falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
43714 Commits
32 Branches
824MB
Tree: 6c180b35c4
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '6c180b35c4'
${ noResults }
FFmpeg/libavcodec/rv40dsp.c

612 lines
26KB
Raw Blame History 

  1. /*

  2.  * RV40 decoder motion compensation functions

  3.  * Copyright (c) 2008 Konstantin Shishkov

  4.  *

  5.  * This file is part of FFmpeg.

  6.  *

  7.  * FFmpeg is free software; you can redistribute it and/or

  8.  * modify it under the terms of the GNU Lesser General Public

  9.  * License as published by the Free Software Foundation; either

  10.  * version 2.1 of the License, or (at your option) any later version.

  11.  *

  12.  * FFmpeg is distributed in the hope that it will be useful,

  13.  * but WITHOUT ANY WARRANTY; without even the implied warranty of

  14.  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU

  15.  * Lesser General Public License for more details.

  16.  *

  17.  * You should have received a copy of the GNU Lesser General Public

  18.  * License along with FFmpeg; if not, write to the Free Software

  19.  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

  20.  */

  21. 

  22. /**

  23.  * @file

  24.  * RV40 decoder motion compensation functions

  25.  */

  26. 

  27. #include "avcodec.h"

  28. #include "dsputil.h"

  29. #include "rv34dsp.h"

  30. #include "libavutil/avassert.h"

  31. #include "libavutil/common.h"

  32. 

  33. #define RV40_LOWPASS(OPNAME, OP) \

  34. static av_unused void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\

  35.                                                      const int h, const int C1, const int C2, const int SHIFT){\

  36.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  37.     int i;\

  38.     for(i = 0; i < h; i++)\

  39.     {\

  40.         OP(dst[0], (src[-2] + src[ 3] - 5*(src[-1]+src[2]) + src[0]*C1 + src[1]*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  41.         OP(dst[1], (src[-1] + src[ 4] - 5*(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  42.         OP(dst[2], (src[ 0] + src[ 5] - 5*(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  43.         OP(dst[3], (src[ 1] + src[ 6] - 5*(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  44.         OP(dst[4], (src[ 2] + src[ 7] - 5*(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  45.         OP(dst[5], (src[ 3] + src[ 8] - 5*(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  46.         OP(dst[6], (src[ 4] + src[ 9] - 5*(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  47.         OP(dst[7], (src[ 5] + src[10] - 5*(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  48.         dst += dstStride;\

  49.         src += srcStride;\

  50.     }\

  51. }\

  52. \

  53. static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\

  54.                                            const int w, const int C1, const int C2, const int SHIFT){\

  55.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  56.     int i;\

  57.     for(i = 0; i < w; i++)\

  58.     {\

  59.         const int srcB  = src[-2*srcStride];\

  60.         const int srcA  = src[-1*srcStride];\

  61.         const int src0  = src[0 *srcStride];\

  62.         const int src1  = src[1 *srcStride];\

  63.         const int src2  = src[2 *srcStride];\

  64.         const int src3  = src[3 *srcStride];\

  65.         const int src4  = src[4 *srcStride];\

  66.         const int src5  = src[5 *srcStride];\

  67.         const int src6  = src[6 *srcStride];\

  68.         const int src7  = src[7 *srcStride];\

  69.         const int src8  = src[8 *srcStride];\

  70.         const int src9  = src[9 *srcStride];\

  71.         const int src10 = src[10*srcStride];\

  72.         OP(dst[0*dstStride], (srcB + src3  - 5*(srcA+src2) + src0*C1 + src1*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  73.         OP(dst[1*dstStride], (srcA + src4  - 5*(src0+src3) + src1*C1 + src2*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  74.         OP(dst[2*dstStride], (src0 + src5  - 5*(src1+src4) + src2*C1 + src3*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  75.         OP(dst[3*dstStride], (src1 + src6  - 5*(src2+src5) + src3*C1 + src4*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  76.         OP(dst[4*dstStride], (src2 + src7  - 5*(src3+src6) + src4*C1 + src5*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  77.         OP(dst[5*dstStride], (src3 + src8  - 5*(src4+src7) + src5*C1 + src6*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  78.         OP(dst[6*dstStride], (src4 + src9  - 5*(src5+src8) + src6*C1 + src7*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  79.         OP(dst[7*dstStride], (src5 + src10 - 5*(src6+src9) + src7*C1 + src8*C2 + (1<<(SHIFT-1))) >> SHIFT);\

