FFmpeg: libavcodec/dsputil.h Source File

00001 /*
00002  * DSP utils
00003  * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
00004  * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
00005  *
00006  * This file is part of FFmpeg.
00007  *
00008  * FFmpeg is free software; you can redistribute it and/or
00009  * modify it under the terms of the GNU Lesser General Public
00010  * License as published by the Free Software Foundation; either
00011  * version 2.1 of the License, or (at your option) any later version.
00012  *
00013  * FFmpeg is distributed in the hope that it will be useful,
00014  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00015  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00016  * Lesser General Public License for more details.
00017  *
00018  * You should have received a copy of the GNU Lesser General Public
00019  * License along with FFmpeg; if not, write to the Free Software
00020  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00021  */
00022 
00030 #ifndef AVCODEC_DSPUTIL_H
00031 #define AVCODEC_DSPUTIL_H
00032 
00033 #include "libavutil/intreadwrite.h"
00034 #include "avcodec.h"
00035 
00036 
00037 //#define DEBUG
00038 /* dct code */
00039 typedef short DCTELEM;
00040 
00041 void fdct_ifast (DCTELEM *data);
00042 void fdct_ifast248 (DCTELEM *data);
00043 void ff_jpeg_fdct_islow (DCTELEM *data);
00044 void ff_fdct248_islow (DCTELEM *data);
00045 
00046 void j_rev_dct (DCTELEM *data);
00047 void j_rev_dct4 (DCTELEM *data);
00048 void j_rev_dct2 (DCTELEM *data);
00049 void j_rev_dct1 (DCTELEM *data);
00050 void ff_wmv2_idct_c(DCTELEM *data);
00051 
00052 void ff_fdct_mmx(DCTELEM *block);
00053 void ff_fdct_mmx2(DCTELEM *block);
00054 void ff_fdct_sse2(DCTELEM *block);
00055 
00056 void ff_h264_idct8_add_c(uint8_t *dst, DCTELEM *block, int stride);
00057 void ff_h264_idct_add_c(uint8_t *dst, DCTELEM *block, int stride);
00058 void ff_h264_idct8_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
00059 void ff_h264_idct_dc_add_c(uint8_t *dst, DCTELEM *block, int stride);
00060 void ff_h264_lowres_idct_add_c(uint8_t *dst, int stride, DCTELEM *block);
00061 void ff_h264_lowres_idct_put_c(uint8_t *dst, int stride, DCTELEM *block);
00062 void ff_h264_idct_add16_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
00063 void ff_h264_idct_add16intra_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
00064 void ff_h264_idct8_add4_c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
00065 void ff_h264_idct_add8_c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);
00066 
00067 void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1,
00068                              const float *win, float add_bias, int len);
00069 void ff_float_to_int16_c(int16_t *dst, const float *src, long len);
00070 void ff_float_to_int16_interleave_c(int16_t *dst, const float **src, long len, int channels);
00071 
00072 /* encoding scans */
00073 extern const uint8_t ff_alternate_horizontal_scan[64];
00074 extern const uint8_t ff_alternate_vertical_scan[64];
00075 extern const uint8_t ff_zigzag_direct[64];
00076 extern const uint8_t ff_zigzag248_direct[64];
00077 
00078 /* pixel operations */
00079 #define MAX_NEG_CROP 1024
00080 
00081 /* temporary */
00082 extern uint32_t ff_squareTbl[512];
00083 extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
00084 
00085 /* VP3 DSP functions */
00086 void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
00087 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00088 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00089 void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
00090 
00091 void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
00092 void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
00093 
00094 /* VP6 DSP functions */
00095 void ff_vp6_filter_diag4_c(uint8_t *dst, uint8_t *src, int stride,
00096                            const int16_t *h_weights, const int16_t *v_weights);
00097 
00098 /* Bink functions */
00099 void ff_bink_idct_c    (DCTELEM *block);
00100 void ff_bink_idct_add_c(uint8_t *dest, int linesize, DCTELEM *block);
00101 void ff_bink_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);
00102 
00103 /* CAVS functions */
00104 void ff_put_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
00105 void ff_avg_cavs_qpel8_mc00_c(uint8_t *dst, uint8_t *src, int stride);
00106 void ff_put_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
00107 void ff_avg_cavs_qpel16_mc00_c(uint8_t *dst, uint8_t *src, int stride);
00108 
00109 /* VC1 functions */
00110 void ff_put_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd);
00111 void ff_avg_vc1_mspel_mc00_c(uint8_t *dst, const uint8_t *src, int stride, int rnd);
00112 
00113 /* EA functions */
00114 void ff_ea_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);
00115 
00116 /* 1/2^n downscaling functions from imgconvert.c */
00117 void ff_img_copy_plane(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00118 void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00119 void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00120 void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00121 
00122 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
00123               int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
00124 
00125 /* minimum alignment rules ;)
00126 If you notice errors in the align stuff, need more alignment for some ASM code
00127 for some CPU or need to use a function with less aligned data then send a mail
00128 to the ffmpeg-devel mailing list, ...
