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 
00031 #define RV40_LOWPASS(OPNAME, OP) \
00032 static av_unused void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
00033                                                      const int h, const int C1, const int C2, const int SHIFT){\
00034     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00035     int i;\
00036     for(i = 0; i < h; i++)\
00037     {\
00038         OP(dst[0], (src[-2] + src[ 3] - 5*(src[-1]+src[2]) + src[0]*C1 + src[1]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00039         OP(dst[1], (src[-1] + src[ 4] - 5*(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00040         OP(dst[2], (src[ 0] + src[ 5] - 5*(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00041         OP(dst[3], (src[ 1] + src[ 6] - 5*(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00042         OP(dst[4], (src[ 2] + src[ 7] - 5*(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00043         OP(dst[5], (src[ 3] + src[ 8] - 5*(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00044         OP(dst[6], (src[ 4] + src[ 9] - 5*(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00045         OP(dst[7], (src[ 5] + src[10] - 5*(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00046         dst += dstStride;\
00047         src += srcStride;\
00048     }\
00049 }\
00050 \
00051 static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
00052                                            const int w, const int C1, const int C2, const int SHIFT){\
00053     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
00054     int i;\
00055     for(i = 0; i < w; i++)\
00056     {\
00057         const int srcB  = src[-2*srcStride];\
00058         const int srcA  = src[-1*srcStride];\
00059         const int src0  = src[0 *srcStride];\
00060         const int src1  = src[1 *srcStride];\
00061         const int src2  = src[2 *srcStride];\
00062         const int src3  = src[3 *srcStride];\
00063         const int src4  = src[4 *srcStride];\
00064         const int src5  = src[5 *srcStride];\
00065         const int src6  = src[6 *srcStride];\
00066         const int src7  = src[7 *srcStride];\
00067         const int src8  = src[8 *srcStride];\
00068         const int src9  = src[9 *srcStride];\
00069         const int src10 = src[10*srcStride];\
00070         OP(dst[0*dstStride], (srcB + src3  - 5*(srcA+src2) + src0*C1 + src1*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00071         OP(dst[1*dstStride], (srcA + src4  - 5*(src0+src3) + src1*C1 + src2*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00072         OP(dst[2*dstStride], (src0 + src5  - 5*(src1+src4) + src2*C1 + src3*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00073         OP(dst[3*dstStride], (src1 + src6  - 5*(src2+src5) + src3*C1 + src4*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00074         OP(dst[4*dstStride], (src2 + src7  - 5*(src3+src6) + src4*C1 + src5*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00075         OP(dst[5*dstStride], (src3 + src8  - 5*(src4+src7) + src5*C1 + src6*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00076         OP(dst[6*dstStride], (src4 + src9  - 5*(src5+src8) + src6*C1 + src7*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00077         OP(dst[7*dstStride], (src5 + src10 - 5*(src6+src9) + src7*C1 + src8*C2 + (1<<(SHIFT-1))) >> SHIFT);\
00078         dst++;\
00079         src++;\
00080     }\
00081 }\
00082 \
00083 static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
00084                                             const int w, const int C1, const int C2, const int SHIFT){\
00085     OPNAME ## rv40_qpel8_v_lowpass(dst  , src  , dstStride, srcStride, 8, C1, C2, SHIFT);\
