?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Move code to avoid forward declarations in top-level .c files
[mplayer.git] / libmpcodecs / vf_spp.c
blob5a4214d87202bdaf79d54a5d98025224201626f4
1 /*
2 * Copyright (C) 2003 Michael Niedermayer <michaelni@gmx.at>
3 *
4 * This file is part of MPlayer.
5 *
6 * MPlayer is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * MPlayer 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
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License along
17 * with MPlayer; if not, write to the Free Software Foundation, Inc.,
18 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
19 */
20
21 /*
22 * This implementation is based on an algorithm described in
23 * "Aria Nosratinia Embedded Post-Processing for
24 * Enhancement of Compressed Images (1999)"
25 * (http://citeseer.nj.nec.com/nosratinia99embedded.html)
26 */
27
28
29 #include <stdio.h>
30 #include <stdlib.h>
31 #include <string.h>
32 #include <inttypes.h>
33 #include <math.h>
34
35 #include "config.h"
36
37 #include "mp_msg.h"
38 #include "cpudetect.h"
39
40 #include "libavutil/intreadwrite.h"
41 #include "libavcodec/avcodec.h"
42 #include "libavcodec/dsputil.h"
43
44 #include "img_format.h"
45 #include "mp_image.h"
46 #include "vf.h"
47 #include "vd_ffmpeg.h"
48 #include "libvo/fastmemcpy.h"
49
50 #define XMIN(a,b) ((a) < (b) ? (a) : (b))
51
52 //===========================================================================//
53 static const uint8_t __attribute__((aligned(8))) dither[8][8]={
54 { 0, 48, 12, 60, 3, 51, 15, 63, },
55 { 32, 16, 44, 28, 35, 19, 47, 31, },
56 { 8, 56, 4, 52, 11, 59, 7, 55, },
57 { 40, 24, 36, 20, 43, 27, 39, 23, },
58 { 2, 50, 14, 62, 1, 49, 13, 61, },
59 { 34, 18, 46, 30, 33, 17, 45, 29, },
60 { 10, 58, 6, 54, 9, 57, 5, 53, },
61 { 42, 26, 38, 22, 41, 25, 37, 21, },
62 };
63
64 static const uint8_t offset[127][2]= {
65 {0,0},
66 {0,0}, {4,4},
67 {0,0}, {2,2}, {6,4}, {4,6},
68 {0,0}, {5,1}, {2,2}, {7,3}, {4,4}, {1,5}, {6,6}, {3,7},
69
70 {0,0}, {4,0}, {1,1}, {5,1}, {3,2}, {7,2}, {2,3}, {6,3},
71 {0,4}, {4,4}, {1,5}, {5,5}, {3,6}, {7,6}, {2,7}, {6,7},
72
73 {0,0}, {0,2}, {0,4}, {0,6}, {1,1}, {1,3}, {1,5}, {1,7},
74 {2,0}, {2,2}, {2,4}, {2,6}, {3,1}, {3,3}, {3,5}, {3,7},
75 {4,0}, {4,2}, {4,4}, {4,6}, {5,1}, {5,3}, {5,5}, {5,7},
76 {6,0}, {6,2}, {6,4}, {6,6}, {7,1}, {7,3}, {7,5}, {7,7},
77
78 {0,0}, {4,4}, {0,4}, {4,0}, {2,2}, {6,6}, {2,6}, {6,2},
79 {0,2}, {4,6}, {0,6}, {4,2}, {2,0}, {6,4}, {2,4}, {6,0},
80 {1,1}, {5,5}, {1,5}, {5,1}, {3,3}, {7,7}, {3,7}, {7,3},
81 {1,3}, {5,7}, {1,7}, {5,3}, {3,1}, {7,5}, {3,5}, {7,1},
82 {0,1}, {4,5}, {0,5}, {4,1}, {2,3}, {6,7}, {2,7}, {6,3},
83 {0,3}, {4,7}, {0,7}, {4,3}, {2,1}, {6,5}, {2,5}, {6,1},
84 {1,0}, {5,4}, {1,4}, {5,0}, {3,2}, {7,6}, {3,6}, {7,2},
85 {1,2}, {5,6}, {1,6}, {5,2}, {3,0}, {7,4}, {3,4}, {7,0},
86 };
87
88 struct vf_priv_s {
89 int log2_count;
90 int qp;