  80.         dst++;\

  81.         src++;\

  82.     }\

  83. }\

  84. \

  85. static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\

  86.                                             const int w, const int C1, const int C2, const int SHIFT){\

  87.     OPNAME ## rv40_qpel8_v_lowpass(dst  , src  , dstStride, srcStride, 8, C1, C2, SHIFT);\

  88.     OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\

  89.     src += 8*srcStride;\

  90.     dst += 8*dstStride;\

  91.     OPNAME ## rv40_qpel8_v_lowpass(dst  , src  , dstStride, srcStride, w-8, C1, C2, SHIFT);\

  92.     OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, w-8, C1, C2, SHIFT);\

  93. }\

  94. \

  95. static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\

  96.                                             const int h, const int C1, const int C2, const int SHIFT){\

  97.     OPNAME ## rv40_qpel8_h_lowpass(dst  , src  , dstStride, srcStride, 8, C1, C2, SHIFT);\

  98.     OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\

  99.     src += 8*srcStride;\

  100.     dst += 8*dstStride;\

  101.     OPNAME ## rv40_qpel8_h_lowpass(dst  , src  , dstStride, srcStride, h-8, C1, C2, SHIFT);\

  102.     OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, h-8, C1, C2, SHIFT);\

  103. }\

  104. \

  105. 

  106. #define RV40_MC(OPNAME, SIZE) \

  107. static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\

  108.     OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\

  109. }\

  110. \

  111. static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\

  112.     OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\

  113. }\

  114. \

  115. static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\

  116.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\

  117. }\

  118. \

  119. static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\

  120.     uint8_t full[SIZE*(SIZE+5)];\

  121.     uint8_t * const full_mid = full + SIZE*2;\

  122.     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\

  123.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\

  124. }\

  125. \

  126. static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\

  127.     uint8_t full[SIZE*(SIZE+5)];\

  128.     uint8_t * const full_mid = full + SIZE*2;\

  129.     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\

  130.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\

  131. }\

  132. \

  133. static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\

  134.     uint8_t full[SIZE*(SIZE+5)];\

  135.     uint8_t * const full_mid = full + SIZE*2;\

  136.     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\

  137.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\

  138. }\

  139. \

  140. static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\

  141.     uint8_t full[SIZE*(SIZE+5)];\

  142.     uint8_t * const full_mid = full + SIZE*2;\

  143.     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\

  144.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\

  145. }\

  146. \

  147. static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\

  148.     uint8_t full[SIZE*(SIZE+5)];\

  149.     uint8_t * const full_mid = full + SIZE*2;\

  150.     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\

  151.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\

  152. }\

  153. \

  154. static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\

  155.     uint8_t full[SIZE*(SIZE+5)];\

  156.     uint8_t * const full_mid = full + SIZE*2;\

  157.     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\

  158.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\

  159. }\

  160. \

  161. static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\

  162.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\

  163. }\

  164. \

  165. static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\

  166.     uint8_t full[SIZE*(SIZE+5)];\

  167.     uint8_t * const full_mid = full + SIZE*2;\

  168.     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\

  169.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\

  170. }\

  171. \

  172. static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\

  173.     uint8_t full[SIZE*(SIZE+5)];\

  174.     uint8_t * const full_mid = full + SIZE*2;\

  175.     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\

  176.     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\

  177. }\

  178. \

  179. 

  180. #define op_avg(a, b)  a = (((a)+cm[b]+1)>>1)

  181. #define op_put(a, b)  a = cm[b]

  182. 

  183. RV40_LOWPASS(put_       , op_put)

  184. RV40_LOWPASS(avg_       , op_avg)

  185. 

  186. #undef op_avg

  187. #undef op_put

  188. 

  189. RV40_MC(put_, 8)

  190. RV40_MC(put_, 16)

  191. RV40_MC(avg_, 8)

  192. RV40_MC(avg_, 16)

  193. 

  194. static const int rv40_bias[4][4] = {

  195.     {  0, 16, 32, 16 },

  196.     { 32, 28, 32, 28 },

  197.     {  0, 32, 16, 32 },

  198.     { 32, 28, 32, 28 }

  199. };

  200. 

