FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
swscale_internal.h
Go to the documentation of this file.
1 /*
2  * Copyright (C) 2001-2011 Michael Niedermayer <michaelni@gmx.at>
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg is free software; you can redistribute it and/or
7  * modify it under the terms of the GNU Lesser General Public
8  * License as published by the Free Software Foundation; either
9  * version 2.1 of the License, or (at your option) any later version.
10  *
11  * FFmpeg is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14  * Lesser General Public License for more details.
15  *
16  * You should have received a copy of the GNU Lesser General Public
17  * License along with FFmpeg; if not, write to the Free Software
18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19  */
20 
21 #ifndef SWSCALE_SWSCALE_INTERNAL_H
22 #define SWSCALE_SWSCALE_INTERNAL_H
23 
24 #include "config.h"
25 
26 #if HAVE_ALTIVEC_H
27 #include <altivec.h>
28 #endif
29 
30 #include "version.h"
31 
32 #include "libavutil/avassert.h"
33 #include "libavutil/avutil.h"
34 #include "libavutil/common.h"
35 #include "libavutil/intreadwrite.h"
36 #include "libavutil/log.h"
37 #include "libavutil/pixfmt.h"
38 #include "libavutil/pixdesc.h"
39 
40 #define STR(s) AV_TOSTRING(s) // AV_STRINGIFY is too long
41 
42 #define YUVRGB_TABLE_HEADROOM 256
43 
44 #define MAX_FILTER_SIZE SWS_MAX_FILTER_SIZE
45 
46 #define DITHER1XBPP
47 
48 #if HAVE_BIGENDIAN
49 #define ALT32_CORR (-1)
50 #else
51 #define ALT32_CORR 1
52 #endif
53 
54 #if ARCH_X86_64
55 # define APCK_PTR2 8
56 # define APCK_COEF 16
57 # define APCK_SIZE 24
58 #else
59 # define APCK_PTR2 4
60 # define APCK_COEF 8
61 # define APCK_SIZE 16
62 #endif
63 
64 #define RETCODE_USE_CASCADE -12345
65 
66 struct SwsContext;
67 
68 typedef enum SwsDither {
69  SWS_DITHER_NONE = 0,
70  SWS_DITHER_AUTO,
71  SWS_DITHER_BAYER,
72  SWS_DITHER_ED,
73  SWS_DITHER_A_DITHER,
74  SWS_DITHER_X_DITHER,
75  NB_SWS_DITHER,
76 } SwsDither;
77 
78 typedef int (*SwsFunc)(struct SwsContext *context, const uint8_t *src[],
79  int srcStride[], int srcSliceY, int srcSliceH,
80  uint8_t *dst[], int dstStride[]);
81 
82 /**
83  * Write one line of horizontally scaled data to planar output
84  * without any additional vertical scaling (or point-scaling).
85  *
86  * @param src scaled source data, 15bit for 8-10bit output,
87  * 19-bit for 16bit output (in int32_t)
88  * @param dest pointer to the output plane. For >8bit
89  * output, this is in uint16_t
90  * @param dstW width of destination in pixels
91  * @param dither ordered dither array of type int16_t and size 8
92  * @param offset Dither offset
93  */
94 typedef void (*yuv2planar1_fn)(const int16_t *src, uint8_t *dest, int dstW,
95  const uint8_t *dither, int offset);
96 
97 /**
98  * Write one line of horizontally scaled data to planar output
99  * with multi-point vertical scaling between input pixels.
100  *
101  * @param filter vertical luma/alpha scaling coefficients, 12bit [0,4096]
102  * @param src scaled luma (Y) or alpha (A) source data, 15bit for 8-10bit output,
103  * 19-bit for 16bit output (in int32_t)
104  * @param filterSize number of vertical input lines to scale
105  * @param dest pointer to output plane. For >8bit
106  * output, this is in uint16_t
107  * @param dstW width of destination pixels
108  * @param offset Dither offset
109  */
110 typedef void (*yuv2planarX_fn)(const int16_t *filter, int filterSize,
111  const int16_t **src, uint8_t *dest, int dstW,
112  const uint8_t *dither, int offset);
113 
114 /**
115  * Write one line of horizontally scaled chroma to interleaved output
116  * with multi-point vertical scaling between input pixels.
117  *
118  * @param c SWS scaling context
119  * @param chrFilter vertical chroma scaling coefficients, 12bit [0,4096]
120  * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output,
121  * 19-bit for 16bit output (in int32_t)
122  * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output,
123  * 19-bit for 16bit output (in int32_t)
124  * @param chrFilterSize number of vertical chroma input lines to scale
125  * @param dest pointer to the output plane. For >8bit
126  * output, this is in uint16_t
127  * @param dstW width of chroma planes
128  */
129 typedef void (*yuv2interleavedX_fn)(struct SwsContext *c,
130  const int16_t *chrFilter,
131  int chrFilterSize,
132  const int16_t **chrUSrc,
133  const int16_t **chrVSrc,
134  uint8_t *dest, int dstW);
135 
136 /**
137  * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
138  * output without any additional vertical scaling (or point-scaling). Note
139  * that this function may do chroma scaling, see the "uvalpha" argument.
