X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fsnow.c;h=b5a381024cb732e69ce0a61b3c4993bef230550f;hb=b6be4acc1fbe78bf3708487465a7211bb1cfb280;hp=e3497ef07d3ae0c813c68cbc58eaedde40879915;hpb=36c32bdddfe9d19d3ea57c6b1fbe30ca84c3d740;p=ffmpeg diff --git a/libavcodec/snow.c b/libavcodec/snow.c index e3497ef07d3..b5a381024cb 100644 --- a/libavcodec/snow.c +++ b/libavcodec/snow.c @@ -1,23 +1,24 @@ /* * Copyright (C) 2004 Michael Niedermayer * - * This library is free software; you can redistribute it and/or + * This file is part of FFmpeg. + * + * FFmpeg is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either - * version 2 of the License, or (at your option) any later version. + * version 2.1 of the License, or (at your option) any later version. * - * This library is distributed in the hope that it will be useful, + * FFmpeg is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with this library; if not, write to the Free Software + * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ #include "avcodec.h" -#include "common.h" #include "dsputil.h" #include "snow.h" @@ -175,57 +176,57 @@ static const int8_t quant13[256]={ #if 0 //64*cubic static const uint8_t obmc32[1024]={ - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, - 0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 2, 2, 1, 1, 1, 1, 0, 0, 0, - 0, 0, 1, 1, 2, 2, 3, 4, 4, 5, 6, 6, 7, 7, 8, 8, 8, 8, 7, 7, 6, 6, 5, 4, 4, 3, 2, 2, 1, 1, 0, 0, - 0, 0, 1, 2, 2, 3, 4, 6, 7, 8, 9,10,11,12,12,12,12,12,12,11,10, 9, 8, 7, 6, 4, 3, 2, 2, 1, 0, 0, - 0, 1, 1, 2, 3, 5, 6, 8,10,11,13,14,15,16,17,18,18,17,16,15,14,13,11,10, 8, 6, 5, 3, 2, 1, 1, 0, - 0, 1, 1, 3, 4, 6, 8,10,13,15,17,19,20,22,22,23,23,22,22,20,19,17,15,13,10, 8, 6, 4, 3, 1, 1, 0, - 0, 1, 2, 4, 6, 8,10,13,16,19,21,23,25,27,28,29,29,28,27,25,23,21,19,16,13,10, 8, 6, 4, 2, 1, 0, - 0, 1, 2, 4, 7,10,13,16,19,22,25,28,31,33,34,35,35,34,33,31,28,25,22,19,16,13,10, 7, 4, 2, 1, 0, - 0, 1, 3, 5, 8,11,15,19,22,26,30,33,36,38,40,41,41,40,38,36,33,30,26,22,19,15,11, 8, 5, 3, 1, 0, - 0, 1, 3, 6, 9,12,17,21,25,30,34,38,41,44,45,46,46,45,44,41,38,34,30,25,21,17,12, 9, 6, 3, 1, 0, - 0, 1, 3, 6,10,14,19,23,28,33,38,42,45,48,51,52,52,51,48,45,42,38,33,28,23,19,14,10, 6, 3, 1, 0, - 0, 1, 4, 7,11,15,20,25,31,36,41,45,49,52,55,56,56,55,52,49,45,41,36,31,25,20,15,11, 7, 4, 1, 0, - 0, 2, 4, 7,12,16,22,27,33,38,44,48,52,56,58,60,60,58,56,52,48,44,38,33,27,22,16,12, 7, 4, 2, 0, - 0, 1, 4, 8,12,17,22,28,34,40,45,51,55,58,61,62,62,61,58,55,51,45,40,34,28,22,17,12, 8, 4, 1, 0, - 0, 2, 4, 8,12,18,23,29,35,41,46,52,56,60,62,64,64,62,60,56,52,46,41,35,29,23,18,12, 8, 4, 2, 0, - 0, 2, 4, 8,12,18,23,29,35,41,46,52,56,60,62,64,64,62,60,56,52,46,41,35,29,23,18,12, 8, 4, 2, 0, - 0, 1, 4, 8,12,17,22,28,34,40,45,51,55,58,61,62,62,61,58,55,51,45,40,34,28,22,17,12, 8, 4, 1, 0, - 0, 2, 4, 7,12,16,22,27,33,38,44,48,52,56,58,60,60,58,56,52,48,44,38,33,27,22,16,12, 7, 4, 2, 0, - 0, 1, 4, 7,11,15,20,25,31,36,41,45,49,52,55,56,56,55,52,49,45,41,36,31,25,20,15,11, 7, 4, 1, 0, - 0, 1, 3, 6,10,14,19,23,28,33,38,42,45,48,51,52,52,51,48,45,42,38,33,28,23,19,14,10, 6, 3, 1, 0, - 0, 1, 3, 6, 9,12,17,21,25,30,34,38,41,44,45,46,46,45,44,41,38,34,30,25,21,17,12, 9, 6, 3, 1, 0, - 0, 1, 3, 5, 8,11,15,19,22,26,30,33,36,38,40,41,41,40,38,36,33,30,26,22,19,15,11, 8, 5, 3, 1, 0, - 0, 1, 2, 4, 7,10,13,16,19,22,25,28,31,33,34,35,35,34,33,31,28,25,22,19,16,13,10, 7, 4, 2, 1, 0, - 0, 1, 2, 4, 6, 8,10,13,16,19,21,23,25,27,28,29,29,28,27,25,23,21,19,16,13,10, 8, 6, 4, 2, 1, 0, - 0, 1, 1, 3, 4, 6, 8,10,13,15,17,19,20,22,22,23,23,22,22,20,19,17,15,13,10, 8, 6, 4, 3, 1, 1, 0, - 0, 1, 1, 2, 3, 5, 6, 8,10,11,13,14,15,16,17,18,18,17,16,15,14,13,11,10, 8, 6, 5, 3, 2, 1, 1, 0, - 0, 0, 1, 2, 2, 3, 4, 6, 7, 8, 9,10,11,12,12,12,12,12,12,11,10, 9, 8, 7, 6, 4, 3, 2, 2, 1, 0, 0, - 0, 0, 1, 1, 2, 2, 3, 4, 4, 5, 6, 6, 7, 7, 8, 8, 8, 8, 7, 7, 6, 6, 5, 4, 4, 3, 2, 2, 1, 1, 0, 0, - 0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4, 3, 3, 3, 2, 2, 1, 1, 1, 1, 0, 0, 0, - 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 4, 4, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, + 0, 0, 0, 4, 4, 4, 4, 8, 8, 12, 12, 12, 16, 16, 16, 16, 16, 16, 16, 16, 12, 12, 12, 8, 8, 4, 4, 4, 4, 0, 0, 0, + 0, 0, 4, 4, 8, 8, 12, 16, 16, 20, 24, 24, 28, 28, 32, 32, 32, 32, 28, 28, 24, 24, 20, 16, 16, 12, 8, 8, 4, 4, 0, 0, + 0, 0, 4, 8, 8, 12, 16, 24, 28, 32, 36, 40, 44, 48, 48, 48, 48, 48, 48, 44, 40, 36, 32, 28, 24, 16, 12, 8, 8, 4, 0, 0, + 0, 4, 4, 8, 12, 20, 24, 32, 40, 44, 52, 56, 60, 64, 68, 72, 72, 68, 64, 60, 56, 52, 44, 40, 32, 24, 20, 12, 8, 4, 4, 0, + 0, 4, 4, 12, 16, 24, 32, 40, 52, 60, 68, 76, 80, 88, 88, 92, 92, 88, 88, 80, 76, 68, 60, 52, 40, 32, 24, 16, 12, 4, 4, 0, + 0, 4, 8, 16, 24, 32, 40, 52, 64, 76, 84, 92,100,108,112,116,116,112,108,100, 92, 84, 76, 64, 52, 40, 32, 24, 16, 8, 4, 0, + 0, 4, 8, 16, 28, 40, 52, 64, 76, 88,100,112,124,132,136,140,140,136,132,124,112,100, 88, 76, 64, 52, 40, 28, 16, 8, 4, 0, + 0, 4, 12, 20, 32, 44, 60, 76, 88,104,120,132,144,152,160,164,164,160,152,144,132,120,104, 88, 76, 60, 44, 32, 20, 12, 4, 0, + 0, 4, 12, 24, 36, 48, 68, 84,100,120,136,152,164,176,180,184,184,180,176,164,152,136,120,100, 84, 68, 48, 36, 24, 12, 4, 0, + 0, 4, 12, 24, 40, 56, 76, 92,112,132,152,168,180,192,204,208,208,204,192,180,168,152,132,112, 92, 76, 56, 40, 24, 12, 4, 0, + 0, 4, 16, 28, 44, 60, 80,100,124,144,164,180,196,208,220,224,224,220,208,196,180,164,144,124,100, 80, 60, 44, 28, 16, 4, 0, + 0, 8, 16, 28, 48, 64, 88,108,132,152,176,192,208,224,232,240,240,232,224,208,192,176,152,132,108, 88, 64, 48, 28, 16, 8, 0, + 0, 4, 16, 32, 48, 68, 88,112,136,160,180,204,220,232,244,248,248,244,232,220,204,180,160,136,112, 88, 68, 48, 32, 16, 4, 0, + 1, 8, 16, 32, 48, 72, 92,116,140,164,184,208,224,240,248,255,255,248,240,224,208,184,164,140,116, 92, 72, 48, 32, 16, 8, 1, + 1, 8, 16, 32, 48, 72, 92,116,140,164,184,208,224,240,248,255,255,248,240,224,208,184,164,140,116, 92, 72, 48, 32, 16, 8, 1, + 0, 4, 16, 32, 48, 68, 88,112,136,160,180,204,220,232,244,248,248,244,232,220,204,180,160,136,112, 88, 68, 48, 32, 16, 4, 0, + 0, 8, 16, 28, 48, 64, 88,108,132,152,176,192,208,224,232,240,240,232,224,208,192,176,152,132,108, 88, 64, 48, 28, 16, 8, 0, + 0, 4, 16, 28, 44, 60, 80,100,124,144,164,180,196,208,220,224,224,220,208,196,180,164,144,124,100, 80, 60, 44, 28, 16, 4, 0, + 0, 4, 12, 24, 40, 56, 76, 92,112,132,152,168,180,192,204,208,208,204,192,180,168,152,132,112, 92, 76, 56, 40, 24, 12, 4, 0, + 0, 4, 12, 24, 36, 48, 68, 84,100,120,136,152,164,176,180,184,184,180,176,164,152,136,120,100, 84, 68, 48, 36, 24, 12, 4, 0, + 0, 4, 12, 20, 32, 44, 60, 76, 88,104,120,132,144,152,160,164,164,160,152,144,132,120,104, 88, 76, 60, 44, 32, 20, 12, 4, 0, + 0, 4, 8, 16, 28, 40, 52, 64, 76, 88,100,112,124,132,136,140,140,136,132,124,112,100, 88, 76, 64, 52, 40, 28, 16, 8, 4, 0, + 0, 4, 8, 16, 24, 32, 40, 52, 64, 76, 84, 92,100,108,112,116,116,112,108,100, 92, 84, 76, 64, 52, 40, 32, 24, 16, 8, 4, 0, + 0, 4, 4, 12, 16, 24, 32, 40, 52, 60, 68, 76, 80, 88, 88, 92, 92, 88, 88, 80, 76, 68, 60, 52, 40, 32, 24, 16, 12, 4, 4, 0, + 0, 4, 4, 8, 12, 20, 24, 32, 40, 44, 52, 56, 60, 64, 68, 72, 72, 68, 64, 60, 56, 52, 44, 40, 32, 24, 20, 12, 8, 4, 4, 0, + 0, 0, 4, 8, 8, 12, 16, 24, 28, 32, 36, 40, 44, 48, 48, 48, 48, 48, 48, 44, 40, 36, 32, 28, 24, 16, 12, 8, 8, 4, 0, 0, + 0, 0, 4, 4, 8, 8, 12, 16, 16, 20, 24, 24, 28, 28, 32, 32, 32, 32, 28, 28, 24, 24, 20, 16, 16, 12, 8, 8, 4, 4, 0, 0, + 0, 0, 0, 4, 4, 4, 4, 8, 8, 12, 12, 12, 16, 16, 16, 16, 16, 16, 16, 16, 12, 12, 12, 8, 8, 4, 4, 4, 4, 0, 0, 0, + 0, 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, 8, 8, 8, 8, 8, 8, 4, 4, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, //error:0.000022 }; static const uint8_t obmc16[256]={ - 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, - 0, 1, 1, 2, 4, 5, 5, 6, 6, 5, 5, 4, 2, 1, 1, 0, - 0, 1, 4, 6, 9,11,13,15,15,13,11, 9, 6, 4, 1, 0, - 0, 2, 6,11,15,20,24,26,26,24,20,15,11, 6, 2, 0, - 0, 4, 9,15,23,29,34,38,38,34,29,23,15, 9, 4, 0, - 0, 5,11,20,29,38,45,49,49,45,38,29,20,11, 5, 0, - 1, 5,13,24,34,45,53,57,57,53,45,34,24,13, 5, 1, - 1, 6,15,26,38,49,57,62,62,57,49,38,26,15, 6, 1, - 1, 6,15,26,38,49,57,62,62,57,49,38,26,15, 6, 1, - 1, 5,13,24,34,45,53,57,57,53,45,34,24,13, 5, 1, - 0, 5,11,20,29,38,45,49,49,45,38,29,20,11, 5, 0, - 0, 4, 9,15,23,29,34,38,38,34,29,23,15, 9, 4, 0, - 0, 2, 6,11,15,20,24,26,26,24,20,15,11, 6, 2, 0, - 0, 1, 4, 6, 9,11,13,15,15,13,11, 9, 6, 4, 1, 0, - 0, 1, 1, 2, 4, 5, 5, 6, 6, 5, 5, 4, 2, 1, 1, 0, - 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, + 0, 4, 4, 8, 16, 20, 20, 24, 24, 20, 20, 16, 8, 4, 4, 0, + 0, 4, 16, 24, 36, 44, 52, 60, 60, 52, 44, 36, 24, 16, 4, 0, + 0, 8, 24, 44, 60, 80, 96,104,104, 96, 80, 60, 44, 24, 8, 0, + 0, 16, 36, 60, 92,116,136,152,152,136,116, 92, 60, 36, 16, 0, + 0, 20, 44, 80,116,152,180,196,196,180,152,116, 80, 44, 20, 0, + 4, 20, 52, 96,136,180,212,228,228,212,180,136, 96, 52, 20, 4, + 4, 24, 60,104,152,196,228,248,248,228,196,152,104, 60, 24, 4, + 4, 24, 60,104,152,196,228,248,248,228,196,152,104, 60, 24, 4, + 4, 20, 52, 96,136,180,212,228,228,212,180,136, 96, 52, 20, 4, + 0, 20, 44, 80,116,152,180,196,196,180,152,116, 80, 44, 20, 0, + 0, 16, 36, 60, 92,116,136,152,152,136,116, 92, 60, 36, 16, 0, + 0, 8, 24, 44, 60, 80, 96,104,104, 96, 80, 60, 44, 24, 8, 0, + 0, 4, 16, 24, 36, 44, 52, 60, 60, 52, 44, 36, 24, 16, 4, 0, + 0, 4, 4, 8, 16, 20, 20, 24, 24, 20, 20, 16, 8, 4, 4, 0, + 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, //error:0.000033 }; #elif 1 // 64*linear @@ -285,60 +286,60 @@ static const uint8_t obmc16[256]={ }; #else //64*cos static const uint8_t obmc32[1024]={ - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 2, 2, 1, 1, 1, 1, 0, 0, 0, - 0, 0, 1, 1, 1, 2, 2, 3, 4, 5, 5, 6, 7, 7, 7, 7, 7, 7, 7, 7, 6, 5, 5, 4, 3, 2, 2, 1, 1, 1, 0, 0, - 0, 0, 1, 1, 2, 3, 4, 5, 6, 7, 9,10,11,11,12,12,12,12,11,11,10, 9, 7, 6, 5, 4, 3, 2, 1, 1, 0, 0, - 0, 0, 1, 2, 3, 5, 6, 8, 9,11,12,14,15,16,17,17,17,17,16,15,14,12,11, 9, 8, 6, 5, 3, 2, 1, 0, 0, - 0, 1, 1, 2, 4, 6, 8,10,12,15,17,19,20,21,22,23,23,22,21,20,19,17,15,12,10, 8, 6, 4, 2, 1, 1, 0, - 0, 1, 2, 3, 5, 8,10,13,16,19,21,24,26,27,28,29,29,28,27,26,24,21,19,16,13,10, 8, 5, 3, 2, 1, 0, - 0, 1, 2, 4, 6, 9,12,16,19,23,26,29,31,33,34,35,35,34,33,31,29,26,23,19,16,12, 9, 6, 4, 2, 1, 0, - 0, 1, 3, 5, 7,11,15,19,23,26,30,34,37,39,40,41,41,40,39,37,34,30,26,23,19,15,11, 7, 5, 3, 1, 0, - 0, 1, 3, 5, 9,12,17,21,26,30,35,38,42,44,46,47,47,46,44,42,38,35,30,26,21,17,12, 9, 5, 3, 1, 0, - 0, 1, 3, 6, 9,14,19,24,29,34,38,43,46,49,51,52,52,51,49,46,43,38,34,29,24,19,14, 9, 6, 3, 1, 0, - 0, 1, 3, 6,11,15,20,26,31,37,42,46,50,53,56,57,57,56,53,50,46,42,37,31,26,20,15,11, 6, 3, 1, 0, - 0, 1, 3, 7,11,16,21,27,33,39,44,49,53,57,59,60,60,59,57,53,49,44,39,33,27,21,16,11, 7, 3, 1, 0, - 0, 1, 4, 7,12,17,22,28,34,40,46,51,56,59,61,63,63,61,59,56,51,46,40,34,28,22,17,12, 7, 4, 1, 0, - 0, 1, 4, 7,12,17,23,29,35,41,47,52,57,60,63,64,64,63,60,57,52,47,41,35,29,23,17,12, 7, 4, 1, 0, - 0, 1, 4, 7,12,17,23,29,35,41,47,52,57,60,63,64,64,63,60,57,52,47,41,35,29,23,17,12, 7, 4, 1, 0, - 0, 1, 4, 7,12,17,22,28,34,40,46,51,56,59,61,63,63,61,59,56,51,46,40,34,28,22,17,12, 7, 4, 1, 0, - 0, 1, 3, 7,11,16,21,27,33,39,44,49,53,57,59,60,60,59,57,53,49,44,39,33,27,21,16,11, 7, 3, 1, 0, - 0, 1, 3, 6,11,15,20,26,31,37,42,46,50,53,56,57,57,56,53,50,46,42,37,31,26,20,15,11, 6, 3, 1, 0, - 0, 1, 3, 6, 9,14,19,24,29,34,38,43,46,49,51,52,52,51,49,46,43,38,34,29,24,19,14, 9, 6, 3, 1, 0, - 0, 1, 3, 5, 9,12,17,21,26,30,35,38,42,44,46,47,47,46,44,42,38,35,30,26,21,17,12, 9, 5, 3, 1, 0, - 0, 1, 3, 5, 7,11,15,19,23,26,30,34,37,39,40,41,41,40,39,37,34,30,26,23,19,15,11, 7, 5, 3, 1, 0, - 0, 1, 2, 4, 6, 9,12,16,19,23,26,29,31,33,34,35,35,34,33,31,29,26,23,19,16,12, 9, 6, 4, 2, 1, 0, - 0, 1, 2, 3, 5, 8,10,13,16,19,21,24,26,27,28,29,29,28,27,26,24,21,19,16,13,10, 8, 5, 3, 2, 1, 0, - 0, 1, 1, 2, 4, 6, 8,10,12,15,17,19,20,21,22,23,23,22,21,20,19,17,15,12,10, 8, 6, 4, 2, 1, 1, 0, - 0, 0, 1, 2, 3, 5, 6, 8, 9,11,12,14,15,16,17,17,17,17,16,15,14,12,11, 9, 8, 6, 5, 3, 2, 1, 0, 0, - 0, 0, 1, 1, 2, 3, 4, 5, 6, 7, 9,10,11,11,12,12,12,12,11,11,10, 9, 7, 6, 5, 4, 3, 2, 1, 1, 0, 0, - 0, 0, 1, 1, 1, 2, 2, 3, 4, 5, 5, 6, 7, 7, 7, 7, 7, 7, 7, 7, 6, 5, 5, 4, 3, 2, 2, 1, 1, 1, 0, 0, - 0, 0, 0, 1, 1, 1, 1, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 3, 3, 3, 3, 2, 2, 1, 1, 1, 1, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1, 1, 2, 1, 1, 2, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0, - 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, 8, 4, 4, 8, 4, 4, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 4, 4, 4, 4, 8, 8, 12, 12, 12, 12, 16, 16, 16, 16, 16, 16, 12, 12, 12, 12, 8, 8, 4, 4, 4, 4, 0, 0, 0, + 0, 0, 4, 4, 4, 8, 8, 12, 16, 20, 20, 24, 28, 28, 28, 28, 28, 28, 28, 28, 24, 20, 20, 16, 12, 8, 8, 4, 4, 4, 0, 0, + 0, 0, 4, 4, 8, 12, 16, 20, 24, 28, 36, 40, 44, 44, 48, 48, 48, 48, 44, 44, 40, 36, 28, 24, 20, 16, 12, 8, 4, 4, 0, 0, + 0, 0, 4, 8, 12, 20, 24, 32, 36, 44, 48, 56, 60, 64, 68, 68, 68, 68, 64, 60, 56, 48, 44, 36, 32, 24, 20, 12, 8, 4, 0, 0, + 0, 4, 4, 8, 16, 24, 32, 40, 48, 60, 68, 76, 80, 84, 88, 92, 92, 88, 84, 80, 76, 68, 60, 48, 40, 32, 24, 16, 8, 4, 4, 0, + 0, 4, 8, 12, 20, 32, 40, 52, 64, 76, 84, 96,104,108,112,116,116,112,108,104, 96, 84, 76, 64, 52, 40, 32, 20, 12, 8, 4, 0, + 0, 4, 8, 16, 24, 36, 48, 64, 76, 92,104,116,124,132,136,140,140,136,132,124,116,104, 92, 76, 64, 48, 36, 24, 16, 8, 4, 0, + 0, 4, 12, 20, 28, 44, 60, 76, 92,104,120,136,148,156,160,164,164,160,156,148,136,120,104, 92, 76, 60, 44, 28, 20, 12, 4, 0, + 0, 4, 12, 20, 36, 48, 68, 84,104,120,140,152,168,176,184,188,188,184,176,168,152,140,120,104, 84, 68, 48, 36, 20, 12, 4, 0, + 0, 4, 12, 24, 36, 56, 76, 96,116,136,152,172,184,196,204,208,208,204,196,184,172,152,136,116, 96, 76, 56, 36, 24, 12, 4, 0, + 0, 4, 12, 24, 44, 60, 80,104,124,148,168,184,200,212,224,228,228,224,212,200,184,168,148,124,104, 80, 60, 44, 24, 12, 4, 0, + 0, 4, 12, 28, 44, 64, 84,108,132,156,176,196,212,228,236,240,240,236,228,212,196,176,156,132,108, 84, 64, 44, 28, 12, 4, 0, + 0, 4, 16, 28, 48, 68, 88,112,136,160,184,204,224,236,244,252,252,244,236,224,204,184,160,136,112, 88, 68, 48, 28, 16, 4, 0, + 1, 4, 16, 28, 48, 68, 92,116,140,164,188,208,228,240,252,255,255,252,240,228,208,188,164,140,116, 92, 68, 48, 28, 16, 4, 1, + 1, 4, 16, 28, 48, 68, 92,116,140,164,188,208,228,240,252,255,255,252,240,228,208,188,164,140,116, 92, 68, 48, 28, 16, 4, 1, + 0, 4, 16, 28, 48, 68, 88,112,136,160,184,204,224,236,244,252,252,244,236,224,204,184,160,136,112, 88, 68, 48, 28, 16, 4, 0, + 0, 4, 12, 28, 44, 64, 84,108,132,156,176,196,212,228,236,240,240,236,228,212,196,176,156,132,108, 84, 64, 44, 28, 12, 4, 0, + 0, 4, 12, 24, 44, 60, 80,104,124,148,168,184,200,212,224,228,228,224,212,200,184,168,148,124,104, 80, 60, 44, 24, 12, 4, 0, + 0, 4, 12, 24, 36, 56, 76, 96,116,136,152,172,184,196,204,208,208,204,196,184,172,152,136,116, 96, 76, 56, 36, 24, 12, 4, 0, + 0, 4, 12, 20, 36, 48, 68, 84,104,120,140,152,168,176,184,188,188,184,176,168,152,140,120,104, 84, 68, 48, 36, 20, 12, 4, 0, + 0, 4, 12, 20, 28, 44, 60, 76, 92,104,120,136,148,156,160,164,164,160,156,148,136,120,104, 92, 76, 60, 44, 28, 20, 12, 4, 0, + 0, 4, 8, 16, 24, 36, 48, 64, 76, 92,104,116,124,132,136,140,140,136,132,124,116,104, 92, 76, 64, 48, 36, 24, 16, 8, 4, 0, + 0, 4, 8, 12, 20, 32, 40, 52, 64, 76, 84, 96,104,108,112,116,116,112,108,104, 96, 84, 76, 64, 52, 40, 32, 20, 12, 8, 4, 0, + 0, 4, 4, 8, 16, 24, 32, 40, 48, 60, 68, 76, 80, 84, 88, 92, 92, 