  201. #define RV40_CHROMA_MC(OPNAME, OP)\

  202. static void OPNAME ## rv40_chroma_mc4_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\

  203.     const int A = (8-x) * (8-y);\

  204.     const int B = (  x) * (8-y);\

  205.     const int C = (8-x) * (  y);\

  206.     const int D = (  x) * (  y);\

  207.     int i;\

  208.     int bias = rv40_bias[y>>1][x>>1];\

  209.     \

  210.     av_assert2(x<8 && y<8 && x>=0 && y>=0);\

  211. \

  212.     if(D){\

  213.         for(i = 0; i < h; i++){\

  214.             OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\

  215.             OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\

  216.             OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\

  217.             OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\

  218.             dst += stride;\

  219.             src += stride;\

  220.         }\

  221.     }else{\

  222.         const int E = B + C;\

  223.         const int step = C ? stride : 1;\

  224.         for(i = 0; i < h; i++){\

  225.             OP(dst[0], (A*src[0] + E*src[step+0] + bias));\

  226.             OP(dst[1], (A*src[1] + E*src[step+1] + bias));\

  227.             OP(dst[2], (A*src[2] + E*src[step+2] + bias));\

  228.             OP(dst[3], (A*src[3] + E*src[step+3] + bias));\

  229.             dst += stride;\

  230.             src += stride;\

  231.         }\

  232.     }\

  233. }\

  234. \

  235. static void OPNAME ## rv40_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\

  236.     const int A = (8-x) * (8-y);\

  237.     const int B = (  x) * (8-y);\

  238.     const int C = (8-x) * (  y);\

  239.     const int D = (  x) * (  y);\

  240.     int i;\

  241.     int bias = rv40_bias[y>>1][x>>1];\

  242.     \

  243.     av_assert2(x<8 && y<8 && x>=0 && y>=0);\

  244. \

  245.     if(D){\

  246.         for(i = 0; i < h; i++){\

  247.             OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\

  248.             OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\

  249.             OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\

  250.             OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\

  251.             OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + bias));\

  252.             OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + bias));\

  253.             OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + bias));\

  254.             OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + bias));\

  255.             dst += stride;\

  256.             src += stride;\

  257.         }\

  258.     }else{\

  259.         const int E = B + C;\

  260.         const int step = C ? stride : 1;\

  261.         for(i = 0; i < h; i++){\

  262.             OP(dst[0], (A*src[0] + E*src[step+0] + bias));\

  263.             OP(dst[1], (A*src[1] + E*src[step+1] + bias));\

  264.             OP(dst[2], (A*src[2] + E*src[step+2] + bias));\

  265.             OP(dst[3], (A*src[3] + E*src[step+3] + bias));\

  266.             OP(dst[4], (A*src[4] + E*src[step+4] + bias));\

  267.             OP(dst[5], (A*src[5] + E*src[step+5] + bias));\

  268.             OP(dst[6], (A*src[6] + E*src[step+6] + bias));\

  269.             OP(dst[7], (A*src[7] + E*src[step+7] + bias));\

  270.             dst += stride;\

  271.             src += stride;\

  272.         }\

  273.     }\

  274. }

  275. 

  276. #define op_avg(a, b) a = (((a)+((b)>>6)+1)>>1)

  277. #define op_put(a, b) a = ((b)>>6)

  278. 

  279. RV40_CHROMA_MC(put_, op_put)

  280. RV40_CHROMA_MC(avg_, op_avg)

  281. 

  282. #define RV40_WEIGHT_FUNC(size) \

  283. static void rv40_weight_func_rnd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\

  284. {\

  285.     int i, j;\

  286. \

  287.     for (j = 0; j < size; j++) {\

  288.         for (i = 0; i < size; i++)\

  289.             dst[i] = (((w2 * src1[i]) >> 9) + ((w1 * src2[i]) >> 9) + 0x10) >> 5;\

  290.         src1 += stride;\

  291.         src2 += stride;\

  292.         dst  += stride;\

  293.     }\

  294. }\

  295. static void rv40_weight_func_nornd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\

  296. {\

  297.     int i, j;\

  298. \

  299.     for (j = 0; j < size; j++) {\

  300.         for (i = 0; i < size; i++)\

  301.             dst[i] = (w2 * src1[i] + w1 * src2[i] + 0x10) >> 5;\

  302.         src1 += stride;\

  303.         src2 += stride;\

  304.         dst  += stride;\

  305.     }\

  306. }

  307. 