00086     OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
00087     src += 8*srcStride;\
00088     dst += 8*dstStride;\
00089     OPNAME ## rv40_qpel8_v_lowpass(dst  , src  , dstStride, srcStride, w-8, C1, C2, SHIFT);\
00090     OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, w-8, C1, C2, SHIFT);\
00091 }\
00092 \
00093 static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
00094                                             const int h, const int C1, const int C2, const int SHIFT){\
00095     OPNAME ## rv40_qpel8_h_lowpass(dst  , src  , dstStride, srcStride, 8, C1, C2, SHIFT);\
00096     OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
00097     src += 8*srcStride;\
00098     dst += 8*dstStride;\
00099     OPNAME ## rv40_qpel8_h_lowpass(dst  , src  , dstStride, srcStride, h-8, C1, C2, SHIFT);\
00100     OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, h-8, C1, C2, SHIFT);\
00101 }\
00102 \
00103 
00104 #define RV40_MC(OPNAME, SIZE) \
00105 static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
00106     OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
00107 }\
00108 \
00109 static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
00110     OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
00111 }\
00112 \
00113 static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
00114     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
00115 }\
00116 \
00117 static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
00118     uint8_t full[SIZE*(SIZE+5)];\
00119     uint8_t * const full_mid = full + SIZE*2;\
00120     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
00121     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
00122 }\
00123 \
00124 static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
00125     uint8_t full[SIZE*(SIZE+5)];\
00126     uint8_t * const full_mid = full + SIZE*2;\
00127     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
00128     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
00129 }\
00130 \
00131 static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
00132     uint8_t full[SIZE*(SIZE+5)];\
00133     uint8_t * const full_mid = full + SIZE*2;\
00134     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
00135     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
00136 }\
00137 \
00138 static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
00139     uint8_t full[SIZE*(SIZE+5)];\
00140     uint8_t * const full_mid = full + SIZE*2;\
00141     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
00142     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
00143 }\
00144 \
00145 static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
00146     uint8_t full[SIZE*(SIZE+5)];\
00147     uint8_t * const full_mid = full + SIZE*2;\
00148     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
00149     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
00150 }\
00151 \
00152 static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
00153     uint8_t full[SIZE*(SIZE+5)];\
00154     uint8_t * const full_mid = full + SIZE*2;\
00155     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
00156     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
00157 }\
00158 \
00159 static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
00160     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
00161 }\
00162 \
00163 static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
00164     uint8_t full[SIZE*(SIZE+5)];\
00165     uint8_t * const full_mid = full + SIZE*2;\
00166     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
00167     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
00168 }\
00169 \
