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_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
00070 
00071 /* encoding scans */
00072 extern const uint8_t ff_alternate_horizontal_scan[64];
00073 extern const uint8_t ff_alternate_vertical_scan[64];
00074 extern const uint8_t ff_zigzag_direct[64];
00075 extern const uint8_t ff_zigzag248_direct[64];
00076 
00077 /* pixel operations */
00078 #define MAX_NEG_CROP 1024
00079 
00080 /* temporary */
00081 extern uint32_t ff_squareTbl[512];
00082 extern uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];
00083 
00084 void ff_put_pixels8x8_c(uint8_t *dst, uint8_t *src, int stride);
00085 void ff_avg_pixels8x8_c(uint8_t *dst, uint8_t *src, int stride);
00086 void ff_put_pixels16x16_c(uint8_t *dst, uint8_t *src, int stride);
00087 void ff_avg_pixels16x16_c(uint8_t *dst, uint8_t *src, int stride);
00088 
00089 /* VP3 DSP functions */
00090 void ff_vp3_idct_c(DCTELEM *block/* align 16*/);
00091 void ff_vp3_idct_put_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00092 void ff_vp3_idct_add_c(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00093 void ff_vp3_idct_dc_add_c(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
00094 
00095 void ff_vp3_v_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
00096 void ff_vp3_h_loop_filter_c(uint8_t *src, int stride, int *bounding_values);
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 /* EA functions */
00104 void ff_ea_idct_put_c(uint8_t *dest, int linesize, DCTELEM *block);
00105 
00106 /* 1/2^n downscaling functions from imgconvert.c */
00107 #if LIBAVCODEC_VERSION_MAJOR < 53
00108 
00111 attribute_deprecated
00112 void ff_img_copy_plane(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00113 #endif
00114 
00115 void ff_shrink22(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00116 void ff_shrink44(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00117 void ff_shrink88(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00118 
00119 void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
00120               int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
00121 
00122 /* minimum alignment rules ;)
00123 If you notice errors in the align stuff, need more alignment for some ASM code
00124 for some CPU or need to use a function with less aligned data then send a mail
00125 to the ffmpeg-devel mailing list, ...
00126 
00127 !warning These alignments might not match reality, (missing attribute((align))
00128 stuff somewhere possible).
00129 I (Michael) did not check them, these are just the alignments which I think
00130 could be reached easily ...
00131 
00132 !future video codecs might need functions with less strict alignment
00133 */
00134 
00135 /*
00136 void get_pixels_c(DCTELEM *block, const uint8_t *pixels, int line_size);
00137 void diff_pixels_c(DCTELEM *block, const uint8_t *s1, const uint8_t *s2, int stride);
00138 void put_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
00139 void add_pixels_clamped_c(const DCTELEM *block, uint8_t *pixels, int line_size);
00140 void clear_blocks_c(DCTELEM *blocks);
00141 */
00142 
00143 /* add and put pixel (decoding) */
00144 // blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
00145 //h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller then 4
00146 typedef void (*op_pixels_func)(uint8_t *block/*align width (8 or 16)*/, const uint8_t *pixels/*align 1*/, int line_size, int h);
00147 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);
00148 typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
00149 typedef void (*h264_chroma_mc_func)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x, int y);
00150 
00151 typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);
00152 
00153 #define DEF_OLD_QPEL(name)\
00154 void ff_put_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
00155 void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);\
00156 void ff_avg_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, int stride);
00157 
00158 DEF_OLD_QPEL(qpel16_mc11_old_c)
00159 DEF_OLD_QPEL(qpel16_mc31_old_c)
00160 DEF_OLD_QPEL(qpel16_mc12_old_c)
00161 DEF_OLD_QPEL(qpel16_mc32_old_c)
00162 DEF_OLD_QPEL(qpel16_mc13_old_c)
00163 DEF_OLD_QPEL(qpel16_mc33_old_c)
00164 DEF_OLD_QPEL(qpel8_mc11_old_c)
00165 DEF_OLD_QPEL(qpel8_mc31_old_c)
00166 DEF_OLD_QPEL(qpel8_mc12_old_c)
00167 DEF_OLD_QPEL(qpel8_mc32_old_c)