88, 84, 80, 76, 68, 60, 48, 40, 32, 24, 16, 8, 4, 4, 0, + 0, 0, 4, 8, 12, 20, 24, 32, 36, 44, 48, 56, 60, 64, 68, 68, 68, 68, 64, 60, 56, 48, 44, 36, 32, 24, 20, 12, 8, 4, 0, 0, + 0, 0, 4, 4, 8, 12, 16, 20, 24, 28, 36, 40, 44, 44, 48, 48, 48, 48, 44, 44, 40, 36, 28, 24, 20, 16, 12, 8, 4, 4, 0, 0, + 0, 0, 4, 4, 4, 8, 8, 12, 16, 20, 20, 24, 28, 28, 28, 28, 28, 28, 28, 28, 24, 20, 20, 16, 12, 8, 8, 4, 4, 4, 0, 0, + 0, 0, 0, 4, 4, 4, 4, 8, 8, 12, 12, 12, 12, 16, 16, 16, 16, 16, 16, 12, 12, 12, 12, 8, 8, 4, 4, 4, 4, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 4, 4, 8, 4, 4, 8, 4, 4, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, //error:0.000022 }; static const uint8_t obmc16[256]={ - 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, - 0, 0, 1, 2, 3, 4, 5, 5, 5, 5, 4, 3, 2, 1, 0, 0, - 0, 1, 3, 6, 8,11,13,14,14,13,11, 8, 6, 3, 1, 0, - 0, 2, 6,10,15,20,24,26,26,24,20,15,10, 6, 2, 0, - 0, 3, 8,16,23,30,35,38,38,35,30,23,16, 8, 3, 0, - 1, 4,11,20,30,39,46,49,49,46,39,30,20,11, 4, 1, - 1, 5,13,24,35,46,54,58,58,54,46,35,24,13, 5, 1, - 0, 5,14,26,38,49,58,63,63,58,49,38,26,14, 5, 0, - 0, 5,14,26,38,49,58,63,63,58,49,38,26,14, 5, 0, - 1, 5,13,24,35,46,54,58,58,54,46,35,24,13, 5, 1, - 1, 4,11,20,30,39,46,49,49,46,39,30,20,11, 4, 1, - 0, 3, 8,16,23,30,35,38,38,35,30,23,16, 8, 3, 0, - 0, 2, 6,10,15,20,24,26,26,24,20,15,10, 6, 2, 0, - 0, 1, 3, 6, 8,11,13,14,14,13,11, 8, 6, 3, 1, 0, - 0, 0, 1, 2, 3, 4, 5, 5, 5, 5, 4, 3, 2, 1, 0, 0, - 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, + 0, 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, + 0, 0, 4, 8, 12, 16, 20, 20, 20, 20, 16, 12, 8, 4, 0, 0, + 0, 4, 12, 24, 32, 44, 52, 56, 56, 52, 44, 32, 24, 12, 4, 0, + 0, 8, 24, 40, 60, 80, 96,104,104, 96, 80, 60, 40, 24, 8, 0, + 0, 12, 32, 64, 92,120,140,152,152,140,120, 92, 64, 32, 12, 0, + 4, 16, 44, 80,120,156,184,196,196,184,156,120, 80, 44, 16, 4, + 4, 20, 52, 96,140,184,216,232,232,216,184,140, 96, 52, 20, 4, + 0, 20, 56,104,152,196,232,252,252,232,196,152,104, 56, 20, 0, + 0, 20, 56,104,152,196,232,252,252,232,196,152,104, 56, 20, 0, + 4, 20, 52, 96,140,184,216,232,232,216,184,140, 96, 52, 20, 4, + 4, 16, 44, 80,120,156,184,196,196,184,156,120, 80, 44, 16, 4, + 0, 12, 32, 64, 92,120,140,152,152,140,120, 92, 64, 32, 12, 0, + 0, 8, 24, 40, 60, 80, 96,104,104, 96, 80, 60, 40, 24, 8, 0, + 0, 4, 12, 24, 32, 44, 52, 56, 56, 52, 44, 32, 24, 12, 4, 0, + 0, 0, 4, 8, 12, 16, 20, 20, 20, 20, 16, 12, 8, 4, 0, 0, + 0, 0, 0, 0, 0, 4, 4, 4, 4, 4, 4, 0, 0, 0, 0, 0, //error:0.000022 }; -#endif +#endif /* 0 */ //linear *64 static const uint8_t obmc8[64]={ @@ -392,6 +393,8 @@ static const BlockNode null_block= { //FIXME add border maybe #define LOG2_MB_SIZE 4 #define MB_SIZE (1<line_count = line_count; buf->line_width = line_width; buf->data_count = max_allocated_lines; - buf->line = (DWTELEM * *) av_mallocz (sizeof(DWTELEM *) * line_count); - buf->data_stack = (DWTELEM * *) av_malloc (sizeof(DWTELEM *) * max_allocated_lines); + buf->line = av_mallocz (sizeof(IDWTELEM *) * line_count); + buf->data_stack = av_malloc (sizeof(IDWTELEM *) * max_allocated_lines); - for (i = 0; i < max_allocated_lines; i++) - { - buf->data_stack[i] = (DWTELEM *) av_malloc (sizeof(DWTELEM) * line_width); + for(i = 0; i < max_allocated_lines; i++){ + buf->data_stack[i] = av_malloc (sizeof(IDWTELEM) * line_width); } buf->data_stack_top = max_allocated_lines - 1; } -static DWTELEM * slice_buffer_load_line(slice_buffer * buf, int line) +static IDWTELEM * slice_buffer_load_line(slice_buffer * buf, int line) { int offset; - DWTELEM * buffer; - -// av_log(NULL, AV_LOG_DEBUG, "Cache hit: %d\n", line); + IDWTELEM * buffer; assert(buf->data_stack_top >= 0); // assert(!buf->line[line]); @@ -518,15 +536,13 @@ static DWTELEM * slice_buffer_load_line(slice_buffer * buf, int line) buf->data_stack_top--; buf->line[line] = buffer; -// av_log(NULL, AV_LOG_DEBUG, "slice_buffer_load_line: line: %d remaining: %d\n", line, buf->data_stack_top + 1); - return buffer; } static void slice_buffer_release(slice_buffer * buf, int line) { int offset; - DWTELEM * buffer; + IDWTELEM * buffer; assert(line >= 0 && line < buf->line_count); assert(buf->line[line]); @@ -536,20 +552,14 @@ static void slice_buffer_release(slice_buffer * buf, int line) buf->data_stack_top++; buf->data_stack[buf->data_stack_top] = buffer; buf->line[line] = NULL; - -// av_log(NULL, AV_LOG_DEBUG, "slice_buffer_release: line: %d remaining: %d\n", line, buf->data_stack_top + 1); } static void slice_buffer_flush(slice_buffer * buf) { int i; - for (i = 0; i < buf->line_count; i++) - { + for(i = 0; i < buf->line_count; i++){ if (buf->line[i]) - { -// av_log(NULL, AV_LOG_DEBUG, "slice_buffer_flush: line: %d \n", i); slice_buffer_release(buf, i); - } } } @@ -558,14 +568,10 @@ static void slice_buffer_destroy(slice_buffer * buf) int i; slice_buffer_flush(buf); - for (i = buf->data_count - 1; i >= 0; i--) - { - assert(buf->data_stack[i]); + for(i = buf->data_count - 1; i >= 0; i--){ av_freep(&buf->data_stack[i]); } - assert(buf->data_stack); av_freep(&buf->data_stack); - assert(buf->line); av_freep(&buf->line); } @@ -588,7 +594,7 @@ static inline void put_symbol(RangeCoder *c, uint8_t *state, int v, int is_signe int i; if(v){ - const int a= ABS(v); + const int a= FFABS(v); const int e= av_log2(a); #if 1 const int el= FFMIN(e, 10); @@ -639,7 +645,7 @@ static inline void put_symbol(RangeCoder *c, uint8_t *state, int v, int is_signe if(is_signed) put_rac(c, state+11 + FFMIN(e,10), v < 0); //11..21 } -#endif +#endif /* 1 */ }else{ put_rac(c, state+0, 1); } @@ -707,7 +713,11 @@ static inline int get_symbol2(RangeCoder *c, uint8_t *state, int log2){ return v; } -static always_inline void lift(DWTELEM *dst, DWTELEM *src, DWTELEM *ref, int dst_step, int src_step, int ref_step, int width, int mul, int add, int shift, int highpass, int inverse){ +static av_always_inline void +lift(DWTELEM *dst, DWTELEM *src, DWTELEM *ref, + int dst_step, int src_step, int ref_step, + int width, int mul, int add, int shift, + int highpass, int inverse){ const int mirror_left= !highpass; const int mirror_right= (width&1) ^ highpass; const int w= (width>>1) - 1 + (highpass & width); @@ -721,55 +731,68 @@ static always_inline void lift(DWTELEM *dst, DWTELEM *src, DWTELEM *ref, int dst } for(i=0; i>shift), inverse); + dst[i*dst_step] = + LIFT(src[i*src_step], + ((mul*(ref[i*ref_step] + ref[(i+1)*ref_step])+add)>>shift), + inverse); } if(mirror_right){ - dst[w*dst_step] = LIFT(src[w*src_step], ((mul*2*ref[w*ref_step]+add)>>shift), inverse); + dst[w*dst_step] = + LIFT(src[w*src_step], + ((mul*2*ref[w*ref_step]+add)>>shift), + inverse); } } -#ifndef lift5 -static always_inline void lift5(DWTELEM *dst, DWTELEM *src, DWTELEM *ref, int dst_step, int src_step, int ref_step, int width, int mul, int add, int shift, int highpass, int inverse){ +static av_always_inline void +inv_lift(IDWTELEM *dst, IDWTELEM *src, IDWTELEM *ref, + int dst_step, int src_step, int ref_step, + int width, int mul, int add, int shift, + int highpass, int inverse){ const int mirror_left= !highpass; const int mirror_right= (width&1) ^ highpass; const int w= (width>>1) - 1 + (highpass & width); int i; +#define LIFT(src, ref, inv) ((src) + ((inv) ? - (ref) : + (ref))) if(mirror_left){ - int r= 3*2*ref[0]; - r += r>>4; - r += r>>8; - dst[0] = LIFT(src[0], ((r+add)>>shift), inverse); + dst[0] = LIFT(src[0], ((mul*2*ref[0]+add)>>shift), inverse); dst += dst_step; src += src_step; } for(i=0; i>4; - r += r>>8; - dst[i*dst_step] = LIFT(src[i*src_step], ((r+add)>>shift), inverse); + dst[i*dst_step] = + LIFT(src[i*src_step], + ((mul*(ref[i*ref_step] + ref[(i+1)*ref_step])+add)>>shift), + inverse); } if(mirror_right){ - int r= 3*2*ref[w*ref_step]; - r += r>>4; - r += r>>8; - dst[w*dst_step] = LIFT(src[w*src_step], ((r+add)>>shift), inverse); + dst[w*dst_step] = + LIFT(src[w*src_step], + ((mul*2*ref[w*ref_step]+add)>>shift), + inverse); } } -#endif #ifndef liftS -static always_inline void liftS(DWTELEM *dst, DWTELEM *src, DWTELEM *ref, int dst_step, int src_step, int ref_step, int width, int mul, int add, int shift, int highpass, int inverse){ +static av_always_inline void +liftS(DWTELEM *dst, DWTELEM *src, DWTELEM *ref, + int dst_step, int src_step, int ref_step, + int width, int mul, int add, int shift, + int highpass, int inverse){ const int mirror_left= !highpass; const int mirror_right= (width&1) ^ highpass; const int w= (width>>1) - 1 + (highpass & width); int i; assert(shift == 4); -#define LIFTS(src, ref, inv) ((inv) ? (src) - (((ref) - 4*(src))>>shift): (16*4*(src) + 4*(ref) + 8 + (5<<27))/(5*16) - (1<<23)) +#define LIFTS(src, ref, inv) \ + ((inv) ? \ + (src) + (((ref) + 4*(src))>>shift): \ + -((-16*(src) + (ref) + add/4 + 1 + (5<<25))/(5*4) - (1<<23))) if(mirror_left){ dst[0] = LIFTS(src[0], mul*2*ref[0]+add, inverse); dst += dst_step; @@ -777,242 +800,51 @@ static always_inline void liftS(DWTELEM *dst, DWTELEM *src, DWTELEM *ref, int ds } for(i=0; i=width) x2= 2*width-x2-2; - sum += coeffs[i]*(int64_t)dst[x2]; - } - if(inverse) dst[x] -= (sum + (1<>shift; - else dst[x] += (sum + (1<>shift; - } -} - -static void inplace_liftV(DWTELEM *dst, int width, int height, int stride, int *coeffs, int n, int shift, int start, int inverse){ - int x, y, i; - for(y=start; y=height) y2= 2*height-y2-2; - sum += coeffs[i]*(int64_t)dst[x + y2*stride]; - } - if(inverse) dst[x + y*stride] -= (sum + (1<>shift; - else dst[x + y*stride] += (sum + (1<>shift; - } - } -} - -#define SCALEX 1 -#define LX0 0 -#define LX1 1 - -#if 0 // more accurate 9/7 -#define N1 2 -#define SHIFT1 14 -#define COEFFS1 (int[]){-25987,-25987} -#define N2 2 -#define SHIFT2 19 -#define COEFFS2 (int[]){-27777,-27777} -#define N3 2 -#define SHIFT3 15 -#define COEFFS3 (int[]){28931,28931} -#define N4 2 -#define SHIFT4 15 -#define COEFFS4 (int[]){14533,14533} -#elif 1 // 13/7 CRF -#define N1 4 -#define SHIFT1 4 -#define COEFFS1 (int[]){1,-9,-9,1} -#define N2 4 -#define SHIFT2 4 -#define COEFFS2 (int[]){-1,5,5,-1} -#define N3 0 -#define SHIFT3 1 -#define COEFFS3 NULL -#define N4 0 -#define SHIFT4 1 -#define COEFFS4 NULL -#elif 1 // 3/5 -#define LX0 1 -#define LX1 0 -#define SCALEX 0.5 -#define N1 2 -#define SHIFT1 1 -#define COEFFS1 (int[]){1,1} -#define N2 2 -#define SHIFT2 2 -#define COEFFS2 (int[]){-1,-1} -#define N3 0 -#define SHIFT3 0 -#define COEFFS3 NULL -#define N4 0 -#define SHIFT4 0 -#define COEFFS4 NULL -#elif 1 // 11/5 -#define N1 0 -#define SHIFT1 1 -#define COEFFS1 NULL -#define N2 2 -#define SHIFT2 2 -#define COEFFS2 (int[]){-1,-1} -#define N3 2 -#define SHIFT3 0 -#define COEFFS3 (int[]){-1,-1} -#define N4 4 -#define SHIFT4 7 -#define COEFFS4 (int[]){-5,29,29,-5} -#define SCALEX 4 -#elif 1 // 9/7 CDF -#define N1 2 -#define SHIFT1 7 -#define COEFFS1 (int[]){-203,-203} -#define N2 2 -#define SHIFT2 12 -#define COEFFS2 (int[]){-217,-217} -#define N3 2 -#define SHIFT3 7 -#define COEFFS3 (int[]){113,113} -#define N4 2 -#define SHIFT4 9 -#define COEFFS4 (int[]){227,227} -#define SCALEX 1 -#elif 1 // 7/5 CDF -#define N1 0 -#define SHIFT1 1 -#define COEFFS1 NULL -#define N2 2 -#define SHIFT2 2 -#define COEFFS2 (int[]){-1,-1} -#define N3 2 -#define SHIFT3 0 -#define COEFFS3 (int[]){-1,-1} -#define N4 2 -#define SHIFT4 4 -#define COEFFS4 (int[]){3,3} -#elif 1 // 9/7 MN -#define N1 4 -#define SHIFT1 4 -#define COEFFS1 (int[]){1,-9,-9,1} -#define N2 2 -#define SHIFT2 2 -#define COEFFS2 (int[]){1,1} -#define N3 0 -#define SHIFT3 1 -#define COEFFS3 NULL -#define N4 0 -#define SHIFT4 1 -#define COEFFS4 NULL -#else // 13/7 CRF -#define N1 4 -#define SHIFT1 4 -#define COEFFS1 (int[]){1,-9,-9,1} -#define N2 4 -#define SHIFT2 4 -#define COEFFS2 (int[]){-1,5,5,-1} -#define N3 0 -#define SHIFT3 1 -#define COEFFS3 NULL -#define N4 0 -#define SHIFT4 1 -#define COEFFS4 NULL -#endif -static void horizontal_decomposeX(DWTELEM *b, int width){ - DWTELEM temp[width]; - const int width2= width>>1; - const int w2= (width+1)>>1; - int x; - - inplace_lift(b, width, COEFFS1, N1, SHIFT1, LX1, 0); - inplace_lift(b, width, COEFFS2, N2, SHIFT2, LX0, 0); - inplace_lift(b, width, COEFFS3, N3, SHIFT3, LX1, 0); - inplace_lift(b, width, COEFFS4, N4, SHIFT4, LX0, 0); - - for(x=0; x>1; - int x; - const int w2= (width+1)>>1; - - memcpy(temp, b, width*sizeof(int)); - for(x=0; x>1) - 1 + (highpass & width); + int i; - for(y=0; y>shift): \ + -((-16*(src) + (ref) + add/4 + 1 + (5<<25))/(5*4) - (1<<23))) + if(mirror_left){ + dst[0] = LIFTS(src[0], mul*2*ref[0]+add, inverse); + dst += dst_step; + src += src_step; } - inplace_liftV(buffer, width, height, stride, COEFFS1, N1, SHIFT1, LX1, 0); - inplace_liftV(buffer, width, height, stride, COEFFS2, N2, SHIFT2, LX0, 0); - inplace_liftV(buffer, width, height, stride, COEFFS3, N3, SHIFT3, LX1, 0); - inplace_liftV(buffer, width, height, stride, COEFFS4, N4, SHIFT4, LX0, 0); -} - -static void spatial_composeX(DWTELEM *buffer, int width, int height, int stride){ - int x, y; - - inplace_liftV(buffer, width, height, stride, COEFFS4, N4, SHIFT4, LX0, 1); - inplace_liftV(buffer, width, height, stride, COEFFS3, N3, SHIFT3, LX1, 1); - inplace_liftV(buffer, width, height, stride, COEFFS2, N2, SHIFT2, LX0, 1); - inplace_liftV(buffer, width, height, stride, COEFFS1, N1, SHIFT1, LX1, 1); - - for(y=0; y>1; - lift (temp+w2, b +1, b , 1, 2, 2, width, -W_AM, W_AO, W_AS, 1, 0); - liftS(temp , b , temp+w2, 1, 2, 1, width, -W_BM, W_BO, W_BS, 0, 0); - lift5(b +w2, temp+w2, temp , 1, 1, 1, width, W_CM, W_CO, W_CS, 1, 0); + lift (temp+w2, b +1, b , 1, 2, 2, width, W_AM, W_AO, W_AS, 1, 1); + liftS(temp , b , temp+w2, 1, 2, 1, width, W_BM, W_BO, W_BS, 0, 0); + lift (b +w2, temp+w2, temp , 1, 1, 1, width, W_CM, W_CO, W_CS, 1, 0); lift (b , temp , b +w2, 1, 1, 1, width, W_DM, W_DO, W_DS, 0, 0); } @@ -1126,14 +954,7 @@ static void vertical_decompose97iH1(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int w int i; for(i=0; i>W_CS; -#else - int r= 3*(b0[i] + b2[i]); - r+= r>>4; - r+= r>>8; - b1[i] += (r+W_CO)>>W_CS; -#endif } } @@ -1144,7 +965,7 @@ static void vertical_decompose97iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int w #ifdef liftS b1[i] -= (W_BM*(b0[i] + b2[i])+W_BO)>>W_BS; #else - b1[i] = (16*4*b1[i] - 4*(b0[i] + b2[i]) + 8*5 + (5<<27)) / (5*16) - (1<<23); + b1[i] = (16*4*b1[i] - 4*(b0[i] + b2[i]) + W_BO*5 + (5<<27)) / (5*16) - (1<<23); #endif } } @@ -1168,23 +989,14 @@ static void spatial_decompose97i(DWTELEM *buffer, int width, int height, int str DWTELEM *b4= buffer + mirror(y+3, height-1)*stride; DWTELEM *b5= buffer + mirror(y+4, height-1)*stride; -{START_TIMER if(y+3<(unsigned)height) horizontal_decompose97i(b4, width); if(y+4<(unsigned)height) horizontal_decompose97i(b5, width); -if(width>400){ -STOP_TIMER("horizontal_decompose97i") -}} -{START_TIMER if(y+3<(unsigned)height) vertical_decompose97iH0(b3, b4, b5, width); if(y+2<(unsigned)height) vertical_decompose97iL0(b2, b3, b4, width); if(y+1<(unsigned)height) vertical_decompose97iH1(b1, b2, b3, width); if(y+0<(unsigned)height) vertical_decompose97iL1(b0, b1, b2, width); -if(width>400){ -STOP_TIMER("vertical_decompose97i") -}} - b0=b2; b1=b3; b2=b4; @@ -1199,13 +1011,12 @@ void ff_spatial_dwt(DWTELEM *buffer, int width, int height, int stride, int type switch(type){ case DWT_97: spatial_decompose97i(buffer, width>>level, height>>level, stride<>level, height>>level, stride<>level, height>>level, stride<>1; const int w2= (width+1)>>1; int x; @@ -1240,9 +1051,9 @@ static void horizontal_compose53i(DWTELEM *b, int width){ b[width -1] = A3; b[width2-1] = A2; #else - lift(temp , b , b+w2, 1, 1, 1, width, 1, 