00170 static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
00171     uint8_t full[SIZE*(SIZE+5)];\
00172     uint8_t * const full_mid = full + SIZE*2;\
00173     put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
00174     OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
00175 }\
00176 \
00177 
00178 #define op_avg(a, b)  a = (((a)+cm[b]+1)>>1)
00179 #define op_put(a, b)  a = cm[b]
00180 
00181 RV40_LOWPASS(put_       , op_put)
00182 RV40_LOWPASS(avg_       , op_avg)
00183 
00184 #undef op_avg
00185 #undef op_put
00186 
00187 RV40_MC(put_, 8)
00188 RV40_MC(put_, 16)
00189 RV40_MC(avg_, 8)
00190 RV40_MC(avg_, 16)
00191 
00192 static const int rv40_bias[4][4] = {
00193     {  0, 16, 32, 16 },
00194     { 32, 28, 32, 28 },
00195     {  0, 32, 16, 32 },
00196     { 32, 28, 32, 28 }
00197 };
00198 
00199 #define RV40_CHROMA_MC(OPNAME, OP)\
00200 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){\
00201     const int A = (8-x) * (8-y);\
00202     const int B = (  x) * (8-y);\
00203     const int C = (8-x) * (  y);\
00204     const int D = (  x) * (  y);\
00205     int i;\
00206     int bias = rv40_bias[y>>1][x>>1];\
00207     \
00208     assert(x<8 && y<8 && x>=0 && y>=0);\
00209 \
00210     if(D){\
00211         for(i = 0; i < h; i++){\
00212             OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
00213             OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
00214             OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
00215             OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
00216             dst += stride;\
00217             src += stride;\
00218         }\
00219     }else{\
00220         const int E = B + C;\
00221         const int step = C ? stride : 1;\
00222         for(i = 0; i < h; i++){\
00223             OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
00224             OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
00225             OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
00226             OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
00227             dst += stride;\
00228             src += stride;\
00229         }\
00230     }\
00231 }\
00232 \
00233 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){\
00234     const int A = (8-x) * (8-y);\
00235     const int B = (  x) * (8-y);\
00236     const int C = (8-x) * (  y);\
00237     const int D = (  x) * (  y);\
00238     int i;\
00239     int bias = rv40_bias[y>>1][x>>1];\
00240     \
00241     assert(x<8 && y<8 && x>=0 && y>=0);\
00242 \
00243     if(D){\
00244         for(i = 0; i < h; i++){\
00245             OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
00246             OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
00247             OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
00248             OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
00249             OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + bias));\
00250             OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + bias));\
00251             OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + bias));\
00252             OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + bias));\
00253             dst += stride;\
00254             src += stride;\
00255         }\
00256     }else{\
00257         const int E = B + C;\
00258         const int step = C ? stride : 1;\
00259         for(i = 0; i < h; i++){\
00260             OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
00261             OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
00262             OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
00263             OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
