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 ff_fdct_ifast (DCTELEM *data);
00042 void ff_fdct_ifast248 (DCTELEM *data);
00043 void ff_jpeg_fdct_islow_8(DCTELEM *data);
00044 void ff_jpeg_fdct_islow_10(DCTELEM *data);
00045 void ff_fdct248_islow_8(DCTELEM *data);
00046 void ff_fdct248_islow_10(DCTELEM *data);
00047 
00048 void ff_j_rev_dct (DCTELEM *data);
00049 void ff_j_rev_dct4 (DCTELEM *data);
00050 void ff_j_rev_dct2 (DCTELEM *data);
00051 void ff_j_rev_dct1 (DCTELEM *data);
00052 void ff_wmv2_idct_c(DCTELEM *data);
00053 
00054 void ff_fdct_mmx(DCTELEM *block);
00055 void ff_fdct_mmx2(DCTELEM *block);
00056 void ff_fdct_sse2(DCTELEM *block);
00057 
00058 #define H264_IDCT(depth) \
00059 void ff_h264_idct8_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
00060 void ff_h264_idct_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
00061 void ff_h264_idct8_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
00062 void ff_h264_idct_dc_add_ ## depth ## _c(uint8_t *dst, DCTELEM *block, int stride);\
00063 void ff_h264_idct_add16_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
00064 void ff_h264_idct_add16intra_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
00065 void ff_h264_idct8_add4_ ## depth ## _c(uint8_t *dst, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
00066 void ff_h264_idct_add8_422_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
00067 void ff_h264_idct_add8_ ## depth ## _c(uint8_t **dest, const int *blockoffset, DCTELEM *block, int stride, const uint8_t nnzc[6*8]);\
00068 void ff_h264_luma_dc_dequant_idct_ ## depth ## _c(DCTELEM *output, DCTELEM *input, int qmul);\
00069 void ff_h264_chroma422_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);\
00070 void ff_h264_chroma_dc_dequant_idct_ ## depth ## _c(DCTELEM *block, int qmul);
00071 
00072 H264_IDCT( 8)
00073 H264_IDCT( 9)
00074 H264_IDCT(10)
00075 
00076 void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp);
00077 void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
00078 
00079 /* encoding scans */
00080 extern const uint8_t ff_alternate_horizontal_scan[64];
00081 extern const uint8_t ff_alternate_vertical_scan[64];
00082 extern const uint8_t ff_zigzag_direct[64];
00083 extern const uint8_t ff_zigzag248_direct[64];
00084 
00085 /* pixel operations */
00086 #define MAX_NEG_CROP 1024
00087 
00088 /* temporary */
00089 extern uint32_t ff_squareTbl[512];
00090 extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
00091 
00092 #define PUTAVG_PIXELS(depth)\
00093 void ff_put_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
00094 void ff_avg_pixels8x8_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
00095 void ff_put_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);\
00096 void ff_avg_pixels16x16_ ## depth ## _c(uint8_t *dst, uint8_t *src, int stride);
00097 
00098 PUTAVG_PIXELS( 8)
00099 PUTAVG_PIXELS( 9)
00100 PUTAVG_PIXELS(10)
00101 
00102 #define ff_put_pixels8x8_c ff_put_pixels8x8_8_c
00103 #define ff_avg_pixels8x8_c ff_avg_pixels8x8_8_c
00104 #define ff_put_pixels16x16_c ff_put_pixels16x16_8_c
00105 #define ff_avg_pixels16x16_c ff_avg_pixels16x16_8_c
00106 
00107 /* VP3 DSP functions */
00108 void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
00109 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00110 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00111 void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
00112 
00113 void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
00114 void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
00115 
00116 /* EA functions */
00117 void ff_ea_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);
00118 
00119 /* RV40 functions */
00120 void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
00121 void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, int stride);
00122 void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);
00123 void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, int stride);
00124 
00125 /* 1/2^n downscaling functions from imgconvert.c */
00126 void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00127 void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00128 void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00129 
00130 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
00131               int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
00132 
00133 /* minimum alignment rules ;)
00134 If you notice errors in the align stuff, need more alignment for some ASM code
00135 for some CPU or need to use a function with less aligned data then send a mail
00136 to the ffmpeg-devel mailing list, ...
