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.
34063 Commits
32 Branches
283MB
Tree: 25b51b2c44
Branches Tags
${ item.name }
Create branch ${ searchTerm }
from '25b51b2c44'
${ noResults }
FFmpeg/libavcodec/rv40dsp.c

610 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 "dsputil.h"

  29. #include "rv34dsp.h"

  30. 

  31. #define RV40_LOWPASS(OPNAME, OP) \

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

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

  34.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  35.     int i;\

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

  37.     {\

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

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

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

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

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

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

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

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

  46.         dst += dstStride;\

  47.         src += srcStride;\

  48.     }\

  49. }\

  50. \

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

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

  53.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\

  54.     int i;\

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

  56.     {\

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  78.         dst++;\

  79.         src++;\

  80.     }\

  81. }\

  82. \

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

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

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

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

  87.     src += 8*srcStride;\

  88.     dst += 8*dstStride;\

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

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

  91. }\

  92. \

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

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

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

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

  97.     src += 8*srcStride;\

  98.     dst += 8*dstStride;\

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

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

  101. }\

  102. \

  103. 

  104. #define RV40_MC(OPNAME, SIZE) \

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

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

  107. }\

  108. \

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

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

  111. }\

  112. \

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

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

  115. }\

  116. \

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

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

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

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

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

  122. }\

  123. \

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

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

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

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

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

  129. }\

  130. \

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

  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, 52, 6);\

  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 ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\

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

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

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

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

  143. }\

  144. \

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

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

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

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

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

  150. }\

  151. \

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

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

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

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

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

  157. }\

  158. \

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

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

  161. }\

  162. \

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

  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, 52, 20, 6);\

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

  168. }\

  169. \

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

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

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

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

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

  175. }\

  176. \

  177. 

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

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

  180. 

  181. RV40_LOWPASS(put_       , op_put)

  182. RV40_LOWPASS(avg_       , op_avg)

  183. 

  184. #undef op_avg

  185. #undef op_put

  186. 

  187. RV40_MC(put_, 8)

  188. RV40_MC(put_, 16)

  189. RV40_MC(avg_, 8)

  190. RV40_MC(avg_, 16)

  191. 

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

  193.     {  0, 16, 32, 16 },

  194.     { 32, 28, 32, 28 },

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

  196.     { 32, 28, 32, 28 }

  197. };

  198. 

  199. #define RV40_CHROMA_MC(OPNAME, OP)\

  200. 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){\

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

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

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

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

  205.     int i;\

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

  207.     \

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

  209. \

  210.     if(D){\

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

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

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

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

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

  216.             dst += stride;\

  217.             src += stride;\

  218.         }\

  219.     }else{\

  220.         const int E = B + C;\

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

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

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

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

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

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

  227.             dst += stride;\

  228.             src += stride;\

  229.         }\

  230.     }\

  231. }\

  232. \

  233. 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){\

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

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

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

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

  238.     int i;\

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

  240.     \

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

  242. \

  243.     if(D){\

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

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

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

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

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

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

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

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

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

  253.             dst += stride;\

  254.             src += stride;\

  255.         }\

  256.     }else{\

  257.         const int E = B + C;\

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

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

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

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

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

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

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

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

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

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

  268.             dst += stride;\

  269.             src += stride;\

  270.         }\

  271.     }\

  272. }

  273. 

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

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

  276. 

  277. RV40_CHROMA_MC(put_, op_put)

  278. RV40_CHROMA_MC(avg_, op_avg)

  279. 

  280. #define RV40_WEIGHT_FUNC(size) \

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

  282. {\

  283.     int i, j;\

  284. \

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

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

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

  288.         src1 += stride;\

  289.         src2 += stride;\

  290.         dst  += stride;\

  291.     }\

  292. }\

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

  294. {\

  295.     int i, j;\

  296. \

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

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

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

  300.         src1 += stride;\

  301.         src2 += stride;\

  302.         dst  += stride;\

  303.     }\

  304. }

  305. 

  306. RV40_WEIGHT_FUNC(16)

  307. RV40_WEIGHT_FUNC(8)

  308. 

  309. /**

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

  311.  */

  312. static const uint8_t rv40_dither_l[16] = {

  313.     0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30,

  314.     0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40

  315. };

  316. 

  317. /**

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

  319.  */

  320. static const uint8_t rv40_dither_r[16] = {

  321.     0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40,

  322.     0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40

  323. };

  324. 