2, 2, 0, 1); - lift(temp+w2, b+w2, temp, 1, 1, 1, width, -1, 0, 1, 1, 1); -#endif + inv_lift(temp , b , b+w2, 1, 1, 1, width, 1, 2, 2, 0, 1); + inv_lift(temp+w2, b+w2, temp, 1, 1, 1, width, -1, 0, 1, 1, 1); +#endif /* 0 */ for(x=0; xy = -1; } -static void spatial_compose53i_init(dwt_compose_t *cs, DWTELEM *buffer, int height, int stride){ +static void spatial_compose53i_init(dwt_compose_t *cs, IDWTELEM *buffer, int height, int stride){ cs->b0 = buffer + mirror(-1-1, height-1)*stride; cs->b1 = buffer + mirror(-1 , height-1)*stride; cs->y = -1; @@ -1282,49 +1093,41 @@ static void spatial_compose53i_init(dwt_compose_t *cs, DWTELEM *buffer, int heig static void spatial_compose53i_dy_buffered(dwt_compose_t *cs, slice_buffer * sb, int width, int height, int stride_line){ int y= cs->y; - DWTELEM *b0= cs->b0; - DWTELEM *b1= cs->b1; - DWTELEM *b2= slice_buffer_get_line(sb, mirror(y+1, height-1) * stride_line); - DWTELEM *b3= slice_buffer_get_line(sb, mirror(y+2, height-1) * stride_line); + IDWTELEM *b0= cs->b0; + IDWTELEM *b1= cs->b1; + IDWTELEM *b2= slice_buffer_get_line(sb, mirror(y+1, height-1) * stride_line); + IDWTELEM *b3= slice_buffer_get_line(sb, mirror(y+2, height-1) * stride_line); -{START_TIMER if(y+1<(unsigned)height) vertical_compose53iL0(b1, b2, b3, width); if(y+0<(unsigned)height) vertical_compose53iH0(b0, b1, b2, width); -STOP_TIMER("vertical_compose53i*")} -{START_TIMER if(y-1<(unsigned)height) horizontal_compose53i(b0, width); if(y+0<(unsigned)height) horizontal_compose53i(b1, width); -STOP_TIMER("horizontal_compose53i")} cs->b0 = b2; cs->b1 = b3; cs->y += 2; } -static void spatial_compose53i_dy(dwt_compose_t *cs, DWTELEM *buffer, int width, int height, int stride){ +static void spatial_compose53i_dy(dwt_compose_t *cs, IDWTELEM *buffer, int width, int height, int stride){ int y= cs->y; - DWTELEM *b0= cs->b0; - DWTELEM *b1= cs->b1; - DWTELEM *b2= buffer + mirror(y+1, height-1)*stride; - DWTELEM *b3= buffer + mirror(y+2, height-1)*stride; + IDWTELEM *b0= cs->b0; + IDWTELEM *b1= cs->b1; + IDWTELEM *b2= buffer + mirror(y+1, height-1)*stride; + IDWTELEM *b3= buffer + mirror(y+2, height-1)*stride; -{START_TIMER if(y+1<(unsigned)height) vertical_compose53iL0(b1, b2, b3, width); if(y+0<(unsigned)height) vertical_compose53iH0(b0, b1, b2, width); -STOP_TIMER("vertical_compose53i*")} -{START_TIMER if(y-1<(unsigned)height) horizontal_compose53i(b0, width); if(y+0<(unsigned)height) horizontal_compose53i(b1, width); -STOP_TIMER("horizontal_compose53i")} cs->b0 = b2; cs->b1 = b3; cs->y += 2; } -static void spatial_compose53i(DWTELEM *buffer, int width, int height, int stride){ +static void av_unused spatial_compose53i(IDWTELEM *buffer, int width, int height, int stride){ dwt_compose_t cs; spatial_compose53i_init(&cs, buffer, height, stride); while(cs.y <= height) @@ -1332,17 +1135,17 @@ static void spatial_compose53i(DWTELEM *buffer, int width, int height, int strid } -void ff_snow_horizontal_compose97i(DWTELEM *b, int width){ - DWTELEM temp[width]; +void ff_snow_horizontal_compose97i(IDWTELEM *b, int width){ + IDWTELEM temp[width]; const int w2= (width+1)>>1; - lift (temp , b , b +w2, 1, 1, 1, width, W_DM, W_DO, W_DS, 0, 1); - lift5(temp+w2, b +w2, temp , 1, 1, 1, width, W_CM, W_CO, W_CS, 1, 1); - liftS(b , temp , temp+w2, 2, 1, 1, width, -W_BM, W_BO, W_BS, 0, 1); - lift (b+1 , temp+w2, b , 2, 1, 2, width, -W_AM, W_AO, W_AS, 1, 1); + inv_lift (temp , b , b +w2, 1, 1, 1, width, W_DM, W_DO, W_DS, 0, 1); + inv_lift (temp+w2, b +w2, temp , 1, 1, 1, width, W_CM, W_CO, W_CS, 1, 1); + inv_liftS(b , temp , temp+w2, 2, 1, 1, width, W_BM, W_BO, W_BS, 0, 1); + inv_lift (b+1 , temp+w2, b , 2, 1, 2, width, W_AM, W_AO, W_AS, 1, 0); } -static void vertical_compose97iH0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){ +static void vertical_compose97iH0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){ int i; for(i=0; i>W_CS; -#else - int r= 3*(b0[i] + b2[i]); - r+= r>>4; - r+= r>>8; - b1[i] -= (r+W_CO)>>W_CS; -#endif } } -static void vertical_compose97iL0(DWTELEM *b0, DWTELEM *b1, DWTELEM *b2, int width){ +static void vertical_compose97iL0(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM *b2, int width){ int i; for(i=0; i>W_DS; -#ifdef lift5 b3[i] -= (W_CM*(b2[i] + b4[i])+W_CO)>>W_CS; -#else - r= 3*(b2[i] + b4[i]); - r+= r>>4; - r+= r>>8; - b3[i] -= (r+W_CO)>>W_CS; -#endif #ifdef liftS b2[i] += (W_BM*(b1[i] + b3[i])+W_BO)>>W_BS; #else @@ -1418,7 +1204,7 @@ static void spatial_compose97i_buffered_init(dwt_compose_t *cs, slice_buffer * s cs->y = -3; } -static void spatial_compose97i_init(dwt_compose_t *cs, DWTELEM *buffer, int height, int stride){ +static void spatial_compose97i_init(dwt_compose_t *cs, IDWTELEM *buffer, int height, int stride){ cs->b0 = buffer + mirror(-3-1, height-1)*stride; cs->b1 = buffer + mirror(-3 , height-1)*stride; cs->b2 = buffer + mirror(-3+1, height-1)*stride; @@ -1429,14 +1215,13 @@ static void spatial_compose97i_init(dwt_compose_t *cs, DWTELEM *buffer, int heig static void spatial_compose97i_dy_buffered(DSPContext *dsp, dwt_compose_t *cs, slice_buffer * sb, int width, int height, int stride_line){ int y = cs->y; - DWTELEM *b0= cs->b0; - DWTELEM *b1= cs->b1; - DWTELEM *b2= cs->b2; - DWTELEM *b3= cs->b3; - DWTELEM *b4= slice_buffer_get_line(sb, mirror(y + 3, height - 1) * stride_line); - DWTELEM *b5= slice_buffer_get_line(sb, mirror(y + 4, height - 1) * stride_line); + IDWTELEM *b0= cs->b0; + IDWTELEM *b1= cs->b1; + IDWTELEM *b2= cs->b2; + IDWTELEM *b3= cs->b3; + IDWTELEM *b4= slice_buffer_get_line(sb, mirror(y + 3, height - 1) * stride_line); + IDWTELEM *b5= slice_buffer_get_line(sb, mirror(y + 4, height - 1) * stride_line); -{START_TIMER if(y>0 && y+4vertical_compose97i(b0, b1, b2, b3, b4, b5, width); }else{ @@ -1445,14 +1230,9 @@ static void spatial_compose97i_dy_buffered(DSPContext *dsp, dwt_compose_t *cs, s if(y+1<(unsigned)height) vertical_compose97iL0(b1, b2, b3, width); if(y+0<(unsigned)height) vertical_compose97iH0(b0, b1, b2, width); } -if(width>400){ -STOP_TIMER("vertical_compose97i")}} -{START_TIMER if(y-1<(unsigned)height) dsp->horizontal_compose97i(b0, width); if(y+0<(unsigned)height) dsp->horizontal_compose97i(b1, width); -if(width>400 && y+0<(unsigned)height){ -STOP_TIMER("horizontal_compose97i")}} cs->b0=b2; cs->b1=b3; @@ -1461,28 +1241,22 @@ STOP_TIMER("horizontal_compose97i")}} cs->y += 2; } -static void spatial_compose97i_dy(dwt_compose_t *cs, DWTELEM *buffer, int width, int height, int stride){ +static void spatial_compose97i_dy(dwt_compose_t *cs, IDWTELEM *buffer, int width, int height, int stride){ int y = cs->y; - DWTELEM *b0= cs->b0; - DWTELEM *b1= cs->b1; - DWTELEM *b2= cs->b2; - DWTELEM *b3= cs->b3; - DWTELEM *b4= buffer + mirror(y+3, height-1)*stride; - DWTELEM *b5= buffer + mirror(y+4, height-1)*stride; - -{START_TIMER + IDWTELEM *b0= cs->b0; + IDWTELEM *b1= cs->b1; + IDWTELEM *b2= cs->b2; + IDWTELEM *b3= cs->b3; + IDWTELEM *b4= buffer + mirror(y+3, height-1)*stride; + IDWTELEM *b5= buffer + mirror(y+4, height-1)*stride; + if(y+3<(unsigned)height) vertical_compose97iL1(b3, b4, b5, width); if(y+2<(unsigned)height) vertical_compose97iH1(b2, b3, b4, width); if(y+1<(unsigned)height) vertical_compose97iL0(b1, b2, b3, width); if(y+0<(unsigned)height) vertical_compose97iH0(b0, b1, b2, width); -if(width>400){ -STOP_TIMER("vertical_compose97i")}} -{START_TIMER if(y-1<(unsigned)height) ff_snow_horizontal_compose97i(b0, width); if(y+0<(unsigned)height) ff_snow_horizontal_compose97i(b1, width); -if(width>400 && b0 <= b2){ -STOP_TIMER("horizontal_compose97i")}} cs->b0=b2; cs->b1=b3; @@ -1491,7 +1265,7 @@ STOP_TIMER("horizontal_compose97i")}} cs->y += 2; } -static void spatial_compose97i(DWTELEM *buffer, int width, int height, int stride){ +static void av_unused spatial_compose97i(IDWTELEM *buffer, int width, int height, int stride){ dwt_compose_t cs; spatial_compose97i_init(&cs, buffer, height, stride); while(cs.y <= height) @@ -1504,26 +1278,21 @@ static void ff_spatial_idwt_buffered_init(dwt_compose_t *cs, slice_buffer * sb, switch(type){ case DWT_97: spatial_compose97i_buffered_init(cs+level, sb, height>>level, stride_line<>level, stride_line<>level, height>>level, stride<=0; level--){ switch(type){ case DWT_97: spatial_compose97i_init(cs+level, buffer, height>>level, stride<>level, stride<>level, height>>level, stride<>level)+support, height>>level)){ switch(type){ case DWT_97: spatial_compose97i_dy(cs+level, buffer, width>>level, height>>level, stride<>level, height>>level, stride<>level)+support, height>>level)){ switch(type){ case DWT_97: spatial_compose97i_dy_buffered(dsp, cs+level, slice_buf, width>>level, height>>level, stride_line<>level, height>>level, stride_line<=0; level--) - spatial_composeX (buffer, width>>level, height>>level, stride<width; const int h= b->height; int x, y; @@ -1664,7 +1425,7 @@ static int encode_subband_c0run(SnowContext *s, SubBand *b, DWTELEM *src, DWTELE p= parent[px + py*2*stride]; } if(/*ll|*/l|lt|t|rt|p){ - int context= av_log2(/*ABS(ll) + */3*ABS(l) + ABS(lt) + 2*ABS(t) + ABS(rt) + ABS(p)); + int context= av_log2(/*FFABS(ll) + */3*FFABS(l) + FFABS(lt) + 2*FFABS(t) + FFABS(rt) + FFABS(p)); put_rac(&s->c, &b->state[0][context], !!v); }else{ @@ -1680,11 +1441,11 @@ static int encode_subband_c0run(SnowContext *s, SubBand *b, DWTELEM *src, DWTELE } } if(v){ - int context= av_log2(/*ABS(ll) + */3*ABS(l) + ABS(lt) + 2*ABS(t) + ABS(rt) + ABS(p)); - int l2= 2*ABS(l) + (l<0); - int t2= 2*ABS(t) + (t<0); + int context= av_log2(/*FFABS(ll) + */3*FFABS(l) + FFABS(lt) + 2*FFABS(t) + FFABS(rt) + FFABS(p)); + int l2= 2*FFABS(l) + (l<0); + int t2= 2*FFABS(t) + (t<0); - put_symbol2(&s->c, b->state[context + 2], ABS(v)-1, context-4); + put_symbol2(&s->c, b->state[context + 2], FFABS(v)-1, context-4); put_rac(&s->c, &b->state[0][16 + 1 + 3 + quant3bA[l2&0xFF] + 3*quant3bA[t2&0xFF]], v<0); } } @@ -1693,7 +1454,7 @@ static int encode_subband_c0run(SnowContext *s, SubBand *b, DWTELEM *src, DWTELE return 0; } -static int encode_subband(SnowContext *s, SubBand *b, DWTELEM *src, DWTELEM *parent, int stride, int orientation){ +static int encode_subband(SnowContext *s, SubBand *b, IDWTELEM *src, IDWTELEM *parent, int stride, int orientation){ // encode_subband_qtree(s, b, src, parent, stride, orientation); // encode_subband_z0run(s, b, src, parent, stride, orientation); return encode_subband_c0run(s, b, src, parent, stride, orientation); @@ -1747,7 +1508,7 @@ static inline void unpack_coeffs(SnowContext *s, SubBand *b, SubBand * parent, i } } if(/*ll|*/l|lt|t|rt|p){ - int context= av_log2(/*ABS(ll) + */3*(l>>1) + (lt>>1) + (t&~1) + (rt>>1) + (p>>1)); + int context= av_log2(/*FFABS(ll) + */3*(l>>1) + (lt>>1) + (t&~1) + (rt>>1) + (p>>1)); v=get_rac(&s->c, &b->state[0][context]); if(v){ @@ -1803,14 +1564,12 @@ static inline void unpack_coeffs(SnowContext *s, SubBand *b, SubBand * parent, i static inline void decode_subband_slice_buffered(SnowContext *s, SubBand *b, slice_buffer * sb, int start_y, int h, int save_state[1]){ const int w= b->width; int y; - const int qlog= clip(s->qlog + b->qlog, 0, QROOT*16); + const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16); int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT); int qadd= (s->qbias*qmul)>>QBIAS_SHIFT; int new_index = 0; - START_TIMER - - if(b->buf == s->spatial_dwt_buffer || s->qlog == LOSSLESS_QLOG){ + if(b->ibuf == s->spatial_idwt_buffer || s->qlog == LOSSLESS_QLOG){ qadd= 0; qmul= 1<stride_line + b->buf_y_offset) + b->buf_x_offset; - memset(line, 0, b->width*sizeof(DWTELEM)); + IDWTELEM * line = slice_buffer_get_line(sb, y * b->stride_line + b->buf_y_offset) + b->buf_x_offset; + memset(line, 0, b->width*sizeof(IDWTELEM)); v = b->x_coeff[new_index].coeff; x = b->x_coeff[new_index++].x; - while(x < w) - { + while(x < w){ register int t= ( (v>>1)*qmul + qadd)>>QEXPSHIFT; register int u= -(v&1); line[x] = (t^u) - u; @@ -1837,9 +1595,6 @@ static inline void decode_subband_slice_buffered(SnowContext *s, SubBand *b, sli x = b->x_coeff[new_index++].x; } } - if(w > 200 && start_y != 0/*level+1 == s->spatial_decomposition_count*/){ - STOP_TIMER("decode_subband") - } /* Save our variables for the next slice. */ save_state[0] = new_index; @@ -1847,11 +1602,11 @@ static inline void decode_subband_slice_buffered(SnowContext *s, SubBand *b, sli return; } -static void reset_contexts(SnowContext *s){ +static void reset_contexts(SnowContext *s){ //FIXME better initial contexts int plane_index, level, orientation; for(plane_index=0; plane_index<3; plane_index++){ - for(level=0; levelspatial_decomposition_count; level++){ + for(level=0; levelplane[plane_index].band[level][orientation].state, MID_STATE, sizeof(s->plane[plane_index].band[level][orientation].state)); } @@ -1900,7 +1655,7 @@ static int pix_sum(uint8_t * pix, int line_size, int w) static int pix_norm1(uint8_t * pix, int line_size, int w) { int s, i, j; - uint32_t *sq = squareTbl + 256; + uint32_t *sq = ff_squareTbl + 256; s = 0; for (i = 0; i < w; i++) { @@ -1952,18 +1707,18 @@ static inline void init_ref(MotionEstContext *c, uint8_t *src[3], uint8_t *ref[3 } static inline void pred_mv(SnowContext *s, int *mx, int *my, int ref, - BlockNode *left, BlockNode *top, BlockNode *tr){ + const BlockNode *left, const BlockNode *top, const BlockNode *tr){ if(s->ref_frames == 1){ *mx = mid_pred(left->mx, top->mx, tr->mx); *my = mid_pred(left->my, top->my, tr->my); }else{ const int *scale = scale_mv_ref[ref]; - *mx = mid_pred(left->mx * scale[left->ref] + 128 >>8, - top ->mx * scale[top ->ref] + 128 >>8, - tr ->mx * scale[tr ->ref] + 128 >>8); - *my = mid_pred(left->my * scale[left->ref] + 128 >>8, - top ->my * scale[top ->ref] + 128 >>8, - tr ->my * scale[tr ->ref] + 128 >>8); + *mx = mid_pred((left->mx * scale[left->ref] + 128) >>8, + (top ->mx * scale[top ->ref] + 128) >>8, + (tr ->mx * scale[tr ->ref] + 128) >>8); + *my = mid_pred((left->my * scale[left->ref] + 128) >>8, + (top ->my * scale[top ->ref] + 128) >>8, + (tr ->my * scale[tr ->ref] + 128) >>8); } } @@ -1983,7 +1738,7 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ RangeCoder pc, ic; uint8_t *pbbak= s->c.bytestream; uint8_t *pbbak_start= s->c.bytestream_start; - int score, score2, iscore, i_len, p_len, block_s, sum; + int score, score2, iscore, i_len, p_len, block_s, sum, base_bits; const int w= s->b_width << s->block_max_depth; const int h= s->b_height << s->block_max_depth; const int rem_depth= s->block_max_depth - level; @@ -1991,12 +1746,12 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ const int block_w= 1<<(LOG2_MB_SIZE - level); int trx= (x+1)<block[index-1] : &null_block; - BlockNode *top = y ? &s->block[index-w] : &null_block; - BlockNode *right = trxblock[index+1] : &null_block; - BlockNode *bottom= tryblock[index+w] : &null_block; - BlockNode *tl = y && x ? &s->block[index-w-1] : left; - BlockNode *tr = y && trxblock[index-w+(1<block[index-1] : &null_block; + const BlockNode *top = y ? &s->block[index-w] : &null_block; + const BlockNode *right = trxblock[index+1] : &null_block; + const BlockNode *bottom= tryblock[index+w] : &null_block; + const BlockNode *tl = y && x ? &s->block[index-w-1] : left; + const BlockNode *tr = y && trxblock[index-w+(1<color[0]; int pcb= left->color[1]; int pcr= left->color[2]; @@ -2014,8 +1769,8 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ const int shift= 1+qpel; MotionEstContext *c= &s->m.me; int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref); - int mx_context= av_log2(2*ABS(left->mx - top->mx)); - int my_context= av_log2(2*ABS(left->my - top->my)); + int