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.
39560 Commits
32 Branches
844MB
Tree: f9279ee749
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from 'f9279ee749'
${ noResults }
FFmpeg/libavcodec/rv40dsp.c

629 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 Libav.

  6.  *

  7.  * Libav 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.  * Libav 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 Libav; 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 "h264qpel.h"

  29. #include "mathops.h"

  30. #include "rv34dsp.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.     const uint8_t *cm = ff_crop_tab + 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.     const uint8_t *cm = ff_crop_tab + 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, ptrdiff_t stride)\

  108. {\

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

  110. }\

  111. \

  112. static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  113. {\

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

  115. }\

  116. \

  117. static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  118. {\

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

  120. }\

  121. \

  122. static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  123. {\

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

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

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

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

  128. }\

  129. \

  130. static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  131. {\

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

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

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

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

  136. }\

  137. \

  138. static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  139. {\

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

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

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

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

  144. }\

  145. \

  146. static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  147. {\

  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, 52, 20, 6);\

  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 ## _mc22_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  155. {\

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

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

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

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

  160. }\

  161. \

  162. static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  163. {\

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

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

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

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

  168. }\

  169. \

  170. static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  171. {\

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

  173. }\

  174. \

  175. static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  176. {\

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

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

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

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

  181. }\

  182. \

  183. static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride)\

  184. {\

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

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

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

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

  189. }\

  190. \

  191. 

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

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

  194. 

  195. RV40_LOWPASS(put_       , op_put)

  196. RV40_LOWPASS(avg_       , op_avg)

  197. 

  198. #undef op_avg

  199. #undef op_put

  200. 

  201. RV40_MC(put_, 8)

  202. RV40_MC(put_, 16)

  203. RV40_MC(avg_, 8)

  204. RV40_MC(avg_, 16)

  205. 

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

  207.     {  0, 16, 32, 16 },

  208.     { 32, 28, 32, 28 },

  209.     {  0, 32, 16, 32 },

  210.     { 32, 28, 32, 28 }

  211. };

  212. 

  213. #define RV40_CHROMA_MC(OPNAME, OP)\

  214. 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){\

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

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

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

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

  219.     int i;\

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

  221.     \

  222.     assert(x<8 && y<8 && x>=0 && y>=0);\

  223. \

  224.     if(D){\

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

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

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

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

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

  230.             dst += stride;\

  231.             src += stride;\

  232.         }\

  233.     }else{\

  234.         const int E = B + C;\

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

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

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

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

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

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

  241.             dst += stride;\

  242.             src += stride;\

  243.         }\

  244.     }\

  245. }\

  246. \

  247. 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){\

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

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

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

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

  252.     int i;\

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

  254.     \

  255.     assert(x<8 && y<8 && x>=0 && y>=0);\

  256. \

  257.     if(D){\

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

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

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

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

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

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

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

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

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

  267.             dst += stride;\

  268.             src += stride;\

  269.         }\

  270.     }else{\

  271.         const int E = B + C;\

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

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

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

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

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

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

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

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

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

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

  282.             dst += stride;\

  283.             src += stride;\

  284.         }\

  285.     }\

  286. }

  287. 

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

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

  290. 

  291. RV40_CHROMA_MC(put_, op_put)

  292. RV40_CHROMA_MC(avg_, op_avg)

  293. 

  294. #define RV40_WEIGHT_FUNC(size) \

  295. static void rv40_weight_func_rnd_ ## 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]) >> 9) + ((w1 * src2[i]) >> 9) + 0x10) >> 5;\

  302.         src1 += stride;\

  303.         src2 += stride;\

  304.         dst  += stride;\

  305.     }\

  306. }\

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

  308. {\

  309.     int i, j;\

  310. \

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

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

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

  314.         src1 += stride;\

  315.         src2 += stride;\

  316.         dst  += stride;\

  317.     }\

  318. }

  319. 

  320. RV40_WEIGHT_FUNC(16)

  321. RV40_WEIGHT_FUNC(8)

  322. 

  323. /**

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

  325.  */

  326. static const uint8_t rv40_dither_l[16] = {

  327.     0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30,

  328.     0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40

  329. };

  330. 

  331. /**

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

  333.  */

  334. static const uint8_t rv40_dither_r[16] = {

  335.     0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40,

  336.     0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40

  337. };

  338. 

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

  340. /**

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

  342.  */

  343. static av_always_inline void rv40_weak_loop_filter(uint8_t *src,

  344.                                                    const int step,

  345.                                                    const ptrdiff_t stride,

  346.                                                    const int filter_p1,

  347.                                                    const int filter_q1,

  348.                                                    const int alpha,

  349.                                                    const int beta,

  350.                                                    const int lim_p0q0,

  351.                                                    const int lim_q1,

  352.                                                    const int lim_p1)

  353. {

  354.     const uint8_t *cm = ff_crop_tab + MAX_NEG_CROP;

  355.     int i, t, u, diff;

  356. 

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

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

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

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

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

  362. 

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

  364.         if (!t)

  365.             continue;

  366. 

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

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

  369.             continue;

  370. 

  371.         t <<= 2;

  372.         if (filter_p1 && filter_q1)

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

  374. 

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

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

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

  378. 

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

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

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

  382.         }

  383. 

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

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

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

  387.         }

  388.     }

  389. }

  390. 

