libavcodec/vp8dsp.c
6168781f
 /*
3b636f21
  * Copyright (C) 2010 David Conrad
  * Copyright (C) 2010 Ronald S. Bultje
b8664c92
  * Copyright (C) 2014 Peter Ross
3b636f21
  *
  * This file is part of FFmpeg.
  *
  * FFmpeg is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
  * License as published by the Free Software Foundation; either
  * version 2.1 of the License, or (at your option) any later version.
  *
  * FFmpeg is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  * Lesser General Public License for more details.
  *
  * You should have received a copy of the GNU Lesser General Public
  * License along with FFmpeg; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
6168781f
 /**
  * @file
  * VP8 compatible video decoder
  */
 
3b636f21
 #include "dsputil.h"
 #include "vp8dsp.h"
1d9c2dc8
 #include "libavutil/common.h"
3b636f21
 
911eb711
 #define MK_IDCT_DC_ADD4_C(name) \
 static void name ## _idct_dc_add4uv_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)\
 {\
     name ## _idct_dc_add_c(dst+stride*0+0, block[0], stride);\
     name ## _idct_dc_add_c(dst+stride*0+4, block[1], stride);\
     name ## _idct_dc_add_c(dst+stride*4+0, block[2], stride);\
     name ## _idct_dc_add_c(dst+stride*4+4, block[3], stride);\
 }\
 \
 static void name ## _idct_dc_add4y_c(uint8_t *dst, int16_t block[4][16], ptrdiff_t stride)\
 {\
     name ## _idct_dc_add_c(dst+ 0, block[0], stride);\
     name ## _idct_dc_add_c(dst+ 4, block[1], stride);\
     name ## _idct_dc_add_c(dst+ 8, block[2], stride);\
     name ## _idct_dc_add_c(dst+12, block[3], stride);\
 }
 
 #if CONFIG_VP7_DECODER
b8664c92
 static void vp7_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
 {
     int i, a1, b1, c1, d1;
     int16_t tmp[16];
 
     for (i = 0; i < 4; i++) {
         a1 = (dc[i*4+0] + dc[i*4+2]) * 23170;
         b1 = (dc[i*4+0] - dc[i*4+2]) * 23170;
         c1 = dc[i*4+1] * 12540 - dc[i*4+3] * 30274;
         d1 = dc[i*4+1] * 30274 + dc[i*4+3] * 12540;
         tmp[i*4+0] = (a1 + d1) >> 14;
         tmp[i*4+3] = (a1 - d1) >> 14;
         tmp[i*4+1] = (b1 + c1) >> 14;
         tmp[i*4+2] = (b1 - c1) >> 14;
     }
 
     for (i = 0; i < 4; i++) {
         a1 = (tmp[i + 0] + tmp[i + 8]) * 23170;
         b1 = (tmp[i + 0] - tmp[i + 8]) * 23170;
         c1 = tmp[i + 4] * 12540 - tmp[i + 12] * 30274;
         d1 = tmp[i + 4] * 30274 + tmp[i + 12] * 12540;
bef6b27f
         AV_ZERO64(dc + i*4);
b8664c92
         block[0][i][0] = (a1 + d1 + 0x20000) >> 18;
         block[3][i][0] = (a1 - d1 + 0x20000) >> 18;
         block[1][i][0] = (b1 + c1 + 0x20000) >> 18;
         block[2][i][0] = (b1 - c1 + 0x20000) >> 18;
     }
 }
 
 static void vp7_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16])
 {
     int i, val = (23170 * (23170 * dc[0] >> 14) + 0x20000) >> 18;
     dc[0] = 0;
 
     for (i = 0; i < 4; i++) {
         block[i][0][0] = val;
         block[i][1][0] = val;
         block[i][2][0] = val;
         block[i][3][0] = val;
     }
 }
 
 static void vp7_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
 {
     int i, a1, b1, c1, d1;
     int16_t tmp[16];
 