mx_context= av_log2(2*FFABS(left->mx - top->mx)); + int my_context= av_log2(2*FFABS(left->my - top->my)); int s_context= 2*left->level + 2*top->level + tl->level + tr->level; int ref, best_ref, ref_score, ref_mx, ref_my; @@ -2044,10 +1799,10 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ s->m.mb_stride=2; s->m.mb_x= s->m.mb_y= 0; - s->m.me.skip= 0; + c->skip= 0; - assert(s->m.me. stride == stride); - assert(s->m.me.uvstride == uvstride); + assert(c-> stride == stride); + assert(c->uvstride == uvstride); c->penalty_factor = get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_cmp); c->sub_penalty_factor= get_penalty_factor(s->lambda, s->lambda2, c->avctx->me_sub_cmp); @@ -2091,7 +1846,7 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ assert(ref_my >= c->ymin); assert(ref_my <= c->ymax); - ref_score= s->m.me.sub_motion_search(&s->m, &ref_mx, &ref_my, ref_score, 0, 0, level-LOG2_MB_SIZE+4, block_w); + ref_score= c->sub_motion_search(&s->m, &ref_mx, &ref_my, ref_score, 0, 0, level-LOG2_MB_SIZE+4, block_w); ref_score= ff_get_mb_score(&s->m, ref_mx, ref_my, 0, 0, level-LOG2_MB_SIZE+4, block_w, 0); ref_score+= 2*av_log2(2*ref)*c->penalty_factor; if(s->ref_mvs[ref]){ @@ -2106,9 +1861,10 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ my= ref_my; } } - //FIXME if mb_cmp != SSE then intra cant be compared currently and mb_penalty vs. lambda2 + //FIXME if mb_cmp != SSE then intra cannot be compared currently and mb_penalty vs. lambda2 // subpel search + base_bits= get_rac_count(&s->c) - 8*(s->c.bytestream - s->c.bytestream_start); pc= s->c; pc.bytestream_start= pc.bytestream= p_buffer; //FIXME end/start? and at the other stoo @@ -2123,10 +1879,7 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ put_symbol(&pc, &p_state[128 + 32*(mx_context + 16*!!best_ref)], mx - pmx, 1); put_symbol(&pc, &p_state[128 + 32*(my_context + 16*!!best_ref)], my - pmy, 1); p_len= pc.bytestream - pc.bytestream_start; - score += (s->lambda2*(p_len*8 - + (pc.outstanding_count - s->c.outstanding_count)*8 - + (-av_log2(pc.range) + av_log2(s->c.range)) - ))>>FF_LAMBDA_SHIFT; + score += (s->lambda2*(get_rac_count(&pc)-base_bits))>>FF_LAMBDA_SHIFT; block_s= block_w*block_w; sum = pix_sum(current_data[0], stride, block_w); @@ -2152,10 +1905,7 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ put_symbol(&ic, &i_state[64], cb-pcb, 1); put_symbol(&ic, &i_state[96], cr-pcr, 1); i_len= ic.bytestream - ic.bytestream_start; - iscore += (s->lambda2*(i_len*8 - + (ic.outstanding_count - s->c.outstanding_count)*8 - + (-av_log2(ic.range) + av_log2(s->c.range)) - ))>>FF_LAMBDA_SHIFT; + iscore += (s->lambda2*(get_rac_count(&ic)-base_bits))>>FF_LAMBDA_SHIFT; // assert(score==256*256*256*64-1); assert(iscore < 255*255*256 + s->lambda2*10); @@ -2204,7 +1954,7 @@ static int encode_q_branch(SnowContext *s, int level, int x, int y){ } } -static always_inline int same_block(BlockNode *a, BlockNode *b){ +static av_always_inline int same_block(BlockNode *a, BlockNode *b){ if((a->type&BLOCK_INTRA) && (b->type&BLOCK_INTRA)){ return !((a->color[0] - b->color[0]) | (a->color[1] - b->color[1]) | (a->color[2] - b->color[2])); }else{ @@ -2218,17 +1968,17 @@ static void encode_q_branch2(SnowContext *s, int level, int x, int y){ const int index= (x + y*w) << rem_depth; int trx= (x+1)<block[index]; - BlockNode *left = x ? &s->block[index-1] : &null_block; - BlockNode *top = y ? &s->block[index-w] : &null_block; - BlockNode *tl = y && x ? &s->block[index-w-1] : left; - BlockNode *tr = y && trxblock[index-w+(1<block[index-1] : &null_block; + const BlockNode *top = y ? &s->block[index-w] : &null_block; + const BlockNode *tl = y && x ? &s->block[index-w-1] : left; + const BlockNode *tr = y && trxblock[index-w+(1<color[0]; int pcb= left->color[1]; int pcr= left->color[2]; int pmx, pmy; int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref); - int mx_context= av_log2(2*ABS(left->mx - top->mx)) + 16*!!b->ref; - int my_context= av_log2(2*ABS(left->my - top->my)) + 16*!!b->ref; + int mx_context= av_log2(2*FFABS(left->mx - top->mx)) + 16*!!b->ref; + int my_context= av_log2(2*FFABS(left->my - top->my)) + 16*!!b->ref; int s_context= 2*left->level + 2*top->level + tl->level + tr->level; if(s->keyframe){ @@ -2271,10 +2021,10 @@ static void decode_q_branch(SnowContext *s, int level, int x, int y){ const int rem_depth= s->block_max_depth - level; const int index= (x + y*w) << rem_depth; int trx= (x+1)<block[index-1] : &null_block; - BlockNode *top = y ? &s->block[index-w] : &null_block; - BlockNode *tl = y && x ? &s->block[index-w-1] : left; - BlockNode *tr = y && trxblock[index-w+(1<block[index-1] : &null_block; + const BlockNode *top = y ? &s->block[index-w] : &null_block; + const BlockNode *tl = y && x ? &s->block[index-w-1] : left; + const BlockNode *tr = y && trxblock[index-w+(1<level + 2*top->level + tl->level + tr->level; if(s->keyframe){ @@ -2283,16 +2033,14 @@ static void decode_q_branch(SnowContext *s, int level, int x, int y){ } if(level==s->block_max_depth || get_rac(&s->c, &s->block_state[4 + s_context])){ - int type; + int type, mx, my; int l = left->color[0]; int cb= left->color[1]; int cr= left->color[2]; - int mx= mid_pred(left->mx, top->mx, tr->mx); - int my= mid_pred(left->my, top->my, tr->my); int ref = 0; int ref_context= av_log2(2*left->ref) + av_log2(2*top->ref); - int mx_context= av_log2(2*ABS(left->mx - top->mx)) + 0*av_log2(2*ABS(tr->mx - top->mx)); - int my_context= av_log2(2*ABS(left->my - top->my)) + 0*av_log2(2*ABS(tr->my - top->my)); + int mx_context= av_log2(2*FFABS(left->mx - top->mx)) + 0*av_log2(2*FFABS(tr->mx - top->mx)); + int my_context= av_log2(2*FFABS(left->my - top->my)) + 0*av_log2(2*FFABS(tr->my - top->my)); type= get_rac(&s->c, &s->block_state[1 + left->type + top->type]) ? BLOCK_INTRA : 0; @@ -2351,80 +2099,195 @@ static void decode_blocks(SnowContext *s){ } } -static void mc_block(uint8_t *dst, uint8_t *src, uint8_t *tmp, int stride, int b_w, int b_h, int dx, int dy){ - int x, y; -START_TIMER - for(y=0; y < b_h+5; y++){ - for(x=0; x < b_w; x++){ - int a0= src[x ]; - int a1= src[x + 1]; - int a2= src[x + 2]; - int a3= src[x + 3]; - int a4= src[x + 4]; - int a5= src[x + 5]; -// int am= 9*(a1+a2) - (a0+a3); - int am= 20*(a2+a3) - 5*(a1+a4) + (a0+a5); -// int am= 18*(a2+a3) - 2*(a1+a4); -// int aL= (-7*a0 + 105*a1 + 35*a2 - 5*a3)>>3; -// int aR= (-7*a3 + 105*a2 + 35*a1 - 5*a0)>>3; - -// if(b_w==16) am= 8*(a1+a2); - - if(dx<8) am = (32*a2*( 8-dx) + am* dx + 128)>>8; - else am = ( am*(16-dx) + 32*a3*(dx-8) + 128)>>8; - - /* FIXME Try increasing tmp buffer to 16 bits and not clipping here. Should give marginally better results. - Robert*/ - if(am&(~255)) am= ~(am>>31); - - tmp[x] = am; - -/* if (dx< 4) tmp[x + y*stride]= (16*a1*( 4-dx) + aL* dx + 32)>>6; - else if(dx< 8) tmp[x + y*stride]= ( aL*( 8-dx) + am*(dx- 4) + 32)>>6; - else if(dx<12) tmp[x + y*stride]= ( am*(12-dx) + aR*(dx- 8) + 32)>>6; - else tmp[x + y*stride]= ( aR*(16-dx) + 16*a2*(dx-12) + 32)>>6;*/ +static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, uint8_t *tmp, int stride, int b_w, int b_h, int dx, int dy){ + const static uint8_t weight[64]={ + 8,7,6,5,4,3,2,1, + 7,7,0,0,0,0,0,1, + 6,0,6,0,0,0,2,0, + 5,0,0,5,0,3,0,0, + 4,0,0,0,4,0,0,0, + 3,0,0,5,0,3,0,0, + 2,0,6,0,0,0,2,0, + 1,7,0,0,0,0,0,1, + }; + + const static uint8_t brane[256]={ + 0x00,0x01,0x01,0x01,0x01,0x01,0x01,0x01,0x11,0x12,0x12,0x12,0x12,0x12,0x12,0x12, + 0x04,0x05,0xcc,0xcc,0xcc,0xcc,0xcc,0x41,0x15,0x16,0xcc,0xcc,0xcc,0xcc,0xcc,0x52, + 0x04,0xcc,0x05,0xcc,0xcc,0xcc,0x41,0xcc,0x15,0xcc,0x16,0xcc,0xcc,0xcc,0x52,0xcc, + 0x04,0xcc,0xcc,0x05,0xcc,0x41,0xcc,0xcc,0x15,0xcc,0xcc,0x16,0xcc,0x52,0xcc,0xcc, + 0x04,0xcc,0xcc,0xcc,0x41,0xcc,0xcc,0xcc,0x15,0xcc,0xcc,0xcc,0x16,0xcc,0xcc,0xcc, + 0x04,0xcc,0xcc,0x41,0xcc,0x05,0xcc,0xcc,0x15,0xcc,0xcc,0x52,0xcc,0x16,0xcc,0xcc, + 0x04,0xcc,0x41,0xcc,0xcc,0xcc,0x05,0xcc,0x15,0xcc,0x52,0xcc,0xcc,0xcc,0x16,0xcc, + 0x04,0x41,0xcc,0xcc,0xcc,0xcc,0xcc,0x05,0x15,0x52,0xcc,0xcc,0xcc,0xcc,0xcc,0x16, + 0x44,0x45,0x45,0x45,0x45,0x45,0x45,0x45,0x55,0x56,0x56,0x56,0x56,0x56,0x56,0x56, + 0x48,0x49,0xcc,0xcc,0xcc,0xcc,0xcc,0x85,0x59,0x5A,0xcc,0xcc,0xcc,0xcc,0xcc,0x96, + 0x48,0xcc,0x49,0xcc,0xcc,0xcc,0x85,0xcc,0x59,0xcc,0x5A,0xcc,0xcc,0xcc,0x96,0xcc, + 0x48,0xcc,0xcc,0x49,0xcc,0x85,0xcc,0xcc,0x59,0xcc,0xcc,0x5A,0xcc,0x96,0xcc,0xcc, + 0x48,0xcc,0xcc,0xcc,0x49,0xcc,0xcc,0xcc,0x59,0xcc,0xcc,0xcc,0x96,0xcc,0xcc,0xcc, + 0x48,0xcc,0xcc,0x85,0xcc,0x49,0xcc,0xcc,0x59,0xcc,0xcc,0x96,0xcc,0x5A,0xcc,0xcc, + 0x48,0xcc,0x85,0xcc,0xcc,0xcc,0x49,0xcc,0x59,0xcc,0x96,0xcc,0xcc,0xcc,0x5A,0xcc, + 0x48,0x85,0xcc,0xcc,0xcc,0xcc,0xcc,0x49,0x59,0x96,0xcc,0xcc,0xcc,0xcc,0xcc,0x5A, + }; + + const static uint8_t needs[16]={ + 0,1,0,0, + 2,4,2,0, + 0,1,0,0, + 15 + }; + + int x, y, b, r, l; + int16_t tmpIt [64*(32+HTAPS_MAX)]; + uint8_t tmp2t[3][stride*(32+HTAPS_MAX)]; + int16_t *tmpI= tmpIt; + uint8_t *tmp2= tmp2t[0]; + const uint8_t *hpel[11]; + assert(dx<16 && dy<16); + r= brane[dx + 16*dy]&15; + l= brane[dx + 16*dy]>>4; + + b= needs[l] | needs[r]; + if(p && !p->diag_mc) + b= 15; + + if(b&5){ + for(y=0; y < b_h+HTAPS_MAX-1; y++){ + for(x=0; x < b_w; x++){ + int a_1=src[x + HTAPS_MAX/2-4]; + int a0= src[x + HTAPS_MAX/2-3]; + int a1= src[x + HTAPS_MAX/2-2]; + int a2= src[x + HTAPS_MAX/2-1]; + int a3= src[x + HTAPS_MAX/2+0]; + int a4= src[x + HTAPS_MAX/2+1]; + int a5= src[x + HTAPS_MAX/2+2]; + int a6= src[x + HTAPS_MAX/2+3]; + int am=0; + if(!p || p->fast_mc){ + am= 20*(a2+a3) - 5*(a1+a4) + (a0+a5); + tmpI[x]= am; + am= (am+16)>>5; + }else{ + am= p->hcoeff[0]*(a2+a3) + p->hcoeff[1]*(a1+a4) + p->hcoeff[2]*(a0+a5) + p->hcoeff[3]*(a_1+a6); + tmpI[x]= am; + am= (am+32)>>6; + } + + if(am&(~255)) am= ~(am>>31); + tmp2[x]= am; + } + tmpI+= 64; + tmp2+= stride; + src += stride; + } + src -= stride*y; + } + src += HTAPS_MAX/2 - 1; + tmp2= tmp2t[1]; + + if(b&2){ + for(y=0; y < b_h; y++){ + for(x=0; x < b_w+1; x++){ + int a_1=src[x + (HTAPS_MAX/2-4)*stride]; + int a0= src[x + (HTAPS_MAX/2-3)*stride]; + int a1= src[x + (HTAPS_MAX/2-2)*stride]; + int a2= src[x + (HTAPS_MAX/2-1)*stride]; + int a3= src[x + (HTAPS_MAX/2+0)*stride]; + int a4= src[x + (HTAPS_MAX/2+1)*stride]; + int a5= src[x + (HTAPS_MAX/2+2)*stride]; + int a6= src[x + (HTAPS_MAX/2+3)*stride]; + int am=0; + if(!p || p->fast_mc) + am= (20*(a2+a3) - 5*(a1+a4) + (a0+a5) + 16)>>5; + else + am= (p->hcoeff[0]*(a2+a3) + p->hcoeff[1]*(a1+a4) + p->hcoeff[2]*(a0+a5) + p->hcoeff[3]*(a_1+a6) + 32)>>6; + + if(am&(~255)) am= ~(am>>31); + tmp2[x]= am; + } + src += stride; + tmp2+= stride; + } + src -= stride*y; + } + src += stride*(HTAPS_MAX/2 - 1); + tmp2= tmp2t[2]; + tmpI= tmpIt; + if(b&4){ + for(y=0; y < b_h; y++){ + for(x=0; x < b_w; x++){ + int a_1=tmpI[x + (HTAPS_MAX/2-4)*64]; + int a0= tmpI[x + (HTAPS_MAX/2-3)*64]; + int a1= tmpI[x + (HTAPS_MAX/2-2)*64]; + int a2= tmpI[x + (HTAPS_MAX/2-1)*64]; + int a3= tmpI[x + (HTAPS_MAX/2+0)*64]; + int a4= tmpI[x + (HTAPS_MAX/2+1)*64]; + int a5= tmpI[x + (HTAPS_MAX/2+2)*64]; + int a6= tmpI[x + (HTAPS_MAX/2+3)*64]; + int am=0; + if(!p || p->fast_mc) + am= (20*(a2+a3) - 5*(a1+a4) + (a0+a5) + 512)>>10; + else + am= (p->hcoeff[0]*(a2+a3) + p->hcoeff[1]*(a1+a4) + p->hcoeff[2]*(a0+a5) + p->hcoeff[3]*(a_1+a6) + 2048)>>12; + if(am&(~255)) am= ~(am>>31); + tmp2[x]= am; + } + tmpI+= 64; + tmp2+= stride; + } + } + + hpel[ 0]= src; + hpel[ 1]= tmp2t[0] + stride*(HTAPS_MAX/2-1); + hpel[ 2]= src + 1; + + hpel[ 4]= tmp2t[1]; + hpel[ 5]= tmp2t[2]; + hpel[ 6]= tmp2t[1] + 1; + + hpel[ 8]= src + stride; + hpel[ 9]= hpel[1] + stride; + hpel[10]= hpel[8] + 1; + + if(b==15){ + const uint8_t *src1= hpel[dx/8 + dy/8*4 ]; + const uint8_t *src2= hpel[dx/8 + dy/8*4+1]; + const uint8_t *src3= hpel[dx/8 + dy/8*4+4]; + const uint8_t *src4= hpel[dx/8 + dy/8*4+5]; + dx&=7; + dy&=7; + for(y=0; y < b_h; y++){ + for(x=0; x < b_w; x++){ + dst[x]= ((8-dx)*(8-dy)*src1[x] + dx*(8-dy)*src2[x]+ + (8-dx)* dy *src3[x] + dx* dy *src4[x]+32)>>6; + } + src1+=stride; + src2+=stride; + src3+=stride; + src4+=stride; + dst +=stride; } - tmp += stride; - src += stride; - } - tmp -= (b_h+5)*stride; - - for(y=0; y < b_h; y++){ - for(x=0; x < b_w; x++){ - int a0= tmp[x + 0*stride]; - int a1= tmp[x + 1*stride]; - int a2= tmp[x + 2*stride]; - int a3= tmp[x + 3*stride]; - int a4= tmp[x + 4*stride]; - int a5= tmp[x + 5*stride]; - int am= 20*(a2+a3) - 5*(a1+a4) + (a0+a5); -// int am= 18*(a2+a3) - 2*(a1+a4); -/* int aL= (-7*a0 + 105*a1 + 35*a2 - 5*a3)>>3; - int aR= (-7*a3 + 105*a2 + 35*a1 - 5*a0)>>3;*/ - -// if(b_w==16) am= 8*(a1+a2); - - if(dy<8) am = (32*a2*( 8-dy) + am* dy + 128)>>8; - else am = ( am*(16-dy) + 32*a3*(dy-8) + 128)>>8; - - if(am&(~255)) am= ~(am>>31); - - dst[x] = am; -/* if (dy< 4) tmp[x + y*stride]= (16*a1*( 4-dy) + aL* dy + 32)>>6; - else if(dy< 8) tmp[x + y*stride]= ( aL*( 8-dy) + am*(dy- 4) + 32)>>6; - else if(dy<12) tmp[x + y*stride]= ( am*(12-dy) + aR*(dy- 8) + 32)>>6; - else tmp[x + y*stride]= ( aR*(16-dy) + 16*a2*(dy-12) + 32)>>6;*/ + }else{ + const uint8_t *src1= hpel[l]; + const uint8_t *src2= hpel[r]; + int a= weight[((dx&7) + (8*(dy&7)))]; + int b= 8-a; + for(y=0; y < b_h; y++){ + for(x=0; x < b_w; x++){ + dst[x]= (a*src1[x] + b*src2[x] + 4)>>3; + } + src1+=stride; + src2+=stride; + dst +=stride; } - dst += stride; - tmp += stride; } -STOP_TIMER("mc_block") } #define mca(dx,dy,b_w)\ -static void mc_block_hpel ## dx ## dy ## b_w(uint8_t *dst, uint8_t *src, int stride, int h){\ - uint8_t tmp[stride*(b_w+5)];\ +static void mc_block_hpel ## dx ## dy ## b_w(uint8_t *dst, const uint8_t *src, int stride, int h){\ + uint8_t tmp[stride*(b_w+HTAPS_MAX-1)];\ assert(h==b_w);\ - mc_block(dst, src-2-2*stride, tmp, stride, b_w, b_w, dx, dy);\ + mc_block(NULL, dst, src-(HTAPS_MAX/2-1)-(HTAPS_MAX/2-1)*stride, tmp, stride, b_w, b_w, dx, dy);\ } mca( 0, 0,16) @@ -2483,49 +2346,49 @@ static void pred_block(SnowContext *s, uint8_t *dst, uint8_t *tmp, int stride, i const int dx= mx&15; const int dy= my&15; const int tab_index= 3 - (b_w>>2) + (b_w>>4); - sx += (mx>>4) - 2; - sy += (my>>4) - 2; + sx += (mx>>4) - (HTAPS_MAX/2-1); + sy += (my>>4) - (HTAPS_MAX/2-1); src += sx + sy*stride; - if( (unsigned)sx >= w - b_w - 4 - || (unsigned)sy >= h - b_h - 4){ - ff_emulated_edge_mc(tmp + MB_SIZE, src, stride, b_w+5, b_h+5, sx, sy, w, h); + if( (unsigned)sx >= w - b_w - (HTAPS_MAX-2) + || (unsigned)sy >= h - b_h - (HTAPS_MAX-2)){ + ff_emulated_edge_mc(tmp + MB_SIZE, src, stride, b_w+HTAPS_MAX-1, b_h+HTAPS_MAX-1, sx, sy, w, h); src= tmp + MB_SIZE; } // assert(b_w == b_h || 2*b_w == b_h || b_w == 2*b_h); // assert(!(b_w&(b_w-1))); assert(b_w>1 && b_h>1); - assert(tab_index>=0 && tab_index<4 || b_w==32); - if((dx&3) || (dy&3) || !(b_w == b_h || 2*b_w == b_h || b_w == 2*b_h) || (b_w&(b_w-1))) - mc_block(dst, src, tmp, stride, b_w, b_h, dx, dy); + assert((tab_index>=0 && tab_index<4) || b_w==32); + if((dx&3) || (dy&3) || !