140  *
141  * @param c SWS scaling context
142  * @param lumSrc scaled luma (Y) source data, 15bit for 8-10bit output,
143  * 19-bit for 16bit output (in int32_t)
144  * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output,
145  * 19-bit for 16bit output (in int32_t)
146  * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output,
147  * 19-bit for 16bit output (in int32_t)
148  * @param alpSrc scaled alpha (A) source data, 15bit for 8-10bit output,
149  * 19-bit for 16bit output (in int32_t)
150  * @param dest pointer to the output plane. For 16bit output, this is
151  * uint16_t
152  * @param dstW width of lumSrc and alpSrc in pixels, number of pixels
153  * to write into dest[]
154  * @param uvalpha chroma scaling coefficient for the second line of chroma
155  * pixels, either 2048 or 0. If 0, one chroma input is used
156  * for 2 output pixels (or if the SWS_FLAG_FULL_CHR_INT flag
157  * is set, it generates 1 output pixel). If 2048, two chroma
158  * input pixels should be averaged for 2 output pixels (this
159  * only happens if SWS_FLAG_FULL_CHR_INT is not set)
160  * @param y vertical line number for this output. This does not need
161  * to be used to calculate the offset in the destination,
162  * but can be used to generate comfort noise using dithering
163  * for some output formats.
164  */
165 typedef void (*yuv2packed1_fn)(struct SwsContext *c, const int16_t *lumSrc,
166  const int16_t *chrUSrc[2],
167  const int16_t *chrVSrc[2],
168  const int16_t *alpSrc, uint8_t *dest,
169  int dstW, int uvalpha, int y);
170 /**
171  * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
172  * output by doing bilinear scaling between two input lines.
173  *
174  * @param c SWS scaling context
175  * @param lumSrc scaled luma (Y) source data, 15bit for 8-10bit output,
176  * 19-bit for 16bit output (in int32_t)
177  * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output,
178  * 19-bit for 16bit output (in int32_t)
179  * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output,
180  * 19-bit for 16bit output (in int32_t)
181  * @param alpSrc scaled alpha (A) source data, 15bit for 8-10bit output,
182  * 19-bit for 16bit output (in int32_t)
183  * @param dest pointer to the output plane. For 16bit output, this is
184  * uint16_t
185  * @param dstW width of lumSrc and alpSrc in pixels, number of pixels
186  * to write into dest[]
187  * @param yalpha luma/alpha scaling coefficients for the second input line.
188  * The first line's coefficients can be calculated by using
189  * 4096 - yalpha
190  * @param uvalpha chroma scaling coefficient for the second input line. The
191  * first line's coefficients can be calculated by using
192  * 4096 - uvalpha
193  * @param y vertical line number for this output. This does not need
194  * to be used to calculate the offset in the destination,
195  * but can be used to generate comfort noise using dithering
196  * for some output formats.
197  */
198 typedef void (*yuv2packed2_fn)(struct SwsContext *c, const int16_t *lumSrc[2],
199  const int16_t *chrUSrc[2],
200  const int16_t *chrVSrc[2],
201  const int16_t *alpSrc[2],
202  uint8_t *dest,
203  int dstW, int yalpha, int uvalpha, int y);
204 /**
205  * Write one line of horizontally scaled Y/U/V/A to packed-pixel YUV/RGB
206  * output by doing multi-point vertical scaling between input pixels.
207  *
208  * @param c SWS scaling context
209  * @param lumFilter vertical luma/alpha scaling coefficients, 12bit [0,4096]
210  * @param lumSrc scaled luma (Y) source data, 15bit for 8-10bit output,
211  * 19-bit for 16bit output (in int32_t)
212  * @param lumFilterSize number of vertical luma/alpha input lines to scale
213  * @param chrFilter vertical chroma scaling coefficients, 12bit [0,4096]
214  * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output,
215  * 19-bit for 16bit output (in int32_t)
216  * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output,
217  * 19-bit for 16bit output (in int32_t)
218  * @param chrFilterSize number of vertical chroma input lines to scale
219  * @param alpSrc scaled alpha (A) source data, 15bit for 8-10bit output,
220  * 19-bit for 16bit output (in int32_t)
221  * @param dest pointer to the output plane. For 16bit output, this is
222  * uint16_t
223  * @param dstW width of lumSrc and alpSrc in pixels, number of pixels
224  * to write into dest[]
225  * @param y vertical line number for this output. This does not need
226  * to be used to calculate the offset in the destination,
227  * but can be used to generate comfort noise using dithering
228  * or some output formats.
229  */
230 typedef void (*yuv2packedX_fn)(struct SwsContext *c, const int16_t *lumFilter,
231  const int16_t **lumSrc, int lumFilterSize,
232  const int16_t *chrFilter,
233  const int16_t **chrUSrc,
234  const int16_t **chrVSrc, int chrFilterSize,
235  const int16_t **alpSrc, uint8_t *dest,
236  int dstW, int y);
237 
238 /**
239  * Write one line of horizontally scaled Y/U/V/A to YUV/RGB
240  * output by doing multi-point vertical scaling between input pixels.
241  *
242  * @param c SWS scaling context
243  * @param lumFilter vertical luma/alpha scaling coefficients, 12bit [0,4096]
244  * @param lumSrc scaled luma (Y) source data, 15bit for 8-10bit output,
245  * 19-bit for 16bit output (in int32_t)
246  * @param lumFilterSize number of vertical luma/alpha input lines to scale
247  * @param chrFilter vertical chroma scaling coefficients, 12bit [0,4096]
248  * @param chrUSrc scaled chroma (U) source data, 15bit for 8-10bit output,
249  * 19-bit for 16bit output (in int32_t)
250  * @param chrVSrc scaled chroma (V) source data, 15bit for 8-10bit output,
251  * 19-bit for 16bit output (in int32_t)
252  * @param chrFilterSize number of vertical chroma input lines to scale
253  * @param alpSrc scaled alpha (A) source data, 15bit for 8-10bit output,
254  * 19-bit for 16bit output (in int32_t)
255  * @param dest pointer to the output planes. For 16bit output, this is
256  * uint16_t
257  * @param dstW width of lumSrc and alpSrc in pixels, number of pixels
258  * to write into dest[]
259  * @param y vertical line number for this output. This does not need
260  * to be used to calculate the offset in the destination,
261  * but can be used to generate comfort noise using dithering
262  * or some output formats.
