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_h264_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qmul);
00068 void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp);
00069 void ff_chroma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qmul);
00070 void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
00071 
00072 void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1,
00073                              const float *win, float add_bias, int len);
00074 void ff_float_to_int16_c(int16_t *dst, const float *src, long len);
00075 void ff_float_to_int16_interleave_c(int16_t *dst, const float **src, long len, int channels);
00076 
00077 /* encoding scans */
00078 extern const uint8_t ff_alternate_horizontal_scan[64];
00079 extern const uint8_t ff_alternate_vertical_scan[64];
00080 extern const uint8_t ff_zigzag_direct[64];
00081 extern const uint8_t ff_zigzag248_direct[64];
00082 
00083 /* pixel operations */
00084 #define MAX_NEG_CROP 1024
00085 
00086 /* temporary */
00087 extern uint32_t ff_squareTbl[512];
00088 extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
00089 
00090 void ff_put_pixels8x8_c(uint8_t *dst, uint8_t *src, int stride);
00091 void ff_avg_pixels8x8_c(uint8_t *dst, uint8_t *src, int stride);
00092 void ff_put_pixels16x16_c(uint8_t *dst, uint8_t *src, int stride);
00093 void ff_avg_pixels16x16_c(uint8_t *dst, uint8_t *src, int stride);
00094 
00095 /* VP3 DSP functions */
00096 void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
00097 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00098 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00099 void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
00100 
00101 void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
00102 void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
00103 
00104 /* Bink functions */
00105 void ff_bink_idct_c    (DCTELEM *block);
00106 void ff_bink_idct_add_c(uint8_t *dest, int linesize, DCTELEM *block);
00107 void ff_bink_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);
00108 
00109 /* EA functions */
00110 void ff_ea_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);
00111 
00112 /* 1/2^n downscaling functions from imgconvert.c */
00113 #if LIBAVCODEC_VERSION_MAJOR < 53
00114 
00117 attribute_deprecated
00118 void ff_img_copy_plane(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00119 #endif
00120 
00121 void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00122 void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00123 void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00124 
00125 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
00126               int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
00127 
00128 /* minimum alignment rules ;)
00129 If you notice errors in the align stuff, need more alignment for some ASM code
00130 for some CPU or need to use a function with less aligned data then send a mail
00131 to the ffmpeg-devel mailing list, ...
00132 
00133 !warning These alignments might not match reality, (missing attribute((align))
00134 stuff somewhere possible).
00135 I (Michael) did not check them, these are just the alignments which I think
00136 could be reached easily ...
00137 
00138 !future video codecs might need functions with less strict alignment
00139 */
00140 
00141 /*
00142 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
00143 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
00144 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
00145 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
00146 void clear_blocks_c(DCTELEM *blocks);
00147 */
00148 
00149 /* add and put pixel (decoding) */
00150 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
00151 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4
00152 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
00153 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);
00154 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
00155 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
00156 
00157 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
00158 
00159 #define DEF_OLD_QPEL(name)\
00160 void ff_put_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
00161 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
00162 void ff_avg_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
00163 
00164 DEF_OLD_QPEL(qpel16_mc11_old_c)
00165 DEF_OLD_QPEL(qpel16_mc31_old_c)
00166 DEF_OLD_QPEL(qpel16_mc12_old_c)
00167 DEF_OLD_QPEL(qpel16_mc32_old_c)
00168 DEF_OLD_QPEL(qpel16_mc13_old_c)
00169 DEF_OLD_QPEL(qpel16_mc33_old_c)
00170 DEF_OLD_QPEL(qpel8_mc11_old_c)