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

  326. /**

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

  328.  */

  329. static av_always_inline void rv40_weak_loop_filter(uint8_t *src,

  330.                                                    const int step,

  331.                                                    const ptrdiff_t stride,

  332.                                                    const int filter_p1,

  333.                                                    const int filter_q1,

  334.                                                    const int alpha,

  335.                                                    const int beta,

  336.                                                    const int lim_p0q0,

  337.                                                    const int lim_q1,

  338.                                                    const int lim_p1)

  339. {

  340.     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;

  341.     int i, t, u, diff;

  342. 

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

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

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

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

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

  348. 

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

  350.         if (!t)

  351.             continue;

  352. 

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

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

  355.             continue;

  356. 

  357.         t <<= 2;

  358.         if (filter_p1 && filter_q1)

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

  360. 

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

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

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

  364. 

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

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

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

  368.         }

  369. 

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

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

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

  373.         }

  374.     }

  375. }

  376. 

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

  378.                                     const int filter_p1, const int filter_q1,

  379.                                     const int alpha, const int beta,

  380.                                     const int lim_p0q0, const int lim_q1,

  381.                                     const int lim_p1)

  382. {

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

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

  385. }

  386. 

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

  388.                                     const int filter_p1, const int filter_q1,

  389.                                     const int alpha, const int beta,

  390.                                     const int lim_p0q0, const int lim_q1,

  391.                                     const int lim_p1)

  392. {

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

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

  395. }

  396. 

  397. static av_always_inline void rv40_strong_loop_filter(uint8_t *src,

  398.                                                      const int step,

  399.                                                      const ptrdiff_t stride,

  400.                                                      const int alpha,

  401.                                                      const int lims,

  402.                                                      const int dmode,

  403.                                                      const int chroma)

  404. {

  405.     int i;

  406. 

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

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

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

  410. 

  411.         if (!t)

  412.             continue;

  413. 

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

  415.         if (sflag > 1)

  416.             continue;

  417. 

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

  419.               26*src[ 0*step] + 25*src[ 1*step] +

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

  421. 

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

  423.               26*src[ 1*step] + 25*src[ 2*step] +

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

  425. 

  426.         if (sflag) {

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

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

  429.         }

  430. 

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

  432.               25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;

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

  434.               25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;

  435. 

  436.         if (sflag) {

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

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

  439.         }

  440. 

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

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

  443.         src[ 0*step] = q0;

  444.         src[ 1*step] = q1;

  445. 

  446.         if(!chroma){

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

  448.                             51*src[-3*step] + 26*src[-4*step] + 64) >> 7;

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

  450.                             51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;

  451.         }

  452.     }

  453. }

  454. 

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

  456.                                       const int alpha, const int lims,

  457.                                       const int dmode, const int chroma)

  458. {

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

  460. }

  461. 

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

  463.                                       const int alpha, const int lims,

  464.                                       const int dmode, const int chroma)

  465. {

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

  467. }

  468. 

  469. static av_always_inline int rv40_loop_filter_strength(uint8_t *src,

  470.                                                       int step, ptrdiff_t stride,

  471.                                                       int beta, int beta2,

  472.                                                       int edge,

  473.                                                       int *p1, int *q1)

  474. {

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

  476.     int strong0 = 0, strong1 = 0;

  477.     uint8_t *ptr;

  478.     int i;

  479. 

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

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

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

  483.     }

  484. 

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

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

  487. 

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

  489.         return 0;

  490. 

  491.     if (!edge)

  492.         return 0;

  493. 

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

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

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

  497.     }

  498. 

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

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

  501. 

  502.     return strong0 && strong1;

  503. }

  504. 

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

  506.                                        int beta, int beta2, int edge,

  507.                                        int *p1, int *q1)

  508. {

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

  510. }

  511. 

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

  513.                                        int beta, int beta2, int edge,

  514.                                        int *p1, int *q1)

  515. {

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

  517. }

  518. 

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

  520. 

  521.     ff_rv34dsp_init(c, dsp);

  522. 

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

  587. 

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

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

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

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

  592. 

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

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

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

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

  597. 

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

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

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

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

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

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

  604. 

  605.     if (HAVE_MMX)

  606.         ff_rv40dsp_init_x86(c, dsp);

  607.     if (HAVE_NEON)

  608.         ff_rv40dsp_init_neon(c, dsp);

  609. }