263  */
264 typedef void (*yuv2anyX_fn)(struct SwsContext *c, const int16_t *lumFilter,
265  const int16_t **lumSrc, int lumFilterSize,
266  const int16_t *chrFilter,
267  const int16_t **chrUSrc,
268  const int16_t **chrVSrc, int chrFilterSize,
269  const int16_t **alpSrc, uint8_t **dest,
270  int dstW, int y);
271 
272 /* This struct should be aligned on at least a 32-byte boundary. */
273 typedef struct SwsContext {
274  /**
275  * info on struct for av_log
276  */
277  const AVClass *av_class;
278 
279  /**
280  * Note that src, dst, srcStride, dstStride will be copied in the
281  * sws_scale() wrapper so they can be freely modified here.
282  */
283  SwsFunc swscale;
284  int srcW; ///< Width of source luma/alpha planes.
285  int srcH; ///< Height of source luma/alpha planes.
286  int dstH; ///< Height of destination luma/alpha planes.
287  int chrSrcW; ///< Width of source chroma planes.
288  int chrSrcH; ///< Height of source chroma planes.
289  int chrDstW; ///< Width of destination chroma planes.
290  int chrDstH; ///< Height of destination chroma planes.
291  int lumXInc, chrXInc;
292  int lumYInc, chrYInc;
293  enum AVPixelFormat dstFormat; ///< Destination pixel format.
294  enum AVPixelFormat srcFormat; ///< Source pixel format.
295  int dstFormatBpp; ///< Number of bits per pixel of the destination pixel format.
296  int srcFormatBpp; ///< Number of bits per pixel of the source pixel format.
297  int dstBpc, srcBpc;
298  int chrSrcHSubSample; ///< Binary logarithm of horizontal subsampling factor between luma/alpha and chroma planes in source image.
299  int chrSrcVSubSample; ///< Binary logarithm of vertical subsampling factor between luma/alpha and chroma planes in source image.
300  int chrDstHSubSample; ///< Binary logarithm of horizontal subsampling factor between luma/alpha and chroma planes in destination image.
301  int chrDstVSubSample; ///< Binary logarithm of vertical subsampling factor between luma/alpha and chroma planes in destination image.
302  int vChrDrop; ///< Binary logarithm of extra vertical subsampling factor in source image chroma planes specified by user.
303  int sliceDir; ///< Direction that slices are fed to the scaler (1 = top-to-bottom, -1 = bottom-to-top).
304  double param[2]; ///< Input parameters for scaling algorithms that need them.
305 
306  /* The cascaded_* fields allow spliting a scaler task into multiple
307  * sequential steps, this is for example used to limit the maximum
308  * downscaling factor that needs to be supported in one scaler.
309  */
310  struct SwsContext *cascaded_context[2];
311  int cascaded_tmpStride[4];
312  uint8_t *cascaded_tmp[4];
313 
314  uint32_t pal_yuv[256];
315  uint32_t pal_rgb[256];
316 
317  /**
318  * @name Scaled horizontal lines ring buffer.
319  * The horizontal scaler keeps just enough scaled lines in a ring buffer
320  * so they may be passed to the vertical scaler. The pointers to the
321  * allocated buffers for each line are duplicated in sequence in the ring
322  * buffer to simplify indexing and avoid wrapping around between lines
323  * inside the vertical scaler code. The wrapping is done before the
324  * vertical scaler is called.
325  */
326  //@{
327  int16_t **lumPixBuf; ///< Ring buffer for scaled horizontal luma plane lines to be fed to the vertical scaler.
328  int16_t **chrUPixBuf; ///< Ring buffer for scaled horizontal chroma plane lines to be fed to the vertical scaler.
329  int16_t **chrVPixBuf; ///< Ring buffer for scaled horizontal chroma plane lines to be fed to the vertical scaler.
330  int16_t **alpPixBuf; ///< Ring buffer for scaled horizontal alpha plane lines to be fed to the vertical scaler.
331  int vLumBufSize; ///< Number of vertical luma/alpha lines allocated in the ring buffer.
332  int vChrBufSize; ///< Number of vertical chroma lines allocated in the ring buffer.
333  int lastInLumBuf; ///< Last scaled horizontal luma/alpha line from source in the ring buffer.
334  int lastInChrBuf; ///< Last scaled horizontal chroma line from source in the ring buffer.
335  int lumBufIndex; ///< Index in ring buffer of the last scaled horizontal luma/alpha line from source.
336  int chrBufIndex; ///< Index in ring buffer of the last scaled horizontal chroma line from source.
337  //@}
338 
339  uint8_t *formatConvBuffer;
340 
341  /**
342  * @name Horizontal and vertical filters.
343  * To better understand the following fields, here is a pseudo-code of
344  * their usage in filtering a horizontal line:
345  * @code
346  * for (i = 0; i < width; i++) {
347  * dst[i] = 0;
348  * for (j = 0; j < filterSize; j++)
349  * dst[i] += src[ filterPos[i] + j ] * filter[ filterSize * i + j ];
350  * dst[i] >>= FRAC_BITS; // The actual implementation is fixed-point.