00171 DEF_OLD_QPEL(qpel8_mc31_old_c)
00172 DEF_OLD_QPEL(qpel8_mc12_old_c)
00173 DEF_OLD_QPEL(qpel8_mc32_old_c)
00174 DEF_OLD_QPEL(qpel8_mc13_old_c)
00175 DEF_OLD_QPEL(qpel8_mc33_old_c)
00176 
00177 #define CALL_2X_PIXELS(a, b, n)\
00178 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
00179     b(block  , pixels  , line_size, h);\
00180     b(block+n, pixels+n, line_size, h);\
00181 }
00182 
00183 /* motion estimation */
00184 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
00185 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
00186 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))*/;
00187 
00191 typedef struct ScanTable{
00192     const uint8_t *scantable;
00193     uint8_t permutated[64];
00194     uint8_t raster_end[64];
00195 #if ARCH_PPC
00196 
00197     DECLARE_ALIGNED(16, uint8_t, inverse)[64];
00198 #endif
00199 } ScanTable;
00200 
00201 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
00202 
00203 void ff_emulated_edge_mc(uint8_t *buf, const uint8_t *src, int linesize,
00204                          int block_w, int block_h,
00205                          int src_x, int src_y, int w, int h);
00206 
00210 typedef struct DSPContext {
00211     /* pixel ops : interface with DCT */
00212     void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
00213     void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
00214     void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00215     void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00216     void (*put_pixels_nonclamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00217     void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00218     void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
00219     void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
00220     int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
00224     void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
00228     void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
00229                     int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
00230     void (*clear_block)(DCTELEM *block/*align 16*/);
00231     void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
00232     int (*pix_sum)(uint8_t * pix, int line_size);
00233     int (*pix_norm1)(uint8_t * pix, int line_size);
00234 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
00235 
00236     me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
00237     me_cmp_func sse[6];
00238     me_cmp_func hadamard8_diff[6];
00239     me_cmp_func dct_sad[6];
00240     me_cmp_func quant_psnr[6];
00241     me_cmp_func bit[6];
00242     me_cmp_func rd[6];
00243     me_cmp_func vsad[6];
00244     me_cmp_func vsse[6];
00245     me_cmp_func nsse[6];
00246     me_cmp_func w53[6];
00247     me_cmp_func w97[6];
00248     me_cmp_func dct_max[6];
00249     me_cmp_func dct264_sad[6];
00250 
00251     me_cmp_func me_pre_cmp[6];
00252     me_cmp_func me_cmp[6];
00253     me_cmp_func me_sub_cmp[6];
00254     me_cmp_func mb_cmp[6];
00255     me_cmp_func ildct_cmp[6]; //only width 16 used
00256     me_cmp_func frame_skip_cmp[6]; //only width 8 used
00257 
00258     int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
00259                              int size);
00260 
00271     op_pixels_func put_pixels_tab[4][4];
00272 
00283     op_pixels_func avg_pixels_tab[4][4];
00284 
00295     op_pixels_func put_no_rnd_pixels_tab[4][4];
00296 
00307     op_pixels_func avg_no_rnd_pixels_tab[4][4];
00308 
00309     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);
00310 
00321     tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00322     tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00323 
00324     qpel_mc_func put_qpel_pixels_tab[2][16];
00325     qpel_mc_func avg_qpel_pixels_tab[2][16];
00326     qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
00327     qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
00328     qpel_mc_func put_mspel_pixels_tab[8];
00329 
00333     h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
00334     h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
00335     /* This is really one func used in VC-1 decoding */
00336     h264_chroma_mc_func put_no_rnd_vc1_chroma_pixels_tab[3];
00337     h264_chroma_mc_func avg_no_rnd_vc1_chroma_pixels_tab[3];
00338 
00339     qpel_mc_func put_h264_qpel_pixels_tab[4][16];
00340     qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
00341 
00342     qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
00343     qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];
00344 
00345     me_cmp_func pix_abs[2][4];