00129 
00130 !warning These alignments might not match reality, (missing attribute((align))
00131 stuff somewhere possible).
00132 I (Michael) did not check them, these are just the alignments which I think
00133 could be reached easily ...
00134 
00135 !future video codecs might need functions with less strict alignment
00136 */
00137 
00138 /*
00139 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
00140 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
00141 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
00142 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
00143 void clear_blocks_c(DCTELEM *blocks);
00144 */
00145 
00146 /* add and put pixel (decoding) */
00147 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
00148 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4
00149 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
00150 typedef void (*tpel_mc_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int w, int h);
00151 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
00152 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
00153 
00154 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
00155 
00156 #define DEF_OLD_QPEL(name)\
00157 void ff_put_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
00158 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
00159 void ff_avg_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
00160 
00161 DEF_OLD_QPEL(qpel16_mc11_old_c)
00162 DEF_OLD_QPEL(qpel16_mc31_old_c)
00163 DEF_OLD_QPEL(qpel16_mc12_old_c)
00164 DEF_OLD_QPEL(qpel16_mc32_old_c)
00165 DEF_OLD_QPEL(qpel16_mc13_old_c)
00166 DEF_OLD_QPEL(qpel16_mc33_old_c)
00167 DEF_OLD_QPEL(qpel8_mc11_old_c)
00168 DEF_OLD_QPEL(qpel8_mc31_old_c)
00169 DEF_OLD_QPEL(qpel8_mc12_old_c)
00170 DEF_OLD_QPEL(qpel8_mc32_old_c)
00171 DEF_OLD_QPEL(qpel8_mc13_old_c)
00172 DEF_OLD_QPEL(qpel8_mc33_old_c)
00173 
00174 #define CALL_2X_PIXELS(a, b, n)\
00175 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
00176     b(block  , pixels  , line_size, h);\
00177     b(block+n, pixels+n, line_size, h);\
00178 }
00179 
00180 /* motion estimation */
00181 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
00182 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
00183 typedef int (*me_cmp_func)(void /*MpegEncContext*/ *s, uint8_t *blk1/*align width (8 or 16)*/, uint8_t *blk2/*align 1*/, int line_size, int h)/* __attribute__ ((const))*/;
00184 
00188 typedef struct ScanTable{
00189     const uint8_t *scantable;
00190     uint8_t permutated[64];
00191     uint8_t raster_end[64];
00192 #if ARCH_PPC
00193 
00194     DECLARE_ALIGNED(16, uint8_t, inverse)[64];
00195 #endif
00196 } ScanTable;
00197 
00198 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
00199 
00200 void ff_emulated_edge_mc(uint8_t *buf, uint8_t *src, int linesize,
00201                          int block_w, int block_h,
00202                          int src_x, int src_y, int w, int h);
00203 
00207 typedef struct DSPContext {
00208     /* pixel ops : interface with DCT */
00209     void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
00210     void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
00211     void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00212     void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00213     void (*put_pixels_nonclamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00214     void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00215     void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
00216     void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
00217     int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
00221     void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
00225     void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
00226                     int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
00227     void (*clear_block)(DCTELEM *block/*align 16*/);
00228     void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
00229     int (*pix_sum)(uint8_t * pix, int line_size);
00230     int (*pix_norm1)(uint8_t * pix, int line_size);
00231 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
00232 
00233     me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
00234     me_cmp_func sse[6];
00235     me_cmp_func hadamard8_diff[6];
00236     me_cmp_func dct_sad[6];
00237     me_cmp_func quant_psnr[6];