(b_w == b_h || 2*b_w == b_h || b_w == 2*b_h) || (b_w&(b_w-1)) || !s->plane[plane_index].fast_mc ) + mc_block(&s->plane[plane_index], dst, src, tmp, stride, b_w, b_h, dx, dy); else if(b_w==32){ int y; for(y=0; ydsp.put_h264_qpel_pixels_tab[0][dy+(dx>>2)](dst + y*stride, src + 2 + (y+2)*stride,stride); - s->dsp.put_h264_qpel_pixels_tab[0][dy+(dx>>2)](dst + 16 + y*stride, src + 18 + (y+2)*stride,stride); + s->dsp.put_h264_qpel_pixels_tab[0][dy+(dx>>2)](dst + y*stride, src + 3 + (y+3)*stride,stride); + s->dsp.put_h264_qpel_pixels_tab[0][dy+(dx>>2)](dst + 16 + y*stride, src + 19 + (y+3)*stride,stride); } }else if(b_w==b_h) - s->dsp.put_h264_qpel_pixels_tab[tab_index ][dy+(dx>>2)](dst,src + 2 + 2*stride,stride); + s->dsp.put_h264_qpel_pixels_tab[tab_index ][dy+(dx>>2)](dst,src + 3 + 3*stride,stride); else if(b_w==2*b_h){ - s->dsp.put_h264_qpel_pixels_tab[tab_index+1][dy+(dx>>2)](dst ,src + 2 + 2*stride,stride); - s->dsp.put_h264_qpel_pixels_tab[tab_index+1][dy+(dx>>2)](dst+b_h,src + 2 + b_h + 2*stride,stride); + s->dsp.put_h264_qpel_pixels_tab[tab_index+1][dy+(dx>>2)](dst ,src + 3 + 3*stride,stride); + s->dsp.put_h264_qpel_pixels_tab[tab_index+1][dy+(dx>>2)](dst+b_h,src + 3 + b_h + 3*stride,stride); }else{ assert(2*b_w==b_h); - s->dsp.put_h264_qpel_pixels_tab[tab_index ][dy+(dx>>2)](dst ,src + 2 + 2*stride ,stride); - s->dsp.put_h264_qpel_pixels_tab[tab_index ][dy+(dx>>2)](dst+b_w*stride,src + 2 + 2*stride+b_w*stride,stride); + s->dsp.put_h264_qpel_pixels_tab[tab_index ][dy+(dx>>2)](dst ,src + 3 + 3*stride ,stride); + s->dsp.put_h264_qpel_pixels_tab[tab_index ][dy+(dx>>2)](dst+b_w*stride,src + 3 + 3*stride+b_w*stride,stride); } } } -void ff_snow_inner_add_yblock(uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, +void ff_snow_inner_add_yblock(const uint8_t *obmc, const int obmc_stride, uint8_t * * block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer * sb, int add, uint8_t * dst8){ int y, x; - DWTELEM * dst; + IDWTELEM * dst; for(y=0; y>1); - uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1); - uint8_t *obmc4= obmc3+ (obmc_stride>>1); + //FIXME ugly misuse of obmc_stride + const uint8_t *obmc1= obmc + y*obmc_stride; + const uint8_t *obmc2= obmc1+ (obmc_stride>>1); + const uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1); + const uint8_t *obmc4= obmc3+ (obmc_stride>>1); dst = slice_buffer_get_line(sb, src_y + y); for(x=0; x>= 8 - FRAC_BITS; } if(add){ @@ -2550,8 +2412,8 @@ void ff_snow_inner_add_yblock(uint8_t *obmc, const int obmc_stride, uint8_t * * } } -//FIXME name clenup (b_w, block_w, b_width stuff) -static always_inline void add_yblock(SnowContext *s, int sliced, slice_buffer *sb, DWTELEM *dst, uint8_t *dst8, const uint8_t *obmc, int src_x, int src_y, int b_w, int b_h, int w, int h, int dst_stride, int src_stride, int obmc_stride, int b_x, int b_y, int add, int offset_dst, int plane_index){ +//FIXME name cleanup (b_w, block_w, b_width stuff) +static av_always_inline void add_yblock(SnowContext *s, int sliced, slice_buffer *sb, IDWTELEM *dst, uint8_t *dst8, const uint8_t *obmc, int src_x, int src_y, int b_w, int b_h, int w, int h, int dst_stride, int src_stride, int obmc_stride, int b_x, int b_y, int add, int offset_dst, int plane_index){ const int b_width = s->b_width << s->block_max_depth; const int b_height= s->b_height << s->block_max_depth; const int b_stride= b_width; @@ -2580,7 +2442,7 @@ static always_inline void add_yblock(SnowContext *s, int sliced, slice_buffer *s rb= rt; } - if(src_x<0){ //FIXME merge with prev & always round internal width upto *16 + if(src_x<0){ //FIXME merge with prev & always round internal width up to *16 obmc -= src_x; b_w += src_x; if(!sliced && !offset_dst) @@ -2601,7 +2463,8 @@ static always_inline void add_yblock(SnowContext *s, int sliced, slice_buffer *s if(b_w<=0 || b_h<=0) return; -assert(src_stride > 2*MB_SIZE + 5); + assert(src_stride > 2*MB_SIZE + 5); + if(!sliced && offset_dst) dst += src_x + src_y*dst_stride; dst8+= src_x + src_y*src_stride; @@ -2675,42 +2538,39 @@ assert(src_stride > 2*MB_SIZE + 5); } #else if(sliced){ - START_TIMER - s->dsp.inner_add_yblock(obmc, obmc_stride, block, b_w, b_h, src_x,src_y, src_stride, sb, add, dst8); - STOP_TIMER("inner_add_yblock") - }else - for(y=0; y>1); - uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1); - uint8_t *obmc4= obmc3+ (obmc_stride>>1); - for(x=0; x>= 8 - FRAC_BITS; - } - if(add){ - v += dst[x + y*dst_stride]; - v = (v + (1<<(FRAC_BITS-1))) >> FRAC_BITS; - if(v&(~255)) v= ~(v>>31); - dst8[x + y*src_stride] = v; - }else{ - dst[x + y*dst_stride] -= v; + }else{ + for(y=0; y>1); + const uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1); + const uint8_t *obmc4= obmc3+ (obmc_stride>>1); + for(x=0; x>= 8 - FRAC_BITS; + } + if(add){ + v += dst[x + y*dst_stride]; + v = (v + (1<<(FRAC_BITS-1))) >> FRAC_BITS; + if(v&(~255)) v= ~(v>>31); + dst8[x + y*src_stride] = v; + }else{ + dst[x + y*dst_stride] -= v; + } } } } -#endif +#endif /* 0 */ } -static always_inline void predict_slice_buffered(SnowContext *s, slice_buffer * sb, DWTELEM * old_buffer, int plane_index, int add, int mb_y){ +static av_always_inline void predict_slice_buffered(SnowContext *s, slice_buffer * sb, IDWTELEM * old_buffer, int plane_index, int add, int mb_y){ Plane *p= &s->plane[plane_index]; const int mb_w= s->b_width << s->block_max_depth; const int mb_h= s->b_height << s->block_max_depth; @@ -2723,19 +2583,16 @@ static always_inline void predict_slice_buffered(SnowContext *s, slice_buffer * uint8_t *dst8= s->current_picture.data[plane_index]; int w= p->width; int h= p->height; - START_TIMER if(s->keyframe || (s->avctx->debug&512)){ if(mb_y==mb_h) return; if(add){ - for(y=block_w*mb_y; yline[y]; - for(x=0; xline[y]; + for(x=0; x>= FRAC_BITS; @@ -2744,12 +2601,10 @@ static always_inline void predict_slice_buffered(SnowContext *s, slice_buffer * } } }else{ - for(y=block_w*mb_y; yline[y]; - for(x=0; xline[y]; + for(x=0; xplane[plane_index]; const int mb_w= s->b_width << s->block_max_depth; const int mb_h= s->b_height << s->block_max_depth; @@ -2790,7 +2639,6 @@ static always_inline void predict_slice(SnowContext *s, DWTELEM *buf, int plane_ uint8_t *dst8= s->current_picture.data[plane_index]; int w= p->width; int h= p->height; - START_TIMER if(s->keyframe || (s->avctx->debug&512)){ if(mb_y==mb_h) @@ -2816,25 +2664,19 @@ static always_inline void predict_slice(SnowContext *s, DWTELEM *buf, int plane_ return; } - for(mb_x=0; mb_x<=mb_w; mb_x++){ - START_TIMER - - add_yblock(s, 0, NULL, buf, dst8, obmc, - block_w*mb_x - block_w/2, - block_w*mb_y - block_w/2, - block_w, block_w, - w, h, - w, ref_stride, obmc_stride, - mb_x - 1, mb_y - 1, - add, 1, plane_index); - - STOP_TIMER("add_yblock") - } - - STOP_TIMER("predict_slice") + for(mb_x=0; mb_x<=mb_w; mb_x++){ + add_yblock(s, 0, NULL, buf, dst8, obmc, + block_w*mb_x - block_w/2, + block_w*mb_y - block_w/2, + block_w, block_w, + w, h, + w, ref_stride, obmc_stride, + mb_x - 1, mb_y - 1, + add, 1, plane_index); + } } -static always_inline void predict_plane(SnowContext *s, DWTELEM *buf, int plane_index, int add){ +static av_always_inline void predict_plane(SnowContext *s, IDWTELEM *buf, int plane_index, int add){ const int mb_h= s->b_height << s->block_max_depth; int mb_y; for(mb_y=0; mb_y<=mb_h; mb_y++) @@ -2850,7 +2692,7 @@ static int get_dc(SnowContext *s, int mb_x, int mb_y, int plane_index){ const int obmc_stride= plane_index ? block_size : 2*block_size; const int ref_stride= s->current_picture.linesize[plane_index]; uint8_t *src= s-> input_picture.data[plane_index]; - DWTELEM *dst= (DWTELEM*)s->m.obmc_scratchpad + plane_index*block_size*block_size*4; + IDWTELEM *dst= (IDWTELEM*)s->m.obmc_scratchpad + plane_index*block_size*block_size*4; //FIXME change to unsigned const int b_stride = s->b_width << s->block_max_depth; const int w= p->width; const int h= p->height; @@ -2862,7 +2704,7 @@ static int get_dc(SnowContext *s, int mb_x, int mb_y, int plane_index){ b->type|= BLOCK_INTRA; b->color[plane_index]= 0; - memset(dst, 0, obmc_stride*obmc_stride*sizeof(DWTELEM)); + memset(dst, 0, obmc_stride*obmc_stride*sizeof(IDWTELEM)); for(i=0; i<4; i++){ int mb_x2= mb_x + (i &1) - 1; @@ -2882,32 +2724,32 @@ static int get_dc(SnowContext *s, int mb_x, int mb_y, int plane_index){ if(x<0) obmc_v += obmc[index + block_w]; if(y+block_w>h) obmc_v += obmc[index - block_w*obmc_stride]; if(x+block_w>w) obmc_v += obmc[index - block_w]; - //FIXME precalc this or simplify it somehow else + //FIXME precalculate this or simplify it somehow else d = -dst[index] + (1<<(FRAC_BITS-1)); dst[index] = d; ab += (src[x2 + y2*ref_stride] - (d>>FRAC_BITS)) * obmc_v; - aa += obmc_v * obmc_v; //FIXME precalclate this + aa += obmc_v * obmc_v; //FIXME precalculate this } } } *b= backup; - return clip(((ab<b_width << s->block_max_depth; const int b_height = s->b_height<< s->block_max_depth; int index= x + y*b_stride; - BlockNode *b = &s->block[index]; - BlockNode *left = x ? &s->block[index-1] : &null_block; - BlockNode *top = y ? &s->block[index-b_stride] : &null_block; - BlockNode *tl = y && x ? &s->block[index-b_stride-1] : left; - BlockNode *tr = y && x+wblock[index-b_stride+w] : tl; + const BlockNode *b = &s->block[index]; + const BlockNode *left = x ? &s->block[index-1] : &null_block; + const BlockNode *top = y ? &s->block[index-b_stride] : &null_block; + const BlockNode *tl = y && x ? &s->block[index-b_stride-1] : left; + const BlockNode *tr = y && x+wblock[index-b_stride+w] : tl; int dmx, dmy; -// int mx_context= av_log2(2*ABS(left->mx - top->mx)); -// int my_context= av_log2(2*ABS(left->my - top->my)); +// int mx_context= av_log2(2*FFABS(left->mx - top->mx)); +// int my_context= av_log2(2*FFABS(left->my - top->my)); if(x<0 || x>=b_stride || y>=b_height) return 0; @@ -2919,17 +2761,17 @@ static inline int get_block_bits(SnowContext *s, int x, int y, int w){ 00001XXXX 15-30 8-15 */ //FIXME try accurate rate -//FIXME intra and inter predictors if surrounding blocks arent the same type +//FIXME intra and inter predictors if surrounding blocks are not the same type if(b->type & BLOCK_INTRA){ - return 3+2*( av_log2(2*ABS(left->color[0] - b->color[0])) - + av_log2(2*ABS(left->color[1] - b->color[1])) - + av_log2(2*ABS(left->color[2] - b->color[2]))); + return 3+2*( av_log2(2*FFABS(left->color[0] - b->color[0])) + + av_log2(2*FFABS(left->color[1] - b->color[1])) + + av_log2(2*FFABS(left->color[2] - b->color[2]))); }else{ pred_mv(s, &dmx, &dmy, b->ref, left, top, tr); dmx-= b->mx; dmy-= b->my; - return 2*(1 + av_log2(2*ABS(dmx)) //FIXME kill the 2* can be merged in lambda - + av_log2(2*ABS(dmy)) + return 2*(1 + av_log2(2*FFABS(dmx)) //FIXME kill the 2* can be merged in lambda + + av_log2(2*FFABS(dmy)) + av_log2(2*b->ref)); } } @@ -2942,9 +2784,9 @@ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, con const int ref_stride= s->current_picture.linesize[plane_index]; uint8_t *dst= s->current_picture.data[plane_index]; uint8_t *src= s-> input_picture.data[plane_index]; - DWTELEM *pred= (DWTELEM*)s->m.obmc_scratchpad + plane_index*block_size*block_size*4; + IDWTELEM *pred= (IDWTELEM*)s->m.obmc_scratchpad + plane_index*block_size*block_size*4; uint8_t cur[ref_stride*2*MB_SIZE]; //FIXME alignment - uint8_t tmp[ref_stride*(2*MB_SIZE+5)]; + uint8_t tmp[ref_stride*(2*MB_SIZE+HTAPS_MAX-1)]; const int b_stride = s->b_width << s->block_max_depth; const int b_height = s->b_height<< s->block_max_depth; const int w= p->width; @@ -2964,11 +2806,15 @@ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, con for(y=y0; y= LOG2_OBMC_MAX int v = (cur1[x] * obmc1[x]) << (FRAC_BITS - LOG2_OBMC_MAX); +#else + int v = (cur1[x] * obmc1[x] + (1<<(LOG2_OBMC_MAX - FRAC_BITS-1))) >> (LOG2_OBMC_MAX - FRAC_BITS); +#endif v = (v + pred1[x]) >> FRAC_BITS; if(v&(~255)) v= ~(v>>31); dst1[x] = v; @@ -2994,10 +2840,10 @@ static int get_block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index, con if(block_w==16){ /* FIXME rearrange dsputil to fit 32x32 cmp functions */ /* FIXME check alignment of the cmp wavelet vs the encoding wavelet */ - /* FIXME cmps overlap but don't cover the wavelet's whole support, - * so improving the score of one block is not strictly guaranteed to - * improve the score of the whole frame, so iterative motion est - * doesn't always converge. */ + /* FIXME cmps overlap but do not cover the wavelet's whole support. + * So improving the score of one block is not strictly guaranteed + * to improve the score of the whole frame, thus iterative motion + * estimation does not always converge. */ if(s->avctx->me_cmp == FF_CMP_W97) distortion = w97_32_c(&s->m, src + sx + sy*ref_stride, dst + sx + sy*ref_stride, ref_stride, 32); else if(s->avctx->me_cmp == FF_CMP_W53) @@ -3038,7 +2884,9 @@ static int get_4block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index){ const int ref_stride= s->current_picture.linesize[plane_index]; uint8_t *dst= s->current_picture.data[plane_index]; uint8_t *src= s-> input_picture.data[plane_index]; - static const DWTELEM zero_dst[4096]; //FIXME + //FIXME zero_dst is const but add_yblock changes dst if add is 0 (this is never the case for dst=zero_dst + // const has only been removed from zero_dst to suppress a warning + static IDWTELEM zero_dst[4096]; //FIXME const int b_stride = s->b_width << s->block_max_depth; const int w= p->width; const int h= p->height; @@ -3092,7 +2940,7 @@ static int get_4block_rd(SnowContext *s, int mb_x, int mb_y, int plane_index){ return distortion + rate*penalty_factor; } -static always_inline int check_block(SnowContext *s, int mb_x, int mb_y, int p[3], int intra, const uint8_t *obmc_edged, int *best_rd){ +static av_always_inline int check_block(SnowContext *s, int mb_x, int mb_y, int p[3], int intra, const uint8_t *obmc_edged, int *best_rd){ const int b_stride= s->b_width << s->block_max_depth; BlockNode *block= &s->block[mb_x + mb_y * b_stride]; BlockNode backup= *block; @@ -3130,13 +2978,14 @@ static always_inline int check_block(SnowContext *s, int mb_x, int mb_y, int p[3 } } -/* special case for int[2] args we discard afterward, fixes compilation prob with gcc 2.95 */ -static always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, const uint8_t *obmc_edged, int *best_rd){ +/* special case for int[2] args we discard afterwards, + * fixes compilation problem with gcc 2.95 */ +static av_always_inline int check_block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, const uint8_t *obmc_edged, int *best_rd){ int p[2] = {p0, p1}; return check_block(s, mb_x, mb_y, p, 0, obmc_edged, best_rd); } -static always_inline int check_4block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, int ref, int *best_rd){ +static av_always_inline int check_4block_inter(SnowContext *s, int mb_x, int mb_y, int p0, int p1, int ref, int *best_rd){ const int b_stride= s->b_width << s->block_max_depth; BlockNode *block= &s->block[mb_x + mb_y * b_stride]; BlockNode backup[4]= {block[0], block[1], block[b_stride], block[b_stride+1]}; @@ -3222,7 +3071,7 @@ static void iterative_me(SnowContext *s){ memset(s->me_cache, 0, sizeof(s->me_cache)); s->me_cache_generation += 1<<22; - //FIXME precalc + //FIXME precalculate { int x, y; memcpy(obmc_edged, obmc_tab[s->block_max_depth], b_w*b_w*4); @@ -3247,8 +3096,7 @@ static void iterative_me(SnowContext *s){ } //skip stuff outside the picture - if(mb_x==0 || mb_y==0 || mb_x==b_width-1 || mb_y==b_height-1) - { + if(mb_x==0 || mb_y==0 || mb_x==b_width-1 || mb_y==b_height-1){ uint8_t *src= s-> input_picture.data[0]; uint8_t *dst= s->current_picture.data[0]; const int stride= s->current_picture.linesize[0]; @@ -3277,7 +3125,7 @@ static void iterative_me(SnowContext *s){ for(i=0; i<3; i++) color[i]= get_dc(s, mb_x, mb_y, i); - // get previous score (cant be cached due to OBMC) + // get previous score (cannot be cached due to OBMC) if(pass > 0 && (block->type&BLOCK_INTRA)){ int color0[3]= {block->color[0], block->color[1], block->color[2]}; check_block(s, mb_x, mb_y, color0, 1, *obmc_edged, &best_rd); @@ -3305,7 +3153,7 @@ static void iterative_me(SnowContext *s){ check_block_inter(s, mb_x, mb_y, mvr[b_stride][0], mvr[b_stride][1], *obmc_edged, &best_rd); /* fullpel ME */ - //FIXME avoid subpel interpol / round to nearest integer + //FIXME avoid subpel interpolation / round to nearest integer do{ dia_change=0; for(i=0; iavctx->dia_size, 1); i++){ @@ -3398,16 +3246,19 @@ static void iterative_me(SnowContext *s){ } } -static void quantize(SnowContext *s, SubBand *b, DWTELEM *src, int stride, int bias){ - const int level= b->level; +static void quantize(SnowContext *s, SubBand *b, IDWTELEM *dst, DWTELEM *src, int stride, int bias){ const int w= b->width; const int h= b->height; - const int qlog= clip(s->qlog + b->qlog, 0, QROOT*16); - const int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT); + const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16); + const int qmul= qexp[qlog&(QROOT-1)]<<((qlog>>QSHIFT) + ENCODER_EXTRA_BITS); int x,y, thres1, thres2; -// START_TIMER - if(s->qlog == LOSSLESS_QLOG) return; + if(s->qlog == LOSSLESS_QLOG){ + for(y=0; y>3; thres1= ((qmul - bias)>>QEXPSHIFT) - 1; @@ -3422,15 +3273,15 @@ static void quantize(SnowContext *s, SubBand *b, DWTELEM *src, int stride, int b if(i>=0){ i<<= QEXPSHIFT; i/= qmul; //FIXME optimize - src[x + y*stride]= i; + dst[x + y*stride]= i; }else{ i= -i; i<<= QEXPSHIFT; i/= qmul; //FIXME optimize - src[x + y*stride]= -i; + dst[x + y*stride]= -i; } }else - src[x + y*stride]= 0; + dst[x + y*stride]= 0; } } }else{ @@ -3442,36 +3293,32 @@ static void quantize(SnowContext *s, SubBand *b, DWTELEM *src, int stride, int b if(i>=0){ i<<= QEXPSHIFT; i= (i + bias) / qmul; //FIXME optimize - src[x + y*stride]= i; + dst[x + y*stride]= i; }else{ i= -i; i<<= QEXPSHIFT; i= (i + bias) / qmul; //FIXME optimize - src[x + y*stride]= -i; + dst[x + y*stride]= -i; } }else - src[x + y*stride]= 0; + dst[x + y*stride]= 0; } } } - if(level+1 == s->spatial_decomposition_count){ -// STOP_TIMER("quantize") - } } -static void dequantize_slice_buffered(SnowContext *s, slice_buffer * sb, SubBand *b, DWTELEM *src, int stride, int start_y, int end_y){ +static void dequantize_slice_buffered(SnowContext *s, slice_buffer * sb, SubBand *b, IDWTELEM *src, int stride, int start_y, int end_y){ const int w= b->width; - const int qlog= clip(s->qlog + b->qlog, 0, QROOT*16); + const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16); const int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT); const int qadd= (s->qbias*qmul)>>QBIAS_SHIFT; int x,y; - START_TIMER if(s->qlog == LOSSLESS_QLOG) return; for(y=start_y; ystride_line) + b->buf_y_offset) + b->buf_x_offset; + IDWTELEM * line = slice_buffer_get_line(sb, (y * b->stride_line) + b->buf_y_offset) + b->buf_x_offset; for(x=0; x 200 /*level+1 == s->spatial_decomposition_count*/){ - STOP_TIMER("dquant") - } } -static void dequantize(SnowContext *s, SubBand *b, DWTELEM *src, int stride){ +static void dequantize(SnowContext *s, SubBand *b, IDWTELEM *src, int stride){ const int w= b->width; const int h= b->height; - const int qlog= clip(s->qlog + b->qlog, 0, QROOT*16); + const int qlog= av_clip(s->qlog + b->qlog, 0, QROOT*16); const int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT); const int qadd= (s->qbias*qmul)>>QBIAS_SHIFT; int x,y; - START_TIMER if(s->qlog == LOSSLESS_QLOG) return; @@ -3507,12 +3350,9 @@ static void dequantize(SnowContext *s, SubBand *b, DWTELEM *src, int stride){ } } } - if(w > 200 /*level+1 == s->spatial_decomposition_count*/){ - STOP_TIMER("dquant") - } } -static void decorrelate(SnowContext *s, SubBand *b, DWTELEM *src, int stride, int inverse, int use_median){ +static void decorrelate(SnowContext *s, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median){ const int w= b->width; const int h= b->height; int x,y; @@ -3536,14 +3376,12 @@ static void decorrelate(SnowContext *s, SubBand *b, DWTELEM *src, int stride, in } } -static void correlate_slice_buffered(SnowContext *s, slice_buffer * sb, SubBand *b, DWTELEM *src, int stride, int inverse, int use_median, int start_y, int end_y){ +static void correlate_slice_buffered(SnowContext *s, slice_buffer * sb, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median, int start_y, int end_y){ const int w= b->width; int x,y; -// START_TIMER - - DWTELEM * line; - DWTELEM * prev; + IDWTELEM * line=0; // silence silly "could be used without having been initialized" warning + IDWTELEM * prev; if (start_y != 0) line = slice_buffer_get_line(sb, ((start_y - 1) * b->stride_line) + b->buf_y_offset) + b->buf_x_offset; @@ -3566,11 +3404,9 @@ static void correlate_slice_buffered(SnowContext *s, slice_buffer * sb, SubBand } } } - -// STOP_TIMER("correlate") } -static void correlate(SnowContext *s, SubBand *b, DWTELEM *src, int stride, int inverse, int use_median){ +static void correlate(SnowContext *s, SubBand *b, IDWTELEM *src, int stride, int inverse, int use_median){ const int w= b->width; const int h= b->height; int x,y; @@ -3594,15 +3430,40 @@ static void correlate(SnowContext *s, SubBand *b, DWTELEM *src, int stride, int } } -static void encode_header(SnowContext *s){ +static void encode_qlogs(SnowContext *s){ int plane_index, level, orientation; + + for(plane_index=0; plane_index<2; plane_index++){ + for(level=0; levelspatial_decomposition_count; level++){ + for(orientation=level ? 1:0; orientation<4; orientation++){ + if(orientation==2) continue; + put_symbol(&s->c, s->header_state, s->plane[plane_index].band[level][orientation].qlog, 1); + } + } + } +} + +static void encode_header(SnowContext *s){ + int plane_index, i; uint8_t kstate[32]; memset(kstate, MID_STATE, sizeof(kstate)); put_rac(&s->c, kstate, s->keyframe); - if(s->keyframe || s->always_reset) + if(s->keyframe || s->always_reset){ reset_contexts(s); + s->last_spatial_decomposition_type= + s->last_qlog= + s->last_qbias= + s->last_mv_scale= + s->last_block_max_depth= 0; + for(plane_index=0; plane_index<2; plane_index++){ + Plane *p= &s->plane[plane_index]; + p->last_htaps=0; + p->last_diag_mc=0; + memset(p->last_hcoeff, 0, sizeof(p->last_hcoeff)); + } + } if(s->keyframe){ put_symbol(&s->c, s->header_state, s->version, 0); put_rac(&s->c, s->header_state, s->always_reset); @@ -3616,31 +3477,94 @@ static void encode_header(SnowContext *s){ // put_rac(&s->c, s->header_state, s->rate_scalability); put_symbol(&s->c, s->header_state, s->max_ref_frames-1, 0); + encode_qlogs(s); + } + + if(!s->keyframe){ + int update_mc=0; for(plane_index=0; plane_index<2; plane_index++){ - for(level=0; levelspatial_decomposition_count; level++){ - for(orientation=level ? 1:0; orientation<4; orientation++){ - if(orientation==2) continue; - put_symbol(&s->c, s->header_state, s->plane[plane_index].band[level][orientation].qlog, 1); - } + Plane *p= &s->plane[plane_index]; + update_mc |= p->last_htaps != p->htaps; + update_mc |= p->last_diag_mc != p->diag_mc; + update_mc |= !!memcmp(p->last_hcoeff, p->hcoeff, sizeof(p->hcoeff)); + } + put_rac(&s->c, s->header_state, update_mc); + if(update_mc){ + for(plane_index=0; plane_index<2; plane_index++){ + Plane *p= &s->plane[plane_index]; + put_rac(&s->c, s->header_state, p->diag_mc); + put_symbol(&s->c, s->header_state, p->htaps/2-1, 0); + for(i= p->htaps/2; i; i--) + put_symbol(&s->c, s->header_state, FFABS(p->hcoeff[i]), 0); } } + if(s->last_spatial_decomposition_count != s->spatial_decomposition_count){ + put_rac(&s->c, s->header_state, 1); + put_symbol(&s->c, s->header_state, s->spatial_decomposition_count, 0); + encode_qlogs(s); + }else + put_rac(&s->c, s->header_state, 0); } - put_symbol(&s->c, s->header_state, s->spatial_decomposition_type, 0); - put_symbol(&s->c, s->header_state, s->qlog, 1); - put_symbol(&s->c, s->header_state, s->mv_scale, 0); - put_symbol(&s->c, s->header_state, s->qbias, 1); - put_symbol(&s->c, s->header_state, s->block_max_depth, 0); + + put_symbol(&s->c, s->header_state, s->spatial_decomposition_type - s->last_spatial_decomposition_type, 1); + put_symbol(&s->c, s->header_state, s->qlog - s->last_qlog , 1); + put_symbol(&s->c, s->header_state, s->mv_scale - s->last_mv_scale, 1); + put_symbol(&s->c, s->header_state, s->qbias - s->last_qbias , 1); + put_symbol(&s->c, s->header_state, s->block_max_depth - s->last_block_max_depth, 1); + } -static int decode_header(SnowContext *s){ +static void update_last_header_values(SnowContext *s){ + int plane_index; + + if(!s->keyframe){ + for(plane_index=0; plane_index<2; plane_index++){ + Plane *p= &s->plane[plane_index]; + p->last_diag_mc= p->diag_mc; + p->last_htaps = p->htaps; + memcpy(p->last_hcoeff, p->hcoeff, sizeof(p->hcoeff)); + } + } + + s->last_spatial_decomposition_type = s->spatial_decomposition_type; + s->last_qlog = s->qlog; + s->last_qbias = s->qbias; + s->last_mv_scale = s->mv_scale; + s->last_block_max_depth = s->block_max_depth; + s->last_spatial_decomposition_count = s->spatial_decomposition_count; +} + +static void decode_qlogs(SnowContext *s){ int plane_index, level, orientation; + + for(plane_index=0; plane_index<3; plane_index++){ + for(level=0; levelspatial_decomposition_count; level++){ + for(orientation=level ? 1:0; orientation<4; orientation++){ + int q; + if (plane_index==2) q= s->plane[1].band[level][orientation].qlog; + else if(orientation==2) q= s->plane[plane_index].band[level][1].qlog; + else q= get_symbol(&s->c, s->header_state, 1); + s->plane[plane_index].band[level][orientation].qlog= q; + } + } + } +} + +static int decode_header(SnowContext *s){ + int plane_index; uint8_t kstate[32]; memset(kstate, MID_STATE, sizeof(kstate)); s->keyframe= get_rac(&s->c, kstate); - if(s->keyframe || s->always_reset) + if(s->keyframe || s->always_reset){ reset_contexts(s); + s->spatial_decomposition_type= + s->qlog= + s->qbias= + s->mv_scale= + s->block_max_depth= 0; + } if(s->keyframe){ s->version= get_symbol(&s->c, s->header_state, 0); if(s->version>0){ @@ -3658,29 +3582,45 @@ static int decode_header(SnowContext *s){ // s->rate_scalability= get_rac(&s->c, s->header_state); s->max_ref_frames= get_symbol(&s->c, s->header_state, 0)+1; - for(plane_index=0; plane_index<3; plane_index++){ - for(level=0; levelspatial_decomposition_count; level++){ - for(orientation=level ? 1:0; orientation<4; orientation++){ - int q; - if (plane_index==2) q= s->plane[1].band[level][orientation].qlog; - else if(orientation==2) q= s->plane[plane_index].band[level][1].qlog; - else q= get_symbol(&s->c, s->header_state, 1); - s->plane[plane_index].band[level][orientation].qlog= q; + decode_qlogs(s); + } + + if(!s->keyframe){ + if(get_rac(&s->c, s->header_state)){ + for(plane_index=0; plane_index<2; plane_index++){ + int htaps, i, sum=0; + Plane *p= &s->plane[plane_index]; + p->diag_mc= get_rac(&s->c, s->header_state); + htaps= get_symbol(&s->c, s->header_state, 0)*2 + 2; + if((unsigned)htaps > HTAPS_MAX || htaps==0) + return -1; + p->htaps= htaps; + for(i= htaps/2; i; i--){ + p->hcoeff[i]= get_symbol(&s->c, s->header_state, 0) * (1-2*(i&1)); + sum += p->hcoeff[i]; } + p->hcoeff[0]= 32-sum; } + s->plane[2].diag_mc= s->plane[1].diag_mc; + s->plane[2].htaps = s->plane[1].htaps; + memcpy(s->plane[2].hcoeff, s->plane[1].hcoeff, sizeof(s->plane[1].hcoeff)); + } + if(get_rac(&s->c, s->header_state)){ + s->spatial_decomposition_count= get_symbol(&s->c, s->header_state, 0); + decode_qlogs(s); } } - s->spatial_decomposition_type= get_symbol(&s->c, s->header_state, 0); - if(s->spatial_decomposition_type > 2){ + s->spatial_decomposition_type+= get_symbol(&s->c, s->header_state, 1); + if(s->spatial_decomposition_type > 1){ av_log(s->avctx, AV_LOG_ERROR, "spatial_decomposition_type %d not supported", s->spatial_decomposition_type); return -1; } - s->qlog= get_symbol(&s->c, s->header_state, 1); - s->mv_scale= get_symbol(&s->c, s->header_state, 0); - s->qbias= get_symbol(&s->c, s->header_state, 1); - s->block_max_depth= get_symbol(&s->c, s->header_state, 0); + s->qlog += get_symbol(&s->c, s->header_state, 1); + s->mv_scale += get_symbol(&s->c, s->header_state, 1); + s->qbias += get_symbol(&s->c, s->header_state, 1); + s->block_max_depth+= get_symbol(&s->c, s->header_state, 1); if(s->block_max_depth > 1 || s->block_max_depth < 0){ av_log(s->avctx, AV_LOG_ERROR, "block_max_depth= %d is too large", s->block_max_depth); s->block_max_depth= 0; @@ -3700,10 +3640,9 @@ static void init_qexp(void){ } } -static int common_init(AVCodecContext *avctx){ +static av_cold int common_init(AVCodecContext *avctx){ SnowContext *s = avctx->priv_data; int width, height; - int level, orientation, plane_index, dec; int i, j; s->avctx= avctx; @@ -3751,21 +3690,26 @@ static int common_init(AVCodecContext *avctx){ if(!qexp[0]) init_qexp(); - dec= s->spatial_decomposition_count= 5; - s->spatial_decomposition_type= avctx->prediction_method; //FIXME add decorrelator type r transform_type - - s->chroma_h_shift= 1; //FIXME XXX - s->chroma_v_shift= 1; - // dec += FFMAX(s->chroma_h_shift, s->chroma_v_shift); width= s->avctx->width; height= s->avctx->height; - s->spatial_dwt_buffer= av_mallocz(width*height*sizeof(DWTELEM)); + s->spatial_idwt_buffer= av_mallocz(width*height*sizeof(IDWTELEM)); + s->spatial_dwt_buffer= av_mallocz(width*height*sizeof(DWTELEM)); //FIXME this does not belong here + + for(i=0; imv_scale= (s->avctx->flags & CODEC_FLAG_QPEL) ? 2 : 4; - s->block_max_depth= (s->avctx->flags & CODEC_FLAG_4MV) ? 1 : 0; + s->avctx->get_buffer(s->avctx, &s->mconly_picture); + + return 0; +} + +static int common_init_after_header(AVCodecContext *avctx){ + SnowContext *s = avctx->priv_data; + int plane_index, level, orientation; for(plane_index=0; plane_index<3; plane_index++){ int w= s->avctx->width; @@ -3777,7 +3721,7 @@ static int common_init(AVCodecContext *avctx){ } s->plane[plane_index].width = w; s->plane[plane_index].height= h; -//av_log(NULL, AV_LOG_DEBUG, "%d %d\n", w, h); + for(level=s->spatial_decomposition_count-1; level>=0; level--){ for(orientation=level ? 1 : 0; orientation<4; orientation++){ SubBand *b= &s->plane[plane_index].band[level][orientation]; @@ -3800,9 +3744,12 @@ static int common_init(AVCodecContext *avctx){ b->buf += b->stride>>1; b->buf_y_offset = b->stride_line >> 1; } + b->ibuf= s->spatial_idwt_buffer + (b->buf - s->spatial_dwt_buffer); if(level) b->parent= &s->plane[plane_index].band[level-1][orientation]; + //FIXME avoid this realloc + av_freep(&b->x_coeff); b->x_coeff=av_mallocz(((b->width+1) * b->height+1)*sizeof(x_and_coeff)); } w= (w+1)>>1; @@ -3810,19 +3757,6 @@ static int common_init(AVCodecContext *avctx){ } } - for(i=0; iwidth= avctx->width; - height= s->height= avctx->height; - - assert(width && height); -*/ - s->avctx->get_buffer(s->avctx, &s->mconly_picture); - return 0; } @@ -3833,7 +3767,7 @@ static int qscale2qlog(int qscale){ static int ratecontrol_1pass(SnowContext *s, AVFrame *pict) { - /* estimate the frame's complexity as a sum of weighted dwt coefs. + /* Estimate the frame's complexity as a sum of weighted dwt coefficients. * FIXME we know exact mv bits at this point, * but ratecontrol isn't set up to include them. */ uint32_t coef_sum= 0; @@ -3842,21 +3776,23 @@ static int ratecontrol_1pass(SnowContext *s, AVFrame *pict) for(level=0; levelspatial_decomposition_count; level++){ for(orientation=level ? 