00137 
00138 !warning These alignments might not match reality, (missing attribute((align))
00139 stuff somewhere possible).
00140 I (Michael) did not check them, these are just the alignments which I think
00141 could be reached easily ...
00142 
00143 !future video codecs might need functions with less strict alignment
00144 */
00145 
00146 /*
00147 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
00148 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
00149 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
00150 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
00151 void clear_blocks_c(DCTELEM *blocks);
00152 */
00153 
00154 /* add and put pixel (decoding) */
00155 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
00156 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller than 4
00157 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
00158 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);
00159 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
00160 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
00161 
00162 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
00163 
00164 #define DEF_OLD_QPEL(name)\
00165 void ff_put_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
00166 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
00167 void ff_avg_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
00168 
00169 DEF_OLD_QPEL(qpel16_mc11_old_c)
00170 DEF_OLD_QPEL(qpel16_mc31_old_c)
00171 DEF_OLD_QPEL(qpel16_mc12_old_c)
00172 DEF_OLD_QPEL(qpel16_mc32_old_c)
00173 DEF_OLD_QPEL(qpel16_mc13_old_c)
00174 DEF_OLD_QPEL(qpel16_mc33_old_c)
00175 DEF_OLD_QPEL(qpel8_mc11_old_c)
00176 DEF_OLD_QPEL(qpel8_mc31_old_c)
00177 DEF_OLD_QPEL(qpel8_mc12_old_c)
00178 DEF_OLD_QPEL(qpel8_mc32_old_c)
00179 DEF_OLD_QPEL(qpel8_mc13_old_c)
00180 DEF_OLD_QPEL(qpel8_mc33_old_c)
00181 
00182 #define CALL_2X_PIXELS(a, b, n)\
00183 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
00184     b(block  , pixels  , line_size, h);\
00185     b(block+n, pixels+n, line_size, h);\
00186 }
00187 
00188 /* motion estimation */
00189 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller than 2
00190 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
00191 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))*/;
00192 
00196 typedef struct ScanTable{
00197     const uint8_t *scantable;
00198     uint8_t permutated[64];
00199     uint8_t raster_end[64];
00200 } ScanTable;
00201 
00202 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
00203 void ff_init_scantable_permutation(uint8_t *idct_permutation,
00204                                    int idct_permutation_type);
00205 
00206 #define EMULATED_EDGE(depth) \
00207 void ff_emulated_edge_mc_ ## depth (uint8_t *buf, const uint8_t *src, int linesize,\
00208                          int block_w, int block_h,\
00209                          int src_x, int src_y, int w, int h);
00210 
00211 EMULATED_EDGE(8)
00212 EMULATED_EDGE(9)
00213 EMULATED_EDGE(10)
00214 
00215 void ff_add_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
00216 void ff_put_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
00217 void ff_put_signed_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
00218 
00222 typedef struct DSPContext {
00226     int dct_bits;
00227 
00228     /* pixel ops : interface with DCT */
00229     void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
00230     void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
00231     void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00232     void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00233     void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00234     void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
00235     void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
00236     int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
00249     void (*emulated_edge_mc)(uint8_t *buf, const uint8_t *src, int linesize,
00250                              int block_w, int block_h,
00251                              int src_x, int src_y, int w, int h);
00255     void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
00259     void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
00260                     int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
00261     void (*clear_block)(DCTELEM *block/*align 16*/);
00262     void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
00263     int (*pix_sum)(uint8_t * pix, int line_size);
00264     int (*pix_norm1)(uint8_t * pix, int line_size);