351  * }
352  * @endcode
353  */
354  //@{
355  int16_t *hLumFilter; ///< Array of horizontal filter coefficients for luma/alpha planes.
356  int16_t *hChrFilter; ///< Array of horizontal filter coefficients for chroma planes.
357  int16_t *vLumFilter; ///< Array of vertical filter coefficients for luma/alpha planes.
358  int16_t *vChrFilter; ///< Array of vertical filter coefficients for chroma planes.
359  int32_t *hLumFilterPos; ///< Array of horizontal filter starting positions for each dst[i] for luma/alpha planes.
360  int32_t *hChrFilterPos; ///< Array of horizontal filter starting positions for each dst[i] for chroma planes.
361  int32_t *vLumFilterPos; ///< Array of vertical filter starting positions for each dst[i] for luma/alpha planes.
362  int32_t *vChrFilterPos; ///< Array of vertical filter starting positions for each dst[i] for chroma planes.
363  int hLumFilterSize; ///< Horizontal filter size for luma/alpha pixels.
364  int hChrFilterSize; ///< Horizontal filter size for chroma pixels.
365  int vLumFilterSize; ///< Vertical filter size for luma/alpha pixels.
366  int vChrFilterSize; ///< Vertical filter size for chroma pixels.
367  //@}
368 
369  int lumMmxextFilterCodeSize; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code size for luma/alpha planes.
370  int chrMmxextFilterCodeSize; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code size for chroma planes.
371  uint8_t *lumMmxextFilterCode; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code for luma/alpha planes.
372  uint8_t *chrMmxextFilterCode; ///< Runtime-generated MMXEXT horizontal fast bilinear scaler code for chroma planes.
373 
374  int canMMXEXTBeUsed;
375 
376  int dstY; ///< Last destination vertical line output from last slice.
377  int flags; ///< Flags passed by the user to select scaler algorithm, optimizations, subsampling, etc...
378  void *yuvTable; // pointer to the yuv->rgb table start so it can be freed()
379  // alignment ensures the offset can be added in a single
380  // instruction on e.g. ARM
381  DECLARE_ALIGNED(16, int, table_gV)[256 + 2*YUVRGB_TABLE_HEADROOM];
382  uint8_t *table_rV[256 + 2*YUVRGB_TABLE_HEADROOM];
383  uint8_t *table_gU[256 + 2*YUVRGB_TABLE_HEADROOM];
384  uint8_t *table_bU[256 + 2*YUVRGB_TABLE_HEADROOM];
385  DECLARE_ALIGNED(16, int32_t, input_rgb2yuv_table)[16+40*4]; // This table can contain both C and SIMD formatted values, the C vales are always at the XY_IDX points
386 #define RY_IDX 0
387 #define GY_IDX 1
388 #define BY_IDX 2
389 #define RU_IDX 3
390 #define GU_IDX 4
391 #define BU_IDX 5
392 #define RV_IDX 6
393 #define GV_IDX 7
394 #define BV_IDX 8
395 #define RGB2YUV_SHIFT 15
396 
397  int *dither_error[4];
398 
399  //Colorspace stuff
400  int contrast, brightness, saturation; // for sws_getColorspaceDetails
401  int srcColorspaceTable[4];
402  int dstColorspaceTable[4];
403  int srcRange; ///< 0 = MPG YUV range, 1 = JPG YUV range (source image).
404  int dstRange; ///< 0 = MPG YUV range, 1 = JPG YUV range (destination image).
405  int src0Alpha;
406  int dst0Alpha;
407  int srcXYZ;
408  int dstXYZ;
409  int src_h_chr_pos;
410  int dst_h_chr_pos;
411  int src_v_chr_pos;
412  int dst_v_chr_pos;
413  int yuv2rgb_y_offset;
414  int yuv2rgb_y_coeff;
415  int yuv2rgb_v2r_coeff;
416  int yuv2rgb_v2g_coeff;
417  int yuv2rgb_u2g_coeff;
418  int yuv2rgb_u2b_coeff;
419 
420 #define RED_DITHER "0*8"
421 #define GREEN_DITHER "1*8"
422 #define BLUE_DITHER "2*8"
423 #define Y_COEFF "3*8"
424 #define VR_COEFF "4*8"
425 #define UB_COEFF "5*8"
426 #define VG_COEFF "6*8"
427 #define UG_COEFF "7*8"
428 #define Y_OFFSET "8*8"
429 #define U_OFFSET "9*8"
430 #define V_OFFSET "10*8"
431 #define LUM_MMX_FILTER_OFFSET "11*8"
432 #define CHR_MMX_FILTER_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)
433 #define DSTW_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2"
434 #define ESP_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+8"
435 #define VROUNDER_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+16"
436 #define U_TEMP "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+24"
437 #define V_TEMP "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+32"
438 #define Y_TEMP "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+40"
439 #define ALP_MMX_FILTER_OFFSET "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*2+48"
440 #define UV_OFF_PX "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+48"
441 #define UV_OFF_BYTE "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+56"
442 #define DITHER16 "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+64"
443 #define DITHER32 "11*8+4*4*"AV_STRINGIFY(MAX_FILTER_SIZE)"*3+80"
444 #define DITHER32_INT (11*8+4*4*MAX_FILTER_SIZE*3+80) // value equal to above, used for checking that the struct hasn't been changed by mistake
445 
446  DECLARE_ALIGNED(8, uint64_t, redDither);
447  DECLARE_ALIGNED(8, uint64_t, greenDither);
448  DECLARE_ALIGNED(8, uint64_t, blueDither);
449 
450  DECLARE_ALIGNED(8, uint64_t, yCoeff);
451  DECLARE_ALIGNED(8, uint64_t, vrCoeff);
452  DECLARE_ALIGNED(8, uint64_t, ubCoeff);
453  DECLARE_ALIGNED(8, uint64_t, vgCoeff);
454  DECLARE_ALIGNED(8, uint64_t, ugCoeff);
455  DECLARE_ALIGNED(8, uint64_t, yOffset);
456  DECLARE_ALIGNED(8, uint64_t, uOffset);
457  DECLARE_ALIGNED(8, uint64_t, vOffset);
458  int32_t lumMmxFilter[4 * MAX_FILTER_SIZE];
459  int32_t chrMmxFilter[4 * MAX_FILTER_SIZE];
460  int dstW; ///< Width of destination luma/alpha planes.