  308. RV40_WEIGHT_FUNC(16)

  309. RV40_WEIGHT_FUNC(8)

  310. 

  311. /**

  312.  * dither values for deblocking filter - left/top values

  313.  */

  314. static const uint8_t rv40_dither_l[16] = {

  315.     0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30,

  316.     0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40

  317. };

  318. 

  319. /**

  320.  * dither values for deblocking filter - right/bottom values

  321.  */

  322. static const uint8_t rv40_dither_r[16] = {

  323.     0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40,

  324.     0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40

  325. };

  326. 

  327. #define CLIP_SYMM(a, b) av_clip(a, -(b), b)

  328. /**

  329.  * weaker deblocking very similar to the one described in 4.4.2 of JVT-A003r1

  330.  */

  331. static av_always_inline void rv40_weak_loop_filter(uint8_t *src,

  332.                                                    const int step,

  333.                                                    const ptrdiff_t stride,

  334.                                                    const int filter_p1,

  335.                                                    const int filter_q1,

  336.                                                    const int alpha,

  337.                                                    const int beta,

  338.                                                    const int lim_p0q0,

  339.                                                    const int lim_q1,

  340.                                                    const int lim_p1)

  341. {

  342.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  343.     int i, t, u, diff;

  344. 

  345.     for (i = 0; i < 4; i++, src += stride) {

  346.         int diff_p1p0 = src[-2*step] - src[-1*step];

  347.         int diff_q1q0 = src[ 1*step] - src[ 0*step];

  348.         int diff_p1p2 = src[-2*step] - src[-3*step];

  349.         int diff_q1q2 = src[ 1*step] - src[ 2*step];

  350. 

  351.         t = src[0*step] - src[-1*step];

  352.         if (!t)

  353.             continue;

  354. 

  355.         u = (alpha * FFABS(t)) >> 7;

  356.         if (u > 3 - (filter_p1 && filter_q1))

  357.             continue;

  358. 

  359.         t <<= 2;

  360.         if (filter_p1 && filter_q1)

  361.             t += src[-2*step] - src[1*step];

  362. 

  363.         diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0);

  364.         src[-1*step] = cm[src[-1*step] + diff];

  365.         src[ 0*step] = cm[src[ 0*step] - diff];

  366. 

  367.         if (filter_p1 && FFABS(diff_p1p2) <= beta) {

  368.             t = (diff_p1p0 + diff_p1p2 - diff) >> 1;

  369.             src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)];

  370.         }

  371. 

  372.         if (filter_q1 && FFABS(diff_q1q2) <= beta) {

  373.             t = (diff_q1q0 + diff_q1q2 + diff) >> 1;

  374.             src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)];

  375.         }

  376.     }

  377. }

  378. 

  379. static void rv40_h_weak_loop_filter(uint8_t *src, const ptrdiff_t stride,

  380.                                     const int filter_p1, const int filter_q1,

  381.                                     const int alpha, const int beta,

  382.                                     const int lim_p0q0, const int lim_q1,

  383.                                     const int lim_p1)

  384. {

  385.     rv40_weak_loop_filter(src, stride, 1, filter_p1, filter_q1,

  386.                           alpha, beta, lim_p0q0, lim_q1, lim_p1);

  387. }

  388. 

  389. static void rv40_v_weak_loop_filter(uint8_t *src, const ptrdiff_t stride,

  390.                                     const int filter_p1, const int filter_q1,

  391.                                     const int alpha, const int beta,

  392.                                     const int lim_p0q0, const int lim_q1,

  393.                                     const int lim_p1)

  394. {

  395.     rv40_weak_loop_filter(src, 1, stride, filter_p1, filter_q1,

  396.                           alpha, beta, lim_p0q0, lim_q1, lim_p1);

  397. }

  398. 

  399. static av_always_inline void rv40_strong_loop_filter(uint8_t *src,

  400.                                                      const int step,

  401.                                                      const ptrdiff_t stride,

  402.                                                      const int alpha,

  403.                                                      const int lims,

  404.                                                      const int dmode,

  405.                                                      const int chroma)

  406. {

  407.     int i;

  408. 

