libavcodec/dsputil.h
Go to the documentation of this 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 Libav.
00007  *
00008  * Libav 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  * Libav 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 Libav; 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_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 j_rev_dct (DCTELEM *data);
00049 void j_rev_dct4 (DCTELEM *data);
00050 void j_rev_dct2 (DCTELEM *data);
00051 void 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 libav-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 #if ARCH_PPC
00201 
00202     DECLARE_ALIGNED(16, uint8_t, inverse)[64];
00203 #endif
00204 } ScanTable;
00205 
00206 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
00207 void ff_init_scantable_permutation(uint8_t *idct_permutation,
00208                                    int idct_permutation_type);
00209 
00210 #define EMULATED_EDGE(depth) \
00211 void ff_emulated_edge_mc_ ## depth (uint8_t *buf, const uint8_t *src, int linesize,\
00212                          int block_w, int block_h,\
00213                          int src_x, int src_y, int w, int h);
00214 
00215 EMULATED_EDGE(8)
00216 EMULATED_EDGE(9)
00217 EMULATED_EDGE(10)
00218 
00219 void ff_add_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
00220 void ff_put_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
00221 void ff_put_signed_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
00222 
00226 typedef struct DSPContext {
00230     int dct_bits;
00231 
00232     /* pixel ops : interface with DCT */
00233     void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
00234     void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
00235     void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00236     void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00237     void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00238     void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
00239     void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
00240     int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
00253     void (*emulated_edge_mc)(uint8_t *buf, const uint8_t *src, int linesize,
00254                              int block_w, int block_h,
00255                              int src_x, int src_y, int w, int h);
00259     void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
00263     void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
00264                     int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
00265     void (*clear_block)(DCTELEM *block/*align 16*/);
00266     void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
00267     int (*pix_sum)(uint8_t * pix, int line_size);
00268     int (*pix_norm1)(uint8_t * pix, int line_size);
00269 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
00270 
00271     me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
00272     me_cmp_func sse[6];
00273     me_cmp_func hadamard8_diff[6];
00274     me_cmp_func dct_sad[6];
00275     me_cmp_func quant_psnr[6];
00276     me_cmp_func bit[6];
00277     me_cmp_func rd[6];
00278     me_cmp_func vsad[6];
00279     me_cmp_func vsse[6];
00280     me_cmp_func nsse[6];
00281     me_cmp_func w53[6];
00282     me_cmp_func w97[6];
00283     me_cmp_func dct_max[6];
00284     me_cmp_func dct264_sad[6];
00285 
00286     me_cmp_func me_pre_cmp[6];
00287     me_cmp_func me_cmp[6];
00288     me_cmp_func me_sub_cmp[6];
00289     me_cmp_func mb_cmp[6];
00290     me_cmp_func ildct_cmp[6]; //only width 16 used
00291     me_cmp_func frame_skip_cmp[6]; //only width 8 used
00292 
00293     int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
00294                              int size);
00295 
00306     op_pixels_func put_pixels_tab[4][4];
00307 
00318     op_pixels_func avg_pixels_tab[4][4];
00319 
00330     op_pixels_func put_no_rnd_pixels_tab[4][4];
00331 
00342     op_pixels_func avg_no_rnd_pixels_tab[4][4];
00343 
00344     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);
00345 
00356     tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00357     tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00358 
00359     qpel_mc_func put_qpel_pixels_tab[2][16];
00360     qpel_mc_func avg_qpel_pixels_tab[2][16];
00361     qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
00362     qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
00363     qpel_mc_func put_mspel_pixels_tab[8];
00364 
00368     h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
00369     h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
00370 
00371     qpel_mc_func put_h264_qpel_pixels_tab[4][16];
00372     qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
00373 
00374     qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
00375     qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];
00376 
00377     me_cmp_func pix_abs[2][4];
00378 
00379     /* huffyuv specific */
00380     void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
00381     void (*add_bytes_l2)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 16*/, int w);
00382     void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
00387     void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
00388     void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
00389     int  (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
00390     void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
00391     /* this might write to dst[w] */
