1); \
yading@10:         dst += dststride; \
yading@10:         src += srcstride; \
yading@10:     } \
yading@10: }
yading@10: #define VP8_EPEL_V(SIZE, TAPS) \
yading@10: 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) \
yading@10: { \
yading@10:     const uint8_t *filter = subpel_filters[my-1]; \
yading@10:     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
yading@10:     int x, y; \
yading@10: \
yading@10:     for (y = 0; y < h; y++) { \
yading@10:         for (x = 0; x < SIZE; x++) \
yading@10:             dst[x] = FILTER_ ## TAPS ## TAP(src, filter, srcstride); \
yading@10:         dst += dststride; \
yading@10:         src += srcstride; \
yading@10:     } \
yading@10: }
yading@10: #define VP8_EPEL_HV(SIZE, HTAPS, VTAPS) \
yading@10: 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) \
yading@10: { \
yading@10:     const uint8_t *filter = subpel_filters[mx-1]; \
yading@10:     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
yading@10:     int x, y; \
yading@10:     uint8_t tmp_array[(2*SIZE+VTAPS-1)*SIZE]; \
yading@10:     uint8_t *tmp = tmp_array; \
yading@10:     src -= (2-(VTAPS==4))*srcstride; \
yading@10: \
yading@10:     for (y = 0; y < h+VTAPS-1; y++) { \
yading@10:         for (x = 0; x < SIZE; x++) \
yading@10:             tmp[x] = FILTER_ ## HTAPS ## TAP(src, filter, 1); \
yading@10:         tmp += SIZE; \
yading@10:         src += srcstride; \
yading@10:     } \
yading@10: \
yading@10:     tmp = tmp_array + (2-(VTAPS==4))*SIZE; \
yading@10:     filter = subpel_filters[my-1]; \
yading@10: \
yading@10:     for (y = 0; y < h; y++) { \
yading@10:         for (x = 0; x < SIZE; x++) \
yading@10:             dst[x] = FILTER_ ## VTAPS ## TAP(tmp, filter, SIZE); \
yading@10:         dst += dststride; \
yading@10:         tmp += SIZE; \
yading@10:     } \
yading@10: }
yading@10: 
yading@10: VP8_EPEL_H(16, 4)
yading@10: VP8_EPEL_H(8,  4)
yading@10: VP8_EPEL_H(4,  4)
yading@10: VP8_EPEL_H(16, 6)
yading@10: VP8_EPEL_H(8,  6)
yading@10: VP8_EPEL_H(4,  6)
yading@10: VP8_EPEL_V(16, 4)
yading@10: VP8_EPEL_V(8,  4)
yading@10: VP8_EPEL_V(4,  4)
yading@10: VP8_EPEL_V(16, 6)
yading@10: VP8_EPEL_V(8,  6)
yading@10: VP8_EPEL_V(4,  6)
yading@10: VP8_EPEL_HV(16, 4, 4)
yading@10: VP8_EPEL_HV(8,  4, 4)
yading@10: VP8_EPEL_HV(4,  4, 4)
yading@10: VP8_EPEL_HV(16, 4, 6)
yading@10: VP8_EPEL_HV(8,  4, 6)
yading@10: VP8_EPEL_HV(4,  4, 6)
yading@10: VP8_EPEL_HV(16, 6, 4)
yading@10: VP8_EPEL_HV(8,  6, 4)
yading@10: VP8_EPEL_HV(4,  6, 4)
yading@10: VP8_EPEL_HV(16, 6, 6)
yading@10: VP8_EPEL_HV(8,  6, 6)
yading@10: VP8_EPEL_HV(4,  6, 6)
yading@10: 
yading@10: #define VP8_BILINEAR(SIZE) \
yading@10: 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) \
yading@10: { \
yading@10:     int a = 8-mx, b = mx; \
yading@10:     int x, y; \
yading@10: \
yading@10:     for (y = 0; y < h; y++) { \
yading@10:         for (x = 0; x < SIZE; x++) \
yading@10:             dst[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
yading@10:         dst += stride; \
yading@10:         src += stride; \
yading@10:     } \
yading@10: } \
yading@10: 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) \
yading@10: { \
yading@10:     int c = 8-my, d = my; \
yading@10:     int x, y; \
yading@10: \
yading@10:     for (y = 0; y < h; y++) { \
yading@10:         for (x = 0; x < SIZE; x++) \
yading@10:             dst[x] = (c*src[x] + d*src[x+stride] + 4) >> 3; \
yading@10:         dst += stride; \
yading@10:         src += stride; \
yading@10:     } \
yading@10: } \
yading@10: \
yading@10: 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) \
yading@10: { \
yading@10:     int a = 8-mx, b = mx; \
yading@10:     int c = 8-my, d = my; \
yading@10:     int x, y; \
yading@10:     uint8_t tmp_array[(2*SIZE+1)*SIZE]; \
yading@10:     uint8_t *tmp = tmp_array; \
yading@10: \
yading@10:     for (y = 0; y < h+1; y++) { \
yading@10:         for (x = 0; x < SIZE; x++) \
yading@10:             tmp[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
