Mercurial > cgi-bin > hgwebdir.cgi > PR > Applications > VSs > VSs__H264__App
diff libavcodec/dsputil.c @ 2:897f711a7157
rearrange to work with autoconf
| author | Nina Engelhardt <nengel@mailbox.tu-berlin.de> |
|---|---|
| date | Tue, 25 Sep 2012 15:55:33 +0200 |
| parents | |
| children |
line diff
1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 1.2 +++ b/libavcodec/dsputil.c Tue Sep 25 15:55:33 2012 +0200 1.3 @@ -0,0 +1,1057 @@ 1.4 +/* 1.5 + * DSP utils 1.6 + * Copyright (c) 2000, 2001 Fabrice Bellard 1.7 + * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at> 1.8 + * 1.9 + * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer <michaelni@gmx.at> 1.10 + * 1.11 + * This file is part of FFmpeg. 1.12 + * 1.13 + * FFmpeg is free software; you can redistribute it and/or 1.14 + * modify it under the terms of the GNU Lesser General Public 1.15 + * License as published by the Free Software Foundation; either 1.16 + * version 2.1 of the License, or (at your option) any later version. 1.17 + * 1.18 + * FFmpeg is distributed in the hope that it will be useful, 1.19 + * but WITHOUT ANY WARRANTY; without even the implied warranty of 1.20 + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 1.21 + * Lesser General Public License for more details. 1.22 + * 1.23 + * You should have received a copy of the GNU Lesser General Public 1.24 + * License along with FFmpeg; if not, write to the Free Software 1.25 + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA 1.26 + */ 1.27 + 1.28 +/** 1.29 + * @file 1.30 + * DSP utils 1.31 + */ 1.32 + 1.33 +#include "libavutil/log.h" 1.34 +#include "dsputil.h" 1.35 +#include "simple_idct.h" 1.36 +#include "mathops.h" 1.37 +#include "config.h" 1.38 + 1.39 +uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP] = {0, }; 1.40 +uint32_t ff_squareTbl[512] = {0, }; 1.41 + 1.42 +const uint8_t ff_zigzag_direct[64] = { 1.43 + 0, 1, 8, 16, 9, 2, 3, 10, 1.44 + 17, 24, 32, 25, 18, 11, 4, 5, 1.45 + 12, 19, 26, 33, 40, 48, 41, 34, 1.46 + 27, 20, 13, 6, 7, 14, 21, 28, 1.47 + 35, 42, 49, 56, 57, 50, 43, 36, 1.48 + 29, 22, 15, 23, 30, 37, 44, 51, 1.49 + 58, 59, 52, 45, 38, 31, 39, 46, 1.50 + 53, 60, 61, 54, 47, 55, 62, 63 1.51 +}; 1.52 + 1.53 + 1.54 +#define PIXOP2(OPNAME, OP) \ 1.55 +static void OPNAME ## _pixels2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.56 + int i;\ 1.57 + for(i=0; i<h; i++){\ 1.58 + OP(*((uint16_t*)(block )), AV_RN16(pixels ));\ 1.59 + pixels+=line_size;\ 1.60 + block +=line_size;\ 1.61 + }\ 1.62 +}\ 1.63 +static void OPNAME ## _pixels4_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.64 + int i;\ 1.65 + for(i=0; i<h; i++){\ 1.66 + OP(*((uint32_t*)(block )), AV_RN32(pixels ));\ 1.67 + pixels+=line_size;\ 1.68 + block +=line_size;\ 1.69 + }\ 1.70 +}\ 1.71 +static void OPNAME ## _pixels8_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.72 + int i;\ 1.73 + for(i=0; i<h; i++){\ 1.74 + OP(*((uint32_t*)(block )), AV_RN32(pixels ));\ 1.75 + OP(*((uint32_t*)(block+4)), AV_RN32(pixels+4));\ 1.76 + pixels+=line_size;\ 1.77 + block +=line_size;\ 1.78 + }\ 1.79 +}\ 1.80 +static inline void OPNAME ## _no_rnd_pixels8_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.81 + OPNAME ## _pixels8_c(block, pixels, line_size, h);\ 1.82 +}\ 1.83 +\ 1.84 +static inline void OPNAME ## _no_rnd_pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \ 1.85 + int src_stride1, int src_stride2, int h){\ 1.86 + int i;\ 1.87 + for(i=0; i<h; i++){\ 1.88 + uint32_t a,b;\ 1.89 + a= AV_RN32(&src1[i*src_stride1 ]);\ 1.90 + b= AV_RN32(&src2[i*src_stride2 ]);\ 1.91 + OP(*((uint32_t*)&dst[i*dst_stride ]), no_rnd_avg32(a, b));\ 1.92 + a= AV_RN32(&src1[i*src_stride1+4]);\ 1.93 + b= AV_RN32(&src2[i*src_stride2+4]);\ 1.94 + OP(*((uint32_t*)&dst[i*dst_stride+4]), no_rnd_avg32(a, b));\ 1.95 + }\ 1.96 +}\ 1.97 +\ 1.98 +static inline void OPNAME ## _pixels8_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \ 1.99 + int src_stride1, int src_stride2, int h){\ 1.100 + int i;\ 1.101 + for(i=0; i<h; i++){\ 1.102 + uint32_t a,b;\ 1.103 + a= AV_RN32(&src1[i*src_stride1 ]);\ 1.104 + b= AV_RN32(&src2[i*src_stride2 ]);\ 1.105 + OP(*((uint32_t*)&dst[i*dst_stride ]), rnd_avg32(a, b));\ 1.106 + a= AV_RN32(&src1[i*src_stride1+4]);\ 1.107 + b= AV_RN32(&src2[i*src_stride2+4]);\ 1.108 + OP(*((uint32_t*)&dst[i*dst_stride+4]), rnd_avg32(a, b));\ 1.109 + }\ 1.110 +}\ 1.111 +\ 1.112 +static inline void OPNAME ## _pixels4_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \ 1.113 + int src_stride1, int src_stride2, int h){\ 1.114 + int i;\ 1.115 + for(i=0; i<h; i++){\ 1.116 + uint32_t a,b;\ 1.117 + a= AV_RN32(&src1[i*src_stride1 ]);\ 1.118 + b= AV_RN32(&src2[i*src_stride2 ]);\ 1.119 + OP(*((uint32_t*)&dst[i*dst_stride ]), rnd_avg32(a, b));\ 1.120 + }\ 1.121 +}\ 1.122 +\ 1.123 +static inline void OPNAME ## _pixels2_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \ 1.124 + int src_stride1, int src_stride2, int h){\ 1.125 + int i;\ 1.126 + for(i=0; i<h; i++){\ 1.127 + uint32_t a,b;\ 1.128 + a= AV_RN16(&src1[i*src_stride1 ]);\ 1.129 + b= AV_RN16(&src2[i*src_stride2 ]);\ 1.130 + OP(*((uint16_t*)&dst[i*dst_stride ]), rnd_avg32(a, b));\ 1.131 + }\ 1.132 +}\ 1.133 +\ 1.134 +static inline void OPNAME ## _pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \ 1.135 + int src_stride1, int src_stride2, int h){\ 1.136 + OPNAME ## _pixels8_l2(dst , src1 , src2 , dst_stride, src_stride1, src_stride2, h);\ 1.137 + OPNAME ## _pixels8_l2(dst+8, src1+8, src2+8, dst_stride, src_stride1, src_stride2, h);\ 1.138 +}\ 1.139 +\ 1.140 +static inline void OPNAME ## _no_rnd_pixels16_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \ 1.141 + int src_stride1, int src_stride2, int h){\ 1.142 + OPNAME ## _no_rnd_pixels8_l2(dst , src1 , src2 , dst_stride, src_stride1, src_stride2, h);\ 1.143 + OPNAME ## _no_rnd_pixels8_l2(dst+8, src1+8, src2+8, dst_stride, src_stride1, src_stride2, h);\ 1.144 +}\ 1.145 +\ 1.146 +static inline void OPNAME ## _no_rnd_pixels8_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.147 + OPNAME ## _no_rnd_pixels8_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\ 1.148 +}\ 1.149 +\ 1.150 +static inline void OPNAME ## _pixels8_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.151 + OPNAME ## _pixels8_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\ 1.152 +}\ 1.153 +\ 1.154 +static inline void OPNAME ## _no_rnd_pixels8_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.155 + OPNAME ## _no_rnd_pixels8_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\ 1.156 +}\ 1.157 +\ 1.158 +static inline void OPNAME ## _pixels8_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.159 + OPNAME ## _pixels8_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\ 1.160 +}\ 1.161 +\ 1.162 +static inline void OPNAME ## _pixels8_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\ 1.163 + int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\ 1.164 + int i;\ 1.165 + for(i=0; i<h; i++){\ 1.166 + uint32_t a, b, c, d, l0, l1, h0, h1;\ 1.167 + a= AV_RN32(&src1[i*src_stride1]);\ 1.168 + b= AV_RN32(&src2[i*src_stride2]);\ 1.169 + c= AV_RN32(&src3[i*src_stride3]);\ 1.170 + d= AV_RN32(&src4[i*src_stride4]);\ 1.171 + l0= (a&0x03030303UL)\ 1.172 + + (b&0x03030303UL)\ 1.173 + + 0x02020202UL;\ 1.174 + h0= ((a&0xFCFCFCFCUL)>>2)\ 1.175 + + ((b&0xFCFCFCFCUL)>>2);\ 1.176 + l1= (c&0x03030303UL)\ 1.177 + + (d&0x03030303UL);\ 1.178 + h1= ((c&0xFCFCFCFCUL)>>2)\ 1.179 + + ((d&0xFCFCFCFCUL)>>2);\ 1.180 + OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.181 + a= AV_RN32(&src1[i*src_stride1+4]);\ 1.182 + b= AV_RN32(&src2[i*src_stride2+4]);\ 1.183 + c= AV_RN32(&src3[i*src_stride3+4]);\ 1.184 + d= AV_RN32(&src4[i*src_stride4+4]);\ 1.185 + l0= (a&0x03030303UL)\ 1.186 + + (b&0x03030303UL)\ 1.187 + + 0x02020202UL;\ 1.188 + h0= ((a&0xFCFCFCFCUL)>>2)\ 1.189 + + ((b&0xFCFCFCFCUL)>>2);\ 1.190 + l1= (c&0x03030303UL)\ 1.191 + + (d&0x03030303UL);\ 1.192 + h1= ((c&0xFCFCFCFCUL)>>2)\ 1.193 + + ((d&0xFCFCFCFCUL)>>2);\ 1.194 + OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.195 + }\ 1.196 +}\ 1.197 +\ 1.198 +static inline void OPNAME ## _pixels4_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.199 + OPNAME ## _pixels4_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\ 1.200 +}\ 1.201 +\ 1.202 +static inline void OPNAME ## _pixels4_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.203 + OPNAME ## _pixels4_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\ 1.204 +}\ 1.205 +\ 1.206 +static inline void OPNAME ## _pixels2_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.207 + OPNAME ## _pixels2_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\ 1.208 +}\ 1.209 +\ 1.210 +static inline void OPNAME ## _pixels2_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ 1.211 + OPNAME ## _pixels2_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\ 1.212 +}\ 1.213 +\ 1.214 +static inline void OPNAME ## _no_rnd_pixels8_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\ 1.215 + int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\ 1.216 + int i;\ 1.217 + for(i=0; i<h; i++){\ 1.218 + uint32_t a, b, c, d, l0, l1, h0, h1;\ 1.219 + a= AV_RN32(&src1[i*src_stride1]);\ 1.220 + b= AV_RN32(&src2[i*src_stride2]);\ 1.221 + c= AV_RN32(&src3[i*src_stride3]);\ 1.222 + d= AV_RN32(&src4[i*src_stride4]);\ 1.223 + l0= (a&0x03030303UL)\ 1.224 + + (b&0x03030303UL)\ 1.225 + + 0x01010101UL;\ 1.226 + h0= ((a&0xFCFCFCFCUL)>>2)\ 1.227 + + ((b&0xFCFCFCFCUL)>>2);\ 1.228 + l1= (c&0x03030303UL)\ 1.229 + + (d&0x03030303UL);\ 1.230 + h1= ((c&0xFCFCFCFCUL)>>2)\ 1.231 + + ((d&0xFCFCFCFCUL)>>2);\ 1.232 + OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.233 + a= AV_RN32(&src1[i*src_stride1+4]);\ 1.234 + b= AV_RN32(&src2[i*src_stride2+4]);\ 1.235 + c= AV_RN32(&src3[i*src_stride3+4]);\ 1.236 + d= AV_RN32(&src4[i*src_stride4+4]);\ 1.237 + l0= (a&0x03030303UL)\ 1.238 + + (b&0x03030303UL)\ 1.239 + + 0x01010101UL;\ 1.240 + h0= ((a&0xFCFCFCFCUL)>>2)\ 1.241 + + ((b&0xFCFCFCFCUL)>>2);\ 1.242 + l1= (c&0x03030303UL)\ 1.243 + + (d&0x03030303UL);\ 1.244 + h1= ((c&0xFCFCFCFCUL)>>2)\ 1.245 + + ((d&0xFCFCFCFCUL)>>2);\ 1.246 + OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.247 + }\ 1.248 +}\ 1.249 +static inline void OPNAME ## _pixels16_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\ 1.250 + int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\ 1.251 + OPNAME ## _pixels8_l4(dst , src1 , src2 , src3 , src4 , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\ 1.252 + OPNAME ## _pixels8_l4(dst+8, src1+8, src2+8, src3+8, src4+8, dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\ 1.253 +}\ 1.254 +static inline void OPNAME ## _no_rnd_pixels16_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\ 1.255 + int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\ 1.256 + OPNAME ## _no_rnd_pixels8_l4(dst , src1 , src2 , src3 , src4 , dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\ 1.257 + OPNAME ## _no_rnd_pixels8_l4(dst+8, src1+8, src2+8, src3+8, src4+8, dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\ 1.258 +}\ 1.259 +\ 1.260 +static inline void OPNAME ## _pixels2_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ 1.261 +{\ 1.262 + int i, a0, b0, a1, b1;\ 1.263 + a0= pixels[0];\ 1.264 + b0= pixels[1] + 2;\ 1.265 + a0 += b0;\ 1.266 + b0 += pixels[2];\ 1.267 +\ 1.268 + pixels+=line_size;\ 1.269 + for(i=0; i<h; i+=2){\ 1.270 + a1= pixels[0];\ 1.271 + b1= pixels[1];\ 1.272 + a1 += b1;\ 1.273 + b1 += pixels[2];\ 1.274 +\ 1.275 + block[0]= (a1+a0)>>2; /* FIXME non put */\ 1.276 + block[1]= (b1+b0)>>2;\ 1.277 +\ 1.278 + pixels+=line_size;\ 1.279 + block +=line_size;\ 1.280 +\ 1.281 + a0= pixels[0];\ 1.282 + b0= pixels[1] + 2;\ 1.283 + a0 += b0;\ 1.284 + b0 += pixels[2];\ 1.285 +\ 1.286 + block[0]= (a1+a0)>>2;\ 1.287 + block[1]= (b1+b0)>>2;\ 1.288 + pixels+=line_size;\ 1.289 + block +=line_size;\ 1.290 + }\ 1.291 +}\ 1.292 +\ 1.293 +static inline void OPNAME ## _pixels4_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ 1.294 +{\ 1.295 + int i;\ 1.296 + const uint32_t a= AV_RN32(pixels );\ 1.297 + const uint32_t b= AV_RN32(pixels+1);\ 1.298 + uint32_t l0= (a&0x03030303UL)\ 1.299 + + (b&0x03030303UL)\ 1.300 + + 0x02020202UL;\ 1.301 + uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\ 1.302 + + ((b&0xFCFCFCFCUL)>>2);\ 1.303 + uint32_t l1,h1;\ 1.304 +\ 1.305 + pixels+=line_size;\ 1.306 + for(i=0; i<h; i+=2){\ 1.307 + uint32_t a= AV_RN32(pixels );\ 1.308 + uint32_t b= AV_RN32(pixels+1);\ 1.309 + l1= (a&0x03030303UL)\ 1.310 + + (b&0x03030303UL);\ 1.311 + h1= ((a&0xFCFCFCFCUL)>>2)\ 1.312 + + ((b&0xFCFCFCFCUL)>>2);\ 1.313 + OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.314 + pixels+=line_size;\ 1.315 + block +=line_size;\ 1.316 + a= AV_RN32(pixels );\ 1.317 + b= AV_RN32(pixels+1);\ 1.318 + l0= (a&0x03030303UL)\ 1.319 + + (b&0x03030303UL)\ 1.320 + + 0x02020202UL;\ 1.321 + h0= ((a&0xFCFCFCFCUL)>>2)\ 1.322 + + ((b&0xFCFCFCFCUL)>>2);\ 1.323 + OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.324 + pixels+=line_size;\ 1.325 + block +=line_size;\ 1.326 + }\ 1.327 +}\ 1.328 +\ 1.329 +static inline void OPNAME ## _pixels8_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ 1.330 +{\ 1.331 + int j;\ 1.332 + for(j=0; j<2; j++){\ 1.333 + int i;\ 1.334 + const uint32_t a= AV_RN32(pixels );\ 1.335 + const uint32_t b= AV_RN32(pixels+1);\ 1.336 + uint32_t l0= (a&0x03030303UL)\ 1.337 + + (b&0x03030303UL)\ 1.338 + + 0x02020202UL;\ 1.339 + uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\ 1.340 + + ((b&0xFCFCFCFCUL)>>2);\ 1.341 + uint32_t l1,h1;\ 1.342 +\ 1.343 + pixels+=line_size;\ 1.344 + for(i=0; i<h; i+=2){\ 1.345 + uint32_t a= AV_RN32(pixels );\ 1.346 + uint32_t b= AV_RN32(pixels+1);\ 1.347 + l1= (a&0x03030303UL)\ 1.348 + + (b&0x03030303UL);\ 1.349 + h1= ((a&0xFCFCFCFCUL)>>2)\ 1.350 + + ((b&0xFCFCFCFCUL)>>2);\ 1.351 + OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.352 + pixels+=line_size;\ 1.353 + block +=line_size;\ 1.354 + a= AV_RN32(pixels );\ 1.355 + b= AV_RN32(pixels+1);\ 1.356 + l0= (a&0x03030303UL)\ 1.357 + + (b&0x03030303UL)\ 1.358 + + 0x02020202UL;\ 1.359 + h0= ((a&0xFCFCFCFCUL)>>2)\ 1.360 + + ((b&0xFCFCFCFCUL)>>2);\ 1.361 + OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.362 + pixels+=line_size;\ 1.363 + block +=line_size;\ 1.364 + }\ 1.365 + pixels+=4-line_size*(h+1);\ 1.366 + block +=4-line_size*h;\ 1.367 + }\ 1.368 +}\ 1.369 +\ 1.370 +static inline void OPNAME ## _no_rnd_pixels8_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ 1.371 +{\ 1.372 + int j;\ 1.373 + for(j=0; j<2; j++){\ 1.374 + int i;\ 1.375 + const uint32_t a= AV_RN32(pixels );\ 1.376 + const uint32_t b= AV_RN32(pixels+1);\ 1.377 + uint32_t l0= (a&0x03030303UL)\ 1.378 + + (b&0x03030303UL)\ 1.379 + + 0x01010101UL;\ 1.380 + uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\ 1.381 + + ((b&0xFCFCFCFCUL)>>2);\ 1.382 + uint32_t l1,h1;\ 1.383 +\ 1.384 + pixels+=line_size;\ 1.385 + for(i=0; i<h; i+=2){\ 1.386 + uint32_t a= AV_RN32(pixels );\ 1.387 + uint32_t b= AV_RN32(pixels+1);\ 1.388 + l1= (a&0x03030303UL)\ 1.389 + + (b&0x03030303UL);\ 1.390 + h1= ((a&0xFCFCFCFCUL)>>2)\ 1.391 + + ((b&0xFCFCFCFCUL)>>2);\ 1.392 + OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.393 + pixels+=line_size;\ 1.394 + block +=line_size;\ 1.395 + a= AV_RN32(pixels );\ 1.396 + b= AV_RN32(pixels+1);\ 1.397 + l0= (a&0x03030303UL)\ 1.398 + + (b&0x03030303UL)\ 1.399 + + 0x01010101UL;\ 1.400 + h0= ((a&0xFCFCFCFCUL)>>2)\ 1.401 + + ((b&0xFCFCFCFCUL)>>2);\ 1.402 + OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ 1.403 + pixels+=line_size;\ 1.404 + block +=line_size;\ 1.405 + }\ 1.406 + pixels+=4-line_size*(h+1);\ 1.407 + block +=4-line_size*h;\ 1.408 + }\ 1.409 +}\ 1.410 +\ 1.411 +CALL_2X_PIXELS(OPNAME ## _pixels16_c , OPNAME ## _pixels8_c , 8)\ 1.412 + 1.413 +#define op_avg(a, b) a = rnd_avg32(a, b) 1.414 + 1.415 +#define op_put(a, b) a = b 1.416 + 1.417 +PIXOP2(avg, op_avg) 1.418 +PIXOP2(put, op_put) 1.419 +#undef op_avg 1.420 +#undef op_put 1.421 + 1.422 + 1.423 +#define H264_CHROMA_MC(OPNAME, OP)\ 1.424 +static void OPNAME ## h264_chroma_mc2_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ 1.425 + const int A=(8-x)*(8-y);\ 1.426 + const int B=( x)*(8-y);\ 1.427 + const int C=(8-x)*( y);\ 1.428 + const int D=( x)*( y);\ 1.429 + int i;\ 1.430 + \ 1.431 + assert(x<8 && y<8 && x>=0 && y>=0);\ 1.432 +\ 1.433 + if(D){\ 1.434 + for(i=0; i<h; i++){\ 1.435 + OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\ 1.436 + OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\ 1.437 + dst+= stride;\ 1.438 + src+= stride;\ 1.439 + }\ 1.440 + }else{\ 1.441 + const int E= B+C;\ 1.442 + const int step= C ? stride : 1;\ 1.443 + for(i=0; i<h; i++){\ 1.444 + OP(dst[0], (A*src[0] + E*src[step+0]));\ 1.445 + OP(dst[1], (A*src[1] + E*src[step+1]));\ 1.446 + dst+= stride;\ 1.447 + src+= stride;\ 1.448 + }\ 1.449 + }\ 1.450 +}\ 1.451 +\ 1.452 +static void OPNAME ## h264_chroma_mc4_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ 1.453 + const int A=(8-x)*(8-y);\ 1.454 + const int B=( x)*(8-y);\ 1.455 + const int C=(8-x)*( y);\ 1.456 + const int D=( x)*( y);\ 1.457 + int i;\ 1.458 + \ 1.459 + assert(x<8 && y<8 && x>=0 && y>=0);\ 1.460 +\ 1.461 + if(D){\ 1.462 + for(i=0; i<h; i++){\ 1.463 + OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\ 1.464 + OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\ 1.465 + OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\ 1.466 + OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\ 1.467 + dst+= stride;\ 1.468 + src+= stride;\ 1.469 + }\ 1.470 + }else{\ 1.471 + const int E= B+C;\ 1.472 + const int step= C ? stride : 1;\ 1.473 + for(i=0; i<h; i++){\ 1.474 + OP(dst[0], (A*src[0] + E*src[step+0]));\ 1.475 + OP(dst[1], (A*src[1] + E*src[step+1]));\ 1.476 + OP(dst[2], (A*src[2] + E*src[step+2]));\ 1.477 + OP(dst[3], (A*src[3] + E*src[step+3]));\ 1.478 + dst+= stride;\ 1.479 + src+= stride;\ 1.480 + }\ 1.481 + }\ 1.482 +}\ 1.483 +\ 1.484 +static void OPNAME ## h264_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ 1.485 + const int A=(8-x)*(8-y);\ 1.486 + const int B=( x)*(8-y);\ 1.487 + const int C=(8-x)*( y);\ 1.488 + const int D=( x)*( y);\ 1.489 + int i;\ 1.490 + \ 1.491 + assert(x<8 && y<8 && x>=0 && y>=0);\ 1.492 +\ 1.493 + if(D){\ 1.494 + for(i=0; i<h; i++){\ 1.495 + OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1]));\ 1.496 + OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2]));\ 1.497 + OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3]));\ 1.498 + OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4]));\ 1.499 + OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5]));\ 1.500 + OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6]));\ 1.501 + OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7]));\ 1.502 + OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8]));\ 1.503 + dst+= stride;\ 1.504 + src+= stride;\ 1.505 + }\ 1.506 + }else{\ 1.507 + const int E= B+C;\ 1.508 + const int step= C ? stride : 1;\ 1.509 + for(i=0; i<h; i++){\ 1.510 + OP(dst[0], (A*src[0] + E*src[step+0]));\ 1.511 + OP(dst[1], (A*src[1] + E*src[step+1]));\ 1.512 + OP(dst[2], (A*src[2] + E*src[step+2]));\ 1.513 + OP(dst[3], (A*src[3] + E*src[step+3]));\ 1.514 + OP(dst[4], (A*src[4] + E*src[step+4]));\ 1.515 + OP(dst[5], (A*src[5] + E*src[step+5]));\ 1.516 + OP(dst[6], (A*src[6] + E*src[step+6]));\ 1.517 + OP(dst[7], (A*src[7] + E*src[step+7]));\ 1.518 + dst+= stride;\ 1.519 + src+= stride;\ 1.520 + }\ 1.521 + }\ 1.522 +} 1.523 + 1.524 +#define op_avg(a, b) a = (((a)+(((b) + 32)>>6)+1)>>1) 1.525 +#define op_put(a, b) a = (((b) + 32)>>6) 1.526 + 1.527 +H264_CHROMA_MC(put_ , op_put) 1.528 +H264_CHROMA_MC(avg_ , op_avg) 1.529 +#undef op_avg 1.530 +#undef op_put 1.531 + 1.532 + 1.533 +#define H264_LOWPASS(OPNAME, OP, OP2) \ 1.534 +static av_unused void OPNAME ## h264_qpel2_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ 1.535 + const int h=2;\ 1.536 + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 1.537 + int i;\ 1.538 + for(i=0; i<h; i++)\ 1.539 + {\ 1.540 + OP(dst[0], (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]));\ 1.541 + OP(dst[1], (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]));\ 1.542 + dst+=dstStride;\ 1.543 + src+=srcStride;\ 1.544 + }\ 1.545 +}\ 1.546 +\ 1.547 +static av_unused void OPNAME ## h264_qpel2_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ 1.548 + const int w=2;\ 1.549 + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 1.550 + int i;\ 1.551 + for(i=0; i<w; i++)\ 1.552 + {\ 1.553 + const int srcB= src[-2*srcStride];\ 1.554 + const int srcA= src[-1*srcStride];\ 1.555 + const int src0= src[0 *srcStride];\ 1.556 + const int src1= src[1 *srcStride];\ 1.557 + const int src2= src[2 *srcStride];\ 1.558 + const int src3= src[3 *srcStride];\ 1.559 + const int src4= src[4 *srcStride];\ 1.560 + OP(dst[0*dstStride], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\ 1.561 + OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\ 1.562 + dst++;\ 1.563 + src++;\ 1.564 + }\ 1.565 +}\ 1.566 +\ 1.567 +static av_unused void OPNAME ## h264_qpel2_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\ 1.568 + const int h=2;\ 1.569 + const int w=2;\ 1.570 + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 1.571 + int i;\ 1.572 + src -= 2*srcStride;\ 1.573 + for(i=0; i<h+5; i++)\ 1.574 + {\ 1.575 + tmp[0]= (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]);\ 1.576 + tmp[1]= (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]);\ 1.577 + tmp+=tmpStride;\ 1.578 + src+=srcStride;\ 1.579 + }\ 1.580 + tmp -= tmpStride*(h+5-2);\ 1.581 + for(i=0; i<w; i++)\ 1.582 + {\ 1.583 + const int tmpB= tmp[-2*tmpStride];\ 1.584 + const int tmpA= tmp[-1*tmpStride];\ 1.585 + const int tmp0= tmp[0 *tmpStride];\ 1.586 + const int tmp1= tmp[1 *tmpStride];\ 1.587 + const int tmp2= tmp[2 *tmpStride];\ 1.588 + const int tmp3= tmp[3 *tmpStride];\ 1.589 + const int tmp4= tmp[4 *tmpStride];\ 1.590 + OP2(dst[0*dstStride], (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));\ 1.591 + OP2(dst[1*dstStride], (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));\ 1.592 + dst++;\ 1.593 + tmp++;\ 1.594 + }\ 1.595 +}\ 1.596 +static void OPNAME ## h264_qpel4_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ 1.597 + const int h=4;\ 1.598 + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 1.599 + int i;\ 1.600 + for(i=0; i<h; i++)\ 1.601 + {\ 1.602 + OP(dst[0], (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]));\ 1.603 + OP(dst[1], (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]));\ 1.604 + OP(dst[2], (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5]));\ 1.605 + OP(dst[3], (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6]));\ 1.606 + dst+=dstStride;\ 1.607 + src+=srcStride;\ 1.608 + }\ 1.609 +}\ 1.610 +\ 1.611 +static void OPNAME ## h264_qpel4_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ 1.612 + const int w=4;\ 1.613 + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 1.614 + int i;\ 1.615 + for(i=0; i<w; i++)\ 1.616 + {\ 1.617 + const int srcB= src[-2*srcStride];\ 1.618 + const int srcA= src[-1*srcStride];\ 1.619 + const int src0= src[0 *srcStride];\ 1.620 + const int src1= src[1 *srcStride];\ 1.621 + const int src2= src[2 *srcStride];\ 1.622 + const int src3= src[3 *srcStride];\ 1.623 + const int src4= src[4 *srcStride];\ 1.624 + const int src5= src[5 *srcStride];\ 1.625 + const int src6= src[6 *srcStride];\ 1.626 + OP(dst[0*dstStride], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\ 1.627 + OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\ 1.628 + OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\ 1.629 + OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\ 1.630 + dst++;\ 1.631 + src++;\ 1.632 + }\ 1.633 +}\ 1.634 +\ 1.635 +static void OPNAME ## h264_qpel4_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\ 1.636 + const int h=4;\ 1.637 + const int w=4;\ 1.638 + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 1.639 + int i;\ 1.640 + src -= 2*srcStride;\ 1.641 + for(i=0; i<h+5; i++)\ 1.642 + {\ 1.643 + tmp[0]= (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3]);\ 1.644 + tmp[1]= (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4]);\ 1.645 + tmp[2]= (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5]);\ 1.646 + tmp[3]= (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6]);\ 1.647 + tmp+=tmpStride;\ 1.648 + src+=srcStride;\ 1.649 + }\ 1.650 + tmp -= tmpStride*(h+5-2);\ 1.651 + for(i=0; i<w; i++)\ 1.652 + {\ 1.653 + const int tmpB= tmp[-2*tmpStride];\ 1.654 + const int tmpA= tmp[-1*tmpStride];\ 1.655 + const int tmp0= tmp[0 *tmpStride];\ 1.656 + const int tmp1= tmp[1 *tmpStride];\ 1.657 + const int tmp2= tmp[2 *tmpStride];\ 1.658 + const int tmp3= tmp[3 *tmpStride];\ 1.659 + const int tmp4= tmp[4 *tmpStride];\ 1.660 + const int tmp5= tmp[5 *tmpStride];\ 1.661 + const int tmp6= tmp[6 *tmpStride];\ 1.662 + OP2(dst[0*dstStride], (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));\ 1.663 + OP2(dst[1*dstStride], (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));\ 1.664 + OP2(dst[2*dstStride], (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));\ 1.665 + OP2(dst[3*dstStride], (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));\ 1.666 + dst++;\ 1.667 + tmp++;\ 1.668 + }\ 1.669 +}\ 1.670 +\ 1.671 +static void OPNAME ## h264_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ 1.672 + const int h=8;\ 1.673 + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 1.674 + int i;\ 1.675 + for(i=0; i<h; i++)\ 1.676 + {\ 1.677 + OP(dst[0], (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3 ]));\ 1.678 + OP(dst[1], (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4 ]));\ 1.679 + OP(dst[2], (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5 ]));\ 1.680 + OP(dst[3], (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6 ]));\ 1.681 + OP(dst[4], (src[4]+src[5])*20 - (src[3 ]+src[6])*5 + (src[2 ]+src[7 ]));\ 1.682 + OP(dst[5], (src[5]+src[6])*20 - (src[4 ]+src[7])*5 + (src[3 ]+src[8 ]));\ 1.683 + OP(dst[6], (src[6]+src[7])*20 - (src[5 ]+src[8])*5 + (src[4 ]+src[9 ]));\ 1.684 + OP(dst[7], (src[7]+src[8])*20 - (src[6 ]+src[9])*5 + (src[5 ]+src[10]));\ 1.685 + dst+=dstStride;\ 1.686 + src+=srcStride;\ 1.687 + }\ 1.688 +}\ 1.689 +\ 1.690 +static void OPNAME ## h264_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ 1.691 + const int w=8;\ 1.692 + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 1.693 + int i;\ 1.694 + for(i=0; i<w; i++)\ 1.695 + {\ 1.696 + const int srcB= src[-2*srcStride];\ 1.697 + const int srcA= src[-1*srcStride];\ 1.698 + const int src0= src[0 *srcStride];\ 1.699 + const int src1= src[1 *srcStride];\ 1.700 + const int src2= src[2 *srcStride];\ 1.701 + const int src3= src[3 *srcStride];\ 1.702 + const int