00392     void (*add_png_paeth_prediction)(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp);
00393     void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
00394     void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);
00395 
00396     void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
00397     void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
00398 
00399     void (*h261_loop_filter)(uint8_t *src, int stride);
00400 
00401     void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale);
00402     void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale);
00403 
00404     void (*vp3_idct_dc_add)(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
00405     void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00406     void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00407 
00408     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
00409     void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
00410     void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
00411     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
00412     void (*vector_fmul)(float *dst, const float *src0, const float *src1, int len);
00413     void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
00414     /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
00415     void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
00416     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
00417     void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, int len);
00418     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
00419     void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
00428     void (*vector_fmul_scalar)(float *dst, const float *src, float mul,
00429                                int len);
00439     void (*vector_fmac_scalar)(float *dst, const float *src, float mul,
00440                                int len);
00447     float (*scalarproduct_float)(const float *v1, const float *v2, int len);
00454     void (*butterflies_float)(float *restrict v1, float *restrict v2, int len);
00455 
00470     void (*butterflies_float_interleave)(float *dst, const float *src0,
00471                                          const float *src1, int len);
00472 
00473     /* (I)DCT */
00474     void (*fdct)(DCTELEM *block/* align 16*/);
00475     void (*fdct248)(DCTELEM *block/* align 16*/);
00476 
00477     /* IDCT really*/
00478     void (*idct)(DCTELEM *block/* align 16*/);
00479 
00485     void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00486 
00491     void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00492 
00505     uint8_t idct_permutation[64];
00506     int idct_permutation_type;
00507 #define FF_NO_IDCT_PERM 1
00508 #define FF_LIBMPEG2_IDCT_PERM 2
00509 #define FF_SIMPLE_IDCT_PERM 3
00510 #define FF_TRANSPOSE_IDCT_PERM 4
00511 #define FF_PARTTRANS_IDCT_PERM 5
00512 #define FF_SSE2_IDCT_PERM 6
00513 
00514     int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
00515     void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
00516 #define BASIS_SHIFT 16
00517 #define RECON_SHIFT 6
00518 
00519     void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
00520 #define EDGE_WIDTH 16
00521 #define EDGE_TOP    1
00522 #define EDGE_BOTTOM 2
00523 
00524     void (*prefetch)(void *mem, int stride, int h);
00525 
00526     void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00527 
00528     /* mlp/truehd functions */
00529     void (*mlp_filter_channel)(int32_t *state, const int32_t *coeff,
00530                                int firorder, int iirorder,
00531                                unsigned int filter_shift, int32_t mask, int blocksize,
00532                                int32_t *sample_buffer);
00533 
00534     /* intrax8 functions */
00535     void (*x8_spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize);
00536     void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
00537            int * range, int * sum,  int edges);
00538 
00544     int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len, int shift);
00545     /* ape functions */
00551     int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
00552 
00564     void (*apply_window_int16)(int16_t *output, const int16_t *input,
00565                                const int16_t *window, unsigned int len);
00566 
00580     void (*vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min,
00581                               int32_t max, unsigned int len);
00582 
00583     op_fill_func fill_block_tab[2];
00584 } DSPContext;
00585 
00586 void dsputil_static_init(void);
00587 void dsputil_init(DSPContext* p, AVCodecContext *avctx);
00588 
00589 int ff_check_alignment(void);
00590 
00595 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
00596 
00597 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
00598 
00599 #define         BYTE_VEC32(c)   ((c)*0x01010101UL)
00600 #define         BYTE_VEC64(c)   ((c)*0x0001000100010001UL)
00601 
00602 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
00603 {
00604     return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00605 }
00606 
00607 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
00608 {
00609     return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00610 }
00611 
00612 static inline uint64_t rnd_avg64(uint64_t a, uint64_t b)
00613 {
00614     return (a | b) - (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
00615 }
00616 
00617 static inline uint64_t no_rnd_avg64(uint64_t a, uint64_t b)
00618 {
00619     return (a & b) + (((a ^ b) & ~BYTE_VEC64(0x01)) >> 1);
00620 }