00238     me_cmp_func bit[6];
00239     me_cmp_func rd[6];
00240     me_cmp_func vsad[6];
00241     me_cmp_func vsse[6];
00242     me_cmp_func nsse[6];
00243     me_cmp_func w53[6];
00244     me_cmp_func w97[6];
00245     me_cmp_func dct_max[6];
00246     me_cmp_func dct264_sad[6];
00247 
00248     me_cmp_func me_pre_cmp[6];
00249     me_cmp_func me_cmp[6];
00250     me_cmp_func me_sub_cmp[6];
00251     me_cmp_func mb_cmp[6];
00252     me_cmp_func ildct_cmp[6]; //only width 16 used
00253     me_cmp_func frame_skip_cmp[6]; //only width 8 used
00254 
00255     int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
00256                              int size);
00257 
00268     op_pixels_func put_pixels_tab[4][4];
00269 
00280     op_pixels_func avg_pixels_tab[4][4];
00281 
00292     op_pixels_func put_no_rnd_pixels_tab[4][4];
00293 
00304     op_pixels_func avg_no_rnd_pixels_tab[4][4];
00305 
00306     void (*put_no_rnd_pixels_l2[2])(uint8_t *block/*align width (8 or 16)*/, const uint8_t *a/*align 1*/, const uint8_t *b/*align 1*/, int line_size, int h);
00307 
00318     tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00319     tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00320 
00321     qpel_mc_func put_qpel_pixels_tab[2][16];
00322     qpel_mc_func avg_qpel_pixels_tab[2][16];
00323     qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
00324     qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
00325     qpel_mc_func put_mspel_pixels_tab[8];
00326 
00330     h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
00331     h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
00332     /* This is really one func used in VC-1 decoding */
00333     h264_chroma_mc_func put_no_rnd_vc1_chroma_pixels_tab[3];
00334     h264_chroma_mc_func avg_no_rnd_vc1_chroma_pixels_tab[3];
00335 
00336     qpel_mc_func put_h264_qpel_pixels_tab[4][16];
00337     qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
00338 
00339     qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
00340     qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];
00341 
00342     /* AVS specific */
00343     qpel_mc_func put_cavs_qpel_pixels_tab[2][16];
00344     qpel_mc_func avg_cavs_qpel_pixels_tab[2][16];
00345     void (*cavs_filter_lv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
00346     void (*cavs_filter_lh)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
00347     void (*cavs_filter_cv)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
00348     void (*cavs_filter_ch)(uint8_t *pix, int stride, int alpha, int beta, int tc, int bs1, int bs2);
00349     void (*cavs_idct8_add)(uint8_t *dst, DCTELEM *block, int stride);
00350 
00351     me_cmp_func pix_abs[2][4];
00352 
00353     /* huffyuv specific */
00354     void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
00355     void (*add_bytes_l2)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 16*/, int w);
00356     void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
00361     void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
00362     void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
00363     int  (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
00364     void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
00365     /* this might write to dst[w] */
00366     void (*add_png_paeth_prediction)(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp);
00367     void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
00368 
00369     void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
00370     void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
00371 
00372     void (*h261_loop_filter)(uint8_t *src, int stride);
00373 
00374     void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale);
00375     void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale);
00376 
00377     void (*vp3_idct_dc_add)(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
00378     void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00379     void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00380 
00381     void (*vp6_filter_diag4)(uint8_t *dst, uint8_t *src, int stride,
00382                              const int16_t *h_weights,const int16_t *v_weights);
00383 
00384     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
00385     void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
