X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fdsputil.c;h=6c734e570e23d2c2a65474986517632691a202d5;hb=d29ee34ce31668703913f9e9c5dc3b8df969addd;hp=fc5afbeb49a37945119800c194295172f7492de7;hpb=db794953762b813d506501b3f6f4796ffd423a65;p=ffmpeg diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c index fc5afbeb49a..6c734e570e2 100644 --- a/libavcodec/dsputil.c +++ b/libavcodec/dsputil.c @@ -1,33 +1,60 @@ /* * DSP utils * Copyright (c) 2000, 2001 Fabrice Bellard. + * Copyright (c) 2002-2004 Michael Niedermayer * - * This library is free software; you can redistribute it and/or + * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer + * + * 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 - * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA - * - * gmc & q-pel & 32/64 bit based MC by Michael Niedermayer + * License along with FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + */ + +/** + * @file dsputil.c + * DSP utils */ + #include "avcodec.h" #include "dsputil.h" -#include "mpegvideo.h" +#include "simple_idct.h" +#include "faandct.h" +#include "faanidct.h" +#include "h263.h" +#include "snow.h" + +/* snow.c */ +void ff_spatial_dwt(int *buffer, int width, int height, int stride, int type, int decomposition_count); + +/* vorbis.c */ +void vorbis_inverse_coupling(float *mag, float *ang, int blocksize); + +/* flacenc.c */ +void ff_flac_compute_autocorr(const int32_t *data, int len, int lag, double *autoc); -int ff_bit_exact=0; +/* pngdec.c */ +void ff_add_png_paeth_prediction(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp); -UINT8 cropTbl[256 + 2 * MAX_NEG_CROP]; -UINT32 squareTbl[512]; +uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP] = {0, }; +uint32_t ff_squareTbl[512] = {0, }; -const UINT8 ff_zigzag_direct[64] = { +// 0x7f7f7f7f or 0x7f7f7f7f7f7f7f7f or whatever, depending on the cpu's native arithmetic size +#define pb_7f (~0UL/255 * 0x7f) +#define pb_80 (~0UL/255 * 0x80) + +const uint8_t ff_zigzag_direct[64] = { 0, 1, 8, 16, 9, 2, 3, 10, 17, 24, 32, 25, 18, 11, 4, 5, 12, 19, 26, 33, 40, 48, 41, 34, @@ -38,120 +65,230 @@ const UINT8 ff_zigzag_direct[64] = { 53, 60, 61, 54, 47, 55, 62, 63 }; +/* Specific zigzag scan for 248 idct. NOTE that unlike the + specification, we interleave the fields */ +const uint8_t ff_zigzag248_direct[64] = { + 0, 8, 1, 9, 16, 24, 2, 10, + 17, 25, 32, 40, 48, 56, 33, 41, + 18, 26, 3, 11, 4, 12, 19, 27, + 34, 42, 49, 57, 50, 58, 35, 43, + 20, 28, 5, 13, 6, 14, 21, 29, + 36, 44, 51, 59, 52, 60, 37, 45, + 22, 30, 7, 15, 23, 31, 38, 46, + 53, 61, 54, 62, 39, 47, 55, 63, +}; + /* not permutated inverse zigzag_direct + 1 for MMX quantizer */ -UINT16 __align8 inv_zigzag_direct16[64]; +DECLARE_ALIGNED_8(uint16_t, inv_zigzag_direct16[64]) = {0, }; -const UINT8 ff_alternate_horizontal_scan[64] = { - 0, 1, 2, 3, 8, 9, 16, 17, +const uint8_t ff_alternate_horizontal_scan[64] = { + 0, 1, 2, 3, 8, 9, 16, 17, 10, 11, 4, 5, 6, 7, 15, 14, - 13, 12, 19, 18, 24, 25, 32, 33, + 13, 12, 19, 18, 24, 25, 32, 33, 26, 27, 20, 21, 22, 23, 28, 29, - 30, 31, 34, 35, 40, 41, 48, 49, + 30, 31, 34, 35, 40, 41, 48, 49, 42, 43, 36, 37, 38, 39, 44, 45, - 46, 47, 50, 51, 56, 57, 58, 59, + 46, 47, 50, 51, 56, 57, 58, 59, 52, 53, 54, 55, 60, 61, 62, 63, }; -const UINT8 ff_alternate_vertical_scan[64] = { - 0, 8, 16, 24, 1, 9, 2, 10, +const uint8_t ff_alternate_vertical_scan[64] = { + 0, 8, 16, 24, 1, 9, 2, 10, 17, 25, 32, 40, 48, 56, 57, 49, - 41, 33, 26, 18, 3, 11, 4, 12, + 41, 33, 26, 18, 3, 11, 4, 12, 19, 27, 34, 42, 50, 58, 35, 43, - 51, 59, 20, 28, 5, 13, 6, 14, + 51, 59, 20, 28, 5, 13, 6, 14, 21, 29, 36, 44, 52, 60, 37, 45, - 53, 61, 22, 30, 7, 15, 23, 31, + 53, 61, 22, 30, 7, 15, 23, 31, 38, 46, 54, 62, 39, 47, 55, 63, }; /* a*inverse[b]>>32 == a/b for all 0<=a<=65536 && 2<=b<=255 */ -const UINT32 inverse[256]={ - 0, 4294967295U,2147483648U,1431655766, 1073741824, 858993460, 715827883, 613566757, - 536870912, 477218589, 429496730, 390451573, 357913942, 330382100, 306783379, 286331154, - 268435456, 252645136, 238609295, 226050911, 214748365, 204522253, 195225787, 186737709, - 178956971, 171798692, 165191050, 159072863, 153391690, 148102321, 143165577, 138547333, - 134217728, 130150525, 126322568, 122713352, 119304648, 116080198, 113025456, 110127367, - 107374183, 104755300, 102261127, 99882961, 97612894, 95443718, 93368855, 91382283, - 89478486, 87652394, 85899346, 84215046, 82595525, 81037119, 79536432, 78090315, - 76695845, 75350304, 74051161, 72796056, 71582789, 70409300, 69273667, 68174085, - 67108864, 66076420, 65075263, 64103990, 63161284, 62245903, 61356676, 60492498, - 59652324, 58835169, 58040099, 57266231, 56512728, 55778797, 55063684, 54366675, - 53687092, 53024288, 52377650, 51746594, 51130564, 50529028, 49941481, 49367441, - 48806447, 48258060, 47721859, 47197443, 46684428, 46182445, 45691142, 45210183, - 44739243, 44278014, 43826197, 43383509, 42949673, 42524429, 42107523, 41698712, - 41297763, 40904451, 40518560, 40139882, 39768216, 39403370, 39045158, 38693400, - 38347923, 38008561, 37675152, 37347542, 37025581, 36709123, 36398028, 36092163, - 35791395, 35495598, 35204650, 34918434, 34636834, 34359739, 34087043, 33818641, - 33554432, 33294321, 33038210, 32786010, 32537632, 32292988, 32051995, 31814573, - 31580642, 31350127, 31122952, 30899046, 30678338, 30460761, 30246249, 30034737, - 29826162, 29620465, 29417585, 29217465, 29020050, 28825284, 28633116, 28443493, - 28256364, 28071682, 27889399, 27709467, 27531842, 27356480, 27183338, 27012373, - 26843546, 26676816, 26512144, 26349493, 26188825, 26030105, 25873297, 25718368, - 25565282, 25414008, 25264514, 25116768, 24970741, 24826401, 24683721, 24542671, - 24403224, 24265352, 24129030, 23994231, 23860930, 23729102, 23598722, 23469767, - 23342214, 23216040, 23091223, 22967740, 22845571, 22724695, 22605092, 22486740, - 22369622, 22253717, 22139007, 22025474, 21913099, 21801865, 21691755, 21582751, - 21474837, 21367997, 21262215, 21157475, 21053762, 20951060, 20849356, 20748635, - 20648882, 20550083, 20452226, 20355296, 20259280, 20164166, 20069941, 19976593, - 19884108, 19792477, 19701685, 19611723, 19522579, 19434242, 19346700, 19259944, - 19173962, 19088744, 19004281, 18920561, 18837576, 18755316, 18673771, 18592933, - 18512791, 18433337, 18354562, 18276457, 18199014, 18122225, 18046082, 17970575, - 17895698, 17821442, 17747799, 17674763, 17602325, 17530479, 17459217, 17388532, +const uint32_t ff_inverse[256]={ + 0, 4294967295U,2147483648U,1431655766, 1073741824, 858993460, 715827883, 613566757, + 536870912, 477218589, 429496730, 390451573, 357913942, 330382100, 306783379, 286331154, + 268435456, 252645136, 238609295, 226050911, 214748365, 204522253, 195225787, 186737709, + 178956971, 171798692, 165191050, 159072863, 153391690, 148102321, 143165577, 138547333, + 134217728, 130150525, 126322568, 122713352, 119304648, 116080198, 113025456, 110127367, + 107374183, 104755300, 102261127, 99882961, 97612894, 95443718, 93368855, 91382283, + 89478486, 87652394, 85899346, 84215046, 82595525, 81037119, 79536432, 78090315, + 76695845, 75350304, 74051161, 72796056, 71582789, 70409300, 69273667, 68174085, + 67108864, 66076420, 65075263, 64103990, 63161284, 62245903, 61356676, 60492498, + 59652324, 58835169, 58040099, 57266231, 56512728, 55778797, 55063684, 54366675, + 53687092, 53024288, 52377650, 51746594, 51130564, 50529028, 49941481, 49367441, + 48806447, 48258060, 47721859, 47197443, 46684428, 46182445, 45691142, 45210183, + 44739243, 44278014, 43826197, 43383509, 42949673, 42524429, 42107523, 41698712, + 41297763, 40904451, 40518560, 40139882, 39768216, 39403370, 39045158, 38693400, + 38347923, 38008561, 37675152, 37347542, 37025581, 36709123, 36398028, 36092163, + 35791395, 35495598, 35204650, 34918434, 34636834, 34359739, 34087043, 33818641, + 33554432, 33294321, 33038210, 32786010, 32537632, 32292988, 32051995, 31814573, + 31580642, 31350127, 31122952, 30899046, 30678338, 30460761, 30246249, 30034737, + 29826162, 29620465, 29417585, 29217465, 29020050, 28825284, 28633116, 28443493, + 28256364, 28071682, 27889399, 27709467, 27531842, 27356480, 27183338, 27012373, + 26843546, 26676816, 26512144, 26349493, 26188825, 26030105, 25873297, 25718368, + 25565282, 25414008, 25264514, 25116768, 24970741, 24826401, 24683721, 24542671, + 24403224, 24265352, 24129030, 23994231, 23860930, 23729102, 23598722, 23469767, + 23342214, 23216040, 23091223, 22967740, 22845571, 22724695, 22605092, 22486740, + 22369622, 22253717, 22139007, 22025474, 21913099, 21801865, 21691755, 21582751, + 21474837, 21367997, 21262215, 21157475, 21053762, 20951060, 20849356, 20748635, + 20648882, 20550083, 20452226, 20355296, 20259280, 20164166, 20069941, 19976593, + 19884108, 19792477, 19701685, 19611723, 19522579, 19434242, 19346700, 19259944, + 19173962, 19088744, 19004281, 18920561, 18837576, 18755316, 18673771, 18592933, + 18512791, 18433337, 18354562, 18276457, 18199014, 18122225, 18046082, 17970575, + 17895698, 17821442, 17747799, 17674763, 17602325, 17530479, 17459217, 17388532, 17318417, 17248865, 17179870, 17111424, 17043522, 16976156, 16909321, 16843010, }; -static int pix_sum_c(UINT8 * pix, int line_size) +/* Input permutation for the simple_idct_mmx */ +static const uint8_t simple_mmx_permutation[64]={ + 0x00, 0x08, 0x04, 0x09, 0x01, 0x0C, 0x05, 0x0D, + 0x10, 0x18, 0x14, 0x19, 0x11, 0x1C, 0x15, 0x1D, + 0x20, 0x28, 0x24, 0x29, 0x21, 0x2C, 0x25, 0x2D, + 0x12, 0x1A, 0x16, 0x1B, 0x13, 0x1E, 0x17, 0x1F, + 0x02, 0x0A, 0x06, 0x0B, 0x03, 0x0E, 0x07, 0x0F, + 0x30, 0x38, 0x34, 0x39, 0x31, 0x3C, 0x35, 0x3D, + 0x22, 0x2A, 0x26, 0x2B, 0x23, 0x2E, 0x27, 0x2F, + 0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F, +}; + +static