src4= src[4 *srcStride];\ 1.703 + const int src5= src[5 *srcStride];\ 1.704 + const int src6= src[6 *srcStride];\ 1.705 + const int src7= src[7 *srcStride];\ 1.706 + const int src8= src[8 *srcStride];\ 1.707 + const int src9= src[9 *srcStride];\ 1.708 + const int src10=src[10*srcStride];\ 1.709 + OP(dst[0*dstStride], (src0+src1)*20 - (srcA+src2)*5 + (srcB+src3));\ 1.710 + OP(dst[1*dstStride], (src1+src2)*20 - (src0+src3)*5 + (srcA+src4));\ 1.711 + OP(dst[2*dstStride], (src2+src3)*20 - (src1+src4)*5 + (src0+src5));\ 1.712 + OP(dst[3*dstStride], (src3+src4)*20 - (src2+src5)*5 + (src1+src6));\ 1.713 + OP(dst[4*dstStride], (src4+src5)*20 - (src3+src6)*5 + (src2+src7));\ 1.714 + OP(dst[5*dstStride], (src5+src6)*20 - (src4+src7)*5 + (src3+src8));\ 1.715 + OP(dst[6*dstStride], (src6+src7)*20 - (src5+src8)*5 + (src4+src9));\ 1.716 + OP(dst[7*dstStride], (src7+src8)*20 - (src6+src9)*5 + (src5+src10));\ 1.717 + dst++;\ 1.718 + src++;\ 1.719 + }\ 1.720 +}\ 1.721 +\ 1.722 +static void OPNAME ## h264_qpel8_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\ 1.723 + const int h=8;\ 1.724 + const int w=8;\ 1.725 + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ 1.726 + int i;\ 1.727 + src -= 2*srcStride;\ 1.728 + for(i=0; i<h+5; i++)\ 1.729 + {\ 1.730 + tmp[0]= (src[0]+src[1])*20 - (src[-1]+src[2])*5 + (src[-2]+src[3 ]);\ 1.731 + tmp[1]= (src[1]+src[2])*20 - (src[0 ]+src[3])*5 + (src[-1]+src[4 ]);\ 1.732 + tmp[2]= (src[2]+src[3])*20 - (src[1 ]+src[4])*5 + (src[0 ]+src[5 ]);\ 1.733 + tmp[3]= (src[3]+src[4])*20 - (src[2 ]+src[5])*5 + (src[1 ]+src[6 ]);\ 1.734 + tmp[4]= (src[4]+src[5])*20 - (src[3 ]+src[6])*5 + (src[2 ]+src[7 ]);\ 1.735 + tmp[5]= (src[5]+src[6])*20 - (src[4 ]+src[7])*5 + (src[3 ]+src[8 ]);\ 1.736 + tmp[6]= (src[6]+src[7])*20 - (src[5 ]+src[8])*5 + (src[4 ]+src[9 ]);\ 1.737 + tmp[7]= (src[7]+src[8])*20 - (src[6 ]+src[9])*5 + (src[5 ]+src[10]);\ 1.738 + tmp+=tmpStride;\ 1.739 + src+=srcStride;\ 1.740 + }\ 1.741 + tmp -= tmpStride*(h+5-2);\ 1.742 + for(i=0; i<w; i++)\ 1.743 + {\ 1.744 + const int tmpB= tmp[-2*tmpStride];\ 1.745 + const int tmpA= tmp[-1*tmpStride];\ 1.746 + const int tmp0= tmp[0 *tmpStride];\ 1.747 + const int tmp1= tmp[1 *tmpStride];\ 1.748 + const int tmp2= tmp[2 *tmpStride];\ 1.749 + const int tmp3= tmp[3 *tmpStride];\ 1.750 + const int tmp4= tmp[4 *tmpStride];\ 1.751 + const int tmp5= tmp[5 *tmpStride];\ 1.752 + const int tmp6= tmp[6 *tmpStride];\ 1.753 + const int tmp7= tmp[7 *tmpStride];\ 1.754 + const int tmp8= tmp[8 *tmpStride];\ 1.755 + const int tmp9= tmp[9 *tmpStride];\ 1.756 + const int tmp10=tmp[10*tmpStride];\ 1.757 + OP2(dst[0*dstStride], (tmp0+tmp1)*20 - (tmpA+tmp2)*5 + (tmpB+tmp3));\ 1.758 + OP2(dst[1*dstStride], (tmp1+tmp2)*20 - (tmp0+tmp3)*5 + (tmpA+tmp4));\ 1.759 + OP2(dst[2*dstStride], (tmp2+tmp3)*20 - (tmp1+tmp4)*5 + (tmp0+tmp5));\ 1.760 + OP2(dst[3*dstStride], (tmp3+tmp4)*20 - (tmp2+tmp5)*5 + (tmp1+tmp6));\ 1.761 + OP2(dst[4*dstStride], (tmp4+tmp5)*20 - (tmp3+tmp6)*5 + (tmp2+tmp7));\ 1.762 + OP2(dst[5*dstStride], (tmp5+tmp6)*20 - (tmp4+tmp7)*5 + (tmp3+tmp8));\ 1.763 + OP2(dst[6*dstStride], (tmp6+tmp7)*20 - (tmp5+tmp8)*5 + (tmp4+tmp9));\ 1.764 + OP2(dst[7*dstStride], (tmp7+tmp8)*20 - (tmp6+tmp9)*5 + (tmp5+tmp10));\ 1.765 + dst++;\ 1.766 + tmp++;\ 1.767 + }\ 1.768 +}\ 1.769 +\ 1.770 +static void OPNAME ## h264_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ 1.771 + OPNAME ## h264_qpel8_v_lowpass(dst , src , dstStride, srcStride);\ 1.772 + OPNAME ## h264_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride);\ 1.773 + src += 8*srcStride;\ 1.774 + dst += 8*dstStride;\ 1.775 + OPNAME ## h264_qpel8_v_lowpass(dst , src , dstStride, srcStride);\ 1.776 + OPNAME ## h264_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride);\ 1.777 +}\ 1.778 +\ 1.779 +static void OPNAME ## h264_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\ 1.780 + OPNAME ## h264_qpel8_h_lowpass(dst , src , dstStride, srcStride);\ 1.781 + OPNAME ## h264_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride);\ 1.782 + src += 8*srcStride;\ 1.783 + dst += 8*dstStride;\ 1.784 + OPNAME ## h264_qpel8_h_lowpass(dst , src , dstStride, srcStride);\ 1.785 + OPNAME ## h264_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride);\ 1.786 +}\ 1.787 +\ 1.788 +static void OPNAME ## h264_qpel16_hv_lowpass(uint8_t *dst, int16_t *tmp, uint8_t *src, int dstStride, int tmpStride, int srcStride){\ 1.789 + OPNAME ## h264_qpel8_hv_lowpass(dst , tmp , src , dstStride, tmpStride, srcStride);\ 1.790 + OPNAME ## h264_qpel8_hv_lowpass(dst+8, tmp+8, src+8, dstStride, tmpStride, srcStride);\ 1.791 + src += 8*srcStride;\ 1.792 + dst += 8*dstStride;\ 1.793 + OPNAME ## h264_qpel8_hv_lowpass(dst , tmp , src , dstStride, tmpStride, srcStride);\ 1.794 + OPNAME ## h264_qpel8_hv_lowpass(dst+8, tmp+8, src+8, dstStride, tmpStride, srcStride);\ 1.795 +}\ 1.796 + 1.797 +#define H264_MC(OPNAME, SIZE) \ 1.798 +static void OPNAME ## h264_qpel ## SIZE ## _mc00_c (uint8_t *dst, uint8_t *src, int stride){\ 1.799 + OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\ 1.800 +}\ 1.801 +\ 1.802 +static void OPNAME ## h264_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\ 1.803 + uint8_t half[SIZE*SIZE];\ 1.804 + put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\ 1.805 + OPNAME ## pixels ## SIZE ## _l2(dst, src, half, stride, stride, SIZE, SIZE);\ 1.806 +}\ 1.807 +\ 1.808 +static void OPNAME ## h264_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\ 1.809 + OPNAME ## h264_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride);\ 1.810 +}\ 1.811 +\ 1.812 +static void OPNAME ## h264_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\ 1.813 + uint8_t half[SIZE*SIZE];\ 1.814 + put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\ 1.815 + OPNAME ## pixels ## SIZE ## _l2(dst, src+1, half, stride, stride, SIZE, SIZE);\ 1.816 +}\ 1.817 +\ 1.818 +static void OPNAME ## h264_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\ 1.819 + uint8_t full[SIZE*(SIZE+5)];\ 1.820 + uint8_t * const full_mid= full + SIZE*2;\ 1.821 + uint8_t half[SIZE*SIZE];\ 1.822 + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ 1.823 + put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\ 1.824 + OPNAME ## pixels ## SIZE ## _l2(dst, full_mid, half, stride, SIZE, SIZE, SIZE);\ 1.825 +}\ 1.826 +\ 1.827 +static void OPNAME ## h264_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\ 1.828 + uint8_t full[SIZE*(SIZE+5)];\ 1.829 + uint8_t * const full_mid= full + SIZE*2;\ 1.830 + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ 1.831 + OPNAME ## h264_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE);\ 1.832 +}\ 1.833 +\ 1.834 +static void OPNAME ## h264_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\ 1.835 + uint8_t full[SIZE*(SIZE+5)];\ 1.836 + uint8_t * const full_mid= full + SIZE*2;\ 1.837 + uint8_t half[SIZE*SIZE];\ 1.838 + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ 1.839 + put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\ 1.840 + OPNAME ## pixels ## SIZE ## _l2(dst, full_mid+SIZE, half, stride, SIZE, SIZE, SIZE);\ 1.841 +}\ 1.842 +\ 1.843 +static void OPNAME ## h264_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\ 1.844 + uint8_t full[SIZE*(SIZE+5)];\ 1.845 + uint8_t * const full_mid= full + SIZE*2;\ 1.846 + uint8_t halfH[SIZE*SIZE];\ 1.847 + uint8_t halfV[SIZE*SIZE];\ 1.848 + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\ 1.849 + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ 1.850 + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ 1.851 + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\ 1.852 +}\ 1.853 +\ 1.854 +static void OPNAME ## h264_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\ 1.855 + uint8_t full[SIZE*(SIZE+5)];\ 1.856 + uint8_t * const full_mid= full + SIZE*2;\ 1.857 + uint8_t halfH[SIZE*SIZE];\ 1.858 + uint8_t halfV[SIZE*SIZE];\ 1.859 + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\ 1.860 + copy_block ## SIZE (full, src - stride*2 + 1, SIZE, stride, SIZE + 5);\ 1.861 + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ 1.862 + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\ 1.863 +}\ 1.864 +\ 1.865 +static void OPNAME ## h264_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\ 1.866 + uint8_t full[SIZE*(SIZE+5)];\ 1.867 + uint8_t * const full_mid= full + SIZE*2;\ 1.868 + uint8_t halfH[SIZE*SIZE];\ 1.869 + uint8_t halfV[SIZE*SIZE];\ 1.870 + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\ 1.871 + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ 1.872 + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ 1.873 + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\ 1.874 +}\ 1.875 +\ 1.876 +static void OPNAME ## h264_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\ 1.877 + uint8_t full[SIZE*(SIZE+5)];\ 1.878 + uint8_t * const full_mid= full + SIZE*2;\ 1.879 + uint8_t halfH[SIZE*SIZE];\ 1.880 + uint8_t halfV[SIZE*SIZE];\ 1.881 + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\ 1.882 + copy_block ## SIZE (full, src - stride*2 + 1, SIZE, stride, SIZE + 5);\ 1.883 + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ 1.884 + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\ 1.885 +}\ 1.886 +\ 1.887 +static void OPNAME ## h264_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\ 1.888 + int16_t tmp[SIZE*(SIZE+5)];\ 1.889 + OPNAME ## h264_qpel ## SIZE ## _hv_lowpass(dst, tmp, src, stride, SIZE, stride);\ 1.890 +}\ 1.891 +\ 1.892 +static void OPNAME ## h264_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\ 1.893 + int16_t tmp[SIZE*(SIZE+5)];\ 1.894 + uint8_t halfH[SIZE*SIZE];\ 1.895 + uint8_t halfHV[SIZE*SIZE];\ 1.896 + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\ 1.897 + put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\ 1.898 + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\ 1.899 +}\ 1.900 +\ 1.901 +static void OPNAME ## h264_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\ 1.902 + int16_t tmp[SIZE*(SIZE+5)];\ 1.903 + uint8_t halfH[SIZE*SIZE];\ 1.904 + uint8_t halfHV[SIZE*SIZE];\ 1.905 + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\ 1.906 + put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\ 1.907 + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\ 1.908 +}\ 1.909 +\ 1.910 +static void OPNAME ## h264_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\ 1.911 + uint8_t full[SIZE*(SIZE+5)];\ 1.912 + uint8_t * const full_mid= full + SIZE*2;\ 1.913 + int16_t tmp[SIZE*(SIZE+5)];\ 1.914 + uint8_t halfV[SIZE*SIZE];\ 1.915 + uint8_t halfHV[SIZE*SIZE];\ 1.916 + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ 1.917 + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ 1.918 + put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\ 