00346 
00347     /* huffyuv specific */
00348     void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
00349     void (*add_bytes_l2)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 16*/, int w);
00350     void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
00355     void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
00356     void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
00357     int  (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
00358     void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
00359     /* this might write to dst[w] */
00360     void (*add_png_paeth_prediction)(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp);
00361     void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
00362 
00363     void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
00364     void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
00365 
00366     void (*h261_loop_filter)(uint8_t *src, int stride);
00367 
00368     void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale);
00369     void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale);
00370 
00371     void (*vp3_idct_dc_add)(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
00372     void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00373     void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00374 
00375     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
00376     void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
00377     void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
00378     /* no alignment needed */
00379     void (*lpc_compute_autocorr)(const int32_t *data, int len, int lag, double *autoc);
00380     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
00381     void (*vector_fmul)(float *dst, const float *src, int len);
00382     void (*vector_fmul_reverse)(float *dst, const float *src0, const float *src1, int len);
00383     /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
00384     void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
00385     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
00386     void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len);
00387     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
00388     void (*int32_to_float_fmul_scalar)(float *dst, const int *src, float mul, int len);
00389     void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
00398     void (*vector_fmul_scalar)(float *dst, const float *src, float mul,
00399                                int len);
00412     void (*vector_fmul_sv_scalar[2])(float *dst, const float *src,
00413                                      const float **sv, float mul, int len);
00424     void (*sv_fmul_scalar[2])(float *dst, const float **sv,
00425                               float mul, int len);
00432     float (*scalarproduct_float)(const float *v1, const float *v2, int len);
00439     void (*butterflies_float)(float *restrict v1, float *restrict v2, int len);
00440 
00441     /* C version: convert floats from the range [384.0,386.0] to ints in [-32768,32767]
00442      * simd versions: convert floats from [-32768.0,32767.0] without rescaling and arrays are 16byte aligned */
00443     void (*float_to_int16)(int16_t *dst, const float *src, long len);
00444     void (*float_to_int16_interleave)(int16_t *dst, const float **src, long len, int channels);
00445 
00446     /* (I)DCT */
00447     void (*fdct)(DCTELEM *block/* align 16*/);
00448     void (*fdct248)(DCTELEM *block/* align 16*/);
00449 
00450     /* IDCT really*/
00451     void (*idct)(DCTELEM *block/* align 16*/);
00452 
00458     void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00459 
00464     void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00465 
00478     uint8_t idct_permutation[64];
00479     int idct_permutation_type;
00480 #define FF_NO_IDCT_PERM 1
00481 #define FF_LIBMPEG2_IDCT_PERM 2
00482 #define FF_SIMPLE_IDCT_PERM 3
00483 #define FF_TRANSPOSE_IDCT_PERM 4
00484 #define FF_PARTTRANS_IDCT_PERM 5
00485 #define FF_SSE2_IDCT_PERM 6
00486 
00487     int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
00488     void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
00489 #define BASIS_SHIFT 16
00490 #define RECON_SHIFT 6
00491 
00492     void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w);
00493 #define EDGE_WIDTH 16
00494 
00495     void (*prefetch)(void *mem, int stride, int h);
00496 
00497     void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00498 
00499     /* mlp/truehd functions */
00500     void (*mlp_filter_channel)(int32_t *state, const int32_t *coeff,