const uint8_t idct_sse2_row_perm[8] = {0, 4, 1, 5, 2, 6, 3, 7}; + +void ff_init_scantable(uint8_t *permutation, ScanTable *st, const uint8_t *src_scantable){ + int i; + int end; + + st->scantable= src_scantable; + + for(i=0; i<64; i++){ + int j; + j = src_scantable[i]; + st->permutated[i] = permutation[j]; +#ifdef ARCH_POWERPC + st->inverse[j] = i; +#endif + } + + end=-1; + for(i=0; i<64; i++){ + int j; + j = st->permutated[i]; + if(j>end) end=j; + st->raster_end[i]= end; + } +} + +static int pix_sum_c(uint8_t * pix, int line_size) { int s, i, j; s = 0; for (i = 0; i < 16; i++) { - for (j = 0; j < 16; j += 8) { - s += pix[0]; - s += pix[1]; - s += pix[2]; - s += pix[3]; - s += pix[4]; - s += pix[5]; - s += pix[6]; - s += pix[7]; - pix += 8; - } - pix += line_size - 16; + for (j = 0; j < 16; j += 8) { + s += pix[0]; + s += pix[1]; + s += pix[2]; + s += pix[3]; + s += pix[4]; + s += pix[5]; + s += pix[6]; + s += pix[7]; + pix += 8; + } + pix += line_size - 16; } return s; } -static int pix_norm1_c(UINT8 * pix, int line_size) +static int pix_norm1_c(uint8_t * pix, int line_size) { int s, i, j; - UINT32 *sq = squareTbl + 256; + uint32_t *sq = ff_squareTbl + 256; s = 0; for (i = 0; i < 16; i++) { - for (j = 0; j < 16; j += 8) { - s += sq[pix[0]]; - s += sq[pix[1]]; - s += sq[pix[2]]; - s += sq[pix[3]]; - s += sq[pix[4]]; - s += sq[pix[5]]; - s += sq[pix[6]]; - s += sq[pix[7]]; - pix += 8; - } - pix += line_size - 16; + for (j = 0; j < 16; j += 8) { +#if 0 + s += sq[pix[0]]; + s += sq[pix[1]]; + s += sq[pix[2]]; + s += sq[pix[3]]; + s += sq[pix[4]]; + s += sq[pix[5]]; + s += sq[pix[6]]; + s += sq[pix[7]]; +#else +#if LONG_MAX > 2147483647 + register uint64_t x=*(uint64_t*)pix; + s += sq[x&0xff]; + s += sq[(x>>8)&0xff]; + s += sq[(x>>16)&0xff]; + s += sq[(x>>24)&0xff]; + s += sq[(x>>32)&0xff]; + s += sq[(x>>40)&0xff]; + s += sq[(x>>48)&0xff]; + s += sq[(x>>56)&0xff]; +#else + register uint32_t x=*(uint32_t*)pix; + s += sq[x&0xff]; + s += sq[(x>>8)&0xff]; + s += sq[(x>>16)&0xff]; + s += sq[(x>>24)&0xff]; + x=*(uint32_t*)(pix+4); + s += sq[x&0xff]; + s += sq[(x>>8)&0xff]; + s += sq[(x>>16)&0xff]; + s += sq[(x>>24)&0xff]; +#endif +#endif + pix += 8; + } + pix += line_size - 16; } return s; } +static void bswap_buf(uint32_t *dst, const uint32_t *src, int w){ + int i; + + for(i=0; i+8<=w; i+=8){ + dst[i+0]= bswap_32(src[i+0]); + dst[i+1]= bswap_32(src[i+1]); + dst[i+2]= bswap_32(src[i+2]); + dst[i+3]= bswap_32(src[i+3]); + dst[i+4]= bswap_32(src[i+4]); + dst[i+5]= bswap_32(src[i+5]); + dst[i+6]= bswap_32(src[i+6]); + dst[i+7]= bswap_32(src[i+7]); + } + for(;i>(dec_count-level); + int sx= (ori&1) ? size : 0; + int stride= 32<<(dec_count-level); + int sy= (ori&2) ? stride>>1 : 0; + + for(i=0; i=0); + return s>>9; +} + +static int w53_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){ + return w_c(v, pix1, pix2, line_size, 8, h, 1); +} + +static int w97_8_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){ + return w_c(v, pix1, pix2, line_size, 8, h, 0); +} + +static int w53_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){ + return w_c(v, pix1, pix2, line_size, 16, h, 1); +} + +static int w97_16_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){ + return w_c(v, pix1, pix2, line_size, 16, h, 0); +} + +int w53_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){ + return w_c(v, pix1, pix2, line_size, 32, h, 1); +} + +int w97_32_c(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h){ + return w_c(v, pix1, pix2, line_size, 32, h, 0); +} +#endif + +/* draw the edges of width 'w' of an image of size width, height */ +//FIXME check that this is ok for mpeg4 interlaced +static void draw_edges_c(uint8_t *buf, int wrap, int width, int height, int w) +{ + uint8_t *ptr, *last_line; + int i; + + last_line = buf + (height - 1) * wrap; + for(i=0;i= h){ + src+= (h-1-src_y)*linesize; + src_y=h-1; + }else if(src_y<=-block_h){ + src+= (1-block_h-src_y)*linesize; + src_y=1-block_h; + } + if(src_x>= w){ + src+= (w-1-src_x); + src_x=w-1; + }else if(src_x<=-block_w){ + src+= (1-block_w-src_x); + src_x=1-block_w; + } + + start_y= FFMAX(0, -src_y); + start_x= FFMAX(0, -src_x); + end_y= FFMIN(block_h, h-src_y); + end_x= FFMIN(block_w, w-src_x); + + // copy existing part + for(y=start_y; y 127) + *pixels = 255; + else + *pixels = (uint8_t)(*block + 128); + block++; + pixels++; + } + pixels += (line_size - 8); + } +} + +static void add_pixels_clamped_c(const DCTELEM *block, uint8_t *restrict pixels, int line_size) { int i; - UINT8 *cm = cropTbl + MAX_NEG_CROP; - + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + /* read the pixels */ for(i=0;i<8;i++) { pixels[0] = cm[pixels[0] + block[0]]; @@ -273,6 +666,77 @@ static void add_pixels_clamped_c(const DCTELEM *block, UINT8 *restrict pixels, block += 8; } } + +static void add_pixels_clamped4_c(const DCTELEM *block, uint8_t *restrict pixels, + int line_size) +{ + int i; + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + /* read the pixels */ + for(i=0;i<4;i++) { + pixels[0] = cm[pixels[0] + block[0]]; + pixels[1] = cm[pixels[1] + block[1]]; + pixels[2] = cm[pixels[2] + block[2]]; + pixels[3] = cm[pixels[3] + block[3]]; + pixels += line_size; + block += 8; + } +} + +static void add_pixels_clamped2_c(const DCTELEM *block, uint8_t *restrict pixels, + int line_size) +{ + int i; + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + /* read the pixels */ + for(i=0;i<2;i++) { + pixels[0] = cm[pixels[0] + block[0]]; + pixels[1] = cm[pixels[1] + block[1]]; + pixels += line_size; + block += 8; + } +} + +static void add_pixels8_c(uint8_t *restrict pixels, DCTELEM *block, int line_size) +{ + int i; + for(i=0;i<8;i++) { + pixels[0] += block[0]; + pixels[1] += block[1]; + pixels[2] += block[2]; + pixels[3] += block[3]; + pixels[4] += block[4]; + pixels[5] += block[5]; + pixels[6] += block[6]; + pixels[7] += block[7]; + pixels += line_size; + block += 8; + } +} + +static void add_pixels4_c(uint8_t *restrict pixels, DCTELEM *block, int line_size) +{ + int i; + for(i=0;i<4;i++) { + pixels[0] += block[0]; + pixels[1] += block[1]; + pixels[2] += block[2]; + pixels[3] += block[3]; + pixels += line_size; + block += 4; + } +} + +static int sum_abs_dctelem_c(DCTELEM *block) +{ + int sum=0, i; + for(i=0; i<64; i++) + sum+= FFABS(block[i]); + return sum; +} + #if 0 #define PIXOP2(OPNAME, OP) \ @@ -280,7 +744,7 @@ static void OPNAME ## _pixels(uint8_t *block, const uint8_t *pixels, int line_si {\ int i;\ for(i=0; i>1));\ pixels+=line_size;\ block +=line_size;\ @@ -302,8 +766,8 @@ static void OPNAME ## _pixels_x2_c(uint8_t *block, const uint8_t *pixels, int li {\ int i;\ for(i=0; i>1));\ pixels+=line_size;\ block +=line_size;\ @@ -314,8 +778,8 @@ static void OPNAME ## _no_rnd_pixels_y2_c(uint8_t *block, const uint8_t *pixels, {\ int i;\ for(i=0; i>1));\ pixels+=line_size;\ block +=line_size;\ @@ -326,8 +790,8 @@ static void OPNAME ## _pixels_y2_c(uint8_t *block, const uint8_t *pixels, int li {\ int i;\ for(i=0; i>1));\ pixels+=line_size;\ block +=line_size;\ @@ -337,8 +801,8 @@ static void OPNAME ## _pixels_y2_c(uint8_t *block, const uint8_t *pixels, int li static void OPNAME ## _pixels_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ {\ int i;\ - const uint64_t a= LD64(pixels );\ - const uint64_t b= LD64(pixels+1);\ + const uint64_t a= AV_RN64(pixels );\ + const uint64_t b= AV_RN64(pixels+1);\ uint64_t l0= (a&0x0303030303030303ULL)\ + (b&0x0303030303030303ULL)\ + 0x0202020202020202ULL;\ @@ -348,8 +812,8 @@ static void OPNAME ## _pixels_xy2_c(uint8_t *block, const uint8_t *pixels, int l \ pixels+=line_size;\ for(i=0; i>2)\ @@ -357,8 +821,8 @@ static void OPNAME ## _pixels_xy2_c(uint8_t *block, const uint8_t *pixels, int l OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\ pixels+=line_size;\ block +=line_size;\ - a= LD64(pixels );\ - b= LD64(pixels+1);\ + a= AV_RN64(pixels );\ + b= AV_RN64(pixels+1);\ l0= (a&0x0303030303030303ULL)\ + (b&0x0303030303030303ULL)\ + 0x0202020202020202ULL;\ @@ -373,8 +837,8 @@ static void OPNAME ## _pixels_xy2_c(uint8_t *block, const uint8_t *pixels, int l static void OPNAME ## _no_rnd_pixels_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ {\ int i;\ - const uint64_t a= LD64(pixels );\ - const uint64_t b= LD64(pixels+1);\ + const uint64_t a= AV_RN64(pixels );\ + const uint64_t b= AV_RN64(pixels+1);\ uint64_t l0= (a&0x0303030303030303ULL)\ + (b&0x0303030303030303ULL)\ + 0x0101010101010101ULL;\ @@ -384,8 +848,8 @@ static void OPNAME ## _no_rnd_pixels_xy2_c(uint8_t *block, const uint8_t *pixels \ pixels+=line_size;\ for(i=0; i>2)\ @@ -393,8 +857,8 @@ static void OPNAME ## _no_rnd_pixels_xy2_c(uint8_t *block, const uint8_t *pixels OP(*((uint64_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0F0F0F0F0FULL));\ pixels+=line_size;\ block +=line_size;\ - a= LD64(pixels );\ - b= LD64(pixels+1);\ + a= AV_RN64(pixels );\ + b= AV_RN64(pixels+1);\ l0= (a&0x0303030303030303ULL)\ + (b&0x0303030303030303ULL)\ + 0x0101010101010101ULL;\ @@ -418,11 +882,27 @@ CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_xy2_c, OPNAME ## _no_rnd_pixels_xy2_c, #else // 64 bit variant #define PIXOP2(OPNAME, OP) \ +static void OPNAME ## _pixels2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ + int i;\ + for(i=0; i>1));\ - a= LD32(&src1[i*src_stride1+4]);\ - b= LD32(&src2[i*src_stride2+4]);\ - OP(*((uint32_t*)&dst[i*dst_stride+4]), (a&b) + (((a^b)&0xFEFEFEFEUL)>>1));\ + a= AV_RN32(&src1[i*src_stride1 ]);\ + b= AV_RN32(&src2[i*src_stride2 ]);\ + OP(*((uint32_t*)&dst[i*dst_stride ]), no_rnd_avg32(a, b));\ + a= AV_RN32(&src1[i*src_stride1+4]);\ + b= AV_RN32(&src2[i*src_stride2+4]);\ + OP(*((uint32_t*)&dst[i*dst_stride+4]), no_rnd_avg32(a, b));\ }\ }\ \ @@ -450,12 +930,34 @@ static inline void OPNAME ## _pixels8_l2(uint8_t *dst, const uint8_t *src1, cons int i;\ for(i=0; i>1));\ - a= LD32(&src1[i*src_stride1+4]);\ - b= LD32(&src2[i*src_stride2+4]);\ - OP(*((uint32_t*)&dst[i*dst_stride+4]), (a|b) - (((a^b)&0xFEFEFEFEUL)>>1));\ + a= AV_RN32(&src1[i*src_stride1 ]);\ + b= AV_RN32(&src2[i*src_stride2 ]);\ + OP(*((uint32_t*)&dst[i*dst_stride ]), rnd_avg32(a, b));\ + a= AV_RN32(&src1[i*src_stride1+4]);\ + b= AV_RN32(&src2[i*src_stride2+4]);\ + OP(*((uint32_t*)&dst[i*dst_stride+4]), rnd_avg32(a, b));\ + }\ +}\ +\ +static inline void OPNAME ## _pixels4_l2(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int dst_stride, \ + int src_stride1, int src_stride2, int h){\ + int i;\ + for(i=0; i>2)\ + ((d&0xFCFCFCFCUL)>>2);\ OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ - a= LD32(&src1[i*src_stride1+4]);\ - b= LD32(&src2[i*src_stride2+4]);\ - c= LD32(&src3[i*src_stride3+4]);\ - d= LD32(&src4[i*src_stride4+4]);\ + a= AV_RN32(&src1[i*src_stride1+4]);\ + b= AV_RN32(&src2[i*src_stride2+4]);\ + c= AV_RN32(&src3[i*src_stride3+4]);\ + d= AV_RN32(&src4[i*src_stride4+4]);\ l0= (a&0x03030303UL)\ + (b&0x03030303UL)\ + 0x02020202UL;\ @@ -522,15 +1024,32 @@ static inline void OPNAME ## _pixels8_l4(uint8_t *dst, const uint8_t *src1, uint OP(*((uint32_t*)&dst[i*dst_stride+4]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ }\ }\ +\ +static inline void OPNAME ## _pixels4_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ + OPNAME ## _pixels4_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\ +}\ +\ +static inline void OPNAME ## _pixels4_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ + OPNAME ## _pixels4_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\ +}\ +\ +static inline void OPNAME ## _pixels2_x2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ + OPNAME ## _pixels2_l2(block, pixels, pixels+1, line_size, line_size, line_size, h);\ +}\ +\ +static inline void OPNAME ## _pixels2_y2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h){\ + OPNAME ## _pixels2_l2(block, pixels, pixels+line_size, line_size, line_size, line_size, h);\ +}\ +\ static inline void OPNAME ## _no_rnd_pixels8_l4(uint8_t *dst, const uint8_t *src1, uint8_t *src2, uint8_t *src3, uint8_t *src4,\ int dst_stride, int src_stride1, int src_stride2,int src_stride3,int src_stride4, int h){\ int i;\ for(i=0; i>2)\ + ((d&0xFCFCFCFCUL)>>2);\ OP(*((uint32_t*)&dst[i*dst_stride]), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ - a= LD32(&src1[i*src_stride1+4]);\ - b= LD32(&src2[i*src_stride2+4]);\ - c= LD32(&src3[i*src_stride3+4]);\ - d= LD32(&src4[i*src_stride4+4]);\ + a= AV_RN32(&src1[i*src_stride1+4]);\ + b= AV_RN32(&src2[i*src_stride2+4]);\ + c= AV_RN32(&src3[i*src_stride3+4]);\ + d= AV_RN32(&src4[i*src_stride4+4]);\ l0= (a&0x03030303UL)\ + (b&0x03030303UL)\ + 0x01010101UL;\ @@ -568,33 +1087,102 @@ static inline void OPNAME ## _no_rnd_pixels16_l4(uint8_t *dst, const uint8_t *sr OPNAME ## _no_rnd_pixels8_l4(dst+8, src1+8, src2+8, src3+8, src4+8, dst_stride, src_stride1, src_stride2, src_stride3, src_stride4, h);\ }\ \ -static inline void OPNAME ## _pixels8_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ +static inline void OPNAME ## _pixels2_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ {\ - int j;\ - for(j=0; j<2; j++){\ - int i;\ - const uint32_t a= LD32(pixels );\ - const uint32_t b= LD32(pixels+1);\ - uint32_t l0= (a&0x03030303UL)\ - + (b&0x03030303UL)\ - + 0x02020202UL;\ - uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\ - + ((b&0xFCFCFCFCUL)>>2);\ - uint32_t l1,h1;\ + int i, a0, b0, a1, b1;\ + a0= pixels[0];\ + b0= pixels[1] + 2;\ + a0 += b0;\ + b0 += pixels[2];\ \ pixels+=line_size;\ for(i=0; i>2)\ - + ((b&0xFCFCFCFCUL)>>2);\ - OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ - pixels+=line_size;\ + a1= pixels[0];\ + b1= pixels[1];\ + a1 += b1;\ + b1 += pixels[2];\ +\ + block[0]= (a1+a0)>>2; /* FIXME non put */\ + block[1]= (b1+b0)>>2;\ +\ + pixels+=line_size;\ + block +=line_size;\ +\ + a0= pixels[0];\ + b0= pixels[1] + 2;\ + a0 += b0;\ + b0 += pixels[2];\ +\ + block[0]= (a1+a0)>>2;\ + block[1]= (b1+b0)>>2;\ + pixels+=line_size;\ + block +=line_size;\ + }\ +}\ +\ +static inline void OPNAME ## _pixels4_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ +{\ + int i;\ + const uint32_t a= AV_RN32(pixels );\ + const uint32_t b= AV_RN32(pixels+1);\ + uint32_t l0= (a&0x03030303UL)\ + + (b&0x03030303UL)\ + + 0x02020202UL;\ + uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\ + + ((b&0xFCFCFCFCUL)>>2);\ + uint32_t l1,h1;\ +\ + pixels+=line_size;\ + for(i=0; i>2)\ + + ((b&0xFCFCFCFCUL)>>2);\ + OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ + pixels+=line_size;\ + block +=line_size;\ + a= AV_RN32(pixels );\ + b= AV_RN32(pixels+1);\ + l0= (a&0x03030303UL)\ + + (b&0x03030303UL)\ + + 0x02020202UL;\ + h0= ((a&0xFCFCFCFCUL)>>2)\ + + ((b&0xFCFCFCFCUL)>>2);\ + OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ + pixels+=line_size;\ + block +=line_size;\ + }\ +}\ +\ +static inline void OPNAME ## _pixels8_xy2_c(uint8_t *block, const uint8_t *pixels, int line_size, int h)\ +{\ + int j;\ + for(j=0; j<2; j++){\ + int i;\ + const uint32_t a= AV_RN32(pixels );\ + const uint32_t b= AV_RN32(pixels+1);\ + uint32_t l0= (a&0x03030303UL)\ + + (b&0x03030303UL)\ + + 0x02020202UL;\ + uint32_t h0= ((a&0xFCFCFCFCUL)>>2)\ + + ((b&0xFCFCFCFCUL)>>2);\ + uint32_t l1,h1;\ +\ + pixels+=line_size;\ + for(i=0; i>2)\ + + ((b&0xFCFCFCFCUL)>>2);\ + OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ + pixels+=line_size;\ block +=line_size;\ - a= LD32(pixels );\ - b= LD32(pixels+1);\ + a= AV_RN32(pixels );\ + b= AV_RN32(pixels+1);\ l0= (a&0x03030303UL)\ + (b&0x03030303UL)\ + 0x02020202UL;\ @@ -614,8 +1202,8 @@ static inline void OPNAME ## _no_rnd_pixels8_xy2_c(uint8_t *block, const uint8_t int j;\ for(j=0; j<2; j++){\ int i;\ - const uint32_t a= LD32(pixels );\ - const uint32_t b= LD32(pixels+1);\ + const uint32_t a= AV_RN32(pixels );\ + const uint32_t b= AV_RN32(pixels+1);\ uint32_t l0= (a&0x03030303UL)\ + (b&0x03030303UL)\ + 0x01010101UL;\ @@ -625,8 +1213,8 @@ static inline void OPNAME ## _no_rnd_pixels8_xy2_c(uint8_t *block, const uint8_t \ pixels+=line_size;\ for(i=0; i>2)\ @@ -634,8 +1222,8 @@ static inline void OPNAME ## _no_rnd_pixels8_xy2_c(uint8_t *block, const uint8_t OP(*((uint32_t*)block), h0+h1+(((l0+l1)>>2)&0x0F0F0F0FUL));\ pixels+=line_size;\ block +=line_size;\ - a= LD32(pixels );\ - b= LD32(pixels+1);\ + a= AV_RN32(pixels );\ + b= AV_RN32(pixels+1);\ l0= (a&0x03030303UL)\ + (b&0x03030303UL)\ + 0x01010101UL;\ @@ -659,7 +1247,7 @@ CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_x2_c , OPNAME ## _no_rnd_pixels8_x2_c CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_y2_c , OPNAME ## _no_rnd_pixels8_y2_c , 8)\ CALL_2X_PIXELS(OPNAME ## _no_rnd_pixels16_xy2_c, OPNAME ## _no_rnd_pixels8_xy2_c, 8)\ -#define op_avg(a, b) a = ( ((a)|(b)) - ((((a)^(b))&0xFEFEFEFEUL)>>1) ) +#define op_avg(a, b) a = rnd_avg32(a, b) #endif #define op_put(a, b) a = b @@ -671,8 +1259,15 @@ PIXOP2(put, op_put) #define avg2(a,b) ((a+b+1)>>1) #define avg4(a,b,c,d) ((a+b+c+d+2)>>2) +static void put_no_rnd_pixels16_l2_c(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h){ + put_no_rnd_pixels16_l2(dst, a, b, stride, stride, stride, h); +} + +static void put_no_rnd_pixels8_l2_c(uint8_t *dst, const uint8_t *a, const uint8_t *b, int stride, int h){ + put_no_rnd_pixels8_l2(dst, a, b, stride, stride, stride, h); +} -static void gmc1_c(UINT8 *dst, UINT8 *src, int stride, int h, int x16, int y16, int rounder) +static void gmc1_c(uint8_t *dst, uint8_t *src, int stride, int h, int x16, int y16, int rounder) { const int A=(16-x16)*(16-y16); const int B=( x16)*(16-y16); @@ -695,12 +1290,12 @@ static void gmc1_c(UINT8 *dst, UINT8 *src, int stride, int h, int x16, int y16, } } -static void gmc_c(UINT8 *dst, UINT8 *src, int stride, int h, int ox, int oy, +void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy, int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height) { int y, vx, vy; const int s= 1<>=shift; src_y>>=shift; - + if((unsigned)src_x < width){ if((unsigned)src_y < height){ index= src_x + src_y*stride; @@ -728,23 +1323,23 @@ static void gmc_c(UINT8 *dst, UINT8 *src, int stride, int h, int ox, int oy, + src[index+stride+1]* frac_x )* frac_y + r)>>(shift*2); }else{ - index= src_x + clip(src_y, 0, height)*stride; - dst[y*stride + x]= ( ( src[index ]*(s-frac_x) + index= src_x + av_clip(src_y, 0, height)*stride; + dst[y*stride + x]= ( ( src[index ]*(s-frac_x) + src[index +1]* frac_x )*s + r)>>(shift*2); } }else{ if((unsigned)src_y < height){ - index= clip(src_x, 0, width) + src_y*stride; - dst[y*stride + x]= ( ( src[index ]*(s-frac_y) + index= av_clip(src_x, 0, width) + src_y*stride; + dst[y*stride + x]= ( ( src[index ]*(s-frac_y) + src[index+stride ]* frac_y )*s + r)>>(shift*2); }else{ - index= clip(src_x, 0, width) + clip(src_y, 0, height)*stride; + index= av_clip(src_x, 0, width) + av_clip(src_y, 0, height)*stride; dst[y*stride + x]= src[index ]; } } - + vx+= dxx; vy+= dyx; } @@ -753,38 +1348,357 @@ static void gmc_c(UINT8 *dst, UINT8 *src, int stride, int h, int ox, int oy, } } -static inline void copy_block17(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h) -{ - int i; - for(i=0; i> 11; + } + src += stride; + dst += stride; + } +} + +static inline void put_tpel_pixels_mc20_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (683*(src[j] + 2*src[j+1] + 1)) >> 11; + } + src += stride; + dst += stride; + } +} + +static inline void put_tpel_pixels_mc01_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (683*(2*src[j] + src[j+stride] + 