1 : 0; orientation<4; orientation++){ SubBand *b= &s->plane[0].band[level][orientation]; - DWTELEM *buf= b->buf; + IDWTELEM *buf= b->ibuf; const int w= b->width; const int h= b->height; const int stride= b->stride; - const int qlog= clip(2*QROOT + b->qlog, 0, QROOT*16); + const int qlog= av_clip(2*QROOT + b->qlog, 0, QROOT*16); const int qmul= qexp[qlog&(QROOT-1)]<<(qlog>>QSHIFT); const int qdiv= (1<<16)/qmul; int x, y; + //FIXME this is ugly + for(y=0; ybuf[x+y*stride]; if(orientation==0) decorrelate(s, b, buf, stride, 1, 0); for(y=0; y> 16; - if(orientation==0) - correlate(s, b, buf, stride, 1, 0); } } @@ -3864,7 +3800,7 @@ static int ratecontrol_1pass(SnowContext *s, AVFrame *pict) coef_sum = (uint64_t)coef_sum * coef_sum >> 16; assert(coef_sum < INT_MAX); - if(pict->pict_type == I_TYPE){ + if(pict->pict_type == FF_I_TYPE){ s->m.current_picture.mb_var_sum= coef_sum; s->m.current_picture.mc_mb_var_sum= 0; }else{ @@ -3874,14 +3810,14 @@ static int ratecontrol_1pass(SnowContext *s, AVFrame *pict) pict->quality= ff_rate_estimate_qscale(&s->m, 1); if (pict->quality < 0) - return -1; + return INT_MIN; s->lambda= pict->quality * 3/2; delta_qlog= qscale2qlog(pict->quality) - s->qlog; s->qlog+= delta_qlog; return delta_qlog; } -static void calculate_vissual_weight(SnowContext *s, Plane *p){ +static void calculate_visual_weight(SnowContext *s, Plane *p){ int width = p->width; int height= p->height; int level, orientation, x, y; @@ -3889,43 +3825,176 @@ static void calculate_vissual_weight(SnowContext *s, Plane *p){ for(level=0; levelspatial_decomposition_count; level++){ for(orientation=level ? 1 : 0; orientation<4; orientation++){ SubBand *b= &p->band[level][orientation]; - DWTELEM *buf= b->buf; + IDWTELEM *ibuf= b->ibuf; int64_t error=0; - memset(s->spatial_dwt_buffer, 0, sizeof(int)*width*height); - buf[b->width/2 + b->height/2*b->stride]= 256*256; - ff_spatial_idwt(s->spatial_dwt_buffer, width, height, width, s->spatial_decomposition_type, s->spatial_decomposition_count); + memset(s->spatial_idwt_buffer, 0, sizeof(*s->spatial_idwt_buffer)*width*height); + ibuf[b->width/2 + b->height/2*b->stride]= 256*16; + ff_spatial_idwt(s->spatial_idwt_buffer, width, height, width, s->spatial_decomposition_type, s->spatial_decomposition_count); for(y=0; yspatial_dwt_buffer[x + y*width]; + int64_t d= s->spatial_idwt_buffer[x + y*width]*16; error += d*d; } } b->qlog= (int)(log(352256.0/sqrt(error)) / log(pow(2.0, 1.0/QROOT))+0.5); -// av_log(NULL, AV_LOG_DEBUG, "%d %d %d\n", level, orientation, b->qlog/*, sqrt(error)*/); } } } -static int encode_init(AVCodecContext *avctx) +#define QUANTIZE2 0 + +#if QUANTIZE2==1 +#define Q2_STEP 8 + +static void find_sse(SnowContext *s, Plane *p, int *score, int score_stride, IDWTELEM *r0, IDWTELEM *r1, int level, int orientation){ + SubBand *b= &p->band[level][orientation]; + int x, y; + int xo=0; + int yo=0; + int step= 1 << (s->spatial_decomposition_count - level); + + if(orientation&1) + xo= step>>1; + if(orientation&2) + yo= step>>1; + + //FIXME bias for nonzero ? + //FIXME optimize + memset(score, 0, sizeof(*score)*score_stride*((p->height + Q2_STEP-1)/Q2_STEP)); + for(y=0; yheight; y++){ + for(x=0; xwidth; x++){ + int sx= (x-xo + step/2) / step / Q2_STEP; + int sy= (y-yo + step/2) / step / Q2_STEP; + int v= r0[x + y*p->width] - r1[x + y*p->width]; + assert(sx>=0 && sy>=0 && sx < score_stride); + v= ((v+8)>>4)<<4; + score[sx + sy*score_stride] += v*v; + assert(score[sx + sy*score_stride] >= 0); + } + } +} + +static void dequantize_all(SnowContext *s, Plane *p, IDWTELEM *buffer, int width, int height){ + int level, orientation; + + for(level=0; levelspatial_decomposition_count; level++){ + for(orientation=level ? 1 : 0; orientation<4; orientation++){ + SubBand *b= &p->band[level][orientation]; + IDWTELEM *dst= buffer + (b->ibuf - s->spatial_idwt_buffer); + + dequantize(s, b, dst, b->stride); + } + } +} + +static void dwt_quantize(SnowContext *s, Plane *p, DWTELEM *buffer, int width, int height, int stride, int type){ + int level, orientation, ys, xs, x, y, pass; + IDWTELEM best_dequant[height * stride]; + IDWTELEM idwt2_buffer[height * stride]; + const int score_stride= (width + 10)/Q2_STEP; + int best_score[(width + 10)/Q2_STEP * (height + 10)/Q2_STEP]; //FIXME size + int score[(width + 10)/Q2_STEP * (height + 10)/Q2_STEP]; //FIXME size + int threshold= (s->m.lambda * s->m.lambda) >> 6; + + //FIXME pass the copy cleanly ? + +// memcpy(dwt_buffer, buffer, height * stride * sizeof(DWTELEM)); + ff_spatial_dwt(buffer, width, height, stride, type, s->spatial_decomposition_count); + + for(level=0; levelspatial_decomposition_count; level++){ + for(orientation=level ? 1 : 0; orientation<4; orientation++){ + SubBand *b= &p->band[level][orientation]; + IDWTELEM *dst= best_dequant + (b->ibuf - s->spatial_idwt_buffer); + DWTELEM *src= buffer + (b-> buf - s->spatial_dwt_buffer); + assert(src == b->buf); // code does not depend on this but it is true currently + + quantize(s, b, dst, src, b->stride, s->qbias); + } + } + for(pass=0; pass<1; pass++){ + if(s->qbias == 0) //keyframe + continue; + for(level=0; levelspatial_decomposition_count; level++){ + for(orientation=level ? 1 : 0; orientation<4; orientation++){ + SubBand *b= &p->band[level][orientation]; + IDWTELEM *dst= idwt2_buffer + (b->ibuf - s->spatial_idwt_buffer); + IDWTELEM *best_dst= best_dequant + (b->ibuf - s->spatial_idwt_buffer); + + for(ys= 0; ysspatial_decomposition_count); + find_sse(s, p, best_score, score_stride, idwt2_buffer, s->spatial_idwt_buffer, level, orientation); + memcpy(idwt2_buffer, best_dequant, height * stride * sizeof(IDWTELEM)); + for(y=ys; yheight; y+= Q2_STEP){ + for(x=xs; xwidth; x+= Q2_STEP){ + if(dst[x + y*b->stride]<0) dst[x + y*b->stride]++; + if(dst[x + y*b->stride]>0) dst[x + y*b->stride]--; + //FIXME try more than just -- + } + } + dequantize_all(s, p, idwt2_buffer, width, height); + ff_spatial_idwt(idwt2_buffer, width, height, stride, type, s->spatial_decomposition_count); + find_sse(s, p, score, score_stride, idwt2_buffer, s->spatial_idwt_buffer, level, orientation); + for(y=ys; yheight; y+= Q2_STEP){ + for(x=xs; xwidth; x+= Q2_STEP){ + int score_idx= x/Q2_STEP + (y/Q2_STEP)*score_stride; + if(score[score_idx] <= best_score[score_idx] + threshold){ + best_score[score_idx]= score[score_idx]; + if(best_dst[x + y*b->stride]<0) best_dst[x + y*b->stride]++; + if(best_dst[x + y*b->stride]>0) best_dst[x + y*b->stride]--; + //FIXME copy instead + } + } + } + } + } + } + } + } + memcpy(s->spatial_idwt_buffer, best_dequant, height * stride * sizeof(IDWTELEM)); //FIXME work with that directly instead of copy at the end +} + +#endif /* QUANTIZE2==1 */ + +static av_cold int encode_init(AVCodecContext *avctx) { SnowContext *s = avctx->priv_data; int plane_index; if(avctx->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL){ - av_log(avctx, AV_LOG_ERROR, "this codec is under development, files encoded with it may not be decodable with future versions!!!\n" - "use vstrict=-2 / -strict -2 to use it anyway\n"); + av_log(avctx, AV_LOG_ERROR, "This codec is under development, files encoded with it may not be decodable with future versions!!!\n" + "Use vstrict=-2 / -strict -2 to use it anyway.\n"); return -1; } if(avctx->prediction_method == DWT_97 && (avctx->flags & CODEC_FLAG_QSCALE) && avctx->global_quality == 0){ - av_log(avctx, AV_LOG_ERROR, "the 9/7 wavelet is incompatible with lossless mode\n"); + av_log(avctx, AV_LOG_ERROR, "The 9/7 wavelet is incompatible with lossless mode.\n"); return -1; } + s->spatial_decomposition_type= avctx->prediction_method; //FIXME add decorrelator type r transform_type + + s->chroma_h_shift= 1; //FIXME XXX + s->chroma_v_shift= 1; + + s->mv_scale = (avctx->flags & CODEC_FLAG_QPEL) ? 2 : 4; + s->block_max_depth= (avctx->flags & CODEC_FLAG_4MV ) ? 1 : 0; + + for(plane_index=0; plane_index<3; plane_index++){ + s->plane[plane_index].diag_mc= 1; + s->plane[plane_index].htaps= 6; + s->plane[plane_index].hcoeff[0]= 40; + s->plane[plane_index].hcoeff[1]= -10; + s->plane[plane_index].hcoeff[2]= 2; + s->plane[plane_index].fast_mc= 1; + } + common_init(avctx); alloc_blocks(s); @@ -3953,11 +4022,6 @@ static int encode_init(AVCodecContext *avctx) } s->pass1_rc= !(avctx->flags & (CODEC_FLAG_QSCALE|CODEC_FLAG_PASS2)); - for(plane_index=0; plane_index<3; plane_index++){ - calculate_vissual_weight(s, &s->plane[plane_index]); - } - - avctx->coded_frame= &s->current_picture; switch(avctx->pix_fmt){ // case PIX_FMT_YUV444P: @@ -3968,11 +4032,11 @@ static int encode_init(AVCodecContext *avctx) // case PIX_FMT_YUV410P: s->colorspace_type= 0; break; -/* case PIX_FMT_RGBA32: +/* case PIX_FMT_RGB32: s->colorspace= 1; break;*/ default: - av_log(avctx, AV_LOG_ERROR, "format not supported\n"); + av_log(avctx, AV_LOG_ERROR, "pixel format not supported\n"); return -1; } // avcodec_get_chroma_sub_sample(avctx->pix_fmt, &s->chroma_h_shift, &s->chroma_v_shift); @@ -3996,19 +4060,68 @@ static int encode_init(AVCodecContext *avctx) return 0; } +#define USE_HALFPEL_PLANE 0 + +static void halfpel_interpol(SnowContext *s, uint8_t *halfpel[4][4], AVFrame *frame){ + int p,x,y; + + assert(!(s->avctx->flags & CODEC_FLAG_EMU_EDGE)); + + for(p=0; p<3; p++){ + int is_chroma= !!p; + int w= s->avctx->width >>is_chroma; + int h= s->avctx->height >>is_chroma; + int ls= frame->linesize[p]; + uint8_t *src= frame->data[p]; + + halfpel[1][p]= (uint8_t*)av_malloc(ls * (h+2*EDGE_WIDTH)) + EDGE_WIDTH*(1+ls); + halfpel[2][p]= (uint8_t*)av_malloc(ls * (h+2*EDGE_WIDTH)) + EDGE_WIDTH*(1+ls); + halfpel[3][p]= (uint8_t*)av_malloc(ls * (h+2*EDGE_WIDTH)) + EDGE_WIDTH*(1+ls); + + halfpel[0][p]= src; + for(y=0; y>5; + } + } + for(y=0; y>5; + } + } + src= halfpel[1][p]; + for(y=0; y>5; + } + } + +//FIXME border! + } +} + static int frame_start(SnowContext *s){ AVFrame tmp; int w= s->avctx->width; //FIXME round up to x16 ? int h= s->avctx->height; if(s->current_picture.data[0]){ - draw_edges(s->current_picture.data[0], s->current_picture.linesize[0], w , h , EDGE_WIDTH ); - draw_edges(s->current_picture.data[1], s->current_picture.linesize[1], w>>1, h>>1, EDGE_WIDTH/2); - draw_edges(s->current_picture.data[2], s->current_picture.linesize[2], w>>1, h>>1, EDGE_WIDTH/2); + s->dsp.draw_edges(s->current_picture.data[0], s->current_picture.linesize[0], w , h , EDGE_WIDTH ); + s->dsp.draw_edges(s->current_picture.data[1], s->current_picture.linesize[1], w>>1, h>>1, EDGE_WIDTH/2); + s->dsp.draw_edges(s->current_picture.data[2], s->current_picture.linesize[2], w>>1, h>>1, EDGE_WIDTH/2); } tmp= s->last_picture[s->max_ref_frames-1]; memmove(s->last_picture+1, s->last_picture, (s->max_ref_frames-1)*sizeof(AVFrame)); + memmove(s->halfpel_plane+1, s->halfpel_plane, (s->max_ref_frames-1)*sizeof(void*)*4*4); + if(USE_HALFPEL_PLANE && s->current_picture.data[0]) + halfpel_interpol(s, s->halfpel_plane[0], &s->current_picture); s->last_picture[0]= s->current_picture; s->current_picture= tmp; @@ -4080,12 +4193,12 @@ static int encode_frame(AVCodecContext *avctx, unsigned char *buf, int buf_size, if(s->qlog < 0 || (!pict->quality && (avctx->flags & CODEC_FLAG_QSCALE))){ s->qlog= LOSSLESS_QLOG; s->lambda = 0; - }//else keep previous frame's qlog until after motion est + }//else keep previous frame's qlog until after motion estimation frame_start(s); s->m.current_picture_ptr= &s->m.current_picture; - if(pict->pict_type == P_TYPE){ + if(pict->pict_type == FF_P_TYPE){ int block_width = (width +15)>>4; int block_height= (height+15)>>4; int stride= s->current_picture.linesize[0]; @@ -4134,8 +4247,21 @@ static int encode_frame(AVCodecContext *avctx, unsigned char *buf, int buf_size, redo_frame: + if(pict->pict_type == FF_I_TYPE) + s->spatial_decomposition_count= 5; + else + s->spatial_decomposition_count= 5; + s->m.pict_type = pict->pict_type; - s->qbias= pict->pict_type == P_TYPE ? 2 : 0; + s->qbias= pict->pict_type == FF_P_TYPE ? 2 : 0; + + common_init_after_header(avctx); + + if(s->last_spatial_decomposition_count != s->spatial_decomposition_count){ + for(plane_index=0; plane_index<3; plane_index++){ + calculate_visual_weight(s, &s->plane[plane_index]); + } + } encode_header(s); s->m.misc_bits = 8*(s->c.bytestream - s->c.bytestream_start); @@ -4149,90 +4275,96 @@ redo_frame: int x, y; // int bits= put_bits_count(&s->c.pb); - if(!(avctx->flags2 & CODEC_FLAG2_MEMC_ONLY)){ - //FIXME optimize - if(pict->data[plane_index]) //FIXME gray hack - for(y=0; yspatial_dwt_buffer[y*w + x]= pict->data[plane_index][y*pict->linesize[plane_index] + x]<flags2 & CODEC_FLAG2_MEMC_ONLY)){ + //FIXME optimize + if(pict->data[plane_index]) //FIXME gray hack + for(y=0; yspatial_idwt_buffer[y*w + x]= pict->data[plane_index][y*pict->linesize[plane_index] + x]<spatial_idwt_buffer, plane_index, 0); + + if( plane_index==0 + && pict->pict_type == FF_P_TYPE + && !(avctx->flags&CODEC_FLAG_PASS2) + && s->m.me.scene_change_score > s->avctx->scenechange_threshold){ + ff_init_range_encoder(c, buf, buf_size); + ff_build_rac_states(c, 0.05*(1LL<<32), 256-8); + pict->pict_type= FF_I_TYPE; + s->keyframe=1; + s->current_picture.key_frame=1; + goto redo_frame; } - } - predict_plane(s, s->spatial_dwt_buffer, plane_index, 0); - - if( plane_index==0 - && pict->pict_type == P_TYPE - && !(avctx->flags&CODEC_FLAG_PASS2) - && s->m.me.scene_change_score > s->avctx->scenechange_threshold){ - ff_init_range_encoder(c, buf, buf_size); - ff_build_rac_states(c, 0.05*(1LL<<32), 256-8); - pict->pict_type= FF_I_TYPE; - s->keyframe=1; - s->current_picture.key_frame=1; - reset_contexts(s); - goto redo_frame; - } - if(s->qlog == LOSSLESS_QLOG){ - for(y=0; yspatial_dwt_buffer[y*w + x]= (s->spatial_dwt_buffer[y*w + x] + (1<<(FRAC_BITS-1))-1)>>FRAC_BITS; + if(s->qlog == LOSSLESS_QLOG){ + for(y=0; yspatial_dwt_buffer[y*w + x]= (s->spatial_idwt_buffer[y*w + x] + (1<<(FRAC_BITS-1))-1)>>FRAC_BITS; + } + } + }else{ + for(y=0; yspatial_dwt_buffer[y*w + x]=s->spatial_idwt_buffer[y*w + x]<spatial_dwt_buffer, w, h, w, s->spatial_decomposition_type, s->spatial_decomposition_count); - if(s->pass1_rc && plane_index==0){ - int delta_qlog = ratecontrol_1pass(s, pict); - if (delta_qlog < 0) - return -1; - if(delta_qlog){ - //reordering qlog in the bitstream would eliminate this reset - ff_init_range_encoder(c, buf, buf_size); - memcpy(s->header_state, rc_header_bak, sizeof(s->header_state)); - memcpy(s->block_state, rc_block_bak, sizeof(s->block_state)); - encode_header(s); - encode_blocks(s, 0); + /* if(QUANTIZE2) + dwt_quantize(s, p, s->spatial_dwt_buffer, w, h, w, s->spatial_decomposition_type); + else*/ + ff_spatial_dwt(s->spatial_dwt_buffer, w, h, w, s->spatial_decomposition_type, s->spatial_decomposition_count); + + if(s->pass1_rc && plane_index==0){ + int delta_qlog = ratecontrol_1pass(s, pict); + if (delta_qlog <= INT_MIN) + return -1; + if(delta_qlog){ + //reordering qlog in the bitstream would eliminate this reset + ff_init_range_encoder(c, buf, buf_size); + memcpy(s->header_state, rc_header_bak, sizeof(s->header_state)); + memcpy(s->block_state, rc_block_bak, sizeof(s->block_state)); + encode_header(s); + encode_blocks(s, 0); + } } - } - for(level=0; levelspatial_decomposition_count; level++){ - for(orientation=level ? 1 : 0; orientation<4; orientation++){ - SubBand *b= &p->band[level][orientation]; - - quantize(s, b, b->buf, b->stride, s->qbias); - if(orientation==0) - decorrelate(s, b, b->buf, b->stride, pict->pict_type == P_TYPE, 0); - encode_subband(s, b, b->buf, b->parent ? b->parent->buf : NULL, b->stride, orientation); - assert(b->parent==NULL || b->parent->stride == b->stride*2); - if(orientation==0) - correlate(s, b, b->buf, b->stride, 1, 0); + for(level=0; levelspatial_decomposition_count; level++){ + for(orientation=level ? 1 : 0; orientation<4; orientation++){ + SubBand *b= &p->band[level][orientation]; + + if(!QUANTIZE2) + quantize(s, b, b->ibuf, b->buf, b->stride, s->qbias); + if(orientation==0) + decorrelate(s, b, b->ibuf, b->stride, pict->pict_type == FF_P_TYPE, 0); + encode_subband(s, b, b->ibuf, b->parent ? b->parent->ibuf : NULL, b->stride, orientation); + assert(b->parent==NULL || b->parent->stride == b->stride*2); + if(orientation==0) + correlate(s, b, b->ibuf, b->stride, 1, 0); + } } - } -// av_log(NULL, AV_LOG_DEBUG, "plane:%d bits:%d\n", plane_index, put_bits_count(&s->c.pb) - bits); - for(level=0; levelspatial_decomposition_count; level++){ - for(orientation=level ? 