00501                                int firorder, int iirorder,
00502                                unsigned int filter_shift, int32_t mask, int blocksize,
00503                                int32_t *sample_buffer);
00504 
00505     /* vc1 functions */
00506     void (*vc1_inv_trans_8x8)(DCTELEM *b);
00507     void (*vc1_inv_trans_8x4)(uint8_t *dest, int line_size, DCTELEM *block);
00508     void (*vc1_inv_trans_4x8)(uint8_t *dest, int line_size, DCTELEM *block);
00509     void (*vc1_inv_trans_4x4)(uint8_t *dest, int line_size, DCTELEM *block);
00510     void (*vc1_inv_trans_8x8_dc)(uint8_t *dest, int line_size, DCTELEM *block);
00511     void (*vc1_inv_trans_8x4_dc)(uint8_t *dest, int line_size, DCTELEM *block);
00512     void (*vc1_inv_trans_4x8_dc)(uint8_t *dest, int line_size, DCTELEM *block);
00513     void (*vc1_inv_trans_4x4_dc)(uint8_t *dest, int line_size, DCTELEM *block);
00514     void (*vc1_v_overlap)(uint8_t* src, int stride);
00515     void (*vc1_h_overlap)(uint8_t* src, int stride);
00516     void (*vc1_v_loop_filter4)(uint8_t *src, int stride, int pq);
00517     void (*vc1_h_loop_filter4)(uint8_t *src, int stride, int pq);
00518     void (*vc1_v_loop_filter8)(uint8_t *src, int stride, int pq);
00519     void (*vc1_h_loop_filter8)(uint8_t *src, int stride, int pq);
00520     void (*vc1_v_loop_filter16)(uint8_t *src, int stride, int pq);
00521     void (*vc1_h_loop_filter16)(uint8_t *src, int stride, int pq);
00522     /* put 8x8 block with bicubic interpolation and quarterpel precision
00523      * last argument is actually round value instead of height
00524      */
00525     op_pixels_func put_vc1_mspel_pixels_tab[16];
00526     op_pixels_func avg_vc1_mspel_pixels_tab[16];
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 
00538     int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len, int shift);
00539     /* ape functions */
00545     int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
00546 
00547     /* rv30 functions */
00548     qpel_mc_func put_rv30_tpel_pixels_tab[4][16];
00549     qpel_mc_func avg_rv30_tpel_pixels_tab[4][16];
00550 
00551     /* rv40 functions */
00552     qpel_mc_func put_rv40_qpel_pixels_tab[4][16];
00553     qpel_mc_func avg_rv40_qpel_pixels_tab[4][16];
00554     h264_chroma_mc_func put_rv40_chroma_pixels_tab[3];
00555     h264_chroma_mc_func avg_rv40_chroma_pixels_tab[3];
00556 
00557     /* bink functions */
00558     op_fill_func fill_block_tab[2];
00559     void (*scale_block)(const uint8_t src[64]/*align 8*/, uint8_t *dst/*align 8*/, int linesize);
00560 } DSPContext;
00561 
00562 void dsputil_static_init(void);
00563 void dsputil_init(DSPContext* p, AVCodecContext *avctx);
00564 
00565 int ff_check_alignment(void);
00566 
00571 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
00572 
00573 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
00574 
00575 #define         BYTE_VEC32(c)   ((c)*0x01010101UL)
00576 
00577 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
00578 {
00579     return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00580 }
00581 
00582 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
00583 {
00584     return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00585 }
00586 
00587 static inline int get_penalty_factor(int lambda, int lambda2, int type){
00588     switch(type&0xFF){
00589     default:
00590     case FF_CMP_SAD:
00591         return lambda>>FF_LAMBDA_SHIFT;
00592     case FF_CMP_DCT:
00593         return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
00594     case FF_CMP_W53:
00595         return (4*lambda)>>(FF_LAMBDA_SHIFT);
00596     case FF_CMP_W97:
00597         return (2*lambda)>>(FF_LAMBDA_SHIFT);
00598     case FF_CMP_SATD:
00599     case FF_CMP_DCT264:
00600         return (2*lambda)>>FF_LAMBDA_SHIFT;
00601     case FF_CMP_RD:
00602     case FF_CMP_PSNR:
00603     case FF_CMP_SSE:
00604     case FF_CMP_NSSE:
00605         return lambda2>>FF_LAMBDA_SHIFT;
00606     case FF_CMP_BIT:
00607         return 1;
00608     }
00609 }
00610 
00616 #define emms_c()
00617 
00618 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
00619 void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
00620 void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
00621 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
00622 void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
00623 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
00624 void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
00625 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
00626 void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