461  DECLARE_ALIGNED(8, uint64_t, esp);
462  DECLARE_ALIGNED(8, uint64_t, vRounder);
463  DECLARE_ALIGNED(8, uint64_t, u_temp);
464  DECLARE_ALIGNED(8, uint64_t, v_temp);
465  DECLARE_ALIGNED(8, uint64_t, y_temp);
466  int32_t alpMmxFilter[4 * MAX_FILTER_SIZE];
467  // alignment of these values is not necessary, but merely here
468  // to maintain the same offset across x8632 and x86-64. Once we
469  // use proper offset macros in the asm, they can be removed.
470  DECLARE_ALIGNED(8, ptrdiff_t, uv_off); ///< offset (in pixels) between u and v planes
471  DECLARE_ALIGNED(8, ptrdiff_t, uv_offx2); ///< offset (in bytes) between u and v planes
472  DECLARE_ALIGNED(8, uint16_t, dither16)[8];
473  DECLARE_ALIGNED(8, uint32_t, dither32)[8];
474 
475  const uint8_t *chrDither8, *lumDither8;
476 
477 #if HAVE_ALTIVEC
478  vector signed short CY;
479  vector signed short CRV;
480  vector signed short CBU;
481  vector signed short CGU;
482  vector signed short CGV;
483  vector signed short OY;
484  vector unsigned short CSHIFT;
485  vector signed short *vYCoeffsBank, *vCCoeffsBank;
486 #endif
487 
488  int use_mmx_vfilter;
489 
490 /* pre defined color-spaces gamma */
491 #define XYZ_GAMMA (2.6f)
492 #define RGB_GAMMA (2.2f)
493  int16_t *xyzgamma;
494  int16_t *rgbgamma;
495  int16_t *xyzgammainv;
496  int16_t *rgbgammainv;
497  int16_t xyz2rgb_matrix[3][4];
498  int16_t rgb2xyz_matrix[3][4];
499 
500  /* function pointers for swscale() */
501  yuv2planar1_fn yuv2plane1;
502  yuv2planarX_fn yuv2planeX;
503  yuv2interleavedX_fn yuv2nv12cX;
504  yuv2packed1_fn yuv2packed1;
505  yuv2packed2_fn yuv2packed2;
506  yuv2packedX_fn yuv2packedX;
507  yuv2anyX_fn yuv2anyX;
508 
509  /// Unscaled conversion of luma plane to YV12 for horizontal scaler.
510  void (*lumToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3,
511  int width, uint32_t *pal);
512  /// Unscaled conversion of alpha plane to YV12 for horizontal scaler.
513  void (*alpToYV12)(uint8_t *dst, const uint8_t *src, const uint8_t *src2, const uint8_t *src3,
514  int width, uint32_t *pal);
515  /// Unscaled conversion of chroma planes to YV12 for horizontal scaler.
516  void (*chrToYV12)(uint8_t *dstU, uint8_t *dstV,
517  const uint8_t *src1, const uint8_t *src2, const uint8_t *src3,
518  int width, uint32_t *pal);
519 
520  /**
521  * Functions to read planar input, such as planar RGB, and convert
522  * internally to Y/UV/A.
523  */
524  /** @{ */
525  void (*readLumPlanar)(uint8_t *dst, const uint8_t *src[4], int width, int32_t *rgb2yuv);
526  void (*readChrPlanar)(uint8_t *dstU, uint8_t *dstV, const uint8_t *src[4],
527  int width, int32_t *rgb2yuv);
528  void (*readAlpPlanar)(uint8_t *dst, const uint8_t *src[4], int width, int32_t *rgb2yuv);
529  /** @} */
530 
531  /**
532  * Scale one horizontal line of input data using a bilinear filter
533  * to produce one line of output data. Compared to SwsContext->hScale(),
534  * please take note of the following caveats when using these:
535  * - Scaling is done using only 7bit instead of 14bit coefficients.
536  * - You can use no more than 5 input pixels to produce 4 output
537  * pixels. Therefore, this filter should not be used for downscaling
538  * by more than ~20% in width (because that equals more than 5/4th
539  * downscaling and thus more than 5 pixels input per 4 pixels output).
540  * - In general, bilinear filters create artifacts during downscaling
541  * (even when <20%), because one output pixel will span more than one
542  * input pixel, and thus some pixels will need edges of both neighbor
543  * pixels to interpolate the output pixel. Since you can use at most
544  * two input pixels per output pixel in bilinear scaling, this is
545  * impossible and thus downscaling by any size will create artifacts.
546  * To enable this type of scaling, set SWS_FLAG_FAST_BILINEAR
547  * in SwsContext->flags.
