FFmpeg: libavcodec/rv40dsp.c Source File

00001 /*
00002  * RV40 decoder motion compensation functions
00003  * Copyright (c) 2008 Konstantin Shishkov
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00027 #include "avcodec.h"
00028 #include "dsputil.h"
00029 #include "rv34dsp.h"
00030 #include "libavutil/avassert.h"
00031 #include "libavutil/common.h"
00032 
00033 #define RV40_LOWPASS(OPNAME, OP) \
00034 static av_unused void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
00035                                                      const int h, const int C1, const int C2, const int SHIFT){\
00036     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00037     int i;\
00038     for(i = 0; i < h; i++)\
00039     {\
00040         OP(dst[0], (src[-2] + src[ 3] - 5*(src[-1]+src[2]) + src[0]*C1 + src[1]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00041         OP(dst[1], (src[-1] + src[ 4] - 5*(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00042         OP(dst[2], (src[ 0] + src[ 5] - 5*(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00043         OP(dst[3], (src[ 1] + src[ 6] - 5*(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00044         OP(dst[4], (src[ 2] + src[ 7] - 5*(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00045         OP(dst[5], (src[ 3] + src[ 8] - 5*(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00046         OP(dst[6], (src[ 4] + src[ 9] - 5*(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00047         OP(dst[7], (src[ 5] + src[10] - 5*(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00048         dst += dstStride;\
00049         src += srcStride;\
00050     }\
00051 }\
00052 \
00053 static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
00054                                            const int w, const int C1, const int C2, const int SHIFT){\
00055     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00056     int i;\
00057     for(i = 0; i < w; i++)\
00058     {\
00059         const int srcB  = src[-2*srcStride];\
00060         const int srcA  = src[-1*srcStride];\
00061         const int src0  = src[0 *srcStride];\
00062         const int src1  = src[1 *srcStride];\
00063         const int src2  = src[2 *srcStride];\
00064         const int src3  = src[3 *srcStride];\
00065         const int src4  = src[4 *srcStride];\
00066         const int src5  = src[5 *srcStride];\
00067         const int src6  = src[6 *srcStride];\
00068         const int src7  = src[7 *srcStride];\
00069         const int src8  = src[8 *srcStride];\
00070         const int src9  = src[9 *srcStride];\
00071         const int src10 = src[10*srcStride];\
00072         OP(dst[0*dstStride], (srcB + src3  - 5*(srcA+src2) + src0*C1 + src1*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00073         OP(dst[1*dstStride], (srcA + src4  - 5*(src0+src3) + src1*C1 + src2*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00074         OP(dst[2*dstStride], (src0 + src5  - 5*(src1+src4) + src2*C1 + src3*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00075         OP(dst[3*dstStride], (src1 + src6  - 5*(src2+src5) + src3*C1 + src4*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00076         OP(dst[4*dstStride], (src2 + src7  - 5*(src3+src6) + src4*C1 + src5*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00077         OP(dst[5*dstStride], (src3 + src8  - 5*(src4+src7) + src5*C1 + src6*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00078         OP(dst[6*dstStride], (src4 + src9  - 5*(src5+src8) + src6*C1 + src7*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00079         OP(dst[7*dstStride], (src5 + src10 - 5*(src6+src9) + src7*C1 + src8*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00080         dst++;\
00081         src++;\
00082     }\
00083 }\
00084 \
00085 static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
00086                                             const int w, const int C1, const int C2, const int SHIFT){\
00087     OPNAME ## rv40_qpel8_v_lowpass(dst  , src  , dstStride, srcStride, 8, C1, C2, SHIFT);\
00088     OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
00089     src += 8*srcStride;\
00090     dst += 8*dstStride;\
00091     OPNAME ## rv40_qpel8_v_lowpass(dst  , src  , dstStride, srcStride, w-8, C1, C2, SHIFT);\
00092     OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, w-8, C1, C2, SHIFT);\
00093 }\
00094 \
00095 static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