00265 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
00266 
00267     me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
00268     me_cmp_func sse[6];
00269     me_cmp_func hadamard8_diff[6];
00270     me_cmp_func dct_sad[6];
00271     me_cmp_func quant_psnr[6];
00272     me_cmp_func bit[6];
00273     me_cmp_func rd[6];
00274     me_cmp_func vsad[6];
00275     me_cmp_func vsse[6];
00276     me_cmp_func nsse[6];
00277     me_cmp_func w53[6];
00278     me_cmp_func w97[6];
00279     me_cmp_func dct_max[6];
00280     me_cmp_func dct264_sad[6];
00281 
00282     me_cmp_func me_pre_cmp[6];
00283     me_cmp_func me_cmp[6];
00284     me_cmp_func me_sub_cmp[6];
00285     me_cmp_func mb_cmp[6];
00286     me_cmp_func ildct_cmp[6]; //only width 16 used
00287     me_cmp_func frame_skip_cmp[6]; //only width 8 used
00288 
00289     int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
00290                              int size);
00291 
00302     op_pixels_func put_pixels_tab[4][4];
00303 
00314     op_pixels_func avg_pixels_tab[4][4];
00315 
00326     op_pixels_func put_no_rnd_pixels_tab[4][4];
00327 
00338     op_pixels_func avg_no_rnd_pixels_tab[4][4];
00339 
00340     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);
00341 
00352     tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00353     tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00354 
00355     qpel_mc_func put_qpel_pixels_tab[2][16];
00356     qpel_mc_func avg_qpel_pixels_tab[2][16];
00357     qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
00358     qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
00359     qpel_mc_func put_mspel_pixels_tab[8];
00360 
00364     h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
00365     h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
00366 
00367     qpel_mc_func put_h264_qpel_pixels_tab[4][16];
00368     qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
00369 
00370     qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
00371     qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];
00372 
00373     me_cmp_func pix_abs[2][4];
00374 
00375     /* huffyuv specific */
00376     void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
00377     void (*diff_bytes)(uint8_t *dst/*align 16*/, const uint8_t *src1/*align 16*/, const uint8_t *src2/*align 1*/,int w);
00382     void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
00383     void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
00384     int  (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
00385     void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
00386     /* this might write to dst[w] */
00387     void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
00388     void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);
00389 
00390     void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
00391     void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
00392 
00393     void (*h261_loop_filter)(uint8_t *src, int stride);
00394 
00395     void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale);
00396     void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale);
00397 
00398     void (*vp3_idct_dc_add)(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
00399     void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00400     void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00401 
00402     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
00403     void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
00404     void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
00405     /* assume len is a multiple of 16, and arrays are 32-byte aligned */
00406     void (*vector_fmul)(float *dst, const float *src0, const float *src1, int len);
00407     void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
00408     /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
00409     void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
00410     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
00411     void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, int len);
00412     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
00413     void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
00422     void (*vector_fmul_scalar)(float *dst, const float *src, float mul,
00423                                int len);
00433     void (*vector_fmac_scalar)(float *dst, const float *src, float mul,
00434                                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 