00386     void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
00387     /* no alignment needed */
00388     void (*lpc_compute_autocorr)(const int32_t *data, int len, int lag, double *autoc);
00389     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
00390     void (*vector_fmul)(float *dst, const float *src, int len);
00391     void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
00392     /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
00393     void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
00394     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
00395     void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len);
00396     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
00397     void (*int32_to_float_fmul_scalar)(float *dst, const int *src, float mul, int len);
00398     void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
00407     void (*vector_fmul_scalar)(float *dst, const float *src, float mul,
00408                                int len);
00421     void (*vector_fmul_sv_scalar[2])(float *dst, const float *src,
00422                                      const float **sv, float mul, int len);
00433     void (*sv_fmul_scalar[2])(float *dst, const float **sv,
00434                               float mul, int len);
00441     float (*scalarproduct_float)(const float *v1, const float *v2, int len);
00448     void (*butterflies_float)(float *restrict v1, float *restrict v2, int len);
00449 
00450     /* C version: convert floats from the range [384.0,386.0] to ints in [-32768,32767]
00451      * simd versions: convert floats from [-32768.0,32767.0] without rescaling and arrays are 16byte aligned */
00452     void (*float_to_int16)(int16_t *dst, const float *src, long len);
00453     void (*float_to_int16_interleave)(int16_t *dst, const float **src, long len, int channels);
00454 
00455     /* (I)DCT */
00456     void (*fdct)(DCTELEM *block/* align 16*/);
00457     void (*fdct248)(DCTELEM *block/* align 16*/);
00458 
00459     /* IDCT really*/
00460     void (*idct)(DCTELEM *block/* align 16*/);
00461 
00467     void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00468 
00473     void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00474 
00487     uint8_t idct_permutation[64];
00488     int idct_permutation_type;
00489 #define FF_NO_IDCT_PERM 1
00490 #define FF_LIBMPEG2_IDCT_PERM 2
00491 #define FF_SIMPLE_IDCT_PERM 3
00492 #define FF_TRANSPOSE_IDCT_PERM 4
00493 #define FF_PARTTRANS_IDCT_PERM 5
00494 #define FF_SSE2_IDCT_PERM 6
00495 
00496     int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
00497     void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
00498 #define BASIS_SHIFT 16
00499 #define RECON_SHIFT 6
00500 
00501     void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
00502 #define EDGE_WIDTH 16
00503 
00504     void (*prefetch)(void *mem, int stride, int h);
00505 
00506     void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00507 
00508     /* mlp/truehd functions */
00509     void (*mlp_filter_channel)(int32_t *state, const int32_t *coeff,
00510                                int firorder, int iirorder,
00511                                unsigned int filter_shift, int32_t mask, int blocksize,
00512                                int32_t *sample_buffer);
00513 
00514     /* vc1 functions */
00515     void (*vc1_inv_trans_8x8)(DCTELEM *b);
00516     void (*vc1_inv_trans_8x4)(uint8_t *dest, int line_size, DCTELEM *block);
00517     void (*vc1_inv_trans_4x8)(uint8_t *dest, int line_size, DCTELEM *block);
00518     void (*vc1_inv_trans_4x4)(uint8_t *dest, int line_size, DCTELEM *block);
00519     void (*vc1_inv_trans_8x8_dc)(uint8_t *dest, int line_size, DCTELEM *block);
00520     void (*vc1_inv_trans_8x4_dc)(uint8_t *dest, int line_size, DCTELEM *block);
00521     void (*vc1_inv_trans_4x8_dc)(uint8_t *dest, int line_size, DCTELEM *block);
00522     void (*vc1_inv_trans_4x4_dc)(uint8_t *dest, int line_size, DCTELEM *block);
00523     void (*vc1_v_overlap)(uint8_t* src, int stride);
00524     void (*vc1_h_overlap)(uint8_t* src, int stride);
00525     void (*vc1_v_loop_filter4)(uint8_t *src, int stride, int pq);
00526     void (*vc1_h_loop_filter4)(uint8_t *src, int stride, int pq);
00527     void (*vc1_v_loop_filter8)(uint8_t *src, int stride, int pq);
00528     void (*vc1_h_loop_filter8)(uint8_t *src, int stride, int pq);