00096                                             const int h, const int C1, const int C2, const int SHIFT){\
00097     OPNAME ## rv40_qpel8_h_lowpass(dst  , src  , dstStride, srcStride, 8, C1, C2, SHIFT);\
00098     OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
00099     src += 8*srcStride;\
00100     dst += 8*dstStride;\
00101     OPNAME ## rv40_qpel8_h_lowpass(dst  , src  , dstStride, srcStride, h-8, C1, C2, SHIFT);\
00102     OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, h-8, C1, C2, SHIFT);\
00103 }\
00104 \
00105 
00106 #define RV40_MC(OPNAME, SIZE) \
00107 static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
00108     OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
00109 }\
00110 \
00111 static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
00112     OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
00113 }\
00114 \
00115 static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
00116     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
00117 }\
00118 \
00119 static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
00120     uint8_t full[SIZE*(SIZE+5)];\
00121     uint8_t * const full_mid = full + SIZE*2;\
00122     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
00123     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
00124 }\
00125 \
00126 static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
00127     uint8_t full[SIZE*(SIZE+5)];\
00128     uint8_t * const full_mid = full + SIZE*2;\
00129     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
00130     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
00131 }\
00132 \
00133 static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
00134     uint8_t full[SIZE*(SIZE+5)];\
00135     uint8_t * const full_mid = full + SIZE*2;\
00136     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
00137     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
00138 }\
00139 \
00140 static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
00141     uint8_t full[SIZE*(SIZE+5)];\
00142     uint8_t * const full_mid = full + SIZE*2;\
00143     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
00144     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
00145 }\
00146 \
00147 static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
00148     uint8_t full[SIZE*(SIZE+5)];\
00149     uint8_t * const full_mid = full + SIZE*2;\
00150     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
00151     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
00152 }\
00153 \
00154 static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
00155     uint8_t full[SIZE*(SIZE+5)];\
00156     uint8_t * const full_mid = full + SIZE*2;\
00157     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
00158     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
00159 }\
00160 \
00161 static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
00162     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
00163 }\
00164 \
00165 static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
00166     uint8_t full[SIZE*(SIZE+5)];\
00167     uint8_t * const full_mid = full + SIZE*2;\
00168     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
00169     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
00170 }\
00171 \
00172 static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
00173     uint8_t full[SIZE*(SIZE+5)];\
00174     uint8_t * const full_mid = full + SIZE*2;\
00175     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
00176     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
00177 }\
00178 \
00179 
00180 #define op_avg(a, b)  a = (((a)+cm[b]+1)>>1)
00181 #define op_put(a, b)  a = cm[b]
00182 
00183 RV40_LOWPASS(put_       , op_put)
00184 RV40_LOWPASS(avg_       , op_avg)
00185 
00186 #undef op_avg
00187 #undef op_put
00188 
00189 RV40_MC(put_, 8)
00190 RV40_MC(put_, 16)
00191 RV40_MC(avg_, 8)
00192 RV40_MC(avg_, 16)
00193 
00194 static const int rv40_bias[4][4] = {
00195     {  0, 16, 32, 16 },
00196     { 32, 28, 32, 28 },
00197     {  0, 32, 16, 32 },
00198     { 32, 28, 32, 28 }
00199 };
00200 
00201 #define RV40_CHROMA_MC(OPNAME, OP)\
00202 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){\
00203     const int A = (8-x) * (8-y);\
00204     const int B = (  x) * (8-y);\
00205     const int C = (8-x) * (  y);\
00206     const int D = (  x) * (  y);\
00207     int i;\
00208     int bias = rv40_bias[y>>1][x>>1];\
00209     \
00210     av_assert2(x<8 && y<8 && x>=0 && y>=0);\
00211 \
00212     if(D){\
00213         for(i = 0; i < h; i++){\
