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.
49470 Commits
32 Branches
535MB
Tree: 9c50e69385
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '9c50e69385'
${ noResults }
FFmpeg/libavcodec/rv40dsp.c

615 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 "h264qpel.h"

  30. #include "rv34dsp.h"

  31. #include "libavutil/avassert.h"

  32. #include "libavutil/common.h"

  33. 

  34. #define RV40_LOWPASS(OPNAME, OP) \

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

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

  37.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  38.     int i;\

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

  40.     {\

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

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

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

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

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

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

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

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

  49.         dst += dstStride;\

  50.         src += srcStride;\

  51.     }\

  52. }\

  53. \

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

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

  56.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  57.     int i;\

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

  59.     {\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  81.         dst++;\

  82.         src++;\

  83.     }\

  84. }\

  85. \

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

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

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

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

  90.     src += 8*srcStride;\

  91.     dst += 8*dstStride;\

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

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

  94. }\

  95. \

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

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

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

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

  100.     src += 8*srcStride;\

  101.     dst += 8*dstStride;\

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

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

  104. }\

  105. \

  106. 

  107. #define RV40_MC(OPNAME, SIZE) \

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

  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, int stride){\

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

  114. }\

  115. \

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

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

  118. }\

  119. \

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

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

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

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

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

  125. }\

  126. \

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

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

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

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

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

  132. }\

  133. \

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

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

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

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

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

  139. }\

  140. \

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

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

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

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

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

  146. }\

  147. \

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

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

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

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

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

  153. }\

  154. \

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

  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, 52, 6);\

  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 ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\

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

  164. }\

  165. \

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

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

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

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

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

  171. }\

  172. \

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

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

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

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

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

  178. }\

  179. \

  180. 

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

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

  183. 

  184. RV40_LOWPASS(put_       , op_put)

  185. RV40_LOWPASS(avg_       , op_avg)

  186. 

  187. #undef op_avg

  188. #undef op_put

  189. 

  190. RV40_MC(put_, 8)

  191. RV40_MC(put_, 16)

  192. RV40_MC(avg_, 8)

  193. RV40_MC(avg_, 16)

  194. 

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

  196.     {  0, 16, 32, 16 },

  197.     { 32, 28, 32, 28 },

  198.     {  0, 32, 16, 32 },

  199.     { 32, 28, 32, 28 }

  200. };

  201. 

  202. #define RV40_CHROMA_MC(OPNAME, OP)\

  203. 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){\

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

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

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

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

  208.     int i;\

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

  210.     \

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

  212. \

  213.     if(D){\

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

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

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

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

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

  219.             dst += stride;\

  220.             src += stride;\

  221.         }\

  222.     }else{\

  223.         const int E = B + C;\

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

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

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

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

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

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

  230.             dst += stride;\

  231.             src += stride;\

  232.         }\

  233.     }\

  234. }\

  235. \

  236. 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){\

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

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

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

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

  241.     int i;\

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

  243.     \

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

  245. \

  246.     if(D){\

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

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

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

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

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

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

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

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

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

  256.             dst += stride;\

  257.             src += stride;\

  258.         }\

  259.     }else{\

  260.         const int E = B + C;\

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

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

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

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

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

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

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

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

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

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

  271.             dst += stride;\

  272.             src += stride;\

  273.         }\

  274.     }\

  275. }

  276. 

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

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

  279. 

  280. RV40_CHROMA_MC(put_, op_put)

  281. RV40_CHROMA_MC(avg_, op_avg)

  282. 

  283. #define RV40_WEIGHT_FUNC(size) \

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

  285. {\

  286.     int i, j;\

  287. \

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

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

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

  291.         src1 += stride;\

  292.         src2 += stride;\

  293.         dst  += stride;\

  294.     }\

  295. }\

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

  297. {\

  298.     int i, j;\

  299. \

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

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

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

  303.         src1 += stride;\

  304.         src2 += stride;\

  305.         dst  += stride;\

  306.     }\

  307. }

  308. 

  309. RV40_WEIGHT_FUNC(16)

  310. RV40_WEIGHT_FUNC(8)

  311. 

  312. /**

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

  314.  */

  315. static const uint8_t rv40_dither_l[16] = {

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

  317.     0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40

  318. };

  319. 

  320. /**

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

  322.  */

  323. static const uint8_t rv40_dither_r[16] = {

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

  325.     0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40

  326. };

  327. 