1.919 + OPNAME ## pixels ## SIZE ## _l2(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\ 1.920 +}\ 1.921 +\ 1.922 +static void OPNAME ## h264_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\ 1.923 + uint8_t full[SIZE*(SIZE+5)];\ 1.924 + uint8_t * const full_mid= full + SIZE*2;\ 1.925 + int16_t tmp[SIZE*(SIZE+5)];\ 1.926 + uint8_t halfV[SIZE*SIZE];\ 1.927 + uint8_t halfHV[SIZE*SIZE];\ 1.928 + copy_block ## SIZE (full, src - stride*2 + 1, SIZE, stride, SIZE + 5);\ 1.929 + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ 1.930 + put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\ 1.931 + OPNAME ## pixels ## SIZE ## _l2(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\ 1.932 +}\ 1.933 + 1.934 +#define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1) 1.935 +#define op_put(a, b) a = cm[((b) + 16)>>5] 1.936 +#define op2_avg(a, b) a = (((a)+cm[((b) + 512)>>10]+1)>>1) 1.937 +#define op2_put(a, b) a = cm[((b) + 512)>>10] 1.938 + 1.939 +H264_LOWPASS(put_ , op_put, op2_put) 1.940 +H264_LOWPASS(avg_ , op_avg, op2_avg) 1.941 +H264_MC(put_, 2) 1.942 +H264_MC(put_, 4) 1.943 +H264_MC(put_, 8) 1.944 +H264_MC(put_, 16) 1.945 +H264_MC(avg_, 4) 1.946 +H264_MC(avg_, 8) 1.947 +H264_MC(avg_, 16) 1.948 + 1.949 +#undef op_avg 1.950 +#undef op_put 1.951 +#undef op2_avg 1.952 +#undef op2_put 1.953 + 1.954 +static void clear_block_c(DCTELEM *block) 1.955 +{ 1.956 + memset(block, 0, sizeof(DCTELEM)*64); 1.957 +} 1.958 + 1.959 +/** 1.960 + * memset(blocks, 0, sizeof(DCTELEM)*6*64) 1.961 + */ 1.962 +static void clear_blocks_c(DCTELEM *blocks) 1.963 +{ 1.964 + memset(blocks, 0, sizeof(DCTELEM)*6*64); 1.965 +} 1.966 + 1.967 +static void just_return(void *mem av_unused, int stride av_unused, int h av_unused) { return; } 1.968 + 1.969 +/* init static data */ 1.970 +av_cold void dsputil_static_init(void) 1.971 +{ 1.972 + int i; 1.973 + 1.974 + for(i=0;i<256;i++) ff_cropTbl[i + MAX_NEG_CROP] = i; 1.975 + for(i=0;i<MAX_NEG_CROP;i++) { 1.976 + ff_cropTbl[i] = 0; 1.977 + ff_cropTbl[i + MAX_NEG_CROP + 256] = 255; 1.978 + } 1.979 + 1.980 + for(i=0;i<512;i++) { 1.981 + ff_squareTbl[i] = (i - 256) * (i - 256); 1.982 + } 1.983 +} 1.984 + 1.985 +int ff_check_alignment(void){ 1.986 + static int did_fail=0; 1.987 + DECLARE_ALIGNED(16, int, aligned); 1.988 + 1.989 + if((intptr_t)&aligned & 15){ 1.990 + if(!did_fail){ 1.991 +#if HAVE_MMX || HAVE_ALTIVEC 1.992 + av_log(AV_LOG_ERROR, 1.993 + "Compiler did not align stack variables. Libavcodec has been miscompiled\n" 1.994 + "and may be very slow or crash. This is not a bug in libavcodec,\n" 1.995 + "but in the compiler. You may try recompiling using gcc >= 4.2.\n" 1.996 + "Do not report crashes to FFmpeg developers.\n"); 1.997 +#endif 1.998 + did_fail=1; 1.999 + } 1.1000 + return -1; 1.1001 + } 1.1002 + return 0; 1.1003 +} 1.1004 + 1.1005 +av_cold void dsputil_init(DSPContext* c) 1.1006 +{ 1.1007 + (void) avg_pixels2_c; // kill a warning, avg_pixels2_c is a macro created function. 1.1008 + ff_check_alignment(); 1.1009 + dsputil_static_init(); 1.1010 + 1.1011 + c->idct_put= ff_simple_idct_put; 1.1012 + c->idct_add= ff_simple_idct_add; 1.1013 + c->idct = ff_simple_idct; 1.1014 + 1.1015 + c->clear_block = clear_block_c; 1.1016 + c->clear_blocks = clear_blocks_c; 1.1017 + 1.1018 +#define dspfunc(PFX, IDX, NUM) \ 1.1019 + c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_c; \ 1.1020 + c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_c; \ 1.1021 + c->PFX ## _pixels_tab[IDX][ 2] = PFX ## NUM ## _mc20_c; \ 1.1022 + c->PFX ## _pixels_tab[IDX][ 3] = PFX ## NUM ## _mc30_c; \ 1.1023 + c->PFX ## _pixels_tab[IDX][ 4] = PFX ## NUM ## _mc01_c; \ 1.1024 + c->PFX ## _pixels_tab[IDX][ 5] = PFX ## NUM ## _mc11_c; \ 1.1025 + c->PFX ## _pixels_tab[IDX][ 6] = PFX ## NUM ## _mc21_c; \ 1.1026 + c->PFX ## _pixels_tab[IDX][ 7] = PFX ## NUM ## _mc31_c; \ 1.1027 + c->PFX ## _pixels_tab[IDX][ 8] = PFX ## NUM ## _mc02_c; \ 1.1028 + c->PFX ## _pixels_tab[IDX][ 9] = PFX ## NUM ## _mc12_c; \ 1.1029 + c->PFX ## _pixels_tab[IDX][10] = PFX ## NUM ## _mc22_c; \ 1.1030 + c->PFX ## _pixels_tab[IDX][11] = PFX ## NUM ## _mc32_c; \ 1.1031 + c->PFX ## _pixels_tab[IDX][12] = PFX ## NUM ## _mc03_c; \ 1.1032 + c->PFX ## _pixels_tab[IDX][13] = PFX ## NUM ## _mc13_c; \ 1.1033 + c->PFX ## _pixels_tab[IDX][14] = PFX ## NUM ## _mc23_c; \ 1.1034 + c->PFX ## _pixels_tab[IDX][15] = PFX ## NUM ## _mc33_c 1.1035 + 1.1036 + 1.1037 + dspfunc(put_h264_qpel, 0, 16); 1.1038 + dspfunc(put_h264_qpel, 1, 8); 1.1039 + dspfunc(put_h264_qpel, 2, 4); 1.1040 + dspfunc(put_h264_qpel, 3, 2); 1.1041 + dspfunc(avg_h264_qpel, 0, 16); 1.1042 + dspfunc(avg_h264_qpel, 1, 8); 1.1043 + dspfunc(avg_h264_qpel, 2, 4); 1.1044 + 1.1045 +#undef dspfunc 1.1046 + c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_c; 1.1047 + c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_c; 1.1048 + c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_c; 1.1049 + c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_c; 1.1050 + c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_c; 1.1051 + c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_c; 1.1052 + 1.1053 + 1.1054 + c->prefetch= just_return; 1.1055 + 1.1056 + if (HAVE_MMX) dsputil_init_mmx (c); 1.1057 + if (ARCH_ARM) dsputil_init_arm (c); 1.1058 + if (HAVE_ALTIVEC) dsputil_init_ppc (c); //fixme PPC prefetch 1.1059 +} 1.1060 +