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

  392.                                     const int filter_p1, const int filter_q1,

  393.                                     const int alpha, const int beta,

  394.                                     const int lim_p0q0, const int lim_q1,

  395.                                     const int lim_p1)

  396. {

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

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

  399. }

  400. 

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

  402.                                     const int filter_p1, const int filter_q1,

  403.                                     const int alpha, const int beta,

  404.                                     const int lim_p0q0, const int lim_q1,

  405.                                     const int lim_p1)

  406. {

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

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

  409. }

  410. 

  411. static av_always_inline void rv40_strong_loop_filter(uint8_t *src,

  412.                                                      const int step,

  413.                                                      const ptrdiff_t stride,

  414.                                                      const int alpha,

  415.                                                      const int lims,

  416.                                                      const int dmode,

  417.                                                      const int chroma)

  418. {

  419.     int i;

  420. 

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

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

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

  424. 

  425.         if (!t)

  426.             continue;

  427. 

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

  429.         if (sflag > 1)

  430.             continue;

  431. 

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

  433.               26*src[ 0*step] + 25*src[ 1*step] +

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

  435. 

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

  437.               26*src[ 1*step] + 25*src[ 2*step] +

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

  439. 

  440.         if (sflag) {

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

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

  443.         }

  444. 

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

  446.               25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;

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

  448.               25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;

  449. 

  450.         if (sflag) {

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

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

  453.         }

  454. 

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

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

  457.         src[ 0*step] = q0;

  458.         src[ 1*step] = q1;

  459. 

  460.         if(!chroma){

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

  462.                             51*src[-3*step] + 26*src[-4*step] + 64) >> 7;

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

  464.                             51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;

  465.         }

  466.     }

  467. }

  468. 

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

  470.                                       const int alpha, const int lims,

  471.                                       const int dmode, const int chroma)

  472. {

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

  474. }

  475. 

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

  477.                                       const int alpha, const int lims,

  478.                                       const int dmode, const int chroma)

  479. {

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

  481. }

  482. 

  483. static av_always_inline int rv40_loop_filter_strength(uint8_t *src,

  484.                                                       int step, ptrdiff_t stride,

  485.                                                       int beta, int beta2,

  486.                                                       int edge,

  487.                                                       int *p1, int *q1)

  488. {

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

  490.     int strong0 = 0, strong1 = 0;

  491.     uint8_t *ptr;

  492.     int i;

  493. 

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

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

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

  497.     }

  498. 

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

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

  501. 

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

  503.         return 0;

  504. 

  505.     if (!edge)

  506.         return 0;

  507. 

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

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

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

  511.     }

  512. 

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

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

  515. 

  516.     return strong0 && strong1;

  517. }

  518. 

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

  520.                                        int beta, int beta2, int edge,

  521.                                        int *p1, int *q1)

  522. {

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

  524. }

  525. 

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

  527.                                        int beta, int beta2, int edge,

  528.                                        int *p1, int *q1)

  529. {

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

  531. }

  532. 

  533. av_cold void ff_rv40dsp_init(RV34DSPContext *c)

  534. {

  535.     H264QpelContext qpel;

  536. 

  537.     ff_rv34dsp_init(c);

  538.     ff_h264qpel_init(&qpel, 8);

  539. 

  540.     c->put_pixels_tab[0][ 0] = qpel.put_h264_qpel_pixels_tab[0][0];

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

  542.     c->put_pixels_tab[0][ 2] = qpel.put_h264_qpel_pixels_tab[0][2];

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

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

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

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

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

  548.     c->put_pixels_tab[0][ 8] = qpel.put_h264_qpel_pixels_tab[0][8];

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

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

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

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

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

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

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

  556.     c->avg_pixels_tab[0][ 0] = qpel.avg_h264_qpel_pixels_tab[0][0];

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

  558.     c->avg_pixels_tab[0][ 2] = qpel.avg_h264_qpel_pixels_tab[0][2];

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

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

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

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

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

  564.     c->avg_pixels_tab[0][ 8] = qpel.avg_h264_qpel_pixels_tab[0][8];

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

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

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

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

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

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

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

  572.     c->put_pixels_tab[1][ 0] = qpel.put_h264_qpel_pixels_tab[1][0];

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

  574.     c->put_pixels_tab[1][ 2] = qpel.put_h264_qpel_pixels_tab[1][2];

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

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

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

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

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

  580.     c->put_pixels_tab[1][ 8] = qpel.put_h264_qpel_pixels_tab[1][8];

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

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

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

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

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

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

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

  588.     c->avg_pixels_tab[1][ 0] = qpel.avg_h264_qpel_pixels_tab[1][0];

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

  590.     c->avg_pixels_tab[1][ 2] = qpel.avg_h264_qpel_pixels_tab[1][2];

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

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

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

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

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

  596.     c->avg_pixels_tab[1][ 8] = qpel.avg_h264_qpel_pixels_tab[1][8];

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

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

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

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

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

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

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

  604. 

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

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

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

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

  609. 

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

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

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

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

  614. 

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

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

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

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

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

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

  621. 

  622.     if (ARCH_AARCH64)

  623.         ff_rv40dsp_init_aarch64(c);

  624.     if (ARCH_ARM)

  625.         ff_rv40dsp_init_arm(c);

  626.     if (ARCH_X86)

  627.         ff_rv40dsp_init_x86(c);

  628. }