00214             OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
00215             OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
00216             OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
00217             OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
00218             dst += stride;\
00219             src += stride;\
00220         }\
00221     }else{\
00222         const int E = B + C;\
00223         const int step = C ? stride : 1;\
00224         for(i = 0; i < h; i++){\
00225             OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
00226             OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
00227             OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
00228             OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
00229             dst += stride;\
00230             src += stride;\
00231         }\
00232     }\
00233 }\
00234 \
00235 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){\
00236     const int A = (8-x) * (8-y);\
00237     const int B = (  x) * (8-y);\
00238     const int C = (8-x) * (  y);\
00239     const int D = (  x) * (  y);\
00240     int i;\
00241     int bias = rv40_bias[y>>1][x>>1];\
00242     \
00243     av_assert2(x<8 && y<8 && x>=0 && y>=0);\
00244 \
00245     if(D){\
00246         for(i = 0; i < h; i++){\
00247             OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
00248             OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
00249             OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
00250             OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
00251             OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + bias));\
00252             OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + bias));\
00253             OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + bias));\
00254             OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + bias));\
00255             dst += stride;\
00256             src += stride;\
00257         }\
00258     }else{\
00259         const int E = B + C;\
00260         const int step = C ? stride : 1;\
00261         for(i = 0; i < h; i++){\
00262             OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
00263             OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
00264             OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
00265             OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
00266             OP(dst[4], (A*src[4] + E*src[step+4] + bias));\
00267             OP(dst[5], (A*src[5] + E*src[step+5] + bias));\
00268             OP(dst[6], (A*src[6] + E*src[step+6] + bias));\
00269             OP(dst[7], (A*src[7] + E*src[step+7] + bias));\
00270             dst += stride;\
00271             src += stride;\
00272         }\
00273     }\
00274 }
00275 
00276 #define op_avg(a, b) a = (((a)+((b)>>6)+1)>>1)
00277 #define op_put(a, b) a = ((b)>>6)
00278 
00279 RV40_CHROMA_MC(put_, op_put)
00280 RV40_CHROMA_MC(avg_, op_avg)
00281 
00282 #define RV40_WEIGHT_FUNC(size) \
00283 static void rv40_weight_func_rnd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\
00284 {\
00285     int i, j;\
00286 \
00287     for (j = 0; j < size; j++) {\
00288         for (i = 0; i < size; i++)\
00289             dst[i] = (((w2 * src1[i]) >> 9) + ((w1 * src2[i]) >> 9) + 0x10) >> 5;\
00290         src1 += stride;\
00291         src2 += stride;\
00292         dst  += stride;\
00293     }\
00294 }\
00295 static void rv40_weight_func_nornd_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, ptrdiff_t stride)\
00296 {\
00297     int i, j;\
00298 \
00299     for (j = 0; j < size; j++) {\
00300         for (i = 0; i < size; i++)\
00301             dst[i] = (w2 * src1[i] + w1 * src2[i] + 0x10) >> 5;\
00302         src1 += stride;\
00303         src2 += stride;\
00304         dst  += stride;\
00305     }\
00306 }
00307 
00308 RV40_WEIGHT_FUNC(16)
00309 RV40_WEIGHT_FUNC(8)
00310 
00314 static const uint8_t rv40_dither_l[16] = {
00315     0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30,
00316     0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40
00317 };
00318 
00322 static const uint8_t rv40_dither_r[16] = {
00323     0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40,
00324     0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40
00325 };
00326 
00327 #define CLIP_SYMM(a, b) av_clip(a, -(b), b)
00328 
00331 static av_always_inline void rv40_weak_loop_filter(uint8_t *src,
00332                                                    const int step,