     for (i = 0; i < 4; i++) {
         a1 = (block[i*4+0] + block[i*4+2]) * 23170;
         b1 = (block[i*4+0] - block[i*4+2]) * 23170;
         c1 = block[i*4+1] * 12540 - block[i*4+3] * 30274;
         d1 = block[i*4+1] * 30274 + block[i*4+3] * 12540;
bef6b27f
         AV_ZERO64(block + i*4);
b8664c92
         tmp[i*4+0] = (a1 + d1) >> 14;
         tmp[i*4+3] = (a1 - d1) >> 14;
         tmp[i*4+1] = (b1 + c1) >> 14;
         tmp[i*4+2] = (b1 - c1) >> 14;
     }
 
     for (i = 0; i < 4; i++) {
         a1 = (tmp[i + 0] + tmp[i + 8]) * 23170;
         b1 = (tmp[i + 0] - tmp[i + 8]) * 23170;
         c1 = tmp[i + 4] * 12540 - tmp[i + 12] * 30274;
         d1 = tmp[i + 4] * 30274 + tmp[i + 12] * 12540;
         dst[0*stride+i] = av_clip_uint8(dst[0*stride+i] + ((a1 + d1 + 0x20000) >> 18));
         dst[3*stride+i] = av_clip_uint8(dst[3*stride+i] + ((a1 - d1 + 0x20000) >> 18));
         dst[1*stride+i] = av_clip_uint8(dst[1*stride+i] + ((b1 + c1 + 0x20000) >> 18));
         dst[2*stride+i] = av_clip_uint8(dst[2*stride+i] + ((b1 - c1 + 0x20000) >> 18));
     }
 }
 
 static void vp7_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
 {
     int i, dc = (23170 * (23170 * block[0] >> 14) + 0x20000) >> 18;
     block[0] = 0;
 
     for (i = 0; i < 4; i++) {
         dst[0] = av_clip_uint8(dst[0] + dc);
         dst[1] = av_clip_uint8(dst[1] + dc);
         dst[2] = av_clip_uint8(dst[2] + dc);
         dst[3] = av_clip_uint8(dst[3] + dc);
         dst += stride;
     }
 }
 
911eb711
 MK_IDCT_DC_ADD4_C(vp7)
 #endif
 
3b636f21
 // TODO: Maybe add dequant
911eb711
 #if CONFIG_VP8_DECODER
88bd7fdc
 static void vp8_luma_dc_wht_c(int16_t block[4][4][16], int16_t dc[16])
3b636f21
 {
     int i, t0, t1, t2, t3;
 
     for (i = 0; i < 4; i++) {
         t0 = dc[0*4+i] + dc[3*4+i];
         t1 = dc[1*4+i] + dc[2*4+i];
         t2 = dc[1*4+i] - dc[2*4+i];
         t3 = dc[0*4+i] - dc[3*4+i];
 
         dc[0*4+i] = t0 + t1;
         dc[1*4+i] = t3 + t2;
         dc[2*4+i] = t0 - t1;
         dc[3*4+i] = t3 - t2;
     }
 
     for (i = 0; i < 4; i++) {
         t0 = dc[i*4+0] + dc[i*4+3] + 3; // rounding
         t1 = dc[i*4+1] + dc[i*4+2];
         t2 = dc[i*4+1] - dc[i*4+2];
         t3 = dc[i*4+0] - dc[i*4+3] + 3; // rounding
bef6b27f
         AV_ZERO64(dc + i*4);
3b636f21
 
f311208c
         block[i][0][0] = (t0 + t1) >> 3;
         block[i][1][0] = (t3 + t2) >> 3;
         block[i][2][0] = (t0 - t1) >> 3;
         block[i][3][0] = (t3 - t2) >> 3;
3b636f21
     }
 }
 
88bd7fdc
 static void vp8_luma_dc_wht_dc_c(int16_t block[4][4][16], int16_t dc[16])
f311208c
 {
     int i, val = (dc[0] + 3) >> 3;
     dc[0] = 0;
 
     for (i = 0; i < 4; i++) {
         block[i][0][0] = val;
         block[i][1][0] = val;
         block[i][2][0] = val;
         block[i][3][0] = val;
     }
 }
3b636f21
 