00621 
00622 static inline int get_penalty_factor(int lambda, int lambda2, int type){
00623     switch(type&0xFF){
00624     default:
00625     case FF_CMP_SAD:
00626         return lambda>>FF_LAMBDA_SHIFT;
00627     case FF_CMP_DCT:
00628         return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
00629     case FF_CMP_W53:
00630         return (4*lambda)>>(FF_LAMBDA_SHIFT);
00631     case FF_CMP_W97:
00632         return (2*lambda)>>(FF_LAMBDA_SHIFT);
00633     case FF_CMP_SATD:
00634     case FF_CMP_DCT264:
00635         return (2*lambda)>>FF_LAMBDA_SHIFT;
00636     case FF_CMP_RD:
00637     case FF_CMP_PSNR:
00638     case FF_CMP_SSE:
00639     case FF_CMP_NSSE:
00640         return lambda2>>FF_LAMBDA_SHIFT;
00641     case FF_CMP_BIT:
00642         return 1;
00643     }
00644 }
00645 
00646 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
00647 void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
00648 void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
00649 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
00650 void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
00651 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
00652 void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
00653 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
00654 void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
00655 
00656 void ff_dsputil_init_dwt(DSPContext *c);
00657 void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx);
00658 void ff_mlp_init(DSPContext* c, AVCodecContext *avctx);
00659 void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx);
00660 
00661 #if ARCH_ARM
00662 
00663 #if HAVE_NEON
00664 #   define STRIDE_ALIGN 16
00665 #endif
00666 
00667 #elif ARCH_PPC
00668 
00669 #define STRIDE_ALIGN 16
00670 
00671 #elif HAVE_MMI
00672 
00673 #define STRIDE_ALIGN 16
00674 
00675 #endif
00676 
00677 #ifndef STRIDE_ALIGN
00678 #   define STRIDE_ALIGN 8
00679 #endif
00680 
00681 #define LOCAL_ALIGNED_A(a, t, v, s, o, ...)             \
00682     uint8_t la_##v[sizeof(t s o) + (a)];                \
00683     t (*v) o = (void *)FFALIGN((uintptr_t)la_##v, a)
00684 
00685 #define LOCAL_ALIGNED_D(a, t, v, s, o, ...) DECLARE_ALIGNED(a, t, v) s o
00686 
00687 #define LOCAL_ALIGNED(a, t, v, ...) LOCAL_ALIGNED_A(a, t, v, __VA_ARGS__,,)
00688 
00689 #if HAVE_LOCAL_ALIGNED_8
00690 #   define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED_D(8, t, v, __VA_ARGS__,,)
00691 #else
00692 #   define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED(8, t, v, __VA_ARGS__)
00693 #endif
00694 
00695 #if HAVE_LOCAL_ALIGNED_16
00696 #   define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED_D(16, t, v, __VA_ARGS__,,)
00697 #else
00698 #   define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED(16, t, v, __VA_ARGS__)
00699 #endif
00700 
00701 #define WRAPPER8_16(name8, name16)\
00702 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00703     return name8(s, dst           , src           , stride, h)\
00704           +name8(s, dst+8         , src+8         , stride, h);\
00705 }
00706 
00707 #define WRAPPER8_16_SQ(name8, name16)\
00708 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00709     int score=0;\
00710     score +=name8(s, dst           , src           , stride, 8);\
00711     score +=name8(s, dst+8         , src+8         , stride, 8);\
00712     if(h==16){\
00713         dst += 8*stride;\
00714         src += 8*stride;\
00715         score +=name8(s, dst           , src           , stride, 8);\
00716         score +=name8(s, dst+8         , src+8         , stride, 8);\
00717     }\
00718     return score;\
00719 }
00720 
00721 
00722 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00723 {
00724     int i;
00725     for(i=0; i<h; i++)
00726     {
00727         AV_WN16(dst   , AV_RN16(src   ));
00728         dst+=dstStride;
00729         src+=srcStride;
00730     }
00731 }
00732 
00733 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00734 {
00735     int i;
00736     for(i=0; i<h; i++)
00737     {
00738         AV_WN32(dst   , AV_RN32(src   ));
00739         dst+=dstStride;
00740         src+=srcStride;
00741     }
00742 }
00743 
00744 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00745 {
00746     int i;
00747     for(i=0; i<h; i++)
00748     {
00749         AV_WN32(dst   , AV_RN32(src   ));
00750         AV_WN32(dst+4 , AV_RN32(src+4 ));
00751         dst+=dstStride;
00752         src+=srcStride;
00753     }
00754 }
00755 
00756 static inline void copy_block9(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[8]= src[8];
00764         dst+=dstStride;
00765         src+=srcStride;
00766     }
00767 }
00768 
00769 static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00770 {
00771     int i;
00772     for(i=0; i<h; i++)
00773     {
00774         AV_WN32(dst   , AV_RN32(src   ));
00775         AV_WN32(dst+4 , AV_RN32(src+4 ));
00776         AV_WN32(dst+8 , AV_RN32(src+8 ));
00777         AV_WN32(dst+12, AV_RN32(src+12));
00778         dst+=dstStride;
00779         src+=srcStride;
00780     }
00781 }
00782 
00783 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00784 {
00785     int i;
00786     for(i=0; i<h; i++)
00787     {
00788         AV_WN32(dst   , AV_RN32(src   ));
00789         AV_WN32(dst+4 , AV_RN32(src+4 ));
00790         AV_WN32(dst+8 , AV_RN32(src+8 ));
00791         AV_WN32(dst+12, AV_RN32(src+12));
00792         dst[16]= src[16];
00793         dst+=dstStride;
00794         src+=srcStride;
00795     }
00796 }
00797 
00798 #endif /* AVCODEC_DSPUTIL_H */