00333                                                    const ptrdiff_t stride,
00334                                                    const int filter_p1,
00335                                                    const int filter_q1,
00336                                                    const int alpha,
00337                                                    const int beta,
00338                                                    const int lim_p0q0,
00339                                                    const int lim_q1,
00340                                                    const int lim_p1)
00341 {
00342     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
00343     int i, t, u, diff;
00344 
00345     for (i = 0; i < 4; i++, src += stride) {
00346         int diff_p1p0 = src[-2*step] - src[-1*step];
00347         int diff_q1q0 = src[ 1*step] - src[ 0*step];
00348         int diff_p1p2 = src[-2*step] - src[-3*step];
00349         int diff_q1q2 = src[ 1*step] - src[ 2*step];
00350 
00351         t = src[0*step] - src[-1*step];
00352         if (!t)
00353             continue;
00354 
00355         u = (alpha * FFABS(t)) >> 7;
00356         if (u > 3 - (filter_p1 && filter_q1))
00357             continue;
00358 
00359         t <<= 2;
00360         if (filter_p1 && filter_q1)
00361             t += src[-2*step] - src[1*step];
00362 
00363         diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0);
00364         src[-1*step] = cm[src[-1*step] + diff];
00365         src[ 0*step] = cm[src[ 0*step] - diff];
00366 
00367         if (filter_p1 && FFABS(diff_p1p2) <= beta) {
00368             t = (diff_p1p0 + diff_p1p2 - diff) >> 1;
00369             src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)];
00370         }
00371 
00372         if (filter_q1 && FFABS(diff_q1q2) <= beta) {
00373             t = (diff_q1q0 + diff_q1q2 + diff) >> 1;
00374             src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)];
00375         }
00376     }
00377 }
00378 
00379 static void rv40_h_weak_loop_filter(uint8_t *src, const ptrdiff_t stride,
00380                                     const int filter_p1, const int filter_q1,
00381                                     const int alpha, const int beta,
00382                                     const int lim_p0q0, const int lim_q1,
00383                                     const int lim_p1)
00384 {
00385     rv40_weak_loop_filter(src, stride, 1, filter_p1, filter_q1,
00386                           alpha, beta, lim_p0q0, lim_q1, lim_p1);
00387 }
00388 
00389 static void rv40_v_weak_loop_filter(uint8_t *src, const ptrdiff_t stride,
00390                                     const int filter_p1, const int filter_q1,
00391                                     const int alpha, const int beta,
00392                                     const int lim_p0q0, const int lim_q1,
00393                                     const int lim_p1)
00394 {
00395     rv40_weak_loop_filter(src, 1, stride, filter_p1, filter_q1,
00396                           alpha, beta, lim_p0q0, lim_q1, lim_p1);
00397 }
00398 
00399 static av_always_inline void rv40_strong_loop_filter(uint8_t *src,
00400                                                      const int step,
00401                                                      const ptrdiff_t stride,
00402                                                      const int alpha,
00403                                                      const int lims,
00404                                                      const int dmode,
00405                                                      const int chroma)
00406 {
00407     int i;
00408 
00409     for(i = 0; i < 4; i++, src += stride){
00410         int sflag, p0, q0, p1, q1;
00411         int t = src[0*step] - src[-1*step];
00412 
00413         if (!t)
00414             continue;
00415 
00416         sflag = (alpha * FFABS(t)) >> 7;
00417         if (sflag > 1)
00418             continue;
00419 
00420         p0 = (25*src[-3*step] + 26*src[-2*step] + 26*src[-1*step] +
00421               26*src[ 0*step] + 25*src[ 1*step] +
00422               rv40_dither_l[dmode + i]) >> 7;
00423 
00424         q0 = (25*src[-2*step] + 26*src[-1*step] + 26*src[ 0*step] +
00425               26*src[ 1*step] + 25*src[ 2*step] +
00426               rv40_dither_r[dmode + i]) >> 7;
00427 
00428         if (sflag) {
00429             p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims);
