aboutsummaryrefslogtreecommitdiffstats
path: root/libavcodec/dsputil.h
blob: 02292faf8e6c5b001a100080a1e225326aee13f9 (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
/*
 * DSP utils
 * Copyright (c) 2000, 2001, 2002 Fabrice Bellard
 * Copyright (c) 2002-2004 Michael Niedermayer <michaelni@gmx.at>
 *
 * This file is part of FFmpeg.
 *
 * FFmpeg is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * FFmpeg is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with FFmpeg; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

/**
 * @file
 * DSP utils.
 * note, many functions in here may use MMX which trashes the FPU state, it is
 * absolutely necessary to call emms_c() between dsp & float/double code
 */

#ifndef AVCODEC_DSPUTIL_H
#define AVCODEC_DSPUTIL_H

#include "avcodec.h"
#include "rnd_avg.h"

/* encoding scans */
extern const uint8_t ff_alternate_horizontal_scan[64];
extern const uint8_t ff_alternate_vertical_scan[64];
extern const uint8_t ff_zigzag_direct[64];
extern const uint8_t ff_zigzag248_direct[64];

/* pixel operations */
#define MAX_NEG_CROP 1024

/* temporary */
extern uint32_t ff_squareTbl[512];
extern const uint8_t ff_cropTbl[256 + 2 * MAX_NEG_CROP];

void ff_put_pixels8x8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
void ff_avg_pixels8x8_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
void ff_put_pixels16x16_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
void ff_avg_pixels16x16_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);

/* RV40 functions */
void ff_put_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
void ff_avg_rv40_qpel16_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
void ff_put_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);
void ff_avg_rv40_qpel8_mc33_c(uint8_t *dst, uint8_t *src, ptrdiff_t stride);

void ff_gmc_c(uint8_t *dst, uint8_t *src, int stride, int h, int ox, int oy,
              int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);

/* minimum alignment rules ;)
If you notice errors in the align stuff, need more alignment for some ASM code
for some CPU or need to use a function with less aligned data then send a mail
to the ffmpeg-devel mailing list, ...

!warning These alignments might not match reality, (missing attribute((align))
stuff somewhere possible).
I (Michael) did not check them, these are just the alignments which I think
could be reached easily ...

!future video codecs might need functions with less strict alignment
*/

/* add and put pixel (decoding) */
// blocksizes for op_pixels_func are 8x4,8x8 16x8 16x16
//h for op_pixels_func is limited to {width/2, width} but never larger than 16 and never smaller than 4
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);
typedef void (*qpel_mc_func)(uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, ptrdiff_t stride);

typedef void (*op_fill_func)(uint8_t *block/*align width (8 or 16)*/, uint8_t value, int line_size, int h);

#define DEF_OLD_QPEL(name)\
void ff_put_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, ptrdiff_t stride);\
void ff_put_no_rnd_ ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, ptrdiff_t stride);\
void ff_avg_        ## name (uint8_t *dst/*align width (8 or 16)*/, uint8_t *src/*align 1*/, ptrdiff_t stride);

DEF_OLD_QPEL(qpel16_mc11_old_c)
DEF_OLD_QPEL(qpel16_mc31_old_c)
DEF_OLD_QPEL(qpel16_mc12_old_c)
DEF_OLD_QPEL(qpel16_mc32_old_c)
DEF_OLD_QPEL(qpel16_mc13_old_c)
DEF_OLD_QPEL(qpel16_mc33_old_c)
DEF_OLD_QPEL(qpel8_mc11_old_c)
DEF_OLD_QPEL(qpel8_mc31_old_c)
DEF_OLD_QPEL(qpel8_mc12_old_c)
DEF_OLD_QPEL(qpel8_mc32_old_c)
DEF_OLD_QPEL(qpel8_mc13_old_c)
DEF_OLD_QPEL(qpel8_mc33_old_c)

/* motion estimation */
// h is limited to {width/2, width, 2*width} but never larger than 16 and never smaller than 2
// although currently h<4 is not used as functions with width <8 are neither used nor implemented
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))*/;

/**
 * Scantable.
 */
typedef struct ScanTable{
    const uint8_t *scantable;
    uint8_t permutated[64];
    uint8_t raster_end[64];
} ScanTable;

void ff_init_scantable(uint8_t *, ScanTable *st, const uint8_t *src_scantable);
void ff_init_scantable_permutation(uint8_t *idct_permutation,
                                   int idct_permutation_type);