00264             OP(dst[4], (A*src[4] + E*src[step+4] + bias));\
00265             OP(dst[5], (A*src[5] + E*src[step+5] + bias));\
00266             OP(dst[6], (A*src[6] + E*src[step+6] + bias));\
00267             OP(dst[7], (A*src[7] + E*src[step+7] + bias));\
00268             dst += stride;\
00269             src += stride;\
00270         }\
00271     }\
00272 }
00273 
00274 #define op_avg(a, b) a = (((a)+((b)>>6)+1)>>1)
00275 #define op_put(a, b) a = ((b)>>6)
00276 
00277 RV40_CHROMA_MC(put_, op_put)
00278 RV40_CHROMA_MC(avg_, op_avg)
00279 
00280 #define RV40_WEIGHT_FUNC(size) \
00281 static void rv40_weight_func_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, int stride)\
00282 {\
00283     int i, j;\
00284 \
00285     for (j = 0; j < size; j++) {\
00286         for (i = 0; i < size; i++)\
00287             dst[i] = (((w2 * src1[i]) >> 9) + ((w1 * src2[i]) >> 9) + 0x10) >> 5;\
00288         src1 += stride;\
00289         src2 += stride;\
00290         dst  += stride;\
00291     }\
00292 }
00293 
00294 RV40_WEIGHT_FUNC(16)
00295 RV40_WEIGHT_FUNC(8)
00296 
00300 static const uint8_t rv40_dither_l[16] = {
00301     0x40, 0x50, 0x20, 0x60, 0x30, 0x50, 0x40, 0x30,
00302     0x50, 0x40, 0x50, 0x30, 0x60, 0x20, 0x50, 0x40
00303 };
00304 
00308 static const uint8_t rv40_dither_r[16] = {
00309     0x40, 0x30, 0x60, 0x20, 0x50, 0x30, 0x30, 0x40,
00310     0x40, 0x40, 0x50, 0x30, 0x20, 0x60, 0x30, 0x40
00311 };
00312 
00313 #define CLIP_SYMM(a, b) av_clip(a, -(b), b)
00314 
00317 static av_always_inline void rv40_weak_loop_filter(uint8_t *src,
00318                                                    const int step,
00319                                                    const int stride,
00320                                                    const int filter_p1,
00321                                                    const int filter_q1,
00322                                                    const int alpha,
00323                                                    const int beta,
00324                                                    const int lim_p0q0,
00325                                                    const int lim_q1,
00326                                                    const int lim_p1)
00327 {
00328     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
00329     int i, t, u, diff;
00330 
00331     for (i = 0; i < 4; i++, src += stride) {
00332         int diff_p1p0 = src[-2*step] - src[-1*step];
00333         int diff_q1q0 = src[ 1*step] - src[ 0*step];
00334         int diff_p1p2 = src[-2*step] - src[-3*step];
00335         int diff_q1q2 = src[ 1*step] - src[ 2*step];
00336 
00337         t = src[0*step] - src[-1*step];
00338         if (!t)
00339             continue;
00340 
00341         u = (alpha * FFABS(t)) >> 7;
00342         if (u > 3 - (filter_p1 && filter_q1))
00343             continue;
00344 
00345         t <<= 2;
00346         if (filter_p1 && filter_q1)
00347             t += src[-2*step] - src[1*step];
00348 
00349         diff = CLIP_SYMM((t + 4) >> 3, lim_p0q0);
00350         src[-1*step] = cm[src[-1*step] + diff];
00351         src[ 0*step] = cm[src[ 0*step] - diff];
00352 
00353         if (filter_p1 && FFABS(diff_p1p2) <= beta) {
00354             t = (diff_p1p0 + diff_p1p2 - diff) >> 1;
00355             src[-2*step] = cm[src[-2*step] - CLIP_SYMM(t, lim_p1)];
00356         }
00357 
00358         if (filter_q1 && FFABS(diff_q1q2) <= beta) {
00359             t = (diff_q1q0 + diff_q1q2 + diff) >> 1;
00360             src[ 1*step] = cm[src[ 1*step] - CLIP_SYMM(t, lim_q1)];
00361         }
00362     }
00363 }
00364 
00365 static void rv40_h_weak_loop_filter(uint8_t *src, const int stride,