00627 
00628 void ff_dsputil_init_dwt(DSPContext *c);
00629 void ff_rv30dsp_init(DSPContext* c, AVCodecContext *avctx);
00630 void ff_rv40dsp_init(DSPContext* c, AVCodecContext *avctx);
00631 void ff_vc1dsp_init(DSPContext* c, AVCodecContext *avctx);
00632 void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx);
00633 void ff_mlp_init(DSPContext* c, AVCodecContext *avctx);
00634 void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx);
00635 
00636 #if HAVE_MMX
00637 
00638 #undef emms_c
00639 
00640 static inline void emms(void)
00641 {
00642     __asm__ volatile ("emms;":::"memory");
00643 }
00644 
00645 #define emms_c() emms()
00646 
00647 #elif ARCH_ARM
00648 
00649 #if HAVE_NEON
00650 #   define STRIDE_ALIGN 16
00651 #endif
00652 
00653 #elif ARCH_PPC
00654 
00655 #define STRIDE_ALIGN 16
00656 
00657 #elif HAVE_MMI
00658 
00659 #define STRIDE_ALIGN 16
00660 
00661 #endif
00662 
00663 #ifndef STRIDE_ALIGN
00664 #   define STRIDE_ALIGN 8
00665 #endif
00666 
00667 #define LOCAL_ALIGNED(a, t, v, s, ...)                          \
00668     uint8_t la_##v[sizeof(t s __VA_ARGS__) + (a)];              \
00669     t (*v) __VA_ARGS__ = (void *)FFALIGN((uintptr_t)la_##v, a)
00670 
00671 #if HAVE_LOCAL_ALIGNED_8
00672 #   define LOCAL_ALIGNED_8(t, v, s, ...) DECLARE_ALIGNED(8, t, v) s __VA_ARGS__
00673 #else
00674 #   define LOCAL_ALIGNED_8(t, v, s, ...) LOCAL_ALIGNED(8, t, v, s, __VA_ARGS__)
00675 #endif
00676 
00677 #if HAVE_LOCAL_ALIGNED_16
00678 #   define LOCAL_ALIGNED_16(t, v, s, ...) DECLARE_ALIGNED(16, t, v) s __VA_ARGS__
00679 #else
00680 #   define LOCAL_ALIGNED_16(t, v, s, ...) LOCAL_ALIGNED(16, t, v, s, __VA_ARGS__)
00681 #endif
00682 
00683 /* PSNR */
00684 void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
00685               int orig_linesize[3], int coded_linesize,
00686               AVCodecContext *avctx);
00687 
00688 #define WRAPPER8_16(name8, name16)\
00689 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00690     return name8(s, dst           , src           , stride, h)\
00691           +name8(s, dst+8         , src+8         , stride, h);\
00692 }
00693 
00694 #define WRAPPER8_16_SQ(name8, name16)\
00695 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00696     int score=0;\
00697     score +=name8(s, dst           , src           , stride, 8);\
00698     score +=name8(s, dst+8         , src+8         , stride, 8);\
00699     if(h==16){\
00700         dst += 8*stride;\
00701         src += 8*stride;\
00702         score +=name8(s, dst           , src           , stride, 8);\
00703         score +=name8(s, dst+8         , src+8         , stride, 8);\
00704     }\
00705     return score;\
00706 }
00707 
00708 
00709 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00710 {
00711     int i;
00712     for(i=0; i<h; i++)
00713     {
00714         AV_WN16(dst   , AV_RN16(src   ));
00715         dst+=dstStride;
00716         src+=srcStride;
00717     }
00718 }
00719 
00720 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00721 {
00722     int i;
00723     for(i=0; i<h; i++)
00724     {
00725         AV_WN32(dst   , AV_RN32(src   ));
00726         dst+=dstStride;
00727         src+=srcStride;
00728     }
00729 }
00730 
00731 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00732 {
00733     int i;
00734     for(i=0; i<h; i++)
00735     {
00736         AV_WN32(dst   , AV_RN32(src   ));
00737         AV_WN32(dst+4 , AV_RN32(src+4 ));
00738         dst+=dstStride;
00739         src+=srcStride;
00740     }
00741 }
00742 
00743 static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00744 {
00745     int i;
00746     for(i=0; i<h; i++)
00747     {
00748         AV_WN32(dst   , AV_RN32(src   ));
00749         AV_WN32(dst+4 , AV_RN32(src+4 ));
00750         dst[8]= src[8];
00751         dst+=dstStride;
00752         src+=srcStride;
00753     }
00754 }
00755 
00756 static inline void copy_block16(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         AV_WN32(dst+8 , AV_RN32(src+8 ));
00764         AV_WN32(dst+12, AV_RN32(src+12));
00765         dst+=dstStride;
00766         src+=srcStride;
00767     }
00768 }
00769 
00770 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00771 {
00772     int i;
00773     for(i=0; i<h; i++)
00774     {
00775         AV_WN32(dst   , AV_RN32(src   ));
00776         AV_WN32(dst+4 , AV_RN32(src+4 ));
00777         AV_WN32(dst+8 , AV_RN32(src+8 ));
00778         AV_WN32(dst+12, AV_RN32(src+12));
00779         dst[16]= src[16];
00780         dst+=dstStride;
00781         src+=srcStride;
00782     }
00783 }
00784 
00785 #endif /* AVCODEC_DSPUTIL_H */