  409.     for(i = 0; i < 4; i++, src += stride){

  410.         int sflag, p0, q0, p1, q1;

  411.         int t = src[0*step] - src[-1*step];

  412. 

  413.         if (!t)

  414.             continue;

  415. 

  416.         sflag = (alpha * FFABS(t)) >> 7;

  417.         if (sflag > 1)

  418.             continue;

  419. 

  420.         p0 = (25*src[-3*step] + 26*src[-2*step] + 26*src[-1*step] +

  421.               26*src[ 0*step] + 25*src[ 1*step] +

  422.               rv40_dither_l[dmode + i]) >> 7;

  423. 

  424.         q0 = (25*src[-2*step] + 26*src[-1*step] + 26*src[ 0*step] +

  425.               26*src[ 1*step] + 25*src[ 2*step] +

  426.               rv40_dither_r[dmode + i]) >> 7;

  427. 

  428.         if (sflag) {

  429.             p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims);

  430.             q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims);

  431.         }

  432. 

  433.         p1 = (25*src[-4*step] + 26*src[-3*step] + 26*src[-2*step] + 26*p0 +

  434.               25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;

  435.         q1 = (25*src[-1*step] + 26*q0 + 26*src[ 1*step] + 26*src[ 2*step] +

  436.               25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;

  437. 

  438.         if (sflag) {

  439.             p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims);

  440.             q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims);

  441.         }

  442. 

  443.         src[-2*step] = p1;

  444.         src[-1*step] = p0;

  445.         src[ 0*step] = q0;

  446.         src[ 1*step] = q1;

  447. 

  448.         if(!chroma){

  449.             src[-3*step] = (25*src[-1*step] + 26*src[-2*step] +

  450.                             51*src[-3*step] + 26*src[-4*step] + 64) >> 7;

  451.             src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] +

  452.                             51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;

  453.         }

  454.     }

  455. }

  456. 

  457. static void rv40_h_strong_loop_filter(uint8_t *src, const ptrdiff_t stride,

  458.                                       const int alpha, const int lims,

  459.                                       const int dmode, const int chroma)

  460. {

  461.     rv40_strong_loop_filter(src, stride, 1, alpha, lims, dmode, chroma);

  462. }

  463. 

  464. static void rv40_v_strong_loop_filter(uint8_t *src, const ptrdiff_t stride,

  465.                                       const int alpha, const int lims,

  466.                                       const int dmode, const int chroma)

  467. {

  468.     rv40_strong_loop_filter(src, 1, stride, alpha, lims, dmode, chroma);

  469. }

  470. 

  471. static av_always_inline int rv40_loop_filter_strength(uint8_t *src,

  472.                                                       int step, ptrdiff_t stride,

  473.                                                       int beta, int beta2,

  474.                                                       int edge,

  475.                                                       int *p1, int *q1)

  476. {

  477.     int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0;

  478.     int strong0 = 0, strong1 = 0;

  479.     uint8_t *ptr;

  480.     int i;

  481. 

  482.     for (i = 0, ptr = src; i < 4; i++, ptr += stride) {

  483.         sum_p1p0 += ptr[-2*step] - ptr[-1*step];

  484.         sum_q1q0 += ptr[ 1*step] - ptr[ 0*step];

  485.     }

  486. 

  487.     *p1 = FFABS(sum_p1p0) < (beta << 2);

  488.     *q1 = FFABS(sum_q1q0) < (beta << 2);

  489. 

  490.     if(!*p1 && !*q1)

  491.         return 0;

  492. 

  493.     if (!edge)

  494.         return 0;

  495. 

  496.     for (i = 0, ptr = src; i < 4; i++, ptr += stride) {

  497.         sum_p1p2 += ptr[-2*step] - ptr[-3*step];

  498.         sum_q1q2 += ptr[ 1*step] - ptr[ 2*step];

  499.     }

  500. 

  501.     strong0 = *p1 && (FFABS(sum_p1p2) < beta2);

  502.     strong1 = *q1 && (FFABS(sum_q1q2) < beta2);

  503. 

  504.     return strong0 && strong1;

  505. }

  506. 

  507. static int rv40_h_loop_filter_strength(uint8_t *src, ptrdiff_t stride,

  508.                                        int beta, int beta2, int edge,

  509.                                        int *p1, int *q1)

  510. {

  511.     return rv40_loop_filter_strength(src, stride, 1, beta, beta2, edge, p1, q1);

  512. }

  513. 