 #define MUL_20091(a) ((((a)*20091) >> 16) + (a))
 #define MUL_35468(a)  (((a)*35468) >> 16)
 
88bd7fdc
 static void vp8_idct_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
3b636f21
 {
     int i, t0, t1, t2, t3;
88bd7fdc
     int16_t tmp[16];
3b636f21
 
     for (i = 0; i < 4; i++) {
         t0 = block[0*4+i] + block[2*4+i];
         t1 = block[0*4+i] - block[2*4+i];
         t2 = MUL_35468(block[1*4+i]) - MUL_20091(block[3*4+i]);
         t3 = MUL_20091(block[1*4+i]) + MUL_35468(block[3*4+i]);
c25c7767
         block[0*4+i] = 0;
         block[1*4+i] = 0;
         block[2*4+i] = 0;
         block[3*4+i] = 0;
3b636f21
 
         tmp[i*4+0] = t0 + t3;
         tmp[i*4+1] = t1 + t2;
         tmp[i*4+2] = t1 - t2;
         tmp[i*4+3] = t0 - t3;
     }
 
     for (i = 0; i < 4; i++) {
         t0 = tmp[0*4+i] + tmp[2*4+i];
         t1 = tmp[0*4+i] - tmp[2*4+i];
         t2 = MUL_35468(tmp[1*4+i]) - MUL_20091(tmp[3*4+i]);
         t3 = MUL_20091(tmp[1*4+i]) + MUL_35468(tmp[3*4+i]);
 
c23acbae
         dst[0] = av_clip_uint8(dst[0] + ((t0 + t3 + 4) >> 3));
         dst[1] = av_clip_uint8(dst[1] + ((t1 + t2 + 4) >> 3));
         dst[2] = av_clip_uint8(dst[2] + ((t1 - t2 + 4) >> 3));
         dst[3] = av_clip_uint8(dst[3] + ((t0 - t3 + 4) >> 3));
3b636f21
         dst += stride;
     }
 }
 
88bd7fdc
 static void vp8_idct_dc_add_c(uint8_t *dst, int16_t block[16], ptrdiff_t stride)
3b636f21
 {
     int i, dc = (block[0] + 4) >> 3;
c25c7767
     block[0] = 0;
3b636f21
 
     for (i = 0; i < 4; i++) {
c23acbae
         dst[0] = av_clip_uint8(dst[0] + dc);
         dst[1] = av_clip_uint8(dst[1] + dc);
         dst[2] = av_clip_uint8(dst[2] + dc);
         dst[3] = av_clip_uint8(dst[3] + dc);
3b636f21
         dst += stride;
     }
 }
 
b8664c92
 MK_IDCT_DC_ADD4_C(vp8)
911eb711
 #endif
3b636f21
 
 // because I like only having two parameters to pass functions...
 #define LOAD_PIXELS\
     int av_unused p3 = p[-4*stride];\
     int av_unused p2 = p[-3*stride];\
     int av_unused p1 = p[-2*stride];\
     int av_unused p0 = p[-1*stride];\
     int av_unused q0 = p[ 0*stride];\
     int av_unused q1 = p[ 1*stride];\
     int av_unused q2 = p[ 2*stride];\
     int av_unused q3 = p[ 3*stride];
 
9942b6a1
 #define clip_int8(n) (cm[n+0x80]-0x80)
 
b8664c92
 static av_always_inline void filter_common(uint8_t *p, ptrdiff_t stride, int is4tap, int vpn)
3b636f21
 {
     LOAD_PIXELS
     int a, f1, f2;
9e0f14f1
     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
3b636f21
 
     a = 3*(q0 - p0);
 
     if (is4tap)
9942b6a1
         a += clip_int8(p1 - q1);
3b636f21
 
9942b6a1
     a = clip_int8(a);
3b636f21
 
     // We deviate from the spec here with c(a+3) >> 3
     // since that's what libvpx does.
     f1 = FFMIN(a+4, 127) >> 3;
b8664c92
 
     if (vpn == 7)
         f2 = f1 - ((a & 7) == 4);
     else
         f2 = FFMIN(a+3, 127) >> 3;
3b636f21
 