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

  329. /**

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

  331.  */

  332. static av_always_inline void rv40_weak_loop_filter(uint8_t *src,

  333.                                                    const int step,

  334.                                                    const ptrdiff_t stride,

  335.                                                    const int filter_p1,

  336.                                                    const int filter_q1,

  337.                                                    const int alpha,

  338.                                                    const int beta,

  339.                                                    const int lim_p0q0,

  340.                                                    const int lim_q1,

  341.                                                    const int lim_p1)

  342. {

  343.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  344.     int i, t, u, diff;

  345. 

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

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

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

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

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

  351. 

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

  353.         if (!t)

  354.             continue;

  355. 

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

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

  358.             continue;

  359. 

  360.         t <<= 2;

  361.         if (filter_p1 && filter_q1)

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

  363. 

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

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

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

  367. 

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

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

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

  371.         }

  372. 

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

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

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

  376.         }

  377.     }

  378. }

  379. 

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

  381.                                     const int filter_p1, const int filter_q1,

  382.                                     const int alpha, const int beta,

  383.                                     const int lim_p0q0, const int lim_q1,

  384.                                     const int lim_p1)

  385. {

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

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

  388. }

  389. 

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

  391.                                     const int filter_p1, const int filter_q1,

  392.                                     const int alpha, const int beta,

  393.                                     const int lim_p0q0, const int lim_q1,

  394.                                     const int lim_p1)

  395. {

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

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

  398. }

  399. 

  400. static av_always_inline void rv40_strong_loop_filter(uint8_t *src,

  401.                                                      const int step,

  402.                                                      const ptrdiff_t stride,

  403.                                                      const int alpha,

  404.                                                      const int lims,

  405.                                                      const int dmode,

  406.                                                      const int chroma)

  407. {

  408.     int i;

  409. 

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

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

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

  413. 

  414.         if (!t)

  415.             continue;

  416. 

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

  418.         if (sflag > 1)

  419.             continue;

  420. 

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

  422.               26*src[ 0*step] + 25*src[ 1*step] +

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

  424. 

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

  426.               26*src[ 1*step] + 25*src[ 2*step] +

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

  428. 

  429.         if (sflag) {

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

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

  432.         }

  433. 

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

  435.               25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;

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

  437.               25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;

  438. 

  439.         if (sflag) {

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

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

  442.         }

  443. 

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

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

  446.         src[ 0*step] = q0;

  447.         src[ 1*step] = q1;

  448. 

  449.         if(!chroma){

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

  451.                             51*src[-3*step] + 26*src[-4*step] + 64) >> 7;

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

  453.                             51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;

  454.         }

  455.     }

  456. }

  457. 

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

  459.                                       const int alpha, const int lims,

  460.                                       const int dmode, const int chroma)

  461. {

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

  463. }

  464. 

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

  466.                                       const int alpha, const int lims,

  467.                                       const int dmode, const int chroma)

  468. {

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

  470. }

  471. 

  472. static av_always_inline int rv40_loop_filter_strength(uint8_t *src,

  473.                                                       int step, ptrdiff_t stride,

  474.                                                       int beta, int beta2,

  475.                                                       int edge,

  476.                                                       int *p1, int *q1)

  477. {

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

  479.     int strong0 = 0, strong1 = 0;

  480.     uint8_t *ptr;

  481.     int i;

  482. 

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

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

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

  486.     }

  487. 

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

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

  490. 

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

  492.         return 0;

  493. 

  494.     if (!edge)

  495.         return 0;

  496. 

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

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

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

  500.     }

  501. 

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

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

  504. 

  505.     return strong0 && strong1;

  506. }

  507. 

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

  509.                                        int beta, int beta2, int edge,

  510.                                        int *p1, int *q1)

  511. {

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

  513. }

  514. 

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

  516.                                        int beta, int beta2, int edge,

  517.                                        int *p1, int *q1)

  518. {

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

  520. }

  521. 

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

  523.     H264QpelContext qpel;

  524. 

  525.     ff_rv34dsp_init(c, dsp);

  526.     ff_h264qpel_init(&qpel, 8);

  527. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  592. 

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

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

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

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

  597. 

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

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

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

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

  602. 

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

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

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

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

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

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

  609. 

  610.     if (ARCH_X86)

  611.         ff_rv40dsp_init_x86(c, dsp);

  612.     if (ARCH_ARM)

  613.         ff_rv40dsp_init_arm(c, dsp);

  614. }