  514. static int rv40_v_loop_filter_strength(uint8_t *src, ptrdiff_t stride,

  515.                                        int beta, int beta2, int edge,

  516.                                        int *p1, int *q1)

  517. {

  518.     return rv40_loop_filter_strength(src, 1, stride, beta, beta2, edge, p1, q1);

  519. }

  520. 

  521. av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {

  522. 

  523.     ff_rv34dsp_init(c, dsp);

  524. 

  525.     c->put_pixels_tab[0][ 0] = dsp->put_h264_qpel_pixels_tab[0][0];

  526.     c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c;

  527.     c->put_pixels_tab[0][ 2] = dsp->put_h264_qpel_pixels_tab[0][2];

  528.     c->put_pixels_tab[0][ 3] = put_rv40_qpel16_mc30_c;

  529.     c->put_pixels_tab[0][ 4] = put_rv40_qpel16_mc01_c;

  530.     c->put_pixels_tab[0][ 5] = put_rv40_qpel16_mc11_c;

  531.     c->put_pixels_tab[0][ 6] = put_rv40_qpel16_mc21_c;

  532.     c->put_pixels_tab[0][ 7] = put_rv40_qpel16_mc31_c;

  533.     c->put_pixels_tab[0][ 8] = dsp->put_h264_qpel_pixels_tab[0][8];

  534.     c->put_pixels_tab[0][ 9] = put_rv40_qpel16_mc12_c;

  535.     c->put_pixels_tab[0][10] = put_rv40_qpel16_mc22_c;

  536.     c->put_pixels_tab[0][11] = put_rv40_qpel16_mc32_c;

  537.     c->put_pixels_tab[0][12] = put_rv40_qpel16_mc03_c;

  538.     c->put_pixels_tab[0][13] = put_rv40_qpel16_mc13_c;

  539.     c->put_pixels_tab[0][14] = put_rv40_qpel16_mc23_c;

  540.     c->put_pixels_tab[0][15] = ff_put_rv40_qpel16_mc33_c;

  541.     c->avg_pixels_tab[0][ 0] = dsp->avg_h264_qpel_pixels_tab[0][0];

  542.     c->avg_pixels_tab[0][ 1] = avg_rv40_qpel16_mc10_c;

  543.     c->avg_pixels_tab[0][ 2] = dsp->avg_h264_qpel_pixels_tab[0][2];

  544.     c->avg_pixels_tab[0][ 3] = avg_rv40_qpel16_mc30_c;

  545.     c->avg_pixels_tab[0][ 4] = avg_rv40_qpel16_mc01_c;

  546.     c->avg_pixels_tab[0][ 5] = avg_rv40_qpel16_mc11_c;

  547.     c->avg_pixels_tab[0][ 6] = avg_rv40_qpel16_mc21_c;

  548.     c->avg_pixels_tab[0][ 7] = avg_rv40_qpel16_mc31_c;

  549.     c->avg_pixels_tab[0][ 8] = dsp->avg_h264_qpel_pixels_tab[0][8];

  550.     c->avg_pixels_tab[0][ 9] = avg_rv40_qpel16_mc12_c;

  551.     c->avg_pixels_tab[0][10] = avg_rv40_qpel16_mc22_c;

  552.     c->avg_pixels_tab[0][11] = avg_rv40_qpel16_mc32_c;

  553.     c->avg_pixels_tab[0][12] = avg_rv40_qpel16_mc03_c;

  554.     c->avg_pixels_tab[0][13] = avg_rv40_qpel16_mc13_c;

  555.     c->avg_pixels_tab[0][14] = avg_rv40_qpel16_mc23_c;

  556.     c->avg_pixels_tab[0][15] = ff_avg_rv40_qpel16_mc33_c;

  557.     c->put_pixels_tab[1][ 0] = dsp->put_h264_qpel_pixels_tab[1][0];

  558.     c->put_pixels_tab[1][ 1] = put_rv40_qpel8_mc10_c;

  559.     c->put_pixels_tab[1][ 2] = dsp->put_h264_qpel_pixels_tab[1][2];

  560.     c->put_pixels_tab[1][ 3] = put_rv40_qpel8_mc30_c;

  561.     c->put_pixels_tab[1][ 4] = put_rv40_qpel8_mc01_c;