     // Despite what the spec says, we do need to clamp here to
     // be bitexact with libvpx.
9942b6a1
     p[-1*stride] = cm[p0 + f2];
     p[ 0*stride] = cm[q0 - f1];
3b636f21
 
     // only used for _inner on blocks without high edge variance
     if (!is4tap) {
         a = (f1+1)>>1;
9942b6a1
         p[-2*stride] = cm[p1 + a];
         p[ 1*stride] = cm[q1 - a];
3b636f21
     }
 }
 
b8664c92
 static av_always_inline int vp7_simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
 {
     LOAD_PIXELS
     return FFABS(p0-q0) <= flim;
 }
 
 static av_always_inline int vp8_simple_limit(uint8_t *p, ptrdiff_t stride, int flim)
3b636f21
 {
     LOAD_PIXELS
     return 2*FFABS(p0-q0) + (FFABS(p1-q1) >> 1) <= flim;
 }
 
 /**
  * E - limit at the macroblock edge
  * I - limit for interior difference
  */
b8664c92
 #define NORMAL_LIMIT(vpn) \
 static av_always_inline int vp ## vpn ## _normal_limit(uint8_t *p, ptrdiff_t stride, int E, int I)\
 {\
     LOAD_PIXELS\
     return vp ## vpn ## _simple_limit(p, stride, E)\
         && FFABS(p3-p2) <= I && FFABS(p2-p1) <= I && FFABS(p1-p0) <= I\
         && FFABS(q3-q2) <= I && FFABS(q2-q1) <= I && FFABS(q1-q0) <= I;\
3b636f21
 }
 
b8664c92
 NORMAL_LIMIT(7)
 NORMAL_LIMIT(8)
 
3b636f21
 // high edge variance
bd66f073
 static av_always_inline int hev(uint8_t *p, ptrdiff_t stride, int thresh)
3b636f21
 {
     LOAD_PIXELS
     return FFABS(p1-p0) > thresh || FFABS(q1-q0) > thresh;
 }
 
bd66f073
 static av_always_inline void filter_mbedge(uint8_t *p, ptrdiff_t stride)
3b636f21
 {
     int a0, a1, a2, w;
9e0f14f1
     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
3b636f21
 
     LOAD_PIXELS
 
9942b6a1
     w = clip_int8(p1-q1);
     w = clip_int8(w + 3*(q0-p0));
3b636f21
 
     a0 = (27*w + 63) >> 7;
     a1 = (18*w + 63) >> 7;
     a2 = ( 9*w + 63) >> 7;
 
9942b6a1
     p[-3*stride] = cm[p2 + a2];
     p[-2*stride] = cm[p1 + a1];
     p[-1*stride] = cm[p0 + a0];
     p[ 0*stride] = cm[q0 - a0];
     p[ 1*stride] = cm[q1 - a1];
     p[ 2*stride] = cm[q2 - a2];
3b636f21
 }
 
b8664c92
 #define LOOP_FILTER(vpn, dir, size, stridea, strideb, maybe_inline) \
 static maybe_inline void vp ## vpn ## _ ## dir ## _loop_filter ## size ## _c(uint8_t *dst, ptrdiff_t stride,\
3b636f21
                                      int flim_E, int flim_I, int hev_thresh)\
 {\
     int i;\
 \
     for (i = 0; i < size; i++)\
b8664c92
         if (vp ## vpn ## _normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
3b636f21
             if (hev(dst+i*stridea, strideb, hev_thresh))\
b8664c92
                 filter_common(dst+i*stridea, strideb, 1, vpn);\
3b636f21
             else\
                 filter_mbedge(dst+i*stridea, strideb);\
         }\
 }\
 \
b8664c92
 static maybe_inline void vp ## vpn ## _ ## dir ## _loop_filter ## size ## _inner_c(uint8_t *dst, ptrdiff_t stride,\
3b636f21
                                       int flim_E, int flim_I, int hev_thresh)\
 {\
95275094
     int i;\
3b636f21
 \
     for (i = 0; i < size; i++)\
b8664c92
         if (vp ## vpn ## _normal_limit(dst+i*stridea, strideb, flim_E, flim_I)) {\
95275094
             int hv = hev(dst+i*stridea, strideb, hev_thresh);\
             if (hv) \
b8664c92
                 filter_common(dst+i*stridea, strideb, 1, vpn);\
95275094
             else \
b8664c92
                 filter_common(dst+i*stridea, strideb, 0, vpn);\
3b636f21
         }\
 }
 