/**
 * DSPContext.
 */
typedef struct DSPContext {
    /* pixel ops : interface with DCT */
    void (*get_pixels)(int16_t *block/*align 16*/, const uint8_t *pixels/*align 8*/, int line_size);
    void (*diff_pixels)(int16_t *block/*align 16*/, const uint8_t *s1/*align 8*/, const uint8_t *s2/*align 8*/, int stride);
    void (*put_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
    void (*put_signed_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
    void (*add_pixels_clamped)(const int16_t *block/*align 16*/, uint8_t *pixels/*align 8*/, int line_size);
    void (*add_pixels8)(uint8_t *pixels, int16_t *block, int line_size);
    int (*sum_abs_dctelem)(int16_t *block/*align 16*/);
    /**
     * translational global motion compensation.
     */
    void (*gmc1)(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int srcStride, int h, int x16, int y16, int rounder);
    /**
     * global motion compensation.
     */
    void (*gmc )(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int ox, int oy,
                    int dxx, int dxy, int dyx, int dyy, int shift, int r, int width, int height);
    void (*clear_block)(int16_t *block/*align 16*/);
    void (*clear_blocks)(int16_t *blocks/*align 16*/);
    int (*pix_sum)(uint8_t * pix, int line_size);
    int (*pix_norm1)(uint8_t * pix, int line_size);
// 16x16 8x8 4x4 2x2 16x8 8x4 4x2 8x16 4x8 2x4

    me_cmp_func sad[6]; /* identical to pix_absAxA except additional void * */
    me_cmp_func sse[6];
    me_cmp_func hadamard8_diff[6];
    me_cmp_func dct_sad[6];
    me_cmp_func quant_psnr[6];
    me_cmp_func bit[6];
    me_cmp_func rd[6];
    me_cmp_func vsad[6];
    me_cmp_func vsse[6];
    me_cmp_func nsse[6];
    me_cmp_func w53[6];
    me_cmp_func w97[6];
    me_cmp_func dct_max[6];
    me_cmp_func dct264_sad[6];

    me_cmp_func me_pre_cmp[6];
    me_cmp_func me_cmp[6];
    me_cmp_func me_sub_cmp[6];
    me_cmp_func mb_cmp[6];
    me_cmp_func ildct_cmp[6]; //only width 16 used
    me_cmp_func frame_skip_cmp[6]; //only width 8 used

    int (*ssd_int8_vs_int16)(const int8_t *pix1, const int16_t *pix2,
                             int size);

    /**
     * Thirdpel motion compensation with rounding (a+b+1)>>1.
     * this is an array[12] of motion compensation functions for the 9 thirdpe
     * positions<br>
     * *pixels_tab[ xthirdpel + 4*ythirdpel ]
     * @param block destination where the result is stored
     * @param pixels source
     * @param line_size number of bytes in a horizontal line of block
     * @param h height
     */
    tpel_mc_func put_tpel_pixels_tab[11]; //FIXME individual func ptr per width?
    tpel_mc_func avg_tpel_pixels_tab[11]; //FIXME individual func ptr per width?

    qpel_mc_func put_qpel_pixels_tab[2][16];
    qpel_mc_func avg_qpel_pixels_tab[2][16];
    qpel_mc_func put_no_rnd_qpel_pixels_tab[2][16];
    qpel_mc_func put_mspel_pixels_tab[8];

    me_cmp_func pix_abs[2][4];

    /* huffyuv specific */
    void (*add_bytes)(uint8_t *dst/*align 16*/, uint8_t *src/*align 16*/, int w);
    void (*diff_bytes)(uint8_t *dst/*align 16*/, const uint8_t *src1/*align 16*/, const uint8_t *src2/*align 1*/,int w);
    /**
     * subtract huffyuv's variant of median prediction
     * note, this might read from src1[-1], src2[-1]
     */
    void (*sub_hfyu_median_prediction)(uint8_t *dst, const uint8_t *src1, const uint8_t *src2, int w, int *left, int *left_top);
    void (*add_hfyu_median_prediction)(uint8_t *dst, const uint8_t *top, const uint8_t *diff, int w, int *left, int *left_top);
    int  (*add_hfyu_left_prediction)(uint8_t *dst, const uint8_t *src, int w, int left);
    void (*add_hfyu_left_prediction_bgr32)(uint8_t *dst, const uint8_t *src, int w, int *red, int *green, int *blue, int *alpha);
    /* this might write to dst[w] */
    void (*bswap_buf)(uint32_t *dst, const uint32_t *src, int w);
    void (*bswap16_buf)(uint16_t *dst, const uint16_t *src, int len);