00168 DEF_OLD_QPEL(qpel8_mc13_old_c)
00169 DEF_OLD_QPEL(qpel8_mc33_old_c)
00170 
00171 #define CALL_2X_PIXELS(a, b, n)\
00172 static void a(uint8_t *block, const uint8_t *pixels, int line_size, int h){\
00173     b(block  , pixels  , line_size, h);\
00174     b(block+n, pixels+n, line_size, h);\
00175 }
00176 
00177 /* motion estimation */
00178 // h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller then 2
00179 // although currently h<4 is not used as functions with width <8 are neither used nor implemented
00180 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))*/;
00181 
00185 typedef struct ScanTable{
00186     const uint8_t *scantable;
00187     uint8_t permutated[64];
00188     uint8_t raster_end[64];
00189 #if ARCH_PPC
00190 
00191     DECLARE_ALIGNED(16, uint8_t, inverse)[64];
00192 #endif
00193 } ScanTable;
00194 
00195 void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
00196 
00197 void ff_emulated_edge_mc(uint8_t *buf, const uint8_t *src, int linesize,
00198                          int block_w, int block_h,
00199                          int src_x, int src_y, int w, int h);
00200 
00201 void ff_add_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
00202 void ff_put_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
00203 void ff_put_signed_pixels_clamped_c(const DCTELEM *block, uint8_t *dest, int linesize);
00204 
00208 typedef struct DSPContext {
00209     /* pixel ops : interface with DCT */
00210     void (*get_pixels)(DCTELEM *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
00211     void (*diff_pixels)(DCTELEM *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
00212     void (*put_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00213     void (*put_signed_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00214     void (*put_pixels_nonclamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00215     void (*add_pixels_clamped)(const DCTELEM *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
00216     void (*add_pixels8)(uint8_t *pixels, DCTELEM *block, int line_size);
00217     void (*add_pixels4)(uint8_t *pixels, DCTELEM *block, int line_size);
00218     int (*sum_abs_dctelem)(DCTELEM *block/*align 16*/);
00231     void (*emulated_edge_mc)(uint8_t *buf, const uint8_t *src, int linesize,
00232                              int block_w, int block_h,
00233                              int src_x, int src_y, int w, int h);
00237     void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
00241     void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
00242                     int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
00243     void (*clear_block)(DCTELEM *block/*align 16*/);
00244     void (*clear_blocks)(DCTELEM *blocks/*align 16*/);
00245     int (*pix_sum)(uint8_t * pix, int line_size);
00246     int (*pix_norm1)(uint8_t * pix, int line_size);
00247 // 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4
00248 
00249     me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
00250     me_cmp_func sse[6];
00251     me_cmp_func hadamard8_diff[6];
00252     me_cmp_func dct_sad[6];
00253     me_cmp_func quant_psnr[6];
00254     me_cmp_func bit[6];
00255     me_cmp_func rd[6];
00256     me_cmp_func vsad[6];
00257     me_cmp_func vsse[6];
00258     me_cmp_func nsse[6];
00259     me_cmp_func w53[6];
00260     me_cmp_func w97[6];
00261     me_cmp_func dct_max[6];
00262     me_cmp_func dct264_sad[6];
00263 
00264     me_cmp_func me_pre_cmp[6];
00265     me_cmp_func me_cmp[6];
00266     me_cmp_func me_sub_cmp[6];
00267     me_cmp_func mb_cmp[6];
00268     me_cmp_func ildct_cmp[6]; //only width 16 used
00269     me_cmp_func frame_skip_cmp[6]; //only width 8 used
00270 
00271     int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
00272                              int size);
00273 
00284     op_pixels_func put_pixels_tab[4][4];
00285 
00296     op_pixels_func avg_pixels_tab[4][4];
00297 
00308     op_pixels_func put_no_rnd_pixels_tab[4][4];
00309 
00320     op_pixels_func avg_no_rnd_pixels_tab[4][4];
00321 
00322     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);
00323 
00334     tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00335     tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
00336 
00337     qpel_mc_func put_qpel_pixels_tab[2][16];
00338     qpel_mc_func avg_qpel_pixels_tab[2][16];