b8664c92
 #define UV_LOOP_FILTER(vpn, dir, stridea, strideb) \
 LOOP_FILTER(vpn, dir, 8, stridea, strideb, av_always_inline) \
 static void vp ## vpn ## _ ## dir ## _loop_filter8uv_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
3facfc99
                                       int fE, int fI, int hev_thresh)\
 {\
b8664c92
   vp ## vpn ## _ ## dir ## _loop_filter8_c(dstU, stride, fE, fI, hev_thresh);\
   vp ## vpn ## _ ## dir ## _loop_filter8_c(dstV, stride, fE, fI, hev_thresh);\
3facfc99
 }\
b8664c92
 static void vp ## vpn ## _ ## dir ## _loop_filter8uv_inner_c(uint8_t *dstU, uint8_t *dstV, ptrdiff_t stride,\
3facfc99
                                       int fE, int fI, int hev_thresh)\
 {\
b8664c92
   vp ## vpn ## _ ## dir ## _loop_filter8_inner_c(dstU, stride, fE, fI, hev_thresh);\
   vp ## vpn ## _ ## dir ## _loop_filter8_inner_c(dstV, stride, fE, fI, hev_thresh);\
3facfc99
 }
 
b8664c92
 #define LOOP_FILTER_SIMPLE(vpn) \
 static void vp ## vpn ## _v_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)\
 {\
     int i;\
 \
     for (i = 0; i < 16; i++)\
         if (vp ## vpn ## _simple_limit(dst+i, stride, flim))\
             filter_common(dst+i, stride, 1, vpn);\
 }\
 \
 static void vp ## vpn ## _h_loop_filter_simple_c(uint8_t *dst, ptrdiff_t stride, int flim)\
 {\
     int i;\
 \
     for (i = 0; i < 16; i++)\
         if (vp ## vpn ## _simple_limit(dst+i*stride, 1, flim))\
             filter_common(dst+i*stride, 1, 1, vpn);\
3b636f21
 }
 
911eb711
 #if CONFIG_VP7_DECODER
 LOOP_FILTER(7, v, 16, 1, stride,)
 LOOP_FILTER(7, h, 16, stride, 1,)
 UV_LOOP_FILTER(7, v, 1, stride)
 UV_LOOP_FILTER(7, h, stride, 1)
b8664c92
 LOOP_FILTER_SIMPLE(7)
911eb711
 #endif
 
 #if CONFIG_VP8_DECODER
 LOOP_FILTER(8, v, 16, 1, stride,)
 LOOP_FILTER(8, h, 16, stride, 1,)
 UV_LOOP_FILTER(8, v, 1, stride)
 UV_LOOP_FILTER(8, h, stride, 1)
b8664c92
 LOOP_FILTER_SIMPLE(8)
911eb711
 #endif
3b636f21
 
 static const uint8_t subpel_filters[7][6] = {
     { 0,   6, 123,  12,   1,   0 },
     { 2,  11, 108,  36,   8,   1 },
     { 0,   9,  93,  50,   6,   0 },
     { 3,  16,  77,  77,  16,   3 },
     { 0,   6,  50,  93,   9,   0 },
     { 1,   8,  36, 108,  11,   2 },
     { 0,   1,  12, 123,   6,   0 },
 };
 
d6f8476b
 #define PUT_PIXELS(WIDTH) \
bd66f073
 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) { \
15d31aa1
     int i; \
     for (i = 0; i < h; i++, dst+= dststride, src+= srcstride) { \
d6f8476b
         memcpy(dst, src, WIDTH); \
     } \
 }
 