548  */
549  /** @{ */
550  void (*hyscale_fast)(struct SwsContext *c,
551  int16_t *dst, int dstWidth,
552  const uint8_t *src, int srcW, int xInc);
553  void (*hcscale_fast)(struct SwsContext *c,
554  int16_t *dst1, int16_t *dst2, int dstWidth,
555  const uint8_t *src1, const uint8_t *src2,
556  int srcW, int xInc);
557  /** @} */
558 
559  /**
560  * Scale one horizontal line of input data using a filter over the input
561  * lines, to produce one (differently sized) line of output data.
562  *
563  * @param dst pointer to destination buffer for horizontally scaled
564  * data. If the number of bits per component of one
565  * destination pixel (SwsContext->dstBpc) is <= 10, data
566  * will be 15bpc in 16bits (int16_t) width. Else (i.e.
567  * SwsContext->dstBpc == 16), data will be 19bpc in
568  * 32bits (int32_t) width.
569  * @param dstW width of destination image
570  * @param src pointer to source data to be scaled. If the number of
571  * bits per component of a source pixel (SwsContext->srcBpc)
572  * is 8, this is 8bpc in 8bits (uint8_t) width. Else
573  * (i.e. SwsContext->dstBpc > 8), this is native depth
574  * in 16bits (uint16_t) width. In other words, for 9-bit
575  * YUV input, this is 9bpc, for 10-bit YUV input, this is
576  * 10bpc, and for 16-bit RGB or YUV, this is 16bpc.
577  * @param filter filter coefficients to be used per output pixel for
578  * scaling. This contains 14bpp filtering coefficients.
579  * Guaranteed to contain dstW * filterSize entries.
580  * @param filterPos position of the first input pixel to be used for
581  * each output pixel during scaling. Guaranteed to
582  * contain dstW entries.
583  * @param filterSize the number of input coefficients to be used (and
584  * thus the number of input pixels to be used) for
585  * creating a single output pixel. Is aligned to 4
586  * (and input coefficients thus padded with zeroes)
587  * to simplify creating SIMD code.
588  */
589  /** @{ */
590  void (*hyScale)(struct SwsContext *c, int16_t *dst, int dstW,
591  const uint8_t *src, const int16_t *filter,
592  const int32_t *filterPos, int filterSize);
593  void (*hcScale)(struct SwsContext *c, int16_t *dst, int dstW,
594  const uint8_t *src, const int16_t *filter,
595  const int32_t *filterPos, int filterSize);
596  /** @} */
597 
598  /// Color range conversion function for luma plane if needed.
599  void (*lumConvertRange)(int16_t *dst, int width);
600  /// Color range conversion function for chroma planes if needed.
601  void (*chrConvertRange)(int16_t *dst1, int16_t *dst2, int width);
602 
603  int needs_hcscale; ///< Set if there are chroma planes to be converted.
604 
605  SwsDither dither;
606 } SwsContext;
607 //FIXME check init (where 0)
608 
609 SwsFunc ff_yuv2rgb_get_func_ptr(SwsContext *c);
610 int ff_yuv2rgb_c_init_tables(SwsContext *c, const int inv_table[4],
611  int fullRange, int brightness,
612  int contrast, int saturation);
613 void ff_yuv2rgb_init_tables_ppc(SwsContext *c, const int inv_table[4],
614  int brightness, int contrast, int saturation);
615 
616 void ff_updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrBufIndex,
617  int lastInLumBuf, int lastInChrBuf);
618 
619 av_cold void ff_sws_init_range_convert(SwsContext *c);
620 
621 SwsFunc ff_yuv2rgb_init_x86(SwsContext *c);
622 SwsFunc ff_yuv2rgb_init_ppc(SwsContext *c);
623 
624 static av_always_inline int is16BPS(enum AVPixelFormat pix_fmt)
625 {
626  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
627  av_assert0(desc);
628  return desc->comp[0].depth_minus1 == 15;
629 }
630 
631 static av_always_inline int is9_OR_10BPS(enum AVPixelFormat pix_fmt)
632 {
633  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
634  av_assert0(desc);
635  return desc->comp[0].depth_minus1 >= 8 && desc->comp[0].depth_minus1 <= 13;
636 }
637 
638 #define isNBPS(x) is9_OR_10BPS(x)
639 
640 static av_always_inline int isBE(enum AVPixelFormat pix_fmt)
641 {
642  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
643  av_assert0(desc);
644  return desc->flags & AV_PIX_FMT_FLAG_BE;
645 }
646 
647 static av_always_inline int isYUV(enum AVPixelFormat pix_fmt)
648 {
649  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