1 : 0; orientation<4; orientation++){ - SubBand *b= &p->band[level][orientation]; + for(level=0; levelspatial_decomposition_count; level++){ + for(orientation=level ? 1 : 0; orientation<4; orientation++){ + SubBand *b= &p->band[level][orientation]; - dequantize(s, b, b->buf, b->stride); + dequantize(s, b, b->ibuf, b->stride); + } } - } - ff_spatial_idwt(s->spatial_dwt_buffer, w, h, w, s->spatial_decomposition_type, s->spatial_decomposition_count); - if(s->qlog == LOSSLESS_QLOG){ - for(y=0; yspatial_dwt_buffer[y*w + x]<<=FRAC_BITS; + ff_spatial_idwt(s->spatial_idwt_buffer, w, h, w, s->spatial_decomposition_type, s->spatial_decomposition_count); + if(s->qlog == LOSSLESS_QLOG){ + for(y=0; yspatial_idwt_buffer[y*w + x]<<=FRAC_BITS; + } } } - } -{START_TIMER - predict_plane(s, s->spatial_dwt_buffer, plane_index, 1); -STOP_TIMER("pred-conv")} - }else{ + predict_plane(s, s->spatial_idwt_buffer, plane_index, 1); + }else{ //ME/MC only - if(pict->pict_type == I_TYPE){ + if(pict->pict_type == FF_I_TYPE){ for(y=0; ycurrent_picture.data[plane_index][y*s->current_picture.linesize[plane_index] + x]= @@ -4240,27 +4372,34 @@ STOP_TIMER("pred-conv")} } } }else{ - memset(s->spatial_dwt_buffer, 0, sizeof(DWTELEM)*w*h); - predict_plane(s, s->spatial_dwt_buffer, plane_index, 1); + memset(s->spatial_idwt_buffer, 0, sizeof(IDWTELEM)*w*h); + predict_plane(s, s->spatial_idwt_buffer, plane_index, 1); } - } + } if(s->avctx->flags&CODEC_FLAG_PSNR){ int64_t error= 0; - if(pict->data[plane_index]) //FIXME gray hack - for(y=0; ycurrent_picture.data[plane_index][y*s->current_picture.linesize[plane_index] + x] - pict->data[plane_index][y*pict->linesize[plane_index] + x]; - error += d*d; + if(pict->data[plane_index]) //FIXME gray hack + for(y=0; ycurrent_picture.data[plane_index][y*s->current_picture.linesize[plane_index] + x] - pict->data[plane_index][y*pict->linesize[plane_index] + x]; + error += d*d; + } } - } s->avctx->error[plane_index] += error; s->current_picture.error[plane_index] = error; } + } - if(s->last_picture[s->max_ref_frames-1].data[0]) + update_last_header_values(s); + + if(s->last_picture[s->max_ref_frames-1].data[0]){ avctx->release_buffer(avctx, &s->last_picture[s->max_ref_frames-1]); + for(i=0; i<9; i++) + if(s->halfpel_plane[s->max_ref_frames-1][1+i/3][i%3]) + av_free(s->halfpel_plane[s->max_ref_frames-1][1+i/3][i%3] - EDGE_WIDTH*(1+s->current_picture.linesize[i%3])); + } s->current_picture.coded_picture_number = avctx->frame_number; s->current_picture.pict_type = pict->pict_type; @@ -4277,16 +4416,21 @@ STOP_TIMER("pred-conv")} if(avctx->flags&CODEC_FLAG_PASS1) ff_write_pass1_stats(&s->m); s->m.last_pict_type = s->m.pict_type; + avctx->frame_bits = s->m.frame_bits; + avctx->mv_bits = s->m.mv_bits; + avctx->misc_bits = s->m.misc_bits; + avctx->p_tex_bits = s->m.p_tex_bits; emms_c(); return ff_rac_terminate(c); } -static void common_end(SnowContext *s){ +static av_cold void common_end(SnowContext *s){ int plane_index, level, orientation, i; av_freep(&s->spatial_dwt_buffer); + av_freep(&s->spatial_idwt_buffer); av_freep(&s->m.me.scratchpad); av_freep(&s->m.me.map); @@ -4313,7 +4457,7 @@ static void common_end(SnowContext *s){ } } -static int encode_end(AVCodecContext *avctx) +static av_cold int encode_end(AVCodecContext *avctx) { SnowContext *s = avctx->priv_data; @@ -4323,37 +4467,45 @@ static int encode_end(AVCodecContext *avctx) return 0; } -static int decode_init(AVCodecContext *avctx) +static av_cold int decode_init(AVCodecContext *avctx) { - SnowContext *s = avctx->priv_data; - int block_size; - avctx->pix_fmt= PIX_FMT_YUV420P; common_init(avctx); - block_size = MB_SIZE >> s->block_max_depth; - slice_buffer_init(&s->sb, s->plane[0].height, (block_size) + (s->spatial_decomposition_count * (s->spatial_decomposition_count + 3)) + 1, s->plane[0].width, s->spatial_dwt_buffer); - return 0; } -static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, uint8_t *buf, int buf_size){ +static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, const uint8_t *buf, int buf_size){ SnowContext *s = avctx->priv_data; RangeCoder * const c= &s->c; int bytes_read; AVFrame *picture = data; - int level, orientation, plane_index; + int level, orientation, plane_index, i; ff_init_range_decoder(c, buf, buf_size); ff_build_rac_states(c, 0.05*(1LL<<32), 256-8); s->current_picture.pict_type= FF_I_TYPE; //FIXME I vs. P - decode_header(s); + if(decode_header(s)<0) + return -1; + common_init_after_header(avctx); + + // realloc slice buffer for the case that spatial_decomposition_count changed + slice_buffer_destroy(&s->sb); + slice_buffer_init(&s->sb, s->plane[0].height, (MB_SIZE >> s->block_max_depth) + s->spatial_decomposition_count * 8 + 1, s->plane[0].width, s->spatial_idwt_buffer); + + for(plane_index=0; plane_index<3; plane_index++){ + Plane *p= &s->plane[plane_index]; + p->fast_mc= p->diag_mc && p->htaps==6 && p->hcoeff[0]==40 + && p->hcoeff[1]==-10 + && p->hcoeff[2]==2; + } + if(!s->block) alloc_blocks(s); frame_start(s); - //keyframe flag dupliaction mess FIXME + //keyframe flag duplication mess FIXME if(avctx->debug&FF_DEBUG_PICT_INFO) av_log(avctx, AV_LOG_ERROR, "keyframe:%d qlog:%d\n", s->keyframe, s->qlog); @@ -4366,122 +4518,118 @@ static int decode_frame(AVCodecContext *avctx, void *data, int *data_size, uint8 int x, y; int decode_state[MAX_DECOMPOSITIONS][4][1]; /* Stored state info for unpack_coeffs. 1 variable per instance. */ -if(s->avctx->debug&2048){ - memset(s->spatial_dwt_buffer, 0, sizeof(DWTELEM)*w*h); - predict_plane(s, s->spatial_dwt_buffer, plane_index, 1); + if(s->avctx->debug&2048){ + memset(s->spatial_dwt_buffer, 0, sizeof(DWTELEM)*w*h); + predict_plane(s, s->spatial_idwt_buffer, plane_index, 1); - for(y=0; ycurrent_picture.data[plane_index][y*s->current_picture.linesize[plane_index] + x]; - s->mconly_picture.data[plane_index][y*s->mconly_picture.linesize[plane_index] + x]= v; + for(y=0; ycurrent_picture.data[plane_index][y*s->current_picture.linesize[plane_index] + x]; + s->mconly_picture.data[plane_index][y*s->mconly_picture.linesize[plane_index] + x]= v; + } } } -} - -{ START_TIMER - for(level=0; levelspatial_decomposition_count; level++){ - for(orientation=level ? 1 : 0; orientation<4; orientation++){ - SubBand *b= &p->band[level][orientation]; - unpack_coeffs(s, b, b->parent, orientation); - } - } - STOP_TIMER("unpack coeffs"); -} - -{START_TIMER - const int mb_h= s->b_height << s->block_max_depth; - const int block_size = MB_SIZE >> s->block_max_depth; - const int block_w = plane_index ? block_size/2 : block_size; - int mb_y; - dwt_compose_t cs[MAX_DECOMPOSITIONS]; - int yd=0, yq=0; - int y; - int end_y; - - ff_spatial_idwt_buffered_init(cs, &s->sb, w, h, 1, s->spatial_decomposition_type, s->spatial_decomposition_count); - for(mb_y=0; mb_y<=mb_h; mb_y++){ - - int slice_starty = block_w*mb_y; - int slice_h = block_w*(mb_y+1); - if (!(s->keyframe || s->avctx->debug&512)){ - slice_starty = FFMAX(0, slice_starty - (block_w >> 1)); - slice_h -= (block_w >> 1); - } { - START_TIMER for(level=0; levelspatial_decomposition_count; level++){ for(orientation=level ? 1 : 0; orientation<4; orientation++){ SubBand *b= &p->band[level][orientation]; - int start_y; - int end_y; - int our_mb_start = mb_y; - int our_mb_end = (mb_y + 1); - const int extra= 3; - start_y = (mb_y ? ((block_w * our_mb_start) >> (s->spatial_decomposition_count - level)) + s->spatial_decomposition_count - level + extra: 0); - end_y = (((block_w * our_mb_end) >> (s->spatial_decomposition_count - level)) + s->spatial_decomposition_count - level + extra); - if (!(s->keyframe || s->avctx->debug&512)){ - start_y = FFMAX(0, start_y - (block_w >> (1+s->spatial_decomposition_count - level))); - end_y = FFMAX(0, end_y - (block_w >> (1+s->spatial_decomposition_count - level))); - } - start_y = FFMIN(b->height, start_y); - end_y = FFMIN(b->height, end_y); - - if (start_y != end_y){ - if (orientation == 0){ - SubBand * correlate_band = &p->band[0][0]; - int correlate_end_y = FFMIN(b->height, end_y + 1); - int correlate_start_y = FFMIN(b->height, (start_y ? start_y + 1 : 0)); - decode_subband_slice_buffered(s, correlate_band, &s->sb, correlate_start_y, correlate_end_y, decode_state[0][0]); - correlate_slice_buffered(s, &s->sb, correlate_band, correlate_band->buf, correlate_band->stride, 1, 0, correlate_start_y, correlate_end_y); - dequantize_slice_buffered(s, &s->sb, correlate_band, correlate_band->buf, correlate_band->stride, start_y, end_y); - } - else - decode_subband_slice_buffered(s, b, &s->sb, start_y, end_y, decode_state[level][orientation]); - } + unpack_coeffs(s, b, b->parent, orientation); } } - STOP_TIMER("decode_subband_slice"); } -{ START_TIMER - for(; yddsp, cs, &s->sb, w, h, 1, s->spatial_decomposition_type, s->spatial_decomposition_count, yd); - } - STOP_TIMER("idwt slice");} + { + const int mb_h= s->b_height << s->block_max_depth; + const int block_size = MB_SIZE >> s->block_max_depth; + const int block_w = plane_index ? block_size/2 : block_size; + int mb_y; + dwt_compose_t cs[MAX_DECOMPOSITIONS]; + int yd=0, yq=0; + int y; + int end_y; + ff_spatial_idwt_buffered_init(cs, &s->sb, w, h, 1, s->spatial_decomposition_type, s->spatial_decomposition_count); + for(mb_y=0; mb_y<=mb_h; mb_y++){ - if(s->qlog == LOSSLESS_QLOG){ - for(; yqsb, yq); - for(x=0; xkeyframe || s->avctx->debug&512)){ + slice_starty = FFMAX(0, slice_starty - (block_w >> 1)); + slice_h -= (block_w >> 1); + } + + for(level=0; levelspatial_decomposition_count; level++){ + for(orientation=level ? 1 : 0; orientation<4; orientation++){ + SubBand *b= &p->band[level][orientation]; + int start_y; + int end_y; + int our_mb_start = mb_y; + int our_mb_end = (mb_y + 1); + const int extra= 3; + start_y = (mb_y ? ((block_w * our_mb_start) >> (s->spatial_decomposition_count - level)) + s->spatial_decomposition_count - level + extra: 0); + end_y = (((block_w * our_mb_end) >> (s->spatial_decomposition_count - level)) + s->spatial_decomposition_count - level + extra); + if (!(s->keyframe || s->avctx->debug&512)){ + start_y = FFMAX(0, start_y - (block_w >> (1+s->spatial_decomposition_count - level))); + end_y = FFMAX(0, end_y - (block_w >> (1+s->spatial_decomposition_count - level))); + } + start_y = FFMIN(b->height, start_y); + end_y = FFMIN(b->height, end_y); + + if (start_y != end_y){ + if (orientation == 0){ + SubBand * correlate_band = &p->band[0][0]; + int correlate_end_y = FFMIN(b->height, end_y + 1); + int correlate_start_y = FFMIN(b->height, (start_y ? start_y + 1 : 0)); + decode_subband_slice_buffered(s, correlate_band, &s->sb, correlate_start_y, correlate_end_y, decode_state[0][0]); + correlate_slice_buffered(s, &s->sb, correlate_band, correlate_band->ibuf, correlate_band->stride, 1, 0, correlate_start_y, correlate_end_y); + dequantize_slice_buffered(s, &s->sb, correlate_band, correlate_band->ibuf, correlate_band->stride, start_y, end_y); + } + else + decode_subband_slice_buffered(s, b, &s->sb, start_y, end_y, decode_state[level][orientation]); + } } } - } - predict_slice_buffered(s, &s->sb, s->spatial_dwt_buffer, plane_index, 1, mb_y); + for(; yddsp, cs, &s->sb, w, h, 1, s->spatial_decomposition_type, s->spatial_decomposition_count, yd); + } - y = FFMIN(p->height, slice_starty); - end_y = FFMIN(p->height, slice_h); - while(y < end_y) - slice_buffer_release(&s->sb, y++); - } + if(s->qlog == LOSSLESS_QLOG){ + for(; yqsb, yq); + for(x=0; xsb, s->spatial_idwt_buffer, plane_index, 1, mb_y); - slice_buffer_flush(&s->sb); + y = FFMIN(p->height, slice_starty); + end_y = FFMIN(p->height, slice_h); + while(y < end_y) + slice_buffer_release(&s->sb, y++); + } + + slice_buffer_flush(&s->sb); + } -STOP_TIMER("idwt + predict_slices")} } emms_c(); - if(s->last_picture[s->max_ref_frames-1].data[0]) + if(s->last_picture[s->max_ref_frames-1].data[0]){ avctx->release_buffer(avctx, &s->last_picture[s->max_ref_frames-1]); + for(i=0; i<9; i++) + if(s->halfpel_plane[s->max_ref_frames-1][1+i/3][i%3]) + av_free(s->halfpel_plane[s->max_ref_frames-1][1+i/3][i%3] - EDGE_WIDTH*(1+s->current_picture.linesize[i%3])); + } -if(!(s->avctx->debug&2048)) - *picture= s->current_picture; -else - *picture= s->mconly_picture; + if(!(s->avctx->debug&2048)) + *picture= s->current_picture; + else + *picture= s->mconly_picture; *data_size = sizeof(AVFrame); @@ -4491,7 +4639,7 @@ else return bytes_read; } -static int decode_end(AVCodecContext *avctx) +static av_cold int decode_end(AVCodecContext *avctx) { SnowContext *s = avctx->priv_data; @@ -4515,7 +4663,7 @@ AVCodec snow_decoder = { NULL }; -#ifdef CONFIG_ENCODERS +#ifdef CONFIG_SNOW_ENCODER AVCodec snow_encoder = { "snow", CODEC_TYPE_VIDEO, @@ -4528,12 +4676,13 @@ AVCodec snow_encoder = { #endif -#if 0 +#ifdef TEST #undef malloc #undef free #undef printf +#undef random -int main(){ +int main(void){ int width=256; int height=256; int buffer[2][width*height]; @@ -4561,7 +4710,7 @@ int main(){ ff_spatial_idwt(buffer[0], width, height, width, s.spatial_decomposition_type, s.spatial_decomposition_count); for(i=0; i20) printf("fsck: %d %d %d\n",i, buffer[0][i], buffer[1][i]); + if(FFABS(buffer[0][i] - buffer[1][i])>20) printf("fsck: %d %d %d\n",i, buffer[0][i], buffer[1][i]); #if 0 printf("testing AC coder\n"); @@ -4570,9 +4719,7 @@ int main(){ ff_init_cabac_states(&s.c, ff_h264_lps_range, ff_h264_mps_state, ff_h264_lps_state, 64); for(i=-256; i<256; i++){ -START_TIMER - put_symbol(&s.c, s.header_state, i*i*i/3*ABS(i), 1); -STOP_TIMER("put_symbol") + put_symbol(&s.c, s.header_state, i*i*i/3*FFABS(i), 1); } ff_rac_terminate(&s.c); @@ -4582,62 +4729,59 @@ STOP_TIMER("put_symbol") for(i=-256; i<256; i++){ int j; -START_TIMER j= get_symbol(&s.c, s.header_state, 1); -STOP_TIMER("get_symbol") - if(j!=i*i*i/3*ABS(i)) printf("fsck: %d != %d\n", i, j); + if(j!=i*i*i/3*FFABS(i)) printf("fsck: %d != %d\n", i, j); } #endif -{ -int level, orientation, x, y; -int64_t errors[8][4]; -int64_t g=0; - - memset(errors, 0, sizeof(errors)); - s.spatial_decomposition_count=3; - s.spatial_decomposition_type=0; - for(level=0; level> (s.spatial_decomposition_count-level); - int h= height >> (s.spatial_decomposition_count-level); - int stride= width << (s.spatial_decomposition_count-level); - DWTELEM *buf= buffer[0]; - int64_t error=0; - - if(orientation&1) buf+=w; - if(orientation>1) buf+=stride>>1; + { + int level, orientation, x, y; + int64_t errors[8][4]; + int64_t g=0; - memset(buffer[0], 0, sizeof(int)*width*height); - buf[w/2 + h/2*stride]= 256*256; - ff_spatial_idwt(buffer[0], width, height, width, s.spatial_decomposition_type, s.spatial_decomposition_count); - for(y=0; y> (s.spatial_decomposition_count-level); + int h= height >> (s.spatial_decomposition_count-level); + int stride= width << (s.spatial_decomposition_count-level); + DWTELEM *buf= buffer[0]; + int64_t error=0; + + if(orientation&1) buf+=w; + if(orientation>1) buf+=stride>>1; + + memset(buffer[0], 0, sizeof(int)*width*height); + buf[w/2 + h/2*stride]= 256*256; + ff_spatial_idwt(buffer[0], width, height, width, s.spatial_decomposition_type, s.spatial_decomposition_count); + for(y=0; y> (s.spatial_decomposition_count-level); - int h= height >> (s.spatial_decomposition_count-level); + //int h= height >> (s.spatial_decomposition_count-level); int stride= width << (s.spatial_decomposition_count-level); DWTELEM *buf= buffer[0]; int64_t error=0; @@ -4667,14 +4811,13 @@ int64_t g=0; for(x=0; x