00464     void (*butterflies_float_interleave)(float *dst, const float *src0,
00465                                          const float *src1, int len);
00466 
00467     /* (I)DCT */
00468     void (*fdct)(DCTELEM *block/* align 16*/);
00469     void (*fdct248)(DCTELEM *block/* align 16*/);
00470 
00471     /* IDCT really*/
00472     void (*idct)(DCTELEM *block/* align 16*/);
00473 
00479     void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00480 
00485     void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00486 
00499     uint8_t idct_permutation[64];
00500     int idct_permutation_type;
00501 #define FF_NO_IDCT_PERM 1
00502 #define FF_LIBMPEG2_IDCT_PERM 2
00503 #define FF_SIMPLE_IDCT_PERM 3
00504 #define FF_TRANSPOSE_IDCT_PERM 4
00505 #define FF_PARTTRANS_IDCT_PERM 5
00506 #define FF_SSE2_IDCT_PERM 6
00507 
00508     int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
00509     void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
00510 #define BASIS_SHIFT 16
00511 #define RECON_SHIFT 6
00512 
00513     void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
00514 #define EDGE_WIDTH 16
00515 #define EDGE_TOP    1
00516 #define EDGE_BOTTOM 2
00517 
00518     void (*prefetch)(void *mem, int stride, int h);
00519 
00520     void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00521 
00522     /* mlp/truehd functions */
00523     void (*mlp_filter_channel)(int32_t *state, const int32_t *coeff,
00524                                int firorder, int iirorder,
00525                                unsigned int filter_shift, int32_t mask, int blocksize,
00526                                int32_t *sample_buffer);
00527 
00528     /* intrax8 functions */
00529     void (*x8_spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize);
00530     void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
00531            int * range, int * sum,  int edges);
00532 
00537     int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len);
00538     /* ape functions */
00544     int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
00545 
00557     void (*apply_window_int16)(int16_t *output, const int16_t *input,
00558                                const int16_t *window, unsigned int len);
00559 
00573     void (*vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min,
00574                               int32_t max, unsigned int len);
00575 
00576     op_fill_func fill_block_tab[2];
00577 } DSPContext;
00578 
00579 void ff_dsputil_static_init(void);
00580 void ff_dsputil_init(DSPContext* p, AVCodecContext *avctx);
00581 attribute_deprecated void dsputil_init(DSPContext* c, AVCodecContext *avctx);
00582 
00583 int ff_check_alignment(void);
00584 
00589 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
00590 
00591 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
00592 
00593 #define         BYTE_VEC32(c)   ((c)*0x01010101UL)
00594 #define         BYTE_VEC64(c)   ((c)*0x0001000100010001UL)
00595 
00596 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
00597 {
00598     return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00599 }
00600 
00601 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
00602 {
00603     return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00604 }
00605 
00606 static inline uint64_t rnd_avg64(uint64_t a, uint64_t b)
00607 {
00608     return (a | b) - (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
00609 }
00610 
00611 static inline uint64_t no_rnd_avg64(uint64_t a, uint64_t b)
00612 {
00613     return (a & b) + (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
00614 }
00615 
00616 static inline int get_penalty_factor(int lambda, int lambda2, int type){
00617     switch(type&0xFF){
00618     default:
00619     case FF_CMP_SAD:
00620         return lambda>>FF_LAMBDA_SHIFT;
00621     case FF_CMP_DCT:
00622         return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
00623     case FF_CMP_W53:
00624         return (4*lambda)>>(FF_LAMBDA_SHIFT);
00625     case FF_CMP_W97:
00626         return (2*lambda)>>(FF_LAMBDA_SHIFT);
00627     case FF_CMP_SATD:
00628     case FF_CMP_DCT264:
00629         return (2*lambda)>>FF_LAMBDA_SHIFT;
00630     case FF_CMP_RD:
00631     case FF_CMP_PSNR:
00632     case FF_CMP_SSE:
00633     case FF_CMP_NSSE:
00634         return lambda2>>FF_LAMBDA_SHIFT;
00635     case FF_CMP_BIT:
00636         return 1;
00637     }
00638 }
00639 
00640 void ff_dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
00641 void ff_dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
00642 void ff_dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