00339     qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
00340     qpel_mc_func avg_no_rnd_qpel_pixels_tab[2][16];
00341     qpel_mc_func put_mspel_pixels_tab[8];
00342 
00346     h264_chroma_mc_func put_h264_chroma_pixels_tab[3];
00347     h264_chroma_mc_func avg_h264_chroma_pixels_tab[3];
00348 
00349     qpel_mc_func put_h264_qpel_pixels_tab[4][16];
00350     qpel_mc_func avg_h264_qpel_pixels_tab[4][16];
00351 
00352     qpel_mc_func put_2tap_qpel_pixels_tab[4][16];
00353     qpel_mc_func avg_2tap_qpel_pixels_tab[4][16];
00354 
00355     me_cmp_func pix_abs[2][4];
00356 
00357     /* huffyuv specific */
00358     void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
00359     void (*add_bytes_l2)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 16*/, int w);
00360     void (*diff_bytes)(uint8_t *dst/*align 16*/, uint8_t *src1/*align 16*/, uint8_t *src2/*align 1*/,int w);
00365     void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
00366     void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
00367     int  (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
00368     void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
00369     /* this might write to dst[w] */
00370     void (*add_png_paeth_prediction)(uint8_t *dst, uint8_t *src, uint8_t *top, int w, int bpp);
00371     void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
00372     void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);
00373 
00374     void (*h263_v_loop_filter)(uint8_t *src, int stride, int qscale);
00375     void (*h263_h_loop_filter)(uint8_t *src, int stride, int qscale);
00376 
00377     void (*h261_loop_filter)(uint8_t *src, int stride);
00378 
00379     void (*x8_v_loop_filter)(uint8_t *src, int stride, int qscale);
00380     void (*x8_h_loop_filter)(uint8_t *src, int stride, int qscale);
00381 
00382     void (*vp3_idct_dc_add)(uint8_t *dest/*align 8*/, int line_size, const DCTELEM *block/*align 16*/);
00383     void (*vp3_v_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00384     void (*vp3_h_loop_filter)(uint8_t *src, int stride, int *bounding_values);
00385 
00386     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
00387     void (*vorbis_inverse_coupling)(float *mag, float *ang, int blocksize);
00388     void (*ac3_downmix)(float (*samples)[256], float (*matrix)[2], int out_ch, int in_ch, int len);
00389     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
00390     void (*vector_fmul)(float *dst, const float *src0, const float *src1, 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, int len);
00396     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
00397     void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);
00406     void (*vector_fmul_scalar)(float *dst, const float *src, float mul,
00407                                int len);
00420     void (*vector_fmul_sv_scalar[2])(float *dst, const float *src,
00421                                      const float **sv, float mul, int len);
00432     void (*sv_fmul_scalar[2])(float *dst, const float **sv,
00433                               float mul, int len);
00440     float (*scalarproduct_float)(const float *v1, const float *v2, int len);
00447     void (*butterflies_float)(float *restrict v1, float *restrict v2, int len);
00448 
00449     /* (I)DCT */
00450     void (*fdct)(DCTELEM *block/* align 16*/);
00451     void (*fdct248)(DCTELEM *block/* align 16*/);
00452 
00453     /* IDCT really*/
00454     void (*idct)(DCTELEM *block/* align 16*/);
00455 
00461     void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00462 
00467     void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, DCTELEM *block/*align 16*/);
00468 
00481     uint8_t idct_permutation[64];
00482     int idct_permutation_type;
00483 #define FF_NO_IDCT_PERM 1
00484 #define FF_LIBMPEG2_IDCT_PERM 2
00485 #define FF_SIMPLE_IDCT_PERM 3
00486 #define FF_TRANSPOSE_IDCT_PERM 4
00487 #define FF_PARTTRANS_IDCT_PERM 5
00488 #define FF_SSE2_IDCT_PERM 6
00489 
00490     int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
00491     void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
00492 #define BASIS_SHIFT 16
00493 #define RECON_SHIFT 6
00494 
00495     void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int sides);
00496 #define EDGE_WIDTH 16
00497 #define EDGE_TOP    1
00498 #define EDGE_BOTTOM 2
00499 