 PUT_PIXELS(16)
 PUT_PIXELS(8)
 PUT_PIXELS(4)
3b636f21
 
 #define FILTER_6TAP(src, F, stride) \
9942b6a1
     cm[(F[2]*src[x+0*stride] - F[1]*src[x-1*stride] + F[0]*src[x-2*stride] + \
         F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + F[5]*src[x+3*stride] + 64) >> 7]
3b636f21
 
 #define FILTER_4TAP(src, F, stride) \
9942b6a1
     cm[(F[2]*src[x+0*stride] - F[1]*src[x-1*stride] + \
         F[3]*src[x+1*stride] - F[4]*src[x+2*stride] + 64) >> 7]
3b636f21
 
e3c53954
 #define VP8_EPEL_H(SIZE, TAPS) \
bd66f073
 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) \
3b636f21
 { \
     const uint8_t *filter = subpel_filters[mx-1]; \
9e0f14f1
     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
3b636f21
     int x, y; \
 \
     for (y = 0; y < h; y++) { \
         for (x = 0; x < SIZE; x++) \
e3c53954
             dst[x] = FILTER_ ## TAPS ## TAP(src, filter, 1); \
d6f8476b
         dst += dststride; \
         src += srcstride; \
3b636f21
     } \
 }
e3c53954
 #define VP8_EPEL_V(SIZE, TAPS) \
bd66f073
 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) \
3b636f21
 { \
     const uint8_t *filter = subpel_filters[my-1]; \
9e0f14f1
     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
3b636f21
     int x, y; \
 \
     for (y = 0; y < h; y++) { \
         for (x = 0; x < SIZE; x++) \
e3c53954
             dst[x] = FILTER_ ## TAPS ## TAP(src, filter, srcstride); \
d6f8476b
         dst += dststride; \
         src += srcstride; \
3b636f21
     } \
 }
e3c53954
 #define VP8_EPEL_HV(SIZE, HTAPS, VTAPS) \
bd66f073
 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) \
3b636f21
 { \
     const uint8_t *filter = subpel_filters[mx-1]; \
9e0f14f1
     const uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; \
3b636f21
     int x, y; \
e3c53954
     uint8_t tmp_array[(2*SIZE+VTAPS-1)*SIZE]; \
3b636f21
     uint8_t *tmp = tmp_array; \
e3c53954
     src -= (2-(VTAPS==4))*srcstride; \
3b636f21
 \
e3c53954
     for (y = 0; y < h+VTAPS-1; y++) { \
3b636f21
         for (x = 0; x < SIZE; x++) \
e3c53954
             tmp[x] = FILTER_ ## HTAPS ## TAP(src, filter, 1); \
3b636f21
         tmp += SIZE; \
d6f8476b
         src += srcstride; \
3b636f21
     } \
 \
e3c53954
     tmp = tmp_array + (2-(VTAPS==4))*SIZE; \
3b636f21
     filter = subpel_filters[my-1]; \
 \
     for (y = 0; y < h; y++) { \
         for (x = 0; x < SIZE; x++) \
e3c53954
             dst[x] = FILTER_ ## VTAPS ## TAP(tmp, filter, SIZE); \
d6f8476b
         dst += dststride; \
3b636f21
         tmp += SIZE; \
     } \
 }
 
e3c53954
 VP8_EPEL_H(16, 4)
 VP8_EPEL_H(8,  4)
 VP8_EPEL_H(4,  4)
 VP8_EPEL_H(16, 6)
 VP8_EPEL_H(8,  6)
 VP8_EPEL_H(4,  6)
 VP8_EPEL_V(16, 4)
 VP8_EPEL_V(8,  4)
 VP8_EPEL_V(4,  4)
 VP8_EPEL_V(16, 6)
 VP8_EPEL_V(8,  6)
 VP8_EPEL_V(4,  6)
 VP8_EPEL_HV(16, 4, 4)
 VP8_EPEL_HV(8,  4, 4)
 VP8_EPEL_HV(4,  4, 4)
 VP8_EPEL_HV(16, 4, 6)
 VP8_EPEL_HV(8,  4, 6)
 VP8_EPEL_HV(4,  4, 6)
 VP8_EPEL_HV(16, 6, 4)
 VP8_EPEL_HV(8,  6, 4)
 VP8_EPEL_HV(4,  6, 4)
 VP8_EPEL_HV(16, 6, 6)
 VP8_EPEL_HV(8,  6, 6)
 VP8_EPEL_HV(4,  6, 6)
3b636f21
 