00529     void (*vc1_v_loop_filter16)(uint8_t *src, int stride, int pq);
00530     void (*vc1_h_loop_filter16)(uint8_t *src, int stride, int pq);
00531     /* put 8x8 block with bicubic interpolation and quarterpel precision
00532      * last argument is actually round value instead of height
00533      */
00534     op_pixels_func put_vc1_mspel_pixels_tab[16];
00535     op_pixels_func avg_vc1_mspel_pixels_tab[16];
00536 
00537     /* intrax8 functions */
00538     void (*x8_spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize);
00539     void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
00540            int * range, int * sum,  int edges);
00541 
00547     int32_t (*scalarproduct_int16)(int16_t *v1, int16_t *v2/*align 16*/, int len, int shift);
00548     /* ape functions */
00554     int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, int16_t *v2, int16_t *v3, int len, int mul);
00555 
00556     /* rv30 functions */
00557     qpel_mc_func put_rv30_tpel_pixels_tab[4][16];
00558     qpel_mc_func avg_rv30_tpel_pixels_tab[4][16];
00559 
00560     /* rv40 functions */
00561     qpel_mc_func put_rv40_qpel_pixels_tab[4][16];
00562     qpel_mc_func avg_rv40_qpel_pixels_tab[4][16];
00563     h264_chroma_mc_func put_rv40_chroma_pixels_tab[3];
00564     h264_chroma_mc_func avg_rv40_chroma_pixels_tab[3];
00565 
00566     /* bink functions */
00567     op_fill_func fill_block_tab[2];
00568     void (*scale_block)(const uint8_t src[64]/*align 8*/, uint8_t *dst/*align 8*/, int linesize);
00569 } DSPContext;
00570 
00571 void dsputil_static_init(void);
00572 void dsputil_init(DSPContext* p, AVCodecContext *avctx);
00573 
00574 int ff_check_alignment(void);
00575 
00580 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
00581 
00582 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
00583 
00584 #define         BYTE_VEC32(c)   ((c)*0x01010101UL)
00585 
00586 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
00587 {
00588     return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00589 }
00590 
00591 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
00592 {
00593     return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00594 }
00595 
00596 static inline int get_penalty_factor(int lambda, int lambda2, int type){
00597     switch(type&0xFF){
00598     default:
00599     case FF_CMP_SAD:
00600         return lambda>>FF_LAMBDA_SHIFT;
00601     case FF_CMP_DCT:
00602         return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
00603     case FF_CMP_W53:
00604         return (4*lambda)>>(FF_LAMBDA_SHIFT);
00605     case FF_CMP_W97:
00606         return (2*lambda)>>(FF_LAMBDA_SHIFT);
00607     case FF_CMP_SATD:
00608     case FF_CMP_DCT264:
00609         return (2*lambda)>>FF_LAMBDA_SHIFT;
00610     case FF_CMP_RD:
00611     case FF_CMP_PSNR:
00612     case FF_CMP_SSE:
00613     case FF_CMP_NSSE:
00614         return lambda2>>FF_LAMBDA_SHIFT;
00615     case FF_CMP_BIT:
00616         return 1;
00617     }
00618 }
00619 
00625 #define emms_c()
00626 
00627 /* should be defined by architectures supporting
00628    one or more MultiMedia extension */
00629 int mm_support(void);
00630 extern int mm_flags;
00631 
00632 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
00633 void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
00634 void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
00635 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
00636 void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
00637 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
00638 void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
00639 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
00640 void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
00641 
00642 void ff_dsputil_init_dwt(DSPContext *c);
00643 void ff_cavsdsp_init(DSPContext* c, AVCodecContext *avctx);
00644 void ff_rv30dsp_init(DSPContext* c, AVCodecContext *avctx);
00645 void ff_rv40dsp_init(DSPContext* c, AVCodecContext *avctx);
00646 void ff_vc1dsp_init(DSPContext* c, AVCodecContext *avctx);
00647 void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx);
00648 void ff_mlp_init(DSPContext* c, AVCodecContext *avctx);
00649 void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx);
00650 
00651 #if HAVE_MMX
00652 
00653 #undef emms_c
00654 
00655 static inline void emms(void)