00500     void (*prefetch)(void *mem, int stride, int h);
00501 
00502     void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);
00503 
00504     /* mlp/truehd functions */
00505     void (*mlp_filter_channel)(int32_t *state, const int32_t *coeff,
00506                                int firorder, int iirorder,
00507                                unsigned int filter_shift, int32_t mask, int blocksize,
00508                                int32_t *sample_buffer);
00509 
00510     /* intrax8 functions */
00511     void (*x8_spatial_compensation[12])(uint8_t *src , uint8_t *dst, int linesize);
00512     void (*x8_setup_spatial_compensation)(uint8_t *src, uint8_t *dst, int linesize,
00513            int * range, int * sum,  int edges);
00514 
00520     int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len, int shift);
00521     /* ape functions */
00527     int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);
00528 
00529     /* rv30 functions */
00530     qpel_mc_func put_rv30_tpel_pixels_tab[4][16];
00531     qpel_mc_func avg_rv30_tpel_pixels_tab[4][16];
00532 
00533     /* rv40 functions */
00534     qpel_mc_func put_rv40_qpel_pixels_tab[4][16];
00535     qpel_mc_func avg_rv40_qpel_pixels_tab[4][16];
00536     h264_chroma_mc_func put_rv40_chroma_pixels_tab[3];
00537     h264_chroma_mc_func avg_rv40_chroma_pixels_tab[3];
00538 
00539     /* bink functions */
00540     op_fill_func fill_block_tab[2];
00541     void (*scale_block)(const uint8_t src[64]/*align 8*/, uint8_t *dst/*align 8*/, int linesize);
00542 } DSPContext;
00543 
00544 void dsputil_static_init(void);
00545 void dsputil_init(DSPContext* p, AVCodecContext *avctx);
00546 
00547 int ff_check_alignment(void);
00548 
00553 void ff_block_permute(DCTELEM *block, uint8_t *permutation, const uint8_t *scantable, int last);
00554 
00555 void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);
00556 
00557 #define         BYTE_VEC32(c)   ((c)*0x01010101UL)
00558 
00559 static inline uint32_t rnd_avg32(uint32_t a, uint32_t b)
00560 {
00561     return (a | b) - (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00562 }
00563 
00564 static inline uint32_t no_rnd_avg32(uint32_t a, uint32_t b)
00565 {
00566     return (a & b) + (((a ^ b) & ~BYTE_VEC32(0x01)) >> 1);
00567 }
00568 
00569 static inline int get_penalty_factor(int lambda, int lambda2, int type){
00570     switch(type&0xFF){
00571     default:
00572     case FF_CMP_SAD:
00573         return lambda>>FF_LAMBDA_SHIFT;
00574     case FF_CMP_DCT:
00575         return (3*lambda)>>(FF_LAMBDA_SHIFT+1);
00576     case FF_CMP_W53:
00577         return (4*lambda)>>(FF_LAMBDA_SHIFT);
00578     case FF_CMP_W97:
00579         return (2*lambda)>>(FF_LAMBDA_SHIFT);
00580     case FF_CMP_SATD:
00581     case FF_CMP_DCT264:
00582         return (2*lambda)>>FF_LAMBDA_SHIFT;
00583     case FF_CMP_RD:
00584     case FF_CMP_PSNR:
00585     case FF_CMP_SSE:
00586     case FF_CMP_NSSE:
00587         return lambda2>>FF_LAMBDA_SHIFT;
00588     case FF_CMP_BIT:
00589         return 1;
00590     }
00591 }
00592 
00598 #define emms_c()
00599 
00600 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
00601 void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
00602 void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
00603 void dsputil_init_mlib(DSPContext* c, AVCodecContext *avctx);
00604 void dsputil_init_mmi(DSPContext* c, AVCodecContext *avctx);
00605 void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx);
00606 void dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
00607 void dsputil_init_sh4(DSPContext* c, AVCodecContext *avctx);
00608 void dsputil_init_vis(DSPContext* c, AVCodecContext *avctx);
00609 
00610 void ff_dsputil_init_dwt(DSPContext *c);
00611 void ff_rv30dsp_init(DSPContext* c, AVCodecContext *avctx);
00612 void ff_rv40dsp_init(DSPContext* c, AVCodecContext *avctx);
00613 void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx);
00614 void ff_mlp_init(DSPContext* c, AVCodecContext *avctx);
00615 void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx);
00616 
00617 #if HAVE_MMX
00618 
00619 #undef emms_c
00620 
00621 static inline void emms(void)
00622 {
00623     __asm__ volatile ("emms;":::"memory");
00624 }
00625 
00626 #define emms_c() emms()
00627 
00628 #elif ARCH_ARM
00629 
00630 #if HAVE_NEON
00631 #   define STRIDE_ALIGN 16
00632 #endif
00633 
00634 #elif ARCH_PPC