0ef1dbed
 #define VP8_BILINEAR(SIZE) \
34b429d5
 static void put_vp8_bilinear ## SIZE ## _h_c(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \
0ef1dbed
 { \
     int a = 8-mx, b = mx; \
     int x, y; \
 \
     for (y = 0; y < h; y++) { \
         for (x = 0; x < SIZE; x++) \
             dst[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
34b429d5
         dst += dstride; \
         src += sstride; \
0ef1dbed
     } \
 } \
34b429d5
 static void put_vp8_bilinear ## SIZE ## _v_c(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \
0ef1dbed
 { \
     int c = 8-my, d = my; \
     int x, y; \
 \
     for (y = 0; y < h; y++) { \
         for (x = 0; x < SIZE; x++) \
34b429d5
             dst[x] = (c*src[x] + d*src[x+sstride] + 4) >> 3; \
         dst += dstride; \
         src += sstride; \
0ef1dbed
     } \
 } \
 \
34b429d5
 static void put_vp8_bilinear ## SIZE ## _hv_c(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) \
0ef1dbed
 { \
     int a = 8-mx, b = mx; \
     int c = 8-my, d = my; \
     int x, y; \
     uint8_t tmp_array[(2*SIZE+1)*SIZE]; \
     uint8_t *tmp = tmp_array; \
 \
     for (y = 0; y < h+1; y++) { \
         for (x = 0; x < SIZE; x++) \
             tmp[x] = (a*src[x] + b*src[x+1] + 4) >> 3; \
         tmp += SIZE; \
34b429d5
         src += sstride; \
0ef1dbed
     } \
 \
     tmp = tmp_array; \
 \
     for (y = 0; y < h; y++) { \
         for (x = 0; x < SIZE; x++) \
             dst[x] = (c*tmp[x] + d*tmp[x+SIZE] + 4) >> 3; \
34b429d5
         dst += dstride; \
0ef1dbed
         tmp += SIZE; \
     } \
 }
 
 VP8_BILINEAR(16)
 VP8_BILINEAR(8)
 VP8_BILINEAR(4)
 
3b636f21
 #define VP8_MC_FUNC(IDX, SIZE) \
d6f8476b
     dsp->put_vp8_epel_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
3b636f21
     dsp->put_vp8_epel_pixels_tab[IDX][0][1] = put_vp8_epel ## SIZE ## _h4_c; \
     dsp->put_vp8_epel_pixels_tab[IDX][0][2] = put_vp8_epel ## SIZE ## _h6_c; \
     dsp->put_vp8_epel_pixels_tab[IDX][1][0] = put_vp8_epel ## SIZE ## _v4_c; \
     dsp->put_vp8_epel_pixels_tab[IDX][1][1] = put_vp8_epel ## SIZE ## _h4v4_c; \
     dsp->put_vp8_epel_pixels_tab[IDX][1][2] = put_vp8_epel ## SIZE ## _h6v4_c; \
     dsp->put_vp8_epel_pixels_tab[IDX][2][0] = put_vp8_epel ## SIZE ## _v6_c; \
     dsp->put_vp8_epel_pixels_tab[IDX][2][1] = put_vp8_epel ## SIZE ## _h4v6_c; \
     dsp->put_vp8_epel_pixels_tab[IDX][2][2] = put_vp8_epel ## SIZE ## _h6v6_c
 