    /* assume len is a multiple of 8, and arrays are 16-byte aligned */
    void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);

    /* (I)DCT */
    void (*fdct)(int16_t *block/* align 16*/);
    void (*fdct248)(int16_t *block/* align 16*/);

    /* IDCT really*/
    void (*idct)(int16_t *block/* align 16*/);

    /**
     * block -> idct -> clip to unsigned 8 bit -> dest.
     * (-1392, 0, 0, ...) -> idct -> (-174, -174, ...) -> put -> (0, 0, ...)
     * @param line_size size in bytes of a horizontal line of dest
     */
    void (*idct_put)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);

    /**
     * block -> idct -> add dest -> clip to unsigned 8 bit -> dest.
     * @param line_size size in bytes of a horizontal line of dest
     */
    void (*idct_add)(uint8_t *dest/*align 8*/, int line_size, int16_t *block/*align 16*/);

    /**
     * idct input permutation.
     * several optimized IDCTs need a permutated input (relative to the normal order of the reference
     * IDCT)
     * this permutation must be performed before the idct_put/add, note, normally this can be merged
     * with the zigzag/alternate scan<br>
     * an example to avoid confusion:
     * - (->decode coeffs -> zigzag reorder -> dequant -> reference idct ->...)
     * - (x -> reference dct -> reference idct -> x)
     * - (x -> reference dct -> simple_mmx_perm = idct_permutation -> simple_idct_mmx -> x)
     * - (->decode coeffs -> zigzag reorder -> simple_mmx_perm -> dequant -> simple_idct_mmx ->...)
     */
    uint8_t idct_permutation[64];
    int idct_permutation_type;
#define FF_NO_IDCT_PERM 1
#define FF_LIBMPEG2_IDCT_PERM 2
#define FF_SIMPLE_IDCT_PERM 3
#define FF_TRANSPOSE_IDCT_PERM 4
#define FF_PARTTRANS_IDCT_PERM 5
#define FF_SSE2_IDCT_PERM 6

    int (*try_8x8basis)(int16_t rem[64], int16_t weight[64], int16_t basis[64], int scale);
    void (*add_8x8basis)(int16_t rem[64], int16_t basis[64], int scale);
#define BASIS_SHIFT 16
#define RECON_SHIFT 6

    void (*draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides);
#define EDGE_WIDTH 16
#define EDGE_TOP    1
#define EDGE_BOTTOM 2

    void (*shrink[4])(uint8_t *dst, int dst_wrap, const uint8_t *src, int src_wrap, int width, int height);

    /**
     * Calculate scalar product of two vectors.
     * @param len length of vectors, should be multiple of 16
     */
    int32_t (*scalarproduct_int16)(const int16_t *v1, const int16_t *v2/*align 16*/, int len);
    /* ape functions */
    /**
     * Calculate scalar product of v1 and v2,
     * and v1[i] += v3[i] * mul
     * @param len length of vectors, should be multiple of 16
     */
    int32_t (*scalarproduct_and_madd_int16)(int16_t *v1/*align 16*/, const int16_t *v2, const int16_t *v3, int len, int mul);

    /**
     * Clip each element in an array of int32_t to a given minimum and maximum value.
     * @param dst  destination array
     *             constraints: 16-byte aligned
     * @param src  source array
     *             constraints: 16-byte aligned
     * @param min  minimum value
     *             constraints: must be in the range [-(1 << 24), 1 << 24]
     * @param max  maximum value
     *             constraints: must be in the range [-(1 << 24), 1 << 24]
     * @param len  number of elements in the array
     *             constraints: multiple of 32 greater than zero
     */
    void (*vector_clip_int32)(int32_t *dst, const int32_t *src, int32_t min,
                              int32_t max, unsigned int len);

    op_fill_func fill_block_tab[2];
} DSPContext;

void ff_dsputil_static_init(void);
void ff_dsputil_init(DSPContext* p, AVCodecContext *avctx);
void avpriv_dsputil_init(DSPContext* p, AVCodecContext *avctx);
attribute_deprecated void dsputil_init(DSPContext* c, AVCodecContext *avctx);

int ff_check_alignment(void);

void ff_set_cmp(DSPContext* c, me_cmp_func *cmp, int type);

void ff_dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_ppc(DSPContext* c, AVCodecContext *avctx);
void ff_dsputil_init_x86(DSPContext* c, AVCodecContext *avctx);

void ff_dsputil_init_dwt(DSPContext *c);

#endif /* AVCODEC_DSPUTIL_H */