650  av_assert0(desc);
651  return !(desc->flags & AV_PIX_FMT_FLAG_RGB) && desc->nb_components >= 2;
652 }
653 
654 static av_always_inline int isPlanarYUV(enum AVPixelFormat pix_fmt)
655 {
656  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
657  av_assert0(desc);
658  return ((desc->flags & AV_PIX_FMT_FLAG_PLANAR) && isYUV(pix_fmt));
659 }
660 
661 static av_always_inline int isRGB(enum AVPixelFormat pix_fmt)
662 {
663  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
664  av_assert0(desc);
665  return (desc->flags & AV_PIX_FMT_FLAG_RGB);
666 }
667 
668 #if 0 // FIXME
669 #define isGray(x) \
670  (!(av_pix_fmt_desc_get(x)->flags & AV_PIX_FMT_FLAG_PAL) && \
671  av_pix_fmt_desc_get(x)->nb_components <= 2)
672 #else
673 #define isGray(x) \
674  ((x) == AV_PIX_FMT_GRAY8 || \
675  (x) == AV_PIX_FMT_YA8 || \
676  (x) == AV_PIX_FMT_GRAY16BE || \
677  (x) == AV_PIX_FMT_GRAY16LE || \
678  (x) == AV_PIX_FMT_YA16BE || \
679  (x) == AV_PIX_FMT_YA16LE)
680 #endif
681 
682 #define isRGBinInt(x) \
683  ( \
684  (x) == AV_PIX_FMT_RGB48BE || \
685  (x) == AV_PIX_FMT_RGB48LE || \
686  (x) == AV_PIX_FMT_RGB32 || \
687  (x) == AV_PIX_FMT_RGB32_1 || \
688  (x) == AV_PIX_FMT_RGB24 || \
689  (x) == AV_PIX_FMT_RGB565BE || \
690  (x) == AV_PIX_FMT_RGB565LE || \
691  (x) == AV_PIX_FMT_RGB555BE || \
692  (x) == AV_PIX_FMT_RGB555LE || \
693  (x) == AV_PIX_FMT_RGB444BE || \
694  (x) == AV_PIX_FMT_RGB444LE || \
695  (x) == AV_PIX_FMT_RGB8 || \
696  (x) == AV_PIX_FMT_RGB4 || \
697  (x) == AV_PIX_FMT_RGB4_BYTE || \
698  (x) == AV_PIX_FMT_RGBA64BE || \
699  (x) == AV_PIX_FMT_RGBA64LE || \
700  (x) == AV_PIX_FMT_MONOBLACK || \
701  (x) == AV_PIX_FMT_MONOWHITE \
702  )
703 #define isBGRinInt(x) \
704  ( \
705  (x) == AV_PIX_FMT_BGR48BE || \
706  (x) == AV_PIX_FMT_BGR48LE || \
707  (x) == AV_PIX_FMT_BGR32 || \
708  (x) == AV_PIX_FMT_BGR32_1 || \
709  (x) == AV_PIX_FMT_BGR24 || \
710  (x) == AV_PIX_FMT_BGR565BE || \
711  (x) == AV_PIX_FMT_BGR565LE || \
712  (x) == AV_PIX_FMT_BGR555BE || \
713  (x) == AV_PIX_FMT_BGR555LE || \
714  (x) == AV_PIX_FMT_BGR444BE || \
715  (x) == AV_PIX_FMT_BGR444LE || \
716  (x) == AV_PIX_FMT_BGR8 || \
717  (x) == AV_PIX_FMT_BGR4 || \
718  (x) == AV_PIX_FMT_BGR4_BYTE || \
719  (x) == AV_PIX_FMT_BGRA64BE || \
720  (x) == AV_PIX_FMT_BGRA64LE || \
721  (x) == AV_PIX_FMT_MONOBLACK || \
722  (x) == AV_PIX_FMT_MONOWHITE \
723  )
724 
725 #define isRGBinBytes(x) ( \
726  (x) == AV_PIX_FMT_RGB48BE \
727  || (x) == AV_PIX_FMT_RGB48LE \
728  || (x) == AV_PIX_FMT_RGBA64BE \
729  || (x) == AV_PIX_FMT_RGBA64LE \
730  || (x) == AV_PIX_FMT_RGBA \
731  || (x) == AV_PIX_FMT_ARGB \
732  || (x) == AV_PIX_FMT_RGB24 \
733  )
734 #define isBGRinBytes(x) ( \
735  (x) == AV_PIX_FMT_BGR48BE \
736  || (x) == AV_PIX_FMT_BGR48LE \
737  || (x) == AV_PIX_FMT_BGRA64BE \
738  || (x) == AV_PIX_FMT_BGRA64LE \
739  || (x) == AV_PIX_FMT_BGRA \
740  || (x) == AV_PIX_FMT_ABGR \
741  || (x) == AV_PIX_FMT_BGR24 \
742  )
743 
744 #define isBayer(x) ( \
745  (x)==AV_PIX_FMT_BAYER_BGGR8 \
746  || (x)==AV_PIX_FMT_BAYER_BGGR16LE \
747  || (x)==AV_PIX_FMT_BAYER_BGGR16BE \
748  || (x)==AV_PIX_FMT_BAYER_RGGB8 \
749  || (x)==AV_PIX_FMT_BAYER_RGGB16LE \
750  || (x)==AV_PIX_FMT_BAYER_RGGB16BE \
751  || (x)==AV_PIX_FMT_BAYER_GBRG8 \
752  || (x)==AV_PIX_FMT_BAYER_GBRG16LE \
753  || (x)==AV_PIX_FMT_BAYER_GBRG16BE \
754  || (x)==AV_PIX_FMT_BAYER_GRBG8 \
755  || (x)==AV_PIX_FMT_BAYER_GRBG16LE \
756  || (x)==AV_PIX_FMT_BAYER_GRBG16BE \
757  )
758 
759 #define isAnyRGB(x) \
760  ( \
761  isBayer(x) || \
762  isRGBinInt(x) || \
763  isBGRinInt(x) || \
764  isRGB(x) \
765  )
766 
767 static av_always_inline int isALPHA(enum AVPixelFormat pix_fmt)
768 {
769  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
770  av_assert0(desc);
771  if (pix_fmt == AV_PIX_FMT_PAL8)
772  return 1;
773  return desc->flags & AV_PIX_FMT_FLAG_ALPHA;
774 }
775 
776 #if 1
777 #define isPacked(x) ( \
778  (x)==AV_PIX_FMT_PAL8 \
779  || (x)==AV_PIX_FMT_YUYV422 \
780  || (x)==AV_PIX_FMT_YVYU422 \
781  || (x)==AV_PIX_FMT_UYVY422 \
782  || (x)==AV_PIX_FMT_YA8 \
783  || (x)==AV_PIX_FMT_YA16LE \
784  || (x)==AV_PIX_FMT_YA16BE \
785  || isRGBinInt(x) \
786  || isBGRinInt(x) \
787  )
788 #else