00366                                     const int filter_p1, const int filter_q1,
00367                                     const int alpha, const int beta,
00368                                     const int lim_p0q0, const int lim_q1,
00369                                     const int lim_p1)
00370 {
00371     rv40_weak_loop_filter(src, stride, 1, filter_p1, filter_q1,
00372                           alpha, beta, lim_p0q0, lim_q1, lim_p1);
00373 }
00374 
00375 static void rv40_v_weak_loop_filter(uint8_t *src, const int stride,
00376                                     const int filter_p1, const int filter_q1,
00377                                     const int alpha, const int beta,
00378                                     const int lim_p0q0, const int lim_q1,
00379                                     const int lim_p1)
00380 {
00381     rv40_weak_loop_filter(src, 1, stride, filter_p1, filter_q1,
00382                           alpha, beta, lim_p0q0, lim_q1, lim_p1);
00383 }
00384 
00385 static av_always_inline void rv40_strong_loop_filter(uint8_t *src,
00386                                                      const int step,
00387                                                      const int stride,
00388                                                      const int alpha,
00389                                                      const int lims,
00390                                                      const int dmode,
00391                                                      const int chroma)
00392 {
00393     int i;
00394 
00395     for(i = 0; i < 4; i++, src += stride){
00396         int sflag, p0, q0, p1, q1;
00397         int t = src[0*step] - src[-1*step];
00398 
00399         if (!t)
00400             continue;
00401 
00402         sflag = (alpha * FFABS(t)) >> 7;
00403         if (sflag > 1)
00404             continue;
00405 
00406         p0 = (25*src[-3*step] + 26*src[-2*step] + 26*src[-1*step] +
00407               26*src[ 0*step] + 25*src[ 1*step] +
00408               rv40_dither_l[dmode + i]) >> 7;
00409 
00410         q0 = (25*src[-2*step] + 26*src[-1*step] + 26*src[ 0*step] +
00411               26*src[ 1*step] + 25*src[ 2*step] +
00412               rv40_dither_r[dmode + i]) >> 7;
00413 
00414         if (sflag) {
00415             p0 = av_clip(p0, src[-1*step] - lims, src[-1*step] + lims);
00416             q0 = av_clip(q0, src[ 0*step] - lims, src[ 0*step] + lims);
00417         }
00418 
00419         p1 = (25*src[-4*step] + 26*src[-3*step] + 26*src[-2*step] + 26*p0 +
00420               25*src[ 0*step] + rv40_dither_l[dmode + i]) >> 7;
00421         q1 = (25*src[-1*step] + 26*q0 + 26*src[ 1*step] + 26*src[ 2*step] +
00422               25*src[ 3*step] + rv40_dither_r[dmode + i]) >> 7;
00423 
00424         if (sflag) {
00425             p1 = av_clip(p1, src[-2*step] - lims, src[-2*step] + lims);
00426             q1 = av_clip(q1, src[ 1*step] - lims, src[ 1*step] + lims);
00427         }
00428 
00429         src[-2*step] = p1;
00430         src[-1*step] = p0;
00431         src[ 0*step] = q0;
00432         src[ 1*step] = q1;
00433 
00434         if(!chroma){
00435             src[-3*step] = (25*src[-1*step] + 26*src[-2*step] +
00436                             51*src[-3*step] + 26*src[-4*step] + 64) >> 7;
00437             src[ 2*step] = (25*src[ 0*step] + 26*src[ 1*step] +
00438                             51*src[ 2*step] + 26*src[ 3*step] + 64) >> 7;
00439         }
00440     }
00441 }
00442 
00443 static void rv40_h_strong_loop_filter(uint8_t *src, const int stride,
00444                                       const int alpha, const int lims,
00445                                       const int dmode, const int chroma)
00446 {
00447     rv40_strong_loop_filter(src, stride, 1, alpha, lims, dmode, chroma);