0ef1dbed
 #define VP8_BILINEAR_MC_FUNC(IDX, SIZE) \
     dsp->put_vp8_bilinear_pixels_tab[IDX][0][0] = put_vp8_pixels ## SIZE ## _c; \
     dsp->put_vp8_bilinear_pixels_tab[IDX][0][1] = put_vp8_bilinear ## SIZE ## _h_c; \
     dsp->put_vp8_bilinear_pixels_tab[IDX][0][2] = put_vp8_bilinear ## SIZE ## _h_c; \
     dsp->put_vp8_bilinear_pixels_tab[IDX][1][0] = put_vp8_bilinear ## SIZE ## _v_c; \
     dsp->put_vp8_bilinear_pixels_tab[IDX][1][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
     dsp->put_vp8_bilinear_pixels_tab[IDX][1][2] = put_vp8_bilinear ## SIZE ## _hv_c; \
     dsp->put_vp8_bilinear_pixels_tab[IDX][2][0] = put_vp8_bilinear ## SIZE ## _v_c; \
     dsp->put_vp8_bilinear_pixels_tab[IDX][2][1] = put_vp8_bilinear ## SIZE ## _hv_c; \
     dsp->put_vp8_bilinear_pixels_tab[IDX][2][2] = put_vp8_bilinear ## SIZE ## _hv_c
 
b8664c92
 av_cold void ff_vp8dsp_init(VP8DSPContext *dsp, int vp7)
3b636f21
 {
911eb711
 #if CONFIG_VP7_DECODER && CONFIG_VP8_DECODER
 #define VPX(f) vp7 ? vp7_ ## f : vp8_ ## f
 #elif CONFIG_VP7_DECODER
 #define VPX(f) vp7_ ## f
 #else // CONFIG_VP8_DECODER
 #define VPX(f) vp8_ ## f
 #endif
b8664c92
 
     dsp->vp8_luma_dc_wht    = VPX(luma_dc_wht_c);
     dsp->vp8_luma_dc_wht_dc = VPX(luma_dc_wht_dc_c);
     dsp->vp8_idct_add       = VPX(idct_add_c);
     dsp->vp8_idct_dc_add    = VPX(idct_dc_add_c);
     dsp->vp8_idct_dc_add4y  = VPX(idct_dc_add4y_c);
     dsp->vp8_idct_dc_add4uv = VPX(idct_dc_add4uv_c);
 
     dsp->vp8_v_loop_filter16y = VPX(v_loop_filter16_c);
     dsp->vp8_h_loop_filter16y = VPX(h_loop_filter16_c);
     dsp->vp8_v_loop_filter8uv = VPX(v_loop_filter8uv_c);
     dsp->vp8_h_loop_filter8uv = VPX(h_loop_filter8uv_c);
 
     dsp->vp8_v_loop_filter16y_inner = VPX(v_loop_filter16_inner_c);
     dsp->vp8_h_loop_filter16y_inner = VPX(h_loop_filter16_inner_c);
     dsp->vp8_v_loop_filter8uv_inner = VPX(v_loop_filter8uv_inner_c);
     dsp->vp8_h_loop_filter8uv_inner = VPX(h_loop_filter8uv_inner_c);
 
     dsp->vp8_v_loop_filter_simple = VPX(v_loop_filter_simple_c);
     dsp->vp8_h_loop_filter_simple = VPX(h_loop_filter_simple_c);
3b636f21
 
     VP8_MC_FUNC(0, 16);
     VP8_MC_FUNC(1, 8);
     VP8_MC_FUNC(2, 4);
0ef1dbed
 
     VP8_BILINEAR_MC_FUNC(0, 16);
     VP8_BILINEAR_MC_FUNC(1, 8);
     VP8_BILINEAR_MC_FUNC(2, 4);
0178d14f
 
ef15d71c
     if (ARCH_ARM)
b8664c92
         ff_vp8dsp_init_arm(dsp, vp7);
38282149
     if (ARCH_PPC)
         ff_vp8dsp_init_ppc(dsp);
67e6a9f5
     if (ARCH_X86)
b8664c92
         ff_vp8dsp_init_x86(dsp, vp7);
3b636f21
 }