yading@10:         tmp += SIZE; \
yading@10:         src += stride; \
yading@10:     } \
yading@10: \
yading@10:     tmp = tmp_array; \
yading@10: \
yading@10:     for (y = 0; y < h; y++) { \
yading@10:         for (x = 0; x < SIZE; x++) \
yading@10:             dst[x] = (c*tmp[x] + d*tmp[x+SIZE] + 4) >> 3; \
yading@10:         dst += stride; \
yading@10:         tmp += SIZE; \
yading@10:     } \
yading@10: }
yading@10: 
yading@10: VP8_BILINEAR(16)
yading@10: VP8_BILINEAR(8)
yading@10: VP8_BILINEAR(4)
yading@10: 
yading@10: #define VP8_MC_FUNC(IDX, SIZE) \
yading@10:     dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
yading@10:     dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c; \
yading@10:     dsp->put_vp8_epel_pixels_tab[IDX][0][2] = put_vp8_epel ## SIZE ## _h6_c; \
yading@10:     dsp->put_vp8_epel_pixels_tab[IDX][1][0] = put_vp8_epel ## SIZE ## _v4_c; \
yading@10:     dsp->put_vp8_epel_pixels_tab[IDX][1][1] = put_vp8_epel ## SIZE ## _h4v4_c; \
yading@10:     dsp->put_vp8_epel_pixels_tab[IDX][1][2] = put_vp8_epel ## SIZE ## _h6v4_c; \
yading@10:     dsp->put_vp8_epel_pixels_tab[IDX][2][0] = put_vp8_epel ## SIZE ## _v6_c; \
yading@10:     dsp->put_vp8_epel_pixels_tab[IDX][2][1] = put_vp8_epel ## SIZE ## _h4v6_c; \
yading@10:     dsp->put_vp8_epel_pixels_tab[IDX][2][2] = put_vp8_epel ## SIZE ## _h6v6_c
yading@10: 
yading@10: #define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \
yading@10:     dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
yading@10:     dsp->put_vp8_bilinear_pixels_tab[IDX][0][1] = put_vp8_bilinear ## SIZE ## _h_c; \
yading@10:     dsp->put_vp8_bilinear_pixels_tab[IDX][0][2] = put_vp8_bilinear ## SIZE ## _h_c; \
yading@10:     dsp->put_vp8_bilinear_pixels_tab[IDX][1][0] = put_vp8_bilinear ## SIZE ## _v_c; \
yading@10:     dsp->put_vp8_bilinear_pixels_tab[IDX][1][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
yading@10:     dsp->put_vp8_bilinear_pixels_tab[IDX][1][2] = put_vp8_bilinear ## SIZE ## _hv_c; \
yading@10:     dsp->put_vp8_bilinear_pixels_tab[IDX][2][0] = put_vp8_bilinear ## SIZE ## _v_c; \
yading@10:     dsp->put_vp8_bilinear_pixels_tab[IDX][2][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
yading@10:     dsp->put_vp8_bilinear_pixels_tab[IDX][2][2] = put_vp8_bilinear ## SIZE ## _hv_c
yading@10: 
yading@10: av_cold void ff_vp8dsp_init(VP8DSPContext *dsp)
yading@10: {
yading@10:     dsp->vp8_luma_dc_wht    = vp8_luma_dc_wht_c;
yading@10:     dsp->vp8_luma_dc_wht_dc = vp8_luma_dc_wht_dc_c;
yading@10:     dsp->vp8_idct_add       = vp8_idct_add_c;
yading@10:     dsp->vp8_idct_dc_add    = vp8_idct_dc_add_c;
yading@10:     dsp->vp8_idct_dc_add4y  = vp8_idct_dc_add4y_c;
yading@10:     dsp->vp8_idct_dc_add4uv = vp8_idct_dc_add4uv_c;
yading@10: 
yading@10:     dsp->vp8_v_loop_filter16y = vp8_v_loop_filter16_c;
yading@10:     dsp->vp8_h_loop_filter16y = vp8_h_loop_filter16_c;
yading@10:     dsp->vp8_v_loop_filter8uv = vp8_v_loop_filter8uv_c;
yading@10:     dsp->vp8_h_loop_filter8uv = vp8_h_loop_filter8uv_c;
yading@10: 
yading@10:     dsp->vp8_v_loop_filter16y_inner = vp8_v_loop_filter16_inner_c;
yading@10:     dsp->vp8_h_loop_filter16y_inner = vp8_h_loop_filter16_inner_c;
yading@10:     dsp->vp8_v_loop_filter8uv_inner = vp8_v_loop_filter8uv_inner_c;
yading@10:     dsp->vp8_h_loop_filter8uv_inner = vp8_h_loop_filter8uv_inner_c;
yading@10: 
yading@10:     dsp->vp8_v_loop_filter_simple = vp8_v_loop_filter_simple_c;
yading@10:     dsp->vp8_h_loop_filter_simple = vp8_h_loop_filter_simple_c;
yading@10: 
yading@10:     VP8_MC_FUNC(0, 16);
yading@10:     VP8_MC_FUNC(1, 8);
yading@10:     VP8_MC_FUNC(2, 4);
yading@10: 
yading@10:     VP8_BILINEAR_MC_FUNC(0, 16);
yading@10:     VP8_BILINEAR_MC_FUNC(1, 8);
yading@10:     VP8_BILINEAR_MC_FUNC(2, 4);
yading@10: 
yading@10:     if (ARCH_X86)
yading@10:         ff_vp8dsp_init_x86(dsp);
yading@10:     if (HAVE_ALTIVEC)
yading@10:         ff_vp8dsp_init_altivec(dsp);
yading@10:     if (ARCH_ARM)
yading@10:         ff_vp8dsp_init_arm(dsp);
yading@10: }