00448 }
00449 
00450 static void rv40_v_strong_loop_filter(uint8_t *src, const int stride,
00451                                       const int alpha, const int lims,
00452                                       const int dmode, const int chroma)
00453 {
00454     rv40_strong_loop_filter(src, 1, stride, alpha, lims, dmode, chroma);
00455 }
00456 
00457 static av_always_inline int rv40_loop_filter_strength(uint8_t *src,
00458                                                       int step, int stride,
00459                                                       int beta, int beta2,
00460                                                       int edge,
00461                                                       int *p1, int *q1)
00462 {
00463     int sum_p1p0 = 0, sum_q1q0 = 0, sum_p1p2 = 0, sum_q1q2 = 0;
00464     int strong0 = 0, strong1 = 0;
00465     uint8_t *ptr;
00466     int i;
00467 
00468     for (i = 0, ptr = src; i < 4; i++, ptr += stride) {
00469         sum_p1p0 += ptr[-2*step] - ptr[-1*step];
00470         sum_q1q0 += ptr[ 1*step] - ptr[ 0*step];
00471     }
00472 
00473     *p1 = FFABS(sum_p1p0) < (beta << 2);
00474     *q1 = FFABS(sum_q1q0) < (beta << 2);
00475 
00476     if(!*p1 && !*q1)
00477         return 0;
00478 
00479     if (!edge)
00480         return 0;
00481 
00482     for (i = 0, ptr = src; i < 4; i++, ptr += stride) {
00483         sum_p1p2 += ptr[-2*step] - ptr[-3*step];
00484         sum_q1q2 += ptr[ 1*step] - ptr[ 2*step];
00485     }
00486 
00487     strong0 = *p1 && (FFABS(sum_p1p2) < beta2);
00488     strong1 = *q1 && (FFABS(sum_q1q2) < beta2);
00489 
00490     return strong0 && strong1;
00491 }
00492 
00493 static int rv40_h_loop_filter_strength(uint8_t *src, int stride,
00494                                        int beta, int beta2, int edge,
00495                                        int *p1, int *q1)
00496 {
00497     return rv40_loop_filter_strength(src, stride, 1, beta, beta2, edge, p1, q1);
00498 }
00499 
00500 static int rv40_v_loop_filter_strength(uint8_t *src, int stride,
00501                                        int beta, int beta2, int edge,
00502                                        int *p1, int *q1)
00503 {
00504     return rv40_loop_filter_strength(src, 1, stride, beta, beta2, edge, p1, q1);
00505 }
00506 
00507 av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {
00508 
00509     ff_rv34dsp_init(c, dsp);
00510 
00511     c->put_pixels_tab[0][ 0] = dsp->put_h264_qpel_pixels_tab[0][0];
00512     c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c;
00513     c->put_pixels_tab[0][ 2] = dsp->put_h264_qpel_pixels_tab[0][2];
00514     c->put_pixels_tab[0][ 3] = put_rv40_qpel16_mc30_c;
00515     c->put_pixels_tab[0][ 4] = put_rv40_qpel16_mc01_c;
00516     c->put_pixels_tab[0][ 5] = put_rv40_qpel16_mc11_c;
00517     c->put_pixels_tab[0][ 6] = put_rv40_qpel16_mc21_c;
00518     c->put_pixels_tab[0][ 7] = put_rv40_qpel16_mc31_c;
00519     c->put_pixels_tab[0][ 8] = dsp->put_h264_qpel_pixels_tab[0][8];
00520     c->put_pixels_tab[0][ 9] = put_rv40_qpel16_mc12_c;
00521     c->put_pixels_tab[0][10] = put_rv40_qpel16_mc22_c;
00522     c->put_pixels_tab[0][11] = put_rv40_qpel16_mc32_c;
00523     c->put_pixels_tab[0][12] = put_rv40_qpel16_mc03_c;
00524     c->put_pixels_tab[0][13] = put_rv40_qpel16_mc13_c;
00525     c->put_pixels_tab[0][14] = put_rv40_qpel16_mc23_c;
00526     c->put_pixels_tab[0][15] = ff_put_rv40_qpel16_mc33_c;
00527     c->avg_pixels_tab[0][ 0] = dsp->avg_h264_qpel_pixels_tab[0][0];
00528     c->avg_pixels_tab[0][ 1] = avg_rv40_qpel16_mc10_c;
00529     c->avg_pixels_tab[0][ 2] = dsp->avg_h264_qpel_pixels_tab[0][2];
00530     c->avg_pixels_tab[0][ 3] = avg_rv40_qpel16_mc30_c;