789 static av_always_inline int isPacked(enum AVPixelFormat pix_fmt)
790 {
791  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
792  av_assert0(desc);
793  return ((desc->nb_components >= 2 && !(desc->flags & AV_PIX_FMT_FLAG_PLANAR)) ||
794  pix_fmt == AV_PIX_FMT_PAL8);
795 }
796 
797 #endif
798 static av_always_inline int isPlanar(enum AVPixelFormat pix_fmt)
799 {
800  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
801  av_assert0(desc);
802  return (desc->nb_components >= 2 && (desc->flags & AV_PIX_FMT_FLAG_PLANAR));
803 }
804 
805 static av_always_inline int isPackedRGB(enum AVPixelFormat pix_fmt)
806 {
807  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
808  av_assert0(desc);
809  return ((desc->flags & (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB)) == AV_PIX_FMT_FLAG_RGB);
810 }
811 
812 static av_always_inline int isPlanarRGB(enum AVPixelFormat pix_fmt)
813 {
814  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
815  av_assert0(desc);
816  return ((desc->flags & (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB)) ==
817  (AV_PIX_FMT_FLAG_PLANAR | AV_PIX_FMT_FLAG_RGB));
818 }
819 
820 static av_always_inline int usePal(enum AVPixelFormat pix_fmt)
821 {
822  const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(pix_fmt);
823  av_assert0(desc);
824  return (desc->flags & AV_PIX_FMT_FLAG_PAL) || (desc->flags & AV_PIX_FMT_FLAG_PSEUDOPAL);
825 }
826 
827 extern const uint64_t ff_dither4[2];
828 extern const uint64_t ff_dither8[2];
829 
830 extern const uint8_t ff_dither_2x2_4[3][8];
831 extern const uint8_t ff_dither_2x2_8[3][8];
832 extern const uint8_t ff_dither_4x4_16[5][8];
833 extern const uint8_t ff_dither_8x8_32[9][8];
834 extern const uint8_t ff_dither_8x8_73[9][8];
835 extern const uint8_t ff_dither_8x8_128[9][8];
836 extern const uint8_t ff_dither_8x8_220[9][8];
837 
838 extern const int32_t ff_yuv2rgb_coeffs[8][4];
839 
840 extern const AVClass sws_context_class;
841 
842 /**
843  * Set c->swscale to an unscaled converter if one exists for the specific
844  * source and destination formats, bit depths, flags, etc.
845  */
846 void ff_get_unscaled_swscale(SwsContext *c);
847 void ff_get_unscaled_swscale_ppc(SwsContext *c);
848 void ff_get_unscaled_swscale_arm(SwsContext *c);
849 
850 /**
851  * Return function pointer to fastest main scaler path function depending
852  * on architecture and available optimizations.
853  */
854 SwsFunc ff_getSwsFunc(SwsContext *c);
855 
856 void ff_sws_init_input_funcs(SwsContext *c);
857 void ff_sws_init_output_funcs(SwsContext *c,
858  yuv2planar1_fn *yuv2plane1,
859  yuv2planarX_fn *yuv2planeX,
860  yuv2interleavedX_fn *yuv2nv12cX,
861  yuv2packed1_fn *yuv2packed1,
862  yuv2packed2_fn *yuv2packed2,
863  yuv2packedX_fn *yuv2packedX,
864  yuv2anyX_fn *yuv2anyX);
865 void ff_sws_init_swscale_ppc(SwsContext *c);
866 void ff_sws_init_swscale_x86(SwsContext *c);
867 
868 void ff_hyscale_fast_c(SwsContext *c, int16_t *dst, int dstWidth,
869  const uint8_t *src, int srcW, int xInc);
870 void ff_hcscale_fast_c(SwsContext *c, int16_t *dst1, int16_t *dst2,
871  int dstWidth, const uint8_t *src1,
872  const uint8_t *src2, int srcW, int xInc);
873 int ff_init_hscaler_mmxext(int dstW, int xInc, uint8_t *filterCode,
874  int16_t *filter, int32_t *filterPos,
875  int numSplits);
876 void ff_hyscale_fast_mmxext(SwsContext *c, int16_t *dst,
877  int dstWidth, const uint8_t *src,
878  int srcW, int xInc);
879 void ff_hcscale_fast_mmxext(SwsContext *c, int16_t *dst1, int16_t *dst2,
880  int dstWidth, const uint8_t *src1,
881  const uint8_t *src2, int srcW, int xInc);
882 
883 static inline void fillPlane16(uint8_t *plane, int stride, int width, int height, int y,
884  int alpha, int bits, const int big_endian)
885 {
886  int i, j;
887  uint8_t *ptr = plane + stride * y;
888  int v = alpha ? 0xFFFF>>(15-bits) : (1<<bits);
889  for (i = 0; i < height; i++) {
890 #define FILL(wfunc) \
891  for (j = 0; j < width; j++) {\
892  wfunc(ptr+2*j, v);\
893  }
894  if (big_endian) {
895  FILL(AV_WB16);
896  } else {
897  FILL(AV_WL16);
898  }
899  ptr += stride;
900  }
901 }
902 
903 #endif /* SWSCALE_SWSCALE_INTERNAL_H */
Generated on Sun Mar 8 2015 02:35:15 for FFmpeg by   doxygen 1.8.2