  562.     c->put_pixels_tab[1][ 5] = put_rv40_qpel8_mc11_c;

  563.     c->put_pixels_tab[1][ 6] = put_rv40_qpel8_mc21_c;

  564.     c->put_pixels_tab[1][ 7] = put_rv40_qpel8_mc31_c;

  565.     c->put_pixels_tab[1][ 8] = dsp->put_h264_qpel_pixels_tab[1][8];

  566.     c->put_pixels_tab[1][ 9] = put_rv40_qpel8_mc12_c;

  567.     c->put_pixels_tab[1][10] = put_rv40_qpel8_mc22_c;

  568.     c->put_pixels_tab[1][11] = put_rv40_qpel8_mc32_c;

  569.     c->put_pixels_tab[1][12] = put_rv40_qpel8_mc03_c;

  570.     c->put_pixels_tab[1][13] = put_rv40_qpel8_mc13_c;

  571.     c->put_pixels_tab[1][14] = put_rv40_qpel8_mc23_c;

  572.     c->put_pixels_tab[1][15] = ff_put_rv40_qpel8_mc33_c;

  573.     c->avg_pixels_tab[1][ 0] = dsp->avg_h264_qpel_pixels_tab[1][0];

  574.     c->avg_pixels_tab[1][ 1] = avg_rv40_qpel8_mc10_c;

  575.     c->avg_pixels_tab[1][ 2] = dsp->avg_h264_qpel_pixels_tab[1][2];

  576.     c->avg_pixels_tab[1][ 3] = avg_rv40_qpel8_mc30_c;

  577.     c->avg_pixels_tab[1][ 4] = avg_rv40_qpel8_mc01_c;

  578.     c->avg_pixels_tab[1][ 5] = avg_rv40_qpel8_mc11_c;

  579.     c->avg_pixels_tab[1][ 6] = avg_rv40_qpel8_mc21_c;

  580.     c->avg_pixels_tab[1][ 7] = avg_rv40_qpel8_mc31_c;

  581.     c->avg_pixels_tab[1][ 8] = dsp->avg_h264_qpel_pixels_tab[1][8];

  582.     c->avg_pixels_tab[1][ 9] = avg_rv40_qpel8_mc12_c;

  583.     c->avg_pixels_tab[1][10] = avg_rv40_qpel8_mc22_c;

  584.     c->avg_pixels_tab[1][11] = avg_rv40_qpel8_mc32_c;

  585.     c->avg_pixels_tab[1][12] = avg_rv40_qpel8_mc03_c;

  586.     c->avg_pixels_tab[1][13] = avg_rv40_qpel8_mc13_c;

  587.     c->avg_pixels_tab[1][14] = avg_rv40_qpel8_mc23_c;

  588.     c->avg_pixels_tab[1][15] = ff_avg_rv40_qpel8_mc33_c;

  589. 

  590.     c->put_chroma_pixels_tab[0] = put_rv40_chroma_mc8_c;

  591.     c->put_chroma_pixels_tab[1] = put_rv40_chroma_mc4_c;

  592.     c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c;

  593.     c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c;

  594. 

  595.     c->rv40_weight_pixels_tab[0][0] = rv40_weight_func_rnd_16;

  596.     c->rv40_weight_pixels_tab[0][1] = rv40_weight_func_rnd_8;

  597.     c->rv40_weight_pixels_tab[1][0] = rv40_weight_func_nornd_16;

  598.     c->rv40_weight_pixels_tab[1][1] = rv40_weight_func_nornd_8;

  599. 

  600.     c->rv40_weak_loop_filter[0]     = rv40_h_weak_loop_filter;

  601.     c->rv40_weak_loop_filter[1]     = rv40_v_weak_loop_filter;

  602.     c->rv40_strong_loop_filter[0]   = rv40_h_strong_loop_filter;

  603.     c->rv40_strong_loop_filter[1]   = rv40_v_strong_loop_filter;

  604.     c->rv40_loop_filter_strength[0] = rv40_h_loop_filter_strength;

  605.     c->rv40_loop_filter_strength[1] = rv40_v_loop_filter_strength;

  606. 

  607.     if (HAVE_MMX)

  608.         ff_rv40dsp_init_x86(c, dsp);

  609.     if (HAVE_NEON)

  610.         ff_rv40dsp_init_neon(c, dsp);

  611. }