00531     c->avg_pixels_tab[0][ 4] = avg_rv40_qpel16_mc01_c;
00532     c->avg_pixels_tab[0][ 5] = avg_rv40_qpel16_mc11_c;
00533     c->avg_pixels_tab[0][ 6] = avg_rv40_qpel16_mc21_c;
00534     c->avg_pixels_tab[0][ 7] = avg_rv40_qpel16_mc31_c;
00535     c->avg_pixels_tab[0][ 8] = dsp->avg_h264_qpel_pixels_tab[0][8];
00536     c->avg_pixels_tab[0][ 9] = avg_rv40_qpel16_mc12_c;
00537     c->avg_pixels_tab[0][10] = avg_rv40_qpel16_mc22_c;
00538     c->avg_pixels_tab[0][11] = avg_rv40_qpel16_mc32_c;
00539     c->avg_pixels_tab[0][12] = avg_rv40_qpel16_mc03_c;
00540     c->avg_pixels_tab[0][13] = avg_rv40_qpel16_mc13_c;
00541     c->avg_pixels_tab[0][14] = avg_rv40_qpel16_mc23_c;
00542     c->avg_pixels_tab[0][15] = ff_avg_rv40_qpel16_mc33_c;
00543     c->put_pixels_tab[1][ 0] = dsp->put_h264_qpel_pixels_tab[1][0];
00544     c->put_pixels_tab[1][ 1] = put_rv40_qpel8_mc10_c;
00545     c->put_pixels_tab[1][ 2] = dsp->put_h264_qpel_pixels_tab[1][2];
00546     c->put_pixels_tab[1][ 3] = put_rv40_qpel8_mc30_c;
00547     c->put_pixels_tab[1][ 4] = put_rv40_qpel8_mc01_c;
00548     c->put_pixels_tab[1][ 5] = put_rv40_qpel8_mc11_c;
00549     c->put_pixels_tab[1][ 6] = put_rv40_qpel8_mc21_c;
00550     c->put_pixels_tab[1][ 7] = put_rv40_qpel8_mc31_c;
00551     c->put_pixels_tab[1][ 8] = dsp->put_h264_qpel_pixels_tab[1][8];
00552     c->put_pixels_tab[1][ 9] = put_rv40_qpel8_mc12_c;
00553     c->put_pixels_tab[1][10] = put_rv40_qpel8_mc22_c;
00554     c->put_pixels_tab[1][11] = put_rv40_qpel8_mc32_c;
00555     c->put_pixels_tab[1][12] = put_rv40_qpel8_mc03_c;
00556     c->put_pixels_tab[1][13] = put_rv40_qpel8_mc13_c;
00557     c->put_pixels_tab[1][14] = put_rv40_qpel8_mc23_c;
00558     c->put_pixels_tab[1][15] = ff_put_rv40_qpel8_mc33_c;
00559     c->avg_pixels_tab[1][ 0] = dsp->avg_h264_qpel_pixels_tab[1][0];
00560     c->avg_pixels_tab[1][ 1] = avg_rv40_qpel8_mc10_c;
00561     c->avg_pixels_tab[1][ 2] = dsp->avg_h264_qpel_pixels_tab[1][2];
00562     c->avg_pixels_tab[1][ 3] = avg_rv40_qpel8_mc30_c;
00563     c->avg_pixels_tab[1][ 4] = avg_rv40_qpel8_mc01_c;
00564     c->avg_pixels_tab[1][ 5] = avg_rv40_qpel8_mc11_c;
00565     c->avg_pixels_tab[1][ 6] = avg_rv40_qpel8_mc21_c;
00566     c->avg_pixels_tab[1][ 7] = avg_rv40_qpel8_mc31_c;
00567     c->avg_pixels_tab[1][ 8] = dsp->avg_h264_qpel_pixels_tab[1][8];
00568     c->avg_pixels_tab[1][ 9] = avg_rv40_qpel8_mc12_c;
00569     c->avg_pixels_tab[1][10] = avg_rv40_qpel8_mc22_c;
00570     c->avg_pixels_tab[1][11] = avg_rv40_qpel8_mc32_c;
00571     c->avg_pixels_tab[1][12] = avg_rv40_qpel8_mc03_c;
00572     c->avg_pixels_tab[1][13] = avg_rv40_qpel8_mc13_c;
00573     c->avg_pixels_tab[1][14] = avg_rv40_qpel8_mc23_c;
00574     c->avg_pixels_tab[1][15] = ff_avg_rv40_qpel8_mc33_c;
00575 
00576     c->put_chroma_pixels_tab[0] = put_rv40_chroma_mc8_c;
00577     c->put_chroma_pixels_tab[1] = put_rv40_chroma_mc4_c;
00578     c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c;
00579     c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c;
00580 
00581     c->rv40_weight_pixels_tab[0] = rv40_weight_func_16;
00582     c->rv40_weight_pixels_tab[1] = rv40_weight_func_8;
00583 
00584     c->rv40_weak_loop_filter[0]     = rv40_h_weak_loop_filter;
00585     c->rv40_weak_loop_filter[1]     = rv40_v_weak_loop_filter;
00586     c->rv40_strong_loop_filter[0]   = rv40_h_strong_loop_filter;
00587     c->rv40_strong_loop_filter[1]   = rv40_v_strong_loop_filter;
00588     c->rv40_loop_filter_strength[0] = rv40_h_loop_filter_strength;
00589     c->rv40_loop_filter_strength[1] = rv40_v_loop_filter_strength;
00590 
00591     if (HAVE_MMX)
00592         ff_rv40dsp_init_x86(c, dsp);
00593     if (HAVE_NEON)
00594         ff_rv40dsp_init_neon(c, dsp);
00595 }