00656 {
00657     __asm__ volatile ("emms;":::"memory");
00658 }
00659 
00660 
00661 #define emms_c() \
00662 {\
00663     if (mm_flags & FF_MM_MMX)\
00664         emms();\
00665 }
00666 
00667 #elif ARCH_ARM
00668 
00669 #if HAVE_NEON
00670 #   define STRIDE_ALIGN 16
00671 #endif
00672 
00673 #elif ARCH_PPC
00674 
00675 #define STRIDE_ALIGN 16
00676 
00677 #elif HAVE_MMI
00678 
00679 #define STRIDE_ALIGN 16
00680 
00681 #else
00682 
00683 #define mm_flags 0
00684 #define mm_support() 0
00685 
00686 #endif
00687 
00688 #ifndef STRIDE_ALIGN
00689 #   define STRIDE_ALIGN 8
00690 #endif
00691 
00692 #define LOCAL_ALIGNED(a, t, v, s, ...)                          \
00693     uint8_t la_##v[sizeof(t s __VA_ARGS__) + (a)];              \
00694     t (*v) __VA_ARGS__ = (void *)FFALIGN((uintptr_t)la_##v, a)
00695 
00696 #if HAVE_LOCAL_ALIGNED_8
00697 #   define LOCAL_ALIGNED_8(t, v, s, ...) DECLARE_ALIGNED(8, t, v) s __VA_ARGS__
00698 #else
00699 #   define LOCAL_ALIGNED_8(t, v, s, ...) LOCAL_ALIGNED(8, t, v, s, __VA_ARGS__)
00700 #endif
00701 
00702 #if HAVE_LOCAL_ALIGNED_16
00703 #   define LOCAL_ALIGNED_16(t, v, s, ...) DECLARE_ALIGNED(16, t, v) s __VA_ARGS__
00704 #else
00705 #   define LOCAL_ALIGNED_16(t, v, s, ...) LOCAL_ALIGNED(16, t, v, s, __VA_ARGS__)
00706 #endif
00707 
00708 /* PSNR */
00709 void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
00710               int orig_linesize[3], int coded_linesize,
00711               AVCodecContext *avctx);
00712 
00713 #define WRAPPER8_16(name8, name16)\
00714 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00715     return name8(s, dst           , src           , stride, h)\
00716           +name8(s, dst+8         , src+8         , stride, h);\
00717 }
00718 
00719 #define WRAPPER8_16_SQ(name8, name16)\
00720 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00721     int score=0;\
00722     score +=name8(s, dst           , src           , stride, 8);\
00723     score +=name8(s, dst+8         , src+8         , stride, 8);\
00724     if(h==16){\
00725         dst += 8*stride;\
00726         src += 8*stride;\
00727         score +=name8(s, dst           , src           , stride, 8);\
00728         score +=name8(s, dst+8         , src+8         , stride, 8);\
00729     }\
00730     return score;\
00731 }
00732 
00733 
00734 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00735 {
00736     int i;
00737     for(i=0; i<h; i++)
00738     {
00739         AV_WN16(dst   , AV_RN16(src   ));
00740         dst+=dstStride;
00741         src+=srcStride;
00742     }
00743 }
00744 
00745 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00746 {
00747     int i;
00748     for(i=0; i<h; i++)
00749     {
00750         AV_WN32(dst   , AV_RN32(src   ));
00751         dst+=dstStride;
00752         src+=srcStride;
00753     }
00754 }
00755 
00756 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00757 {
00758     int i;
00759     for(i=0; i<h; i++)
00760     {
00761         AV_WN32(dst   , AV_RN32(src   ));
00762         AV_WN32(dst+4 , AV_RN32(src+4 ));
00763         dst+=dstStride;
00764         src+=srcStride;
00765     }
00766 }
00767 
00768 static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00769 {
00770     int i;
00771     for(i=0; i<h; i++)
00772     {
00773         AV_WN32(dst   , AV_RN32(src   ));
00774         AV_WN32(dst+4 , AV_RN32(src+4 ));
00775         dst[8]= src[8];
00776         dst+=dstStride;
00777         src+=srcStride;
00778     }
00779 }
00780 
00781 static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00782 {
00783     int i;
00784     for(i=0; i<h; i++)
00785     {
00786         AV_WN32(dst   , AV_RN32(src   ));
00787         AV_WN32(dst+4 , AV_RN32(src+4 ));
00788         AV_WN32(dst+8 , AV_RN32(src+8 ));
00789         AV_WN32(dst+12, AV_RN32(src+12));
00790         dst+=dstStride;
00791         src+=srcStride;
00792     }
00793 }
00794 
00795 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00796 {
00797     int i;
00798     for(i=0; i<h; i++)
00799     {
00800         AV_WN32(dst   , AV_RN32(src   ));
00801         AV_WN32(dst+4 , AV_RN32(src+4 ));
00802         AV_WN32(dst+8 , AV_RN32(src+8 ));
00803         AV_WN32(dst+12, AV_RN32(src+12));
00804         dst[16]= src[16];
00805         dst+=dstStride;
00806         src+=srcStride;
00807     }
00808 }
00809 
00810 #endif /* AVCODEC_DSPUTIL_H */