00430             q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims);
00431         }
00432 
00433         p1 = (25*src[-4*step] + 26*src[-3*step] + 26*src[-2*step] + 26*p0 +
00434               25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;
00435         q1 = (25*src[-1*step] + 26*q0 + 26*src[ 1*step] + 26*src[ 2*step] +
00436               25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;
00437 
00438         if (sflag) {
00439             p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims);
00440             q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims);
00441         }
00442 
00443         src[-2*step] = p1;
00444         src[-1*step] = p0;
00445         src[ 0*step] = q0;
00446         src[ 1*step] = q1;
00447 
00448         if(!chroma){
00449             src[-3*step] = (25*src[-1*step] + 26*src[-2*step] +
00450                             51*src[-3*step] + 26*src[-4*step] + 64) >> 7;
00451             src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] +
00452                             51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;
00453         }
00454     }
00455 }
00456 
00457 static void rv40_h_strong_loop_filter(uint8_t *src, const ptrdiff_t stride,
00458                                       const int alpha, const int lims,
00459                                       const int dmode, const int chroma)
00460 {
00461     rv40_strong_loop_filter(src, stride, 1, alpha, lims, dmode, chroma);
00462 }
00463 
00464 static void rv40_v_strong_loop_filter(uint8_t *src, const ptrdiff_t stride,
00465                                       const int alpha, const int lims,
00466                                       const int dmode, const int chroma)
00467 {
00468     rv40_strong_loop_filter(src, 1, stride, alpha, lims, dmode, chroma);
00469 }
00470 
00471 static av_always_inline int rv40_loop_filter_strength(uint8_t *src,
00472                                                       int step, ptrdiff_t stride,
00473                                                       int beta, int beta2,
00474                                                       int edge,
00475                                                       int *p1, int *q1)
00476 {
00477     int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0;
00478     int strong0 = 0, strong1 = 0;
00479     uint8_t *ptr;
00480     int i;
00481 
00482     for (i = 0, ptr = src; i < 4; i++, ptr += stride) {
00483         sum_p1p0 += ptr[-2*step] - ptr[-1*step];
00484         sum_q1q0 += ptr[ 1*step] - ptr[ 0*step];
00485     }
00486 
00487     *p1 = FFABS(sum_p1p0) < (beta << 2);
00488     *q1 = FFABS(sum_q1q0) < (beta << 2);
00489 
00490     if(!*p1 && !*q1)
00491         return 0;
00492 
00493     if (!edge)
00494         return 0;
00495 
00496     for (i = 0, ptr = src; i < 4; i++, ptr += stride) {
00497         sum_p1p2 += ptr[-2*step] - ptr[-3*step];
00498         sum_q1q2 += ptr[ 1*step] - ptr[ 2*step];
00499     }
00500 
00501     strong0 = *p1 && (FFABS(sum_p1p2) < beta2);
00502     strong1 = *q1 && (FFABS(sum_q1q2) < beta2);
00503 
00504     return strong0 && strong1;
00505 }
00506 
00507 static int rv40_h_loop_filter_strength(uint8_t *src, ptrdiff_t stride,
00508                                        int beta, int beta2, int edge,
00509                                        int *p1, int *q1)
00510 {
00511     return rv40_loop_filter_strength(src, stride, 1, beta, beta2, edge, p1, q1);
00512 }
00513 
00514 static int rv40_v_loop_filter_strength(uint8_t *src, ptrdiff_t stride,
00515                                        int beta, int beta2, int edge,
00516                                        int *p1, int *q1)
00517 {
00518     return rv40_loop_filter_strength(src, 1, stride, beta, beta2, edge, p1, q1);
00519 }
00520 
00521 av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {
00522 
00523     ff_rv34dsp_init(c, dsp);
00524 
00525     c->put_pixels_tab[0][ 0] = dsp->put_h264_qpel_pixels_tab[0][0];
00526     c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c;
00527     c->put_pixels_tab[0][ 2] = dsp->put_h264_qpel_pixels_tab[0][2];
00528     c->put_pixels_tab[0][ 3] = put_rv40_qpel16_mc30_c;
00529     c->put_pixels_tab[0][ 4] = put_rv40_qpel16_mc01_c;
00530     c->put_pixels_tab[0][ 5] = put_rv40_qpel16_mc11_c;
00531     c->put_pixels_tab[0][ 6] = put_rv40_qpel16_mc21_c;
00532     c->put_pixels_tab[0][ 7] = put_rv40_qpel16_mc31_c;
00533     c->put_pixels_tab[0][ 8] = dsp->put_h264_qpel_pixels_tab[0][8];
00534     c->put_pixels_tab[0][ 9] = put_rv40_qpel16_mc12_c;