00635 
00636 #define STRIDE_ALIGN 16
00637 
00638 #elif HAVE_MMI
00639 
00640 #define STRIDE_ALIGN 16
00641 
00642 #endif
00643 
00644 #ifndef STRIDE_ALIGN
00645 #   define STRIDE_ALIGN 8
00646 #endif
00647 
00648 #define LOCAL_ALIGNED_A(a, t, v, s, o, ...)             \
00649     uint8_t la_##v[sizeof(t s o) + (a)];                \
00650     t (*v) o = (void *)FFALIGN((uintptr_t)la_##v, a)
00651 
00652 #define LOCAL_ALIGNED_D(a, t, v, s, o, ...) DECLARE_ALIGNED(a, t, v) s o
00653 
00654 #define LOCAL_ALIGNED(a, t, v, ...) LOCAL_ALIGNED_A(a, t, v, __VA_ARGS__,,)
00655 
00656 #if HAVE_LOCAL_ALIGNED_8
00657 #   define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED_D(8, t, v, __VA_ARGS__,,)
00658 #else
00659 #   define LOCAL_ALIGNED_8(t, v, ...) LOCAL_ALIGNED(8, t, v, __VA_ARGS__)
00660 #endif
00661 
00662 #if HAVE_LOCAL_ALIGNED_16
00663 #   define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED_D(16, t, v, __VA_ARGS__,,)
00664 #else
00665 #   define LOCAL_ALIGNED_16(t, v, ...) LOCAL_ALIGNED(16, t, v, __VA_ARGS__)
00666 #endif
00667 
00668 /* PSNR */
00669 void get_psnr(uint8_t *orig_image[3], uint8_t *coded_image[3],
00670               int orig_linesize[3], int coded_linesize,
00671               AVCodecContext *avctx);
00672 
00673 #define WRAPPER8_16(name8, name16)\
00674 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00675     return name8(s, dst           , src           , stride, h)\
00676           +name8(s, dst+8         , src+8         , stride, h);\
00677 }
00678 
00679 #define WRAPPER8_16_SQ(name8, name16)\
00680 static int name16(void /*MpegEncContext*/ *s, uint8_t *dst, uint8_t *src, int stride, int h){\
00681     int score=0;\
00682     score +=name8(s, dst           , src           , stride, 8);\
00683     score +=name8(s, dst+8         , src+8         , stride, 8);\
00684     if(h==16){\
00685         dst += 8*stride;\
00686         src += 8*stride;\
00687         score +=name8(s, dst           , src           , stride, 8);\
00688         score +=name8(s, dst+8         , src+8         , stride, 8);\
00689     }\
00690     return score;\
00691 }
00692 
00693 
00694 static inline void copy_block2(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00695 {
00696     int i;
00697     for(i=0; i<h; i++)
00698     {
00699         AV_WN16(dst   , AV_RN16(src   ));
00700         dst+=dstStride;
00701         src+=srcStride;
00702     }
00703 }
00704 
00705 static inline void copy_block4(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00706 {
00707     int i;
00708     for(i=0; i<h; i++)
00709     {
00710         AV_WN32(dst   , AV_RN32(src   ));
00711         dst+=dstStride;
00712         src+=srcStride;
00713     }
00714 }
00715 
00716 static inline void copy_block8(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00717 {
00718     int i;
00719     for(i=0; i<h; i++)
00720     {
00721         AV_WN32(dst   , AV_RN32(src   ));
00722         AV_WN32(dst+4 , AV_RN32(src+4 ));
00723         dst+=dstStride;
00724         src+=srcStride;
00725     }
00726 }
00727 
00728 static inline void copy_block9(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00729 {
00730     int i;
00731     for(i=0; i<h; i++)
00732     {
00733         AV_WN32(dst   , AV_RN32(src   ));
00734         AV_WN32(dst+4 , AV_RN32(src+4 ));
00735         dst[8]= src[8];
00736         dst+=dstStride;
00737         src+=srcStride;
00738     }
00739 }
00740 
00741 static inline void copy_block16(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00742 {
00743     int i;
00744     for(i=0; i<h; i++)
00745     {
00746         AV_WN32(dst   , AV_RN32(src   ));
00747         AV_WN32(dst+4 , AV_RN32(src+4 ));
00748         AV_WN32(dst+8 , AV_RN32(src+8 ));
00749         AV_WN32(dst+12, AV_RN32(src+12));
00750         dst+=dstStride;
00751         src+=srcStride;
00752     }
00753 }
00754 
00755 static inline void copy_block17(uint8_t *dst, const uint8_t *src, int dstStride, int srcStride, int h)
00756 {
00757     int i;
00758     for(i=0; i<h; i++)
00759     {
00760         AV_WN32(dst   , AV_RN32(src   ));
00761         AV_WN32(dst+4 , AV_RN32(src+4 ));
00762         AV_WN32(dst+8 , AV_RN32(src+8 ));
00763         AV_WN32(dst+12, AV_RN32(src+12));
00764         dst[16]= src[16];
00765         dst+=dstStride;
00766         src+=srcStride;
00767     }
00768 }
00769 
00770 #endif /* AVCODEC_DSPUTIL_H */