1)) >> 11; + } + src += stride; + dst += stride; + } +} + +static inline void put_tpel_pixels_mc11_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (2731*(4*src[j] + 3*src[j+1] + 3*src[j+stride] + 2*src[j+stride+1] + 6)) >> 15; + } + src += stride; + dst += stride; + } +} + +static inline void put_tpel_pixels_mc12_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (2731*(3*src[j] + 2*src[j+1] + 4*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15; + } + src += stride; + dst += stride; + } +} + +static inline void put_tpel_pixels_mc02_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (683*(src[j] + 2*src[j+stride] + 1)) >> 11; + } + src += stride; + dst += stride; + } +} + +static inline void put_tpel_pixels_mc21_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (2731*(3*src[j] + 4*src[j+1] + 2*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15; + } + src += stride; + dst += stride; + } +} + +static inline void put_tpel_pixels_mc22_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (2731*(2*src[j] + 3*src[j+1] + 3*src[j+stride] + 4*src[j+stride+1] + 6)) >> 15; + } + src += stride; + dst += stride; + } +} + +static inline void avg_tpel_pixels_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + switch(width){ + case 2: avg_pixels2_c (dst, src, stride, height); break; + case 4: avg_pixels4_c (dst, src, stride, height); break; + case 8: avg_pixels8_c (dst, src, stride, height); break; + case 16:avg_pixels16_c(dst, src, stride, height); break; + } +} + +static inline void avg_tpel_pixels_mc10_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (dst[j] + ((683*(2*src[j] + src[j+1] + 1)) >> 11) + 1) >> 1; + } + src += stride; + dst += stride; + } +} + +static inline void avg_tpel_pixels_mc20_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (dst[j] + ((683*(src[j] + 2*src[j+1] + 1)) >> 11) + 1) >> 1; + } + src += stride; + dst += stride; + } +} + +static inline void avg_tpel_pixels_mc01_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (dst[j] + ((683*(2*src[j] + src[j+stride] + 1)) >> 11) + 1) >> 1; + } + src += stride; + dst += stride; + } +} + +static inline void avg_tpel_pixels_mc11_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (dst[j] + ((2731*(4*src[j] + 3*src[j+1] + 3*src[j+stride] + 2*src[j+stride+1] + 6)) >> 15) + 1) >> 1; + } + src += stride; + dst += stride; + } +} + +static inline void avg_tpel_pixels_mc12_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (dst[j] + ((2731*(3*src[j] + 2*src[j+1] + 4*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15) + 1) >> 1; + } + src += stride; + dst += stride; + } +} + +static inline void avg_tpel_pixels_mc02_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (dst[j] + ((683*(src[j] + 2*src[j+stride] + 1)) >> 11) + 1) >> 1; + } + src += stride; + dst += stride; + } +} + +static inline void avg_tpel_pixels_mc21_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (dst[j] + ((2731*(3*src[j] + 4*src[j+1] + 2*src[j+stride] + 3*src[j+stride+1] + 6)) >> 15) + 1) >> 1; + } + src += stride; + dst += stride; + } +} + +static inline void avg_tpel_pixels_mc22_c(uint8_t *dst, const uint8_t *src, int stride, int width, int height){ + int i,j; + for (i=0; i < height; i++) { + for (j=0; j < width; j++) { + dst[j] = (dst[j] + ((2731*(2*src[j] + 3*src[j+1] + 3*src[j+stride] + 4*src[j+stride+1] + 6)) >> 15) + 1) >> 1; + } + src += stride; + dst += stride; + } +} +#if 0 +#define TPEL_WIDTH(width)\ +static void put_tpel_pixels ## width ## _mc00_c(uint8_t *dst, const uint8_t *src, int stride, int height){\ + void put_tpel_pixels_mc00_c(dst, src, stride, width, height);}\ +static void put_tpel_pixels ## width ## _mc10_c(uint8_t *dst, const uint8_t *src, int stride, int height){\ + void put_tpel_pixels_mc10_c(dst, src, stride, width, height);}\ +static void put_tpel_pixels ## width ## _mc20_c(uint8_t *dst, const uint8_t *src, int stride, int height){\ + void put_tpel_pixels_mc20_c(dst, src, stride, width, height);}\ +static void put_tpel_pixels ## width ## _mc01_c(uint8_t *dst, const uint8_t *src, int stride, int height){\ + void put_tpel_pixels_mc01_c(dst, src, stride, width, height);}\ +static void put_tpel_pixels ## width ## _mc11_c(uint8_t *dst, const uint8_t *src, int stride, int height){\ + void put_tpel_pixels_mc11_c(dst, src, stride, width, height);}\ +static void put_tpel_pixels ## width ## _mc21_c(uint8_t *dst, const uint8_t *src, int stride, int height){\ + void put_tpel_pixels_mc21_c(dst, src, stride, width, height);}\ +static void put_tpel_pixels ## width ## _mc02_c(uint8_t *dst, const uint8_t *src, int stride, int height){\ + void put_tpel_pixels_mc02_c(dst, src, stride, width, height);}\ +static void put_tpel_pixels ## width ## _mc12_c(uint8_t *dst, const uint8_t *src, int stride, int height){\ + void put_tpel_pixels_mc12_c(dst, src, stride, width, height);}\ +static void put_tpel_pixels ## width ## _mc22_c(uint8_t *dst, const uint8_t *src, int stride, int height){\ + void put_tpel_pixels_mc22_c(dst, src, stride, width, height);} +#endif + +#define H264_CHROMA_MC(OPNAME, OP)\ +static void OPNAME ## h264_chroma_mc2_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\ + const int A=(8-x)*(8-y);\ + const int B=( x)*(8-y);\ + const int C=(8-x)*( y);\ + const int D=( x)*( y);\ + int i;\ + \ + assert(x<8 && y<8 && x>=0 && y>=0);\ +\ + if(D){\ + for(i=0; i=0 && y>=0);\ +\ + if(D){\ + for(i=0; i=0 && y>=0);\ +\ + if(D){\ + for(i=0; i>6)+1)>>1) +#define op_put(a, b) a = (((b) + 32)>>6) + +H264_CHROMA_MC(put_ , op_put) +H264_CHROMA_MC(avg_ , op_avg) +#undef op_avg +#undef op_put + +static void put_no_rnd_h264_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){ + const int A=(8-x)*(8-y); + const int B=( x)*(8-y); + const int C=(8-x)*( y); + const int D=( x)*( y); int i; + + assert(x<8 && y<8 && x>=0 && y>=0); + for(i=0; i> 6; + dst[1] = (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + 32 - 4) >> 6; + dst[2] = (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + 32 - 4) >> 6; + dst[3] = (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + 32 - 4) >> 6; + dst[4] = (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + 32 - 4) >> 6; + dst[5] = (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + 32 - 4) >> 6; + dst[6] = (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + 32 - 4) >> 6; + dst[7] = (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + 32 - 4) >> 6; + dst+= stride; + src+= stride; } } - #define QPEL_MC(r, OPNAME, RND, OP) \ -static void OPNAME ## mpeg4_qpel8_h_lowpass(UINT8 *dst, UINT8 *src, int dstStride, int srcStride, int h){\ - UINT8 *cm = cropTbl + MAX_NEG_CROP;\ +static void OPNAME ## mpeg4_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\ + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\ int i;\ for(i=0; i>4]; - dst[1]= cm[(9*(src[1] + src[2]) - (src[ 0] + src[3]) + 8)>>4]; - dst[2]= cm[(9*(src[2] + src[3]) - (src[ 1] + src[4]) + 8)>>4]; - dst[3]= cm[(9*(src[3] + src[4]) - (src[ 2] + src[5]) + 8)>>4]; - dst[4]= cm[(9*(src[4] + src[5]) - (src[ 3] + src[6]) + 8)>>4]; - dst[5]= cm[(9*(src[5] + src[6]) - (src[ 4] + src[7]) + 8)>>4]; - dst[6]= cm[(9*(src[6] + src[7]) - (src[ 5] + src[8]) + 8)>>4]; - dst[7]= cm[(9*(src[7] + src[8]) - (src[ 6] + src[9]) + 8)>>4]; +#define H264_MC(OPNAME, SIZE) \ +static void OPNAME ## h264_qpel ## SIZE ## _mc00_c (uint8_t *dst, uint8_t *src, int stride){\ + OPNAME ## pixels ## SIZE ## _c(dst, src, stride, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t half[SIZE*SIZE];\ + put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\ + OPNAME ## pixels ## SIZE ## _l2(dst, src, half, stride, stride, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc20_c(uint8_t *dst, uint8_t *src, int stride){\ + OPNAME ## h264_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t half[SIZE*SIZE];\ + put_h264_qpel ## SIZE ## _h_lowpass(half, src, SIZE, stride);\ + OPNAME ## pixels ## SIZE ## _l2(dst, src+1, half, stride, stride, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t full[SIZE*(SIZE+5)];\ + uint8_t * const full_mid= full + SIZE*2;\ + uint8_t half[SIZE*SIZE];\ + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ + put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\ + OPNAME ## pixels ## SIZE ## _l2(dst, full_mid, half, stride, SIZE, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc02_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t full[SIZE*(SIZE+5)];\ + uint8_t * const full_mid= full + SIZE*2;\ + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ + OPNAME ## h264_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t full[SIZE*(SIZE+5)];\ + uint8_t * const full_mid= full + SIZE*2;\ + uint8_t half[SIZE*SIZE];\ + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ + put_h264_qpel ## SIZE ## _v_lowpass(half, full_mid, SIZE, SIZE);\ + OPNAME ## pixels ## SIZE ## _l2(dst, full_mid+SIZE, half, stride, SIZE, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t full[SIZE*(SIZE+5)];\ + uint8_t * const full_mid= full + SIZE*2;\ + uint8_t halfH[SIZE*SIZE];\ + uint8_t halfV[SIZE*SIZE];\ + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\ + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t full[SIZE*(SIZE+5)];\ + uint8_t * const full_mid= full + SIZE*2;\ + uint8_t halfH[SIZE*SIZE];\ + uint8_t halfV[SIZE*SIZE];\ + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\ + copy_block ## SIZE (full, src - stride*2 + 1, SIZE, stride, SIZE + 5);\ + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t full[SIZE*(SIZE+5)];\ + uint8_t * const full_mid= full + SIZE*2;\ + uint8_t halfH[SIZE*SIZE];\ + uint8_t halfV[SIZE*SIZE];\ + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\ + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc33_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t full[SIZE*(SIZE+5)];\ + uint8_t * const full_mid= full + SIZE*2;\ + uint8_t halfH[SIZE*SIZE];\ + uint8_t halfV[SIZE*SIZE];\ + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\ + copy_block ## SIZE (full, src - stride*2 + 1, SIZE, stride, SIZE + 5);\ + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfV, stride, SIZE, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\ + int16_t tmp[SIZE*(SIZE+5)];\ + OPNAME ## h264_qpel ## SIZE ## _hv_lowpass(dst, tmp, src, stride, SIZE, stride);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\ + int16_t tmp[SIZE*(SIZE+5)];\ + uint8_t halfH[SIZE*SIZE];\ + uint8_t halfHV[SIZE*SIZE];\ + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src, SIZE, stride);\ + put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\ + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\ + int16_t tmp[SIZE*(SIZE+5)];\ + uint8_t halfH[SIZE*SIZE];\ + uint8_t halfHV[SIZE*SIZE];\ + put_h264_qpel ## SIZE ## _h_lowpass(halfH, src + stride, SIZE, stride);\ + put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\ + OPNAME ## pixels ## SIZE ## _l2(dst, halfH, halfHV, stride, SIZE, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t full[SIZE*(SIZE+5)];\ + uint8_t * const full_mid= full + SIZE*2;\ + int16_t tmp[SIZE*(SIZE+5)];\ + uint8_t halfV[SIZE*SIZE];\ + uint8_t halfHV[SIZE*SIZE];\ + copy_block ## SIZE (full, src - stride*2, SIZE, stride, SIZE + 5);\ + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ + put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\ + OPNAME ## pixels ## SIZE ## _l2(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\ +}\ +\ +static void OPNAME ## h264_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\ + uint8_t full[SIZE*(SIZE+5)];\ + uint8_t * const full_mid= full + SIZE*2;\ + int16_t tmp[SIZE*(SIZE+5)];\ + uint8_t halfV[SIZE*SIZE];\ + uint8_t halfHV[SIZE*SIZE];\ + copy_block ## SIZE (full, src - stride*2 + 1, SIZE, stride, SIZE + 5);\ + put_h264_qpel ## SIZE ## _v_lowpass(halfV, full_mid, SIZE, SIZE);\ + put_h264_qpel ## SIZE ## _hv_lowpass(halfHV, tmp, src, SIZE, SIZE, stride);\ + OPNAME ## pixels ## SIZE ## _l2(dst, halfV, halfHV, stride, SIZE, SIZE, SIZE);\ +}\ + +#define op_avg(a, b) a = (((a)+cm[((b) + 16)>>5]+1)>>1) +//#define op_avg2(a, b) a = (((a)*w1+cm[((b) + 16)>>5]*w2 + o + 64)>>7) +#define op_put(a, b) a = cm[((b) + 16)>>5] +#define op2_avg(a, b) a = (((a)+cm[((b) + 512)>>10]+1)>>1) +#define op2_put(a, b) a = cm[((b) + 512)>>10] + +H264_LOWPASS(put_ , op_put, op2_put) +H264_LOWPASS(avg_ , op_avg, op2_avg) +H264_MC(put_, 2) +H264_MC(put_, 4) +H264_MC(put_, 8) +H264_MC(put_, 16) +H264_MC(avg_, 4) +H264_MC(avg_, 8) +H264_MC(avg_, 16) + +#undef op_avg +#undef op_put +#undef op2_avg +#undef op2_put +#endif + +#define op_scale1(x) block[x] = av_clip_uint8( (block[x]*weight + offset) >> log2_denom ) +#define op_scale2(x) dst[x] = av_clip_uint8( (src[x]*weights + dst[x]*weightd + offset) >> (log2_denom+1)) +#define H264_WEIGHT(W,H) \ +static void weight_h264_pixels ## W ## x ## H ## _c(uint8_t *block, int stride, int log2_denom, int weight, int offset){ \ + int y; \ + offset <<= log2_denom; \ + if(log2_denom) offset += 1<<(log2_denom-1); \ + for(y=0; y>4]; + dst[1]= cm[(9*(src[1] + src[2]) - (src[ 0] + src[3]) + 8)>>4]; + dst[2]= cm[(9*(src[2] + src[3]) - (src[ 1] + src[4]) + 8)>>4]; + dst[3]= cm[(9*(src[3] + src[4]) - (src[ 2] + src[5]) + 8)>>4]; + dst[4]= cm[(9*(src[4] + src[5]) - (src[ 3] + src[6]) + 8)>>4]; + dst[5]= cm[(9*(src[5] + src[6]) - (src[ 4] + src[7]) + 8)>>4]; + dst[6]= cm[(9*(src[6] + src[7]) - (src[ 5] + src[8]) + 8)>>4]; + dst[7]= cm[(9*(src[7] + src[8]) - (src[ 6] + src[9]) + 8)>>4]; dst+=dstStride; - src+=srcStride; + src+=srcStride; } } +#ifdef CONFIG_CAVS_DECODER +/* AVS specific */ +void ff_cavsdsp_init(DSPContext* c, AVCodecContext *avctx); + +void ff_put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride) { + put_pixels8_c(dst, src, stride, 8); +} +void ff_avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride) { + avg_pixels8_c(dst, src, stride, 8); +} +void ff_put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride) { + put_pixels16_c(dst, src, stride, 16); +} +void ff_avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride) { + avg_pixels16_c(dst, src, stride, 16); +} +#endif /* CONFIG_CAVS_DECODER */ + +#if defined(CONFIG_VC1_DECODER) || defined(CONFIG_WMV3_DECODER) +/* VC-1 specific */ +void ff_vc1dsp_init(DSPContext* c, AVCodecContext *avctx); + +void ff_put_vc1_mspel_mc00_c(uint8_t *dst, uint8_t *src, int stride, int rnd) { + put_pixels8_c(dst, src, stride, 8); +} +#endif /* CONFIG_VC1_DECODER||CONFIG_WMV3_DECODER */ + +void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx); + +/* H264 specific */ +void ff_h264dspenc_init(DSPContext* c, AVCodecContext *avctx); + static void wmv2_mspel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int w){ - uint8_t *cm = cropTbl + MAX_NEG_CROP; + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; int i; for(i=0; i>31); + if(p2&256) p2= ~(p2>>31); + + src[x-1*stride] = p1; + src[x+0*stride] = p2; + + ad1= FFABS(d1)>>1; + + d2= av_clip((p0-p3)/4, -ad1, ad1); + + src[x-2*stride] = p0 - d2; + src[x+ stride] = p3 + d2; + } + } +} + +static void h263_h_loop_filter_c(uint8_t *src, int stride, int qscale){ + if(ENABLE_ANY_H263) { + int y; + const int strength= ff_h263_loop_filter_strength[qscale]; + + for(y=0; y<8; y++){ + int d1, d2, ad1; + int p0= src[y*stride-2]; + int p1= src[y*stride-1]; + int p2= src[y*stride+0]; + int p3= src[y*stride+1]; + int d = (p0 - p3 + 4*(p2 - p1)) / 8; + + if (d<-2*strength) d1= 0; + else if(d<- strength) d1=-2*strength - d; + else if(d< strength) d1= d; + else if(d< 2*strength) d1= 2*strength - d; + else d1= 0; + + p1 += d1; + p2 -= d1; + if(p1&256) p1= ~(p1>>31); + if(p2&256) p2= ~(p2>>31); + + src[y*stride-1] = p1; + src[y*stride+0] = p2; + + ad1= FFABS(d1)>>1; + + d2= av_clip((p0-p3)/4, -ad1, ad1); + + src[y*stride-2] = p0 - d2; + src[y*stride+1] = p3 + d2; + } + } +} + +static void h261_loop_filter_c(uint8_t *src, int stride){ + int x,y,xy,yz; + int temp[64]; + + for(x=0; x<8; x++){ + temp[x ] = 4*src[x ]; + temp[x + 7*8] = 4*src[x + 7*stride]; + } + for(y=1; y<7; y++){ + for(x=0; x<8; x++){ + xy = y * stride + x; + yz = y * 8 + x; + temp[yz] = src[xy - stride] + 2*src[xy] + src[xy + stride]; + } + } -static inline int pix_abs16x16_c(UINT8 *pix1, UINT8 *pix2, int line_size) + for(y=0; y<8; y++){ + src[ y*stride] = (temp[ y*8] + 2)>>2; + src[7+y*stride] = (temp[7+y*8] + 2)>>2; + for(x=1; x<7; x++){ + xy = y * stride + x; + yz = y * 8 + x; + src[xy] = (temp[yz-1] + 2*temp[yz] + temp[yz+1] + 8)>>4; + } + } +} + +static inline void h264_loop_filter_luma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0) +{ + int i, d; + for( i = 0; i < 4; i++ ) { + if( tc0[i] < 0 ) { + pix += 4*ystride; + continue; + } + for( d = 0; d < 4; d++ ) { + const int p0 = pix[-1*xstride]; + const int p1 = pix[-2*xstride]; + const int p2 = pix[-3*xstride]; + const int q0 = pix[0]; + const int q1 = pix[1*xstride]; + const int q2 = pix[2*xstride]; + + if( FFABS( p0 - q0 ) < alpha && + FFABS( p1 - p0 ) < beta && + FFABS( q1 - q0 ) < beta ) { + + int tc = tc0[i]; + int i_delta; + + if( FFABS( p2 - p0 ) < beta ) { + pix[-2*xstride] = p1 + av_clip( (( p2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - p1, -tc0[i], tc0[i] ); + tc++; + } + if( FFABS( q2 - q0 ) < beta ) { + pix[ xstride] = q1 + av_clip( (( q2 + ( ( p0 + q0 + 1 ) >> 1 ) ) >> 1) - q1, -tc0[i], tc0[i] ); + tc++; + } + + i_delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc ); + pix[-xstride] = av_clip_uint8( p0 + i_delta ); /* p0' */ + pix[0] = av_clip_uint8( q0 - i_delta ); /* q0' */ + } + pix += ystride; + } + } +} +static void h264_v_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) +{ + h264_loop_filter_luma_c(pix, stride, 1, alpha, beta, tc0); +} +static void h264_h_loop_filter_luma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) +{ + h264_loop_filter_luma_c(pix, 1, stride, alpha, beta, tc0); +} + +static inline void h264_loop_filter_chroma_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta, int8_t *tc0) +{ + int i, d; + for( i = 0; i < 4; i++ ) { + const int tc = tc0[i]; + if( tc <= 0 ) { + pix += 2*ystride; + continue; + } + for( d = 0; d < 2; d++ ) { + const int p0 = pix[-1*xstride]; + const int p1 = pix[-2*xstride]; + const int q0 = pix[0]; + const int q1 = pix[1*xstride]; + + if( FFABS( p0 - q0 ) < alpha && + FFABS( p1 - p0 ) < beta && + FFABS( q1 - q0 ) < beta ) { + + int delta = av_clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc ); + + pix[-xstride] = av_clip_uint8( p0 + delta ); /* p0' */ + pix[0] = av_clip_uint8( q0 - delta ); /* q0' */ + } + pix += ystride; + } + } +} +static void h264_v_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) +{ + h264_loop_filter_chroma_c(pix, stride, 1, alpha, beta, tc0); +} +static void h264_h_loop_filter_chroma_c(uint8_t *pix, int stride, int alpha, int beta, int8_t *tc0) +{ + h264_loop_filter_chroma_c(pix, 1, stride, alpha, beta, tc0); +} + +static inline void h264_loop_filter_chroma_intra_c(uint8_t *pix, int xstride, int ystride, int alpha, int beta) +{ + int d; + for( d = 0; d < 8; d++ ) { + const int p0 = pix[-1*xstride]; + const int p1 = pix[-2*xstride]; + const int q0 = pix[0]; + const int q1 = pix[1*xstride]; + + if( FFABS( p0 - q0 ) < alpha && + FFABS( p1 - p0 ) < beta && + FFABS( q1 - q0 ) < beta ) { + + pix[-xstride] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */ + pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */ + } + pix += ystride; + } +} +static void h264_v_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta) +{ + h264_loop_filter_chroma_intra_c(pix, stride, 1, alpha, beta); +} +static void h264_h_loop_filter_chroma_intra_c(uint8_t *pix, int stride, int alpha, int beta) +{ + h264_loop_filter_chroma_intra_c(pix, 1, stride, alpha, beta); +} + +static inline int pix_abs16_c(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, int h) { int s, i; s = 0; - for(i=0;i<16;i++) { + for(i=0;iavctx->nsse_weight; + else return score1 + FFABS(score2)*8; } -static int sad8x8_c(void *s, uint8_t *a, uint8_t *b, int stride){ - return pix_abs8x8_c(a,b,stride); +static int nsse8_c(void *v, uint8_t *s1, uint8_t *s2, int stride, int h){ + MpegEncContext *c = v; + int score1=0; + int score2=0; + int x,y; + + for(y=0; yavctx->nsse_weight; + else return score1 + FFABS(score2)*8; } -void ff_block_permute(DCTELEM *block, UINT8 *permutation, const UINT8 *scantable, int last) +static int try_8x8basis_c(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale){ + int i; + unsigned int sum=0; + + for(i=0; i<8*8; i++){ + int b= rem[i] + ((basis[i]*scale + (1<<(BASIS_SHIFT - RECON_SHIFT-1)))>>(BASIS_SHIFT - RECON_SHIFT)); + int w= weight[i]; + b>>= RECON_SHIFT; + assert(-512>4; + } + return sum>>2; +} + +static void add_8x8basis_c(int16_t rem[64], int16_t basis[64], int scale){ + int i; + + for(i=0; i<8*8; i++){ + rem[i] += (basis[i]*scale + (1<<(BASIS_SHIFT - RECON_SHIFT-1)))>>(BASIS_SHIFT - RECON_SHIFT); + } +} + +/** + * permutes an 8x8 block. + * @param block the block which will be permuted according to the given permutation vector + * @param permutation the permutation vector + * @param last the last non zero coefficient in scantable order, used to speed the permutation up + * @param scantable the used scantable, this is only used to speed the permutation up, the block is not + * (inverse) permutated to scantable order! + */ +void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last) { int i; DCTELEM temp[64]; - + if(last<=0) return; - //if(permutation[1]==1) return; //FIXME its ok but not clean and might fail for some perms + //if(permutation[1]==1) return; //FIXME it is ok but not clean and might fail for some permutations for(i=0; i<=last; i++){ const int j= scantable[i]; temp[j]= block[j]; block[j]=0; } - + for(i=0; i<=last; i++){ const int j= scantable[i]; const int perm_j= permutation[j]; @@ -1604,43 +3333,143 @@ void ff_block_permute(DCTELEM *block, UINT8 *permutation, const UINT8 *scantable } } +static int zero_cmp(void *s, uint8_t *a, uint8_t *b, int stride, int h){ + return 0; +} + +void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type){ + int i; + + memset(cmp, 0, sizeof(void*)*5); + + for(i=0; i<5; i++){ + switch(type&0xFF){ + case FF_CMP_SAD: + cmp[i]= c->sad[i]; + break; + case FF_CMP_SATD: + cmp[i]= c->hadamard8_diff[i]; + break; + case FF_CMP_SSE: + cmp[i]= c->sse[i]; + break; + case FF_CMP_DCT: + cmp[i]= c->dct_sad[i]; + break; + case FF_CMP_DCT264: + cmp[i]= c->dct264_sad[i]; + break; + case FF_CMP_DCTMAX: + cmp[i]= c->dct_max[i]; + break; + case FF_CMP_PSNR: + cmp[i]= c->quant_psnr[i]; + break; + case FF_CMP_BIT: + cmp[i]= c->bit[i]; + break; + case FF_CMP_RD: + cmp[i]= c->rd[i]; + break; + case FF_CMP_VSAD: + cmp[i]= c->vsad[i]; + break; + case FF_CMP_VSSE: + cmp[i]= c->vsse[i]; + break; + case FF_CMP_ZERO: + cmp[i]= zero_cmp; + break; + case FF_CMP_NSSE: + cmp[i]= c->nsse[i]; + break; +#ifdef CONFIG_SNOW_ENCODER + case FF_CMP_W53: + cmp[i]= c->w53[i]; + break; + case FF_CMP_W97: + cmp[i]= c->w97[i]; + break; +#endif + default: + av_log(NULL, AV_LOG_ERROR,"internal error in cmp function selection\n"); + } + } +} + +/** + * memset(blocks, 0, sizeof(DCTELEM)*6*64) + */ static void clear_blocks_c(DCTELEM *blocks) { memset(blocks, 0, sizeof(DCTELEM)*6*64); } static void add_bytes_c(uint8_t *dst, uint8_t *src, int w){ - int i; - for(i=0; i+7maxi){ return sum; } -static int hadamard8_abs_c(uint8_t *src, int stride, int mean){ +static int hadamard8_intra8x8_c(/*MpegEncContext*/ void *s, uint8_t *src, uint8_t *dummy, int stride, int h){ int i; int temp[64]; int sum=0; -//FIXME OOOPS ignore 0 term instead of mean mess + + assert(h==8); + for(i=0; i<8; i++){ //FIXME try pointer walks - BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0]-mean,src[stride*i+1]-mean); - BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2]-mean,src[stride*i+3]-mean); - BUTTERFLY2(temp[8*i+4], temp[8*i+5], src[stride*i+4]-mean,src[stride*i+5]-mean); - BUTTERFLY2(temp[8*i+6], temp[8*i+7], src[stride*i+6]-mean,src[stride*i+7]-mean); - + BUTTERFLY2(temp[8*i+0], temp[8*i+1], src[stride*i+0],src[stride*i+1]); + BUTTERFLY2(temp[8*i+2], temp[8*i+3], src[stride*i+2],src[stride*i+3]); + BUTTERFLY2(temp[8*i+4], temp[8*i+5], src[stride*i+4],src[stride*i+5]); + BUTTERFLY2(temp[8*i+6], temp[8*i+7], src[stride*i+6],src[stride*i+7]); + BUTTERFLY1(temp[8*i+0], temp[8*i+2]); BUTTERFLY1(temp[8*i+1], temp[8*i+3]); BUTTERFLY1(temp[8*i+4], temp[8*i+6]); BUTTERFLY1(temp[8*i+5], temp[8*i+7]); - + BUTTERFLY1(temp[8*i+0], temp[8*i+4]); BUTTERFLY1(temp[8*i+1], temp[8*i+5]); BUTTERFLY1(temp[8*i+2], temp[8*i+6]); @@ -1734,105 +3567,654 @@ static int hadamard8_abs_c(uint8_t *src, int stride, int mean){ BUTTERFLY1(temp[8*2+i], temp[8*3+i]); BUTTERFLY1(temp[8*4+i], temp[8*5+i]); BUTTERFLY1(temp[8*6+i], temp[8*7+i]); - + BUTTERFLY1(temp[8*0+i], temp[8*2+i]); BUTTERFLY1(temp[8*1+i], temp[8*3+i]); BUTTERFLY1(temp[8*4+i], temp[8*6+i]); BUTTERFLY1(temp[8*5+i], temp[8*7+i]); - - sum += + + sum += BUTTERFLYA(temp[8*0+i], temp[8*4+i]) +BUTTERFLYA(temp[8*1+i], temp[8*5+i]) +BUTTERFLYA(temp[8*2+i], temp[8*6+i]) +BUTTERFLYA(temp[8*3+i], temp[8*7+i]); } - + + sum -= FFABS(temp[8*0] + temp[8*4]); // -mean + return sum; } -static int dct_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){ +static int dct_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){ MpegEncContext * const s= (MpegEncContext *)c; - DCTELEM temp[64]; + DECLARE_ALIGNED_16(uint64_t, aligned_temp[sizeof(DCTELEM)*64/8]); + DCTELEM * const temp= (DCTELEM*)aligned_temp; + + assert(h==8); + + s->dsp.diff_pixels(temp, src1, src2, stride); + s->dsp.fdct(temp); + return s->dsp.sum_abs_dctelem(temp); +} + +#ifdef CONFIG_GPL +#define DCT8_1D {\ + const int s07 = SRC(0) + SRC(7);\ + const int s16 = SRC(1) + SRC(6);\ + const int s25 = SRC(2) + SRC(5);\ + const int s34 = SRC(3) + SRC(4);\ + const int a0 = s07 + s34;\ + const int a1 = s16 + s25;\ + const int a2 = s07 - s34;\ + const int a3 = s16 - s25;\ + const int d07 = SRC(0) - SRC(7);\ + const int d16 = SRC(1) - SRC(6);\ + const int d25 = SRC(2) - SRC(5);\ + const int d34 = SRC(3) - SRC(4);\ + const int a4 = d16 + d25 + (d07 + (d07>>1));\ + const int a5 = d07 - d34 - (d25 + (d25>>1));\ + const int a6 = d07 + d34 - (d16 + (d16>>1));\ + const int a7 = d16 - d25 + (d34 + (d34>>1));\ + DST(0, a0 + a1 ) ;\ + DST(1, a4 + (a7>>2)) ;\ + DST(2, a2 + (a3>>1)) ;\ + DST(3, a5 + (a6>>2)) ;\ + DST(4, a0 - a1 ) ;\ + DST(5, a6 - (a5>>2)) ;\ + DST(6, (a2>>1) - a3 ) ;\ + DST(7, (a4>>2) - a7 ) ;\ +} + +static int dct264_sad8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){ + MpegEncContext * const s= (MpegEncContext *)c; + DCTELEM dct[8][8]; + int i; + int sum=0; + + s->dsp.diff_pixels(dct[0], src1, src2, stride); + +#define SRC(x) dct[i][x] +#define DST(x,v) dct[i][x]= v + for( i = 0; i < 8; i++ ) + DCT8_1D +#undef SRC +#undef DST + +#define SRC(x) dct[x][i] +#define DST(x,v) sum += FFABS(v) + for( i = 0; i < 8; i++ ) + DCT8_1D +#undef SRC +#undef DST + return sum; +} +#endif + +static int dct_max8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){ + MpegEncContext * const s= (MpegEncContext *)c; + DECLARE_ALIGNED_8(uint64_t, aligned_temp[sizeof(DCTELEM)*64/8]); + DCTELEM * const temp= (DCTELEM*)aligned_temp; int sum=0, i; + assert(h==8); + s->dsp.diff_pixels(temp, src1, src2, stride); - s->fdct(temp); + s->dsp.fdct(temp); for(i=0; i<64; i++) - sum+= ABS(temp[i]); - + sum= FFMAX(sum, FFABS(temp[i])); + return sum; } -void simple_idct(INT16 *block); //FIXME - -static int quant_psnr8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride){ +static int quant_psnr8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){ MpegEncContext * const s= (MpegEncContext *)c; - DCTELEM temp[64], bak[64]; + DECLARE_ALIGNED_8 (uint64_t, aligned_temp[sizeof(DCTELEM)*64*2/8]); + DCTELEM * const temp= (DCTELEM*)aligned_temp; + DCTELEM * const bak = ((DCTELEM*)aligned_temp)+64; int sum=0, i; + assert(h==8); s->mb_intra=0; - + s->dsp.diff_pixels(temp, src1, src2, stride); - + memcpy(bak, temp, 64*sizeof(DCTELEM)); - - s->dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i); - s->dct_unquantize(s, temp, 0, s->qscale); - simple_idct(temp); //FIXME - + + s->block_last_index[0/*FIXME*/]= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i); + s->dct_unquantize_inter(s, temp, 0, s->qscale); + ff_simple_idct(temp); //FIXME + for(i=0; i<64; i++) sum+= (temp[i]-bak[i])*(temp[i]-bak[i]); - + return sum; } -WARPER88_1616(hadamard8_diff_c, hadamard8_diff16_c) -WARPER88_1616(dct_sad8x8_c, dct_sad16x16_c) -WARPER88_1616(quant_psnr8x8_c, quant_psnr16x16_c) +static int rd8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){ + MpegEncContext * const s= (MpegEncContext *)c; + const uint8_t *scantable= s->intra_scantable.permutated; + DECLARE_ALIGNED_8 (uint64_t, aligned_temp[sizeof(DCTELEM)*64/8]); + DECLARE_ALIGNED_8 (uint64_t, aligned_bak[stride]); + DCTELEM * const temp= (DCTELEM*)aligned_temp; + uint8_t * const bak= (uint8_t*)aligned_bak; + int i, last, run, bits, level, distoration, start_i; + const int esc_length= s->ac_esc_length; + uint8_t * length; + uint8_t * last_length; + + assert(h==8); + + for(i=0; i<8; i++){ + ((uint32_t*)(bak + i*stride))[0]= ((uint32_t*)(src2 + i*stride))[0]; + ((uint32_t*)(bak + i*stride))[1]= ((uint32_t*)(src2 + i*stride))[1]; + } + + s->dsp.diff_pixels(temp, src1, src2, stride); + + s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i); + + bits=0; + + if (s->mb_intra) { + start_i = 1; + length = s->intra_ac_vlc_length; + last_length= s->intra_ac_vlc_last_length; + bits+= s->luma_dc_vlc_length[temp[0] + 256]; //FIXME chroma + } else { + start_i = 0; + length = s->inter_ac_vlc_length; + last_length= s->inter_ac_vlc_last_length; + } + + if(last>=start_i){ + run=0; + for(i=start_i; i=0){ + if(s->mb_intra) + s->dct_unquantize_intra(s, temp, 0, s->qscale); + else + s->dct_unquantize_inter(s, temp, 0, s->qscale); + } + + s->dsp.idct_add(bak, stride, temp); + + distoration= s->dsp.sse[1](NULL, bak, src1, stride, 8); + + return distoration + ((bits*s->qscale*s->qscale*109 + 64)>>7); +} + +static int bit8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){ + MpegEncContext * const s= (MpegEncContext *)c; + const uint8_t *scantable= s->intra_scantable.permutated; + DECLARE_ALIGNED_8 (uint64_t, aligned_temp[sizeof(DCTELEM)*64/8]); + DCTELEM * const temp= (DCTELEM*)aligned_temp; + int i, last, run, bits, level, start_i; + const int esc_length= s->ac_esc_length; + uint8_t * length; + uint8_t * last_length; + + assert(h==8); + + s->dsp.diff_pixels(temp, src1, src2, stride); + + s->block_last_index[0/*FIXME*/]= last= s->fast_dct_quantize(s, temp, 0/*FIXME*/, s->qscale, &i); + + bits=0; + + if (s->mb_intra) { + start_i = 1; + length = s->intra_ac_vlc_length; + last_length= s->intra_ac_vlc_last_length; + bits+= s->luma_dc_vlc_length[temp[0] + 256]; //FIXME chroma + } else { + start_i = 0; + length = s->inter_ac_vlc_length; + last_length= s->inter_ac_vlc_last_length; + } + + if(last>=start_i){ + run=0; + for(i=start_i; i>31; + // is this faster on some gcc/cpu combinations? +// if(tmp > 0x43c0ffff) tmp = 0xFFFF; +// else tmp = 0; + } + dst[i] = tmp - 0x8000; + } +} + +#define W0 2048 +#define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */ +#define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */ +#define W3 2408 /* 2048*sqrt (2)*cos (3*pi/16) */ +#define W4 2048 /* 2048*sqrt (2)*cos (4*pi/16) */ +#define W5 1609 /* 2048*sqrt (2)*cos (5*pi/16) */ +#define W6 1108 /* 2048*sqrt (2)*cos (6*pi/16) */ +#define W7 565 /* 2048*sqrt (2)*cos (7*pi/16) */ + +static void wmv2_idct_row(short * b) +{ + int s1,s2; + int a0,a1,a2,a3,a4,a5,a6,a7; + /*step 1*/ + a1 = W1*b[1]+W7*b[7]; + a7 = W7*b[1]-W1*b[7]; + a5 = W5*b[5]+W3*b[3]; + a3 = W3*b[5]-W5*b[3]; + a2 = W2*b[2]+W6*b[6]; + a6 = W6*b[2]-W2*b[6]; + a0 = W0*b[0]+W0*b[4]; + a4 = W0*b[0]-W0*b[4]; + /*step 2*/ + s1 = (181*(a1-a5+a7-a3)+128)>>8;//1,3,5,7, + s2 = (181*(a1-a5-a7+a3)+128)>>8; + /*step 3*/ + b[0] = (a0+a2+a1+a5 + (1<<7))>>8; + b[1] = (a4+a6 +s1 + (1<<7))>>8; + b[2] = (a4-a6 +s2 + (1<<7))>>8; + b[3] = (a0-a2+a7+a3 + (1<<7))>>8; + b[4] = (a0-a2-a7-a3 + (1<<7))>>8; + b[5] = (a4-a6 -s2 + (1<<7))>>8; + b[6] = (a4+a6 -s1 + (1<<7))>>8; + b[7] = (a0+a2-a1-a5 + (1<<7))>>8; +} +static void wmv2_idct_col(short * b) +{ + int s1,s2; + int a0,a1,a2,a3,a4,a5,a6,a7; + /*step 1, with extended precision*/ + a1 = (W1*b[8*1]+W7*b[8*7] + 4)>>3; + a7 = (W7*b[8*1]-W1*b[8*7] + 4)>>3; + a5 = (W5*b[8*5]+W3*b[8*3] + 4)>>3; + a3 = (W3*b[8*5]-W5*b[8*3] + 4)>>3; + a2 = (W2*b[8*2]+W6*b[8*6] + 4)>>3; + a6 = (W6*b[8*2]-W2*b[8*6] + 4)>>3; + a0 = (W0*b[8*0]+W0*b[8*4] )>>3; + a4 = (W0*b[8*0]-W0*b[8*4] )>>3; + /*step 2*/ + s1 = (181*(a1-a5+a7-a3)+128)>>8; + s2 = (181*(a1-a5-a7+a3)+128)>>8; + /*step 3*/ + b[8*0] = (a0+a2+a1+a5 + (1<<13))>>14; + b[8*1] = (a4+a6 +s1 + (1<<13))>>14; + b[8*2] = (a4-a6 +s2 + (1<<13))>>14; + b[8*3] = (a0-a2+a7+a3 + (1<<13))>>14; + + b[8*4] = (a0-a2-a7-a3 + (1<<13))>>14; + b[8*5] = (a4-a6 -s2 + (1<<13))>>14; + b[8*6] = (a4+a6 -s1 + (1<<13))>>14; + b[8*7] = (a0+a2-a1-a5 + (1<<13))>>14; +} +void ff_wmv2_idct_c(short * block){ + int i; + + for(i=0;i<64;i+=8){ + wmv2_idct_row(block+i); + } + for(i=0;i<8;i++){ + wmv2_idct_col(block+i); + } +} +/* XXX: those functions should be suppressed ASAP when all IDCTs are + converted */ +static void ff_wmv2_idct_put_c(uint8_t *dest, int line_size, DCTELEM *block) +{ + ff_wmv2_idct_c(block); + put_pixels_clamped_c(block, dest, line_size); +} +static void ff_wmv2_idct_add_c(uint8_t *dest, int line_size, DCTELEM *block) +{ + ff_wmv2_idct_c(block); + add_pixels_clamped_c(block, dest, line_size); +} +static void ff_jref_idct_put(uint8_t *dest, int line_size, DCTELEM *block) +{ + j_rev_dct (block); + put_pixels_clamped_c(block, dest, line_size); +} +static void ff_jref_idct_add(uint8_t *dest, int line_size, DCTELEM *block) +{ + j_rev_dct (block); + add_pixels_clamped_c(block, dest, line_size); +} + +static void ff_jref_idct4_put(uint8_t *dest, int line_size, DCTELEM *block) +{ + j_rev_dct4 (block); + put_pixels_clamped4_c(block, dest, line_size); +} +static void ff_jref_idct4_add(uint8_t *dest, int line_size, DCTELEM *block) +{ + j_rev_dct4 (block); + add_pixels_clamped4_c(block, dest, line_size); +} + +static void ff_jref_idct2_put(uint8_t *dest, int line_size, DCTELEM *block) +{ + j_rev_dct2 (block); + put_pixels_clamped2_c(block, dest, line_size); +} +static void ff_jref_idct2_add(uint8_t *dest, int line_size, DCTELEM *block) +{ + j_rev_dct2 (block); + add_pixels_clamped2_c(block, dest, line_size); +} + +static void ff_jref_idct1_put(uint8_t *dest, int line_size, DCTELEM *block) +{ + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + dest[0] = cm[(block[0] + 4)>>3]; +} +static void ff_jref_idct1_add(uint8_t *dest, int line_size, DCTELEM *block) +{ + uint8_t *cm = ff_cropTbl + MAX_NEG_CROP; + + dest[0] = cm[dest[0] + ((block[0] + 4)>>3)]; +} + +static void just_return(void *mem av_unused, int stride av_unused, int h av_unused) { return; } + +/* init static data */ +void dsputil_static_init(void) { - static int init_done = 0; int i; - if (!init_done) { - for(i=0;i<256;i++) cropTbl[i + MAX_NEG_CROP] = i; - for(i=0;i= 4.2.\n" + "Do not report crashes to FFmpeg developers.\n"); +#endif + did_fail=1; + } + return -1; + } + return 0; +} + +void dsputil_init(DSPContext* c, AVCodecContext *avctx) +{ + int i; + + ff_check_alignment(); + +#ifdef CONFIG_ENCODERS + if(avctx->dct_algo==FF_DCT_FASTINT) { + c->fdct = fdct_ifast; + c->fdct248 = fdct_ifast248; + } + else if(avctx->dct_algo==FF_DCT_FAAN) { + c->fdct = ff_faandct; + c->fdct248 = ff_faandct248; + } + else { + c->fdct = ff_jpeg_fdct_islow; //slow/accurate/default + c->fdct248 = ff_fdct248_islow; + } +#endif //CONFIG_ENCODERS + + if(avctx->lowres==1){ + if(avctx->idct_algo==FF_IDCT_INT || avctx->idct_algo==FF_IDCT_AUTO || !ENABLE_H264_DECODER){ + c->idct_put= ff_jref_idct4_put; + c->idct_add= ff_jref_idct4_add; + }else{ + c->idct_put= ff_h264_lowres_idct_put_c; + c->idct_add= ff_h264_lowres_idct_add_c; + } + c->idct = j_rev_dct4; + c->idct_permutation_type= FF_NO_IDCT_PERM; + }else if(avctx->lowres==2){ + c->idct_put= ff_jref_idct2_put; + c->idct_add= ff_jref_idct2_add; + c->idct = j_rev_dct2; + c->idct_permutation_type= FF_NO_IDCT_PERM; + }else if(avctx->lowres==3){ + c->idct_put= ff_jref_idct1_put; + c->idct_add= ff_jref_idct1_add; + c->idct = j_rev_dct1; + c->idct_permutation_type= FF_NO_IDCT_PERM; + }else{ + if(avctx->idct_algo==FF_IDCT_INT){ + c->idct_put= ff_jref_idct_put; + c->idct_add= ff_jref_idct_add; + c->idct = j_rev_dct; + c->idct_permutation_type= FF_LIBMPEG2_IDCT_PERM; + }else if((ENABLE_VP3_DECODER || ENABLE_VP5_DECODER || ENABLE_VP6_DECODER || ENABLE_THEORA_DECODER ) && + avctx->idct_algo==FF_IDCT_VP3){ + c->idct_put= ff_vp3_idct_put_c; + c->idct_add= ff_vp3_idct_add_c; + c->idct = ff_vp3_idct_c; + c->idct_permutation_type= FF_NO_IDCT_PERM; + }else if(avctx->idct_algo==FF_IDCT_WMV2){ + c->idct_put= ff_wmv2_idct_put_c; + c->idct_add= ff_wmv2_idct_add_c; + c->idct = ff_wmv2_idct_c; + c->idct_permutation_type= FF_NO_IDCT_PERM; + }else if(avctx->idct_algo==FF_IDCT_FAAN){ + c->idct_put= ff_faanidct_put; + c->idct_add= ff_faanidct_add; + c->idct = ff_faanidct; + c->idct_permutation_type= FF_NO_IDCT_PERM; + }else{ //accurate/default + c->idct_put= ff_simple_idct_put; + c->idct_add= ff_simple_idct_add; + c->idct = ff_simple_idct; + c->idct_permutation_type= FF_NO_IDCT_PERM; + } + } + + if (ENABLE_H264_DECODER) { + c->h264_idct_add= ff_h264_idct_add_c; + c->h264_idct8_add= ff_h264_idct8_add_c; + c->h264_idct_dc_add= ff_h264_idct_dc_add_c; + c->h264_idct8_dc_add= ff_h264_idct8_dc_add_c; } c->get_pixels = get_pixels_c; c->diff_pixels = diff_pixels_c; c->put_pixels_clamped = put_pixels_clamped_c; + c->put_signed_pixels_clamped = put_signed_pixels_clamped_c; c->add_pixels_clamped = add_pixels_clamped_c; + c->add_pixels8 = add_pixels8_c; + c->add_pixels4 = add_pixels4_c; + c->sum_abs_dctelem = sum_abs_dctelem_c; c->gmc1 = gmc1_c; - c->gmc = gmc_c; + c->gmc = ff_gmc_c; c->clear_blocks = clear_blocks_c; c->pix_sum = pix_sum_c; c->pix_norm1 = pix_norm1_c; - c->sse[0]= sse16_c; - c->sse[1]= sse8_c; /* TODO [0] 16 [1] 8 */ - c->pix_abs16x16 = pix_abs16x16_c; - c->pix_abs16x16_x2 = pix_abs16x16_x2_c; - c->pix_abs16x16_y2 = pix_abs16x16_y2_c; - c->pix_abs16x16_xy2 = pix_abs16x16_xy2_c; - c->pix_abs8x8 = pix_abs8x8_c; - c->pix_abs8x8_x2 = pix_abs8x8_x2_c; - c->pix_abs8x8_y2 = pix_abs8x8_y2_c; - c->pix_abs8x8_xy2 = pix_abs8x8_xy2_c; + c->pix_abs[0][0] = pix_abs16_c; + c->pix_abs[0][1] = pix_abs16_x2_c; + c->pix_abs[0][2] = pix_abs16_y2_c; + c->pix_abs[0][3] = pix_abs16_xy2_c; + c->pix_abs[1][0] = pix_abs8_c; + c->pix_abs[1][1] = pix_abs8_x2_c; + c->pix_abs[1][2] = pix_abs8_y2_c; + c->pix_abs[1][3] = pix_abs8_xy2_c; #define