00535     c->put_pixels_tab[0][10] = put_rv40_qpel16_mc22_c;
00536     c->put_pixels_tab[0][11] = put_rv40_qpel16_mc32_c;
00537     c->put_pixels_tab[0][12] = put_rv40_qpel16_mc03_c;
00538     c->put_pixels_tab[0][13] = put_rv40_qpel16_mc13_c;
00539     c->put_pixels_tab[0][14] = put_rv40_qpel16_mc23_c;
00540     c->put_pixels_tab[0][15] = ff_put_rv40_qpel16_mc33_c;
00541     c->avg_pixels_tab[0][ 0] = dsp->avg_h264_qpel_pixels_tab[0][0];
00542     c->avg_pixels_tab[0][ 1] = avg_rv40_qpel16_mc10_c;
00543     c->avg_pixels_tab[0][ 2] = dsp->avg_h264_qpel_pixels_tab[0][2];
00544     c->avg_pixels_tab[0][ 3] = avg_rv40_qpel16_mc30_c;
00545     c->avg_pixels_tab[0][ 4] = avg_rv40_qpel16_mc01_c;
00546     c->avg_pixels_tab[0][ 5] = avg_rv40_qpel16_mc11_c;
00547     c->avg_pixels_tab[0][ 6] = avg_rv40_qpel16_mc21_c;
00548     c->avg_pixels_tab[0][ 7] = avg_rv40_qpel16_mc31_c;
00549     c->avg_pixels_tab[0][ 8] = dsp->avg_h264_qpel_pixels_tab[0][8];
00550     c->avg_pixels_tab[0][ 9] = avg_rv40_qpel16_mc12_c;
00551     c->avg_pixels_tab[0][10] = avg_rv40_qpel16_mc22_c;
00552     c->avg_pixels_tab[0][11] = avg_rv40_qpel16_mc32_c;
00553     c->avg_pixels_tab[0][12] = avg_rv40_qpel16_mc03_c;
00554     c->avg_pixels_tab[0][13] = avg_rv40_qpel16_mc13_c;
00555     c->avg_pixels_tab[0][14] = avg_rv40_qpel16_mc23_c;
00556     c->avg_pixels_tab[0][15] = ff_avg_rv40_qpel16_mc33_c;
00557     c->put_pixels_tab[1][ 0] = dsp->put_h264_qpel_pixels_tab[1][0];
00558     c->put_pixels_tab[1][ 1] = put_rv40_qpel8_mc10_c;
00559     c->put_pixels_tab[1][ 2] = dsp->put_h264_qpel_pixels_tab[1][2];
00560     c->put_pixels_tab[1][ 3] = put_rv40_qpel8_mc30_c;
00561     c->put_pixels_tab[1][ 4] = put_rv40_qpel8_mc01_c;
00562     c->put_pixels_tab[1][ 5] = put_rv40_qpel8_mc11_c;
00563     c->put_pixels_tab[1][ 6] = put_rv40_qpel8_mc21_c;
00564     c->put_pixels_tab[1][ 7] = put_rv40_qpel8_mc31_c;
00565     c->put_pixels_tab[1][ 8] = dsp->put_h264_qpel_pixels_tab[1][8];
00566     c->put_pixels_tab[1][ 9] = put_rv40_qpel8_mc12_c;
00567     c->put_pixels_tab[1][10] = put_rv40_qpel8_mc22_c;
00568     c->put_pixels_tab[1][11] = put_rv40_qpel8_mc32_c;
00569     c->put_pixels_tab[1][12] = put_rv40_qpel8_mc03_c;
00570     c->put_pixels_tab[1][13] = put_rv40_qpel8_mc13_c;
00571     c->put_pixels_tab[1][14] = put_rv40_qpel8_mc23_c;
00572     c->put_pixels_tab[1][15] = ff_put_rv40_qpel8_mc33_c;
00573     c->avg_pixels_tab[1][ 0] = dsp->avg_h264_qpel_pixels_tab[1][0];
00574     c->avg_pixels_tab[1][ 1] = avg_rv40_qpel8_mc10_c;
00575     c->avg_pixels_tab[1][ 2] = dsp->avg_h264_qpel_pixels_tab[1][2];
00576     c->avg_pixels_tab[1][ 3] = avg_rv40_qpel8_mc30_c;
00577     c->avg_pixels_tab[1][ 4] = avg_rv40_qpel8_mc01_c;
00578     c->avg_pixels_tab[1][ 5] = avg_rv40_qpel8_mc11_c;
00579     c->avg_pixels_tab[1][ 6] = avg_rv40_qpel8_mc21_c;
00580     c->avg_pixels_tab[1][ 7] = avg_rv40_qpel8_mc31_c;
00581     c->avg_pixels_tab[1][ 8] = dsp->avg_h264_qpel_pixels_tab[1][8];
00582     c->avg_pixels_tab[1][ 9] = avg_rv40_qpel8_mc12_c;
00583     c->avg_pixels_tab[1][10] = avg_rv40_qpel8_mc22_c;
00584     c->avg_pixels_tab[1][11] = avg_rv40_qpel8_mc32_c;
00585     c->avg_pixels_tab[1][12] = avg_rv40_qpel8_mc03_c;
00586     c->avg_pixels_tab[1][13] = avg_rv40_qpel8_mc13_c;
00587     c->avg_pixels_tab[1][14] = avg_rv40_qpel8_mc23_c;
00588     c->avg_pixels_tab[1][15] = ff_avg_rv40_qpel8_mc33_c;
00589 
00590     c->put_chroma_pixels_tab[0] = put_rv40_chroma_mc8_c;
00591     c->put_chroma_pixels_tab[1] = put_rv40_chroma_mc4_c;
00592     c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c;
00593     c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c;
00594 
00595     c->rv40_weight_pixels_tab[0][0] = rv40_weight_func_rnd_16;
00596     c->rv40_weight_pixels_tab[0][1] = rv40_weight_func_rnd_8;
00597     c->rv40_weight_pixels_tab[1][0] = rv40_weight_func_nornd_16;
00598     c->rv40_weight_pixels_tab[1][1] = rv40_weight_func_nornd_8;
00599 
00600     c->rv40_weak_loop_filter[0]     = rv40_h_weak_loop_filter;
00601     c->rv40_weak_loop_filter[1]     = rv40_v_weak_loop_filter;
00602     c->rv40_strong_loop_filter[0]   = rv40_h_strong_loop_filter;
00603     c->rv40_strong_loop_filter[1]   = rv40_v_strong_loop_filter;
00604     c->rv40_loop_filter_strength[0] = rv40_h_loop_filter_strength;
00605     c->rv40_loop_filter_strength[1] = rv40_v_loop_filter_strength;
00606 
00607     if (HAVE_MMX)
00608         ff_rv40dsp_init_x86(c, dsp);
00609     if (HAVE_NEON)
00610         ff_rv40dsp_init_neon(c, dsp);
00611 }