00643 void ff_dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
00644 void ff_dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
00645 void ff_dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
00646 void ff_dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
00647 void ff_dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
00648 
00649 void ff_dsputil_init_dwt(DSPContext *c);
00650 void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx);
00651 void ff_mlp_init(DSPContext* c, AVCodecContext *avctx);
00652 void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx);
00653 
00654 #if (ARCH_ARM && HAVE_NEON) || ARCH_PPC || HAVE_MMI || HAVE_MMX
00655 #   define STRIDE_ALIGN 16
00656 #else
00657 #   define STRIDE_ALIGN 8
00658 #endif
00659 
00660 #define LOCAL_ALIGNED_A(a, t, v, s, o, ...)             \
00661     uint8_t la_##v[sizeof(t s o) + (a)];                \
00662     t (*v) o = (void *)FFALIGN((uintptr_t)la_##v, a)
00663 
00664 #define LOCAL_ALIGNED_D(a, t, v, s, o, ...) DECLARE_ALIGNED(a, t, v) s o
00665 
00666 #define LOCAL_ALIGNED(a, t, v, ...) LOCAL_ALIGNED_A(a, t, v, __VA_ARGS__,,)
00667 
00668 #if HAVE_LOCAL_ALIGNED_8
00669 #   define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED_D(8, t, v, __VA_ARGS__,,)
00670 #else
00671 #   define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED(8, t, v, __VA_ARGS__)
00672 #endif
00673 
00674 #if HAVE_LOCAL_ALIGNED_16
00675 #   define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED_D(16, t, v, __VA_ARGS__,,)
00676 #else
00677 #   define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED(16, t, v, __VA_ARGS__)
00678 #endif
00679 
00680 #define WRAPPER8_16(name8, name16)\
00681 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00682     return name8(s, dst           , src           , stride, h)\
00683           +name8(s, dst+8         , src+8         , stride, h);\
00684 }
00685 
00686 #define WRAPPER8_16_SQ(name8, name16)\
00687 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00688     int score=0;\
00689     score +=name8(s, dst           , src           , stride, 8);\
00690     score +=name8(s, dst+8         , src+8         , stride, 8);\
00691     if(h==16){\
00692         dst += 8*stride;\
00693         src += 8*stride;\
00694         score +=name8(s, dst           , src           , stride, 8);\
00695         score +=name8(s, dst+8         , src+8         , stride, 8);\
00696     }\
00697     return score;\
00698 }
00699 
00700 
00701 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00702 {
00703     int i;
00704     for(i=0; i<h; i++)
00705     {
00706         AV_WN16(dst   , AV_RN16(src   ));
00707         dst+=dstStride;
00708         src+=srcStride;
00709     }
00710 }
00711 
00712 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00713 {
00714     int i;
00715     for(i=0; i<h; i++)
00716     {
00717         AV_WN32(dst   , AV_RN32(src   ));
00718         dst+=dstStride;
00719         src+=srcStride;
00720     }
00721 }
00722 
00723 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00724 {
00725     int i;
00726     for(i=0; i<h; i++)
00727     {
00728         AV_WN32(dst   , AV_RN32(src   ));
00729         AV_WN32(dst+4 , AV_RN32(src+4 ));
00730         dst+=dstStride;
00731         src+=srcStride;
00732     }
00733 }
00734 
00735 static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00736 {
00737     int i;
00738     for(i=0; i<h; i++)
00739     {
00740         AV_WN32(dst   , AV_RN32(src   ));
00741         AV_WN32(dst+4 , AV_RN32(src+4 ));
00742         dst[8]= src[8];
00743         dst+=dstStride;
00744         src+=srcStride;
00745     }
00746 }
00747 
00748 static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00749 {
00750     int i;
00751     for(i=0; i<h; i++)
00752     {
00753         AV_WN32(dst   , AV_RN32(src   ));
00754         AV_WN32(dst+4 , AV_RN32(src+4 ));
00755         AV_WN32(dst+8 , AV_RN32(src+8 ));
00756         AV_WN32(dst+12, AV_RN32(src+12));
00757         dst+=dstStride;
00758         src+=srcStride;
00759     }
00760 }
00761 
00762 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00763 {
00764     int i;
00765     for(i=0; i<h; i++)
00766     {
00767         AV_WN32(dst   , AV_RN32(src   ));
00768         AV_WN32(dst+4 , AV_RN32(src+4 ));
00769         AV_WN32(dst+8 , AV_RN32(src+8 ));
00770         AV_WN32(dst+12, AV_RN32(src+12));
00771         dst[16]= src[16];
00772         dst+=dstStride;
00773         src+=srcStride;
00774     }
00775 }
00776 
00777 #endif /* AVCODEC_DSPUTIL_H */