dspfunc(PFX, IDX, NUM) \ c->PFX ## _pixels_tab[IDX][0] = PFX ## _pixels ## NUM ## _c; \ @@ -1844,13 +4226,40 @@ void dsputil_init(DSPContext* c, unsigned mask) dspfunc(put_no_rnd, 0, 16); dspfunc(put, 1, 8); dspfunc(put_no_rnd, 1, 8); + dspfunc(put, 2, 4); + dspfunc(put, 3, 2); dspfunc(avg, 0, 16); dspfunc(avg_no_rnd, 0, 16); dspfunc(avg, 1, 8); dspfunc(avg_no_rnd, 1, 8); + dspfunc(avg, 2, 4); + dspfunc(avg, 3, 2); #undef dspfunc + c->put_no_rnd_pixels_l2[0]= put_no_rnd_pixels16_l2_c; + c->put_no_rnd_pixels_l2[1]= put_no_rnd_pixels8_l2_c; + + c->put_tpel_pixels_tab[ 0] = put_tpel_pixels_mc00_c; + c->put_tpel_pixels_tab[ 1] = put_tpel_pixels_mc10_c; + c->put_tpel_pixels_tab[ 2] = put_tpel_pixels_mc20_c; + c->put_tpel_pixels_tab[ 4] = put_tpel_pixels_mc01_c; + c->put_tpel_pixels_tab[ 5] = put_tpel_pixels_mc11_c; + c->put_tpel_pixels_tab[ 6] = put_tpel_pixels_mc21_c; + c->put_tpel_pixels_tab[ 8] = put_tpel_pixels_mc02_c; + c->put_tpel_pixels_tab[ 9] = put_tpel_pixels_mc12_c; + c->put_tpel_pixels_tab[10] = put_tpel_pixels_mc22_c; + + c->avg_tpel_pixels_tab[ 0] = avg_tpel_pixels_mc00_c; + c->avg_tpel_pixels_tab[ 1] = avg_tpel_pixels_mc10_c; + c->avg_tpel_pixels_tab[ 2] = avg_tpel_pixels_mc20_c; + c->avg_tpel_pixels_tab[ 4] = avg_tpel_pixels_mc01_c; + c->avg_tpel_pixels_tab[ 5] = avg_tpel_pixels_mc11_c; + c->avg_tpel_pixels_tab[ 6] = avg_tpel_pixels_mc21_c; + c->avg_tpel_pixels_tab[ 8] = avg_tpel_pixels_mc02_c; + c->avg_tpel_pixels_tab[ 9] = avg_tpel_pixels_mc12_c; + c->avg_tpel_pixels_tab[10] = avg_tpel_pixels_mc22_c; + #define dspfunc(PFX, IDX, NUM) \ c->PFX ## _pixels_tab[IDX][ 0] = PFX ## NUM ## _mc00_c; \ c->PFX ## _pixels_tab[IDX][ 1] = PFX ## NUM ## _mc10_c; \ @@ -1880,7 +4289,59 @@ void dsputil_init(DSPContext* c, unsigned mask) dspfunc(avg_qpel, 1, 8); /* dspfunc(avg_no_rnd_qpel, 1, 8); */ + + dspfunc(put_h264_qpel, 0, 16); + dspfunc(put_h264_qpel, 1, 8); + dspfunc(put_h264_qpel, 2, 4); + dspfunc(put_h264_qpel, 3, 2); + dspfunc(avg_h264_qpel, 0, 16); + dspfunc(avg_h264_qpel, 1, 8); + dspfunc(avg_h264_qpel, 2, 4); + #undef dspfunc + c->put_h264_chroma_pixels_tab[0]= put_h264_chroma_mc8_c; + c->put_h264_chroma_pixels_tab[1]= put_h264_chroma_mc4_c; + c->put_h264_chroma_pixels_tab[2]= put_h264_chroma_mc2_c; + c->avg_h264_chroma_pixels_tab[0]= avg_h264_chroma_mc8_c; + c->avg_h264_chroma_pixels_tab[1]= avg_h264_chroma_mc4_c; + c->avg_h264_chroma_pixels_tab[2]= avg_h264_chroma_mc2_c; + c->put_no_rnd_h264_chroma_pixels_tab[0]= put_no_rnd_h264_chroma_mc8_c; + + c->weight_h264_pixels_tab[0]= weight_h264_pixels16x16_c; + c->weight_h264_pixels_tab[1]= weight_h264_pixels16x8_c; + c->weight_h264_pixels_tab[2]= weight_h264_pixels8x16_c; + c->weight_h264_pixels_tab[3]= weight_h264_pixels8x8_c; + c->weight_h264_pixels_tab[4]= weight_h264_pixels8x4_c; + c->weight_h264_pixels_tab[5]= weight_h264_pixels4x8_c; + c->weight_h264_pixels_tab[6]= weight_h264_pixels4x4_c; + c->weight_h264_pixels_tab[7]= weight_h264_pixels4x2_c; + c->weight_h264_pixels_tab[8]= weight_h264_pixels2x4_c; + c->weight_h264_pixels_tab[9]= weight_h264_pixels2x2_c; + c->biweight_h264_pixels_tab[0]= biweight_h264_pixels16x16_c; + c->biweight_h264_pixels_tab[1]= biweight_h264_pixels16x8_c; + c->biweight_h264_pixels_tab[2]= biweight_h264_pixels8x16_c; + c->biweight_h264_pixels_tab[3]= biweight_h264_pixels8x8_c; + c->biweight_h264_pixels_tab[4]= biweight_h264_pixels8x4_c; + c->biweight_h264_pixels_tab[5]= biweight_h264_pixels4x8_c; + c->biweight_h264_pixels_tab[6]= biweight_h264_pixels4x4_c; + c->biweight_h264_pixels_tab[7]= biweight_h264_pixels4x2_c; + c->biweight_h264_pixels_tab[8]= biweight_h264_pixels2x4_c; + c->biweight_h264_pixels_tab[9]= biweight_h264_pixels2x2_c; + + c->draw_edges = draw_edges_c; + +#ifdef CONFIG_CAVS_DECODER + ff_cavsdsp_init(c,avctx); +#endif +#if defined(CONFIG_VC1_DECODER) || defined(CONFIG_WMV3_DECODER) + ff_vc1dsp_init(c,avctx); +#endif +#if defined(CONFIG_WMV2_DECODER) || defined(CONFIG_VC1_DECODER) || defined(CONFIG_WMV3_DECODER) + ff_intrax8dsp_init(c,avctx); +#endif +#if defined(CONFIG_H264_ENCODER) + ff_h264dspenc_init(c,avctx); +#endif c->put_mspel_pixels_tab[0]= put_mspel8_mc00_c; c->put_mspel_pixels_tab[1]= put_mspel8_mc10_c; @@ -1890,55 +4351,139 @@ void dsputil_init(DSPContext* c, unsigned mask) c->put_mspel_pixels_tab[5]= put_mspel8_mc12_c; c->put_mspel_pixels_tab[6]= put_mspel8_mc22_c; c->put_mspel_pixels_tab[7]= put_mspel8_mc32_c; - - c->hadamard8_diff[0]= hadamard8_diff16_c; - c->hadamard8_diff[1]= hadamard8_diff_c; - c->hadamard8_abs = hadamard8_abs_c; - - c->dct_sad[0]= dct_sad16x16_c; - c->dct_sad[1]= dct_sad8x8_c; - - c->sad[0]= sad16x16_c; - c->sad[1]= sad8x8_c; - - c->quant_psnr[0]= quant_psnr16x16_c; - c->quant_psnr[1]= quant_psnr8x8_c; - - c->add_bytes= add_bytes_c; - c->diff_bytes= diff_bytes_c; -#ifdef HAVE_MMX - dsputil_init_mmx(c, mask); - if (ff_bit_exact) - { - /* FIXME - AVCodec context should have flag for bitexact match */ - /* fprintf(stderr, "\n\n\nff_bit_exact %d\n\n\n\n", ff_bit_exact); */ - dsputil_set_bit_exact_mmx(c, mask); - } +#define SET_CMP_FUNC(name) \ + c->name[0]= name ## 16_c;\ + c->name[1]= name ## 8x8_c; + + SET_CMP_FUNC(hadamard8_diff) + c->hadamard8_diff[4]= hadamard8_intra16_c; + SET_CMP_FUNC(dct_sad) + SET_CMP_FUNC(dct_max) +#ifdef CONFIG_GPL + SET_CMP_FUNC(dct264_sad) #endif -#ifdef ARCH_ARMV4L - dsputil_init_armv4l(c, mask); + c->sad[0]= pix_abs16_c; + c->sad[1]= pix_abs8_c; + c->sse[0]= sse16_c; + c->sse[1]= sse8_c; + c->sse[2]= sse4_c; + SET_CMP_FUNC(quant_psnr) + SET_CMP_FUNC(rd) + SET_CMP_FUNC(bit) + c->vsad[0]= vsad16_c; + c->vsad[4]= vsad_intra16_c; + c->vsse[0]= vsse16_c; + c->vsse[4]= vsse_intra16_c; + c->nsse[0]= nsse16_c; + c->nsse[1]= nsse8_c; +#ifdef CONFIG_SNOW_ENCODER + c->w53[0]= w53_16_c; + c->w53[1]= w53_8_c; + c->w97[0]= w97_16_c; + c->w97[1]= w97_8_c; #endif -#ifdef HAVE_MLIB - dsputil_init_mlib(c, mask); + + c->ssd_int8_vs_int16 = ssd_int8_vs_int16_c; + + c->add_bytes= add_bytes_c; + c->add_bytes_l2= add_bytes_l2_c; + c->diff_bytes= diff_bytes_c; + c->sub_hfyu_median_prediction= sub_hfyu_median_prediction_c; + c->bswap_buf= bswap_buf; +#ifdef CONFIG_PNG_DECODER + c->add_png_paeth_prediction= ff_add_png_paeth_prediction; #endif -#ifdef ARCH_ALPHA - dsputil_init_alpha(c, mask); + + c->h264_v_loop_filter_luma= h264_v_loop_filter_luma_c; + c->h264_h_loop_filter_luma= h264_h_loop_filter_luma_c; + c->h264_v_loop_filter_chroma= h264_v_loop_filter_chroma_c; + c->h264_h_loop_filter_chroma= h264_h_loop_filter_chroma_c; + c->h264_v_loop_filter_chroma_intra= h264_v_loop_filter_chroma_intra_c; + c->h264_h_loop_filter_chroma_intra= h264_h_loop_filter_chroma_intra_c; + c->h264_loop_filter_strength= NULL; + + if (ENABLE_ANY_H263) { + c->h263_h_loop_filter= h263_h_loop_filter_c; + c->h263_v_loop_filter= h263_v_loop_filter_c; + } + + c->h261_loop_filter= h261_loop_filter_c; + + c->try_8x8basis= try_8x8basis_c; + c->add_8x8basis= add_8x8basis_c; + +#ifdef CONFIG_SNOW_DECODER + c->vertical_compose97i = ff_snow_vertical_compose97i; + c->horizontal_compose97i = ff_snow_horizontal_compose97i; + c->inner_add_yblock = ff_snow_inner_add_yblock; #endif -#ifdef ARCH_POWERPC - dsputil_init_ppc(c, mask); + +#ifdef CONFIG_VORBIS_DECODER + c->vorbis_inverse_coupling = vorbis_inverse_coupling; #endif -#ifdef HAVE_MMI - dsputil_init_mmi(c, mask); +#ifdef CONFIG_FLAC_ENCODER + c->flac_compute_autocorr = ff_flac_compute_autocorr; #endif -} + c->vector_fmul = vector_fmul_c; + c->vector_fmul_reverse = vector_fmul_reverse_c; + c->vector_fmul_add_add = ff_vector_fmul_add_add_c; + c->float_to_int16 = ff_float_to_int16_c; -/* remove any non bit exact operation (testing purpose) */ -void avcodec_set_bit_exact(void) -{ - ff_bit_exact=1; -#ifdef HAVE_MMX -// FIXME - better set_bit_exact -// dsputil_set_bit_exact_mmx(); -#endif + c->shrink[0]= ff_img_copy_plane; + c->shrink[1]= ff_shrink22; + c->shrink[2]= ff_shrink44; + c->shrink[3]= ff_shrink88; + + c->prefetch= just_return; + + memset(c->put_2tap_qpel_pixels_tab, 0, sizeof(c->put_2tap_qpel_pixels_tab)); + memset(c->avg_2tap_qpel_pixels_tab, 0, sizeof(c->avg_2tap_qpel_pixels_tab)); + + if (ENABLE_MMX) dsputil_init_mmx (c, avctx); + if (ENABLE_ARMV4L) dsputil_init_armv4l(c, avctx); + if (ENABLE_MLIB) dsputil_init_mlib (c, avctx); + if (ENABLE_VIS) dsputil_init_vis (c, avctx); + if (ENABLE_ALPHA) dsputil_init_alpha (c, avctx); + if (ENABLE_POWERPC) dsputil_init_ppc (c, avctx); + if (ENABLE_MMI) dsputil_init_mmi (c, avctx); + if (ENABLE_SH4) dsputil_init_sh4 (c, avctx); + if (ENABLE_BFIN) dsputil_init_bfin (c, avctx); + + for(i=0; i<64; i++){ + if(!c->put_2tap_qpel_pixels_tab[0][i]) + c->put_2tap_qpel_pixels_tab[0][i]= c->put_h264_qpel_pixels_tab[0][i]; + if(!c->avg_2tap_qpel_pixels_tab[0][i]) + c->avg_2tap_qpel_pixels_tab[0][i]= c->avg_h264_qpel_pixels_tab[0][i]; + } + + switch(c->idct_permutation_type){ + case FF_NO_IDCT_PERM: + for(i=0; i<64; i++) + c->idct_permutation[i]= i; + break; + case FF_LIBMPEG2_IDCT_PERM: + for(i=0; i<64; i++) + c->idct_permutation[i]= (i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2); + break; + case FF_SIMPLE_IDCT_PERM: + for(i=0; i<64; i++) + c->idct_permutation[i]= simple_mmx_permutation[i]; + break; + case FF_TRANSPOSE_IDCT_PERM: + for(i=0; i<64; i++) + c->idct_permutation[i]= ((i&7)<<3) | (i>>3); + break; + case FF_PARTTRANS_IDCT_PERM: + for(i=0; i<64; i++) + c->idct_permutation[i]= (i&0x24) | ((i&3)<<3) | ((i>>3)&3); + break; + case FF_SSE2_IDCT_PERM: + for(i=0; i<64; i++) + c->idct_permutation[i]= (i&0x38) | idct_sse2_row_perm[i&7]; + break; + default: + av_log(avctx, AV_LOG_ERROR, "Internal error, IDCT permutation not set\n"); + } } +