91 int mode;
92 int mpeg2;
93 int temp_stride;
94 uint8_t *src;
95 int16_t *temp;
96 AVCodecContext *avctx;
97 DSPContext dsp;
98 char *non_b_qp;
99 };
100
101 #define SHIFT 22
102
103 static void hardthresh_c(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation){
104 int i;
105 int bias= 0; //FIXME
106 unsigned int threshold1, threshold2;
107
108 threshold1= qp*((1<<4) - bias) - 1;
109 threshold2= (threshold1<<1);
110
111 memset(dst, 0, 64*sizeof(DCTELEM));
112 dst[0]= (src[0] + 4)>>3;
113
114 for(i=1; i<64; i++){
115 int level= src[i];
116 if(((unsigned)(level+threshold1))>threshold2){
117 const int j= permutation[i];
118 dst[j]= (level + 4)>>3;
119 }
120 }
121 }
122
123 static void softthresh_c(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation){
124 int i;
125 int bias= 0; //FIXME
126 unsigned int threshold1, threshold2;
127
128 threshold1= qp*((1<<4) - bias) - 1;
129 threshold2= (threshold1<<1);
130
131 memset(dst, 0, 64*sizeof(DCTELEM));
132 dst[0]= (src[0] + 4)>>3;
133
134 for(i=1; i<64; i++){
135 int level= src[i];
136 if(((unsigned)(level+threshold1))>threshold2){
137 const int j= permutation[i];
138 if(level>0)
139 dst[j]= (level - threshold1 + 4)>>3;
140 else
141 dst[j]= (level + threshold1 + 4)>>3;
142 }
143 }
144 }
145
146 #if HAVE_MMX
147 static void hardthresh_mmx(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation){
148 int bias= 0; //FIXME
149 unsigned int threshold1;
150
151 threshold1= qp*((1<<4) - bias) - 1;
152
153 __asm__ volatile(
154 #define REQUANT_CORE(dst0, dst1, dst2, dst3, src0, src1, src2, src3) \
155 "movq " #src0 ", %%mm0 \n\t"\
156 "movq " #src1 ", %%mm1 \n\t"\
157 "movq " #src2 ", %%mm2 \n\t"\
158 "movq " #src3 ", %%mm3 \n\t"\
159 "psubw %%mm4, %%mm0 \n\t"\
160 "psubw %%mm4, %%mm1 \n\t"\
161 "psubw %%mm4, %%mm2 \n\t"\
162 "psubw %%mm4, %%mm3 \n\t"\
163 "paddusw %%mm5, %%mm0 \n\t"\
164 "paddusw %%mm5, %%mm1 \n\t"\
165 "paddusw %%mm5, %%mm2 \n\t"\
166 "paddusw %%mm5, %%mm3 \n\t"\
167 "paddw %%mm6, %%mm0 \n\t"\
168 "paddw %%mm6, %%mm1 \n\t"\
169 "paddw %%mm6, %%mm2 \n\t"\
170 "paddw %%mm6, %%mm3 \n\t"\
171 "psubusw %%mm6, %%mm0 \n\t"\
172 "psubusw %%mm6, %%mm1 \n\t"\
173 "psubusw %%mm6, %%mm2 \n\t"\
174 "psubusw %%mm6, %%mm3 \n\t"\
175 "psraw $3, %%mm0 \n\t"\
176 "psraw $3, %%mm1 \n\t"\
177 "psraw $3, %%mm2 \n\t"\
178 "psraw $3, %%mm3 \n\t"\
179 \
180 "movq %%mm0, %%mm7 \n\t"\
181 "punpcklwd %%mm2, %%mm0 \n\t" /*A*/\
182 "punpckhwd %%mm2, %%mm7 \n\t" /*C*/\
183 "movq %%mm1, %%mm2 \n\t"\
184 "punpcklwd %%mm3, %%mm1 \n\t" /*B*/\
185 "punpckhwd %%mm3, %%mm2 \n\t" /*D*/\
186 "movq %%mm0, %%mm3 \n\t"\
187 "punpcklwd %%mm1, %%mm0 \n\t" /*A*/\
188 "punpckhwd %%mm7, %%mm3 \n\t" /*C*/\
189 "punpcklwd %%mm2, %%mm7 \n\t" /*B*/\
190 "punpckhwd %%mm2, %%mm1 \n\t" /*D*/\
191 \
192 "movq %%mm0, " #dst0 " \n\t"\
193 "movq %%mm7, " #dst1 " \n\t"\
194 "movq %%mm3, " #dst2 " \n\t"\
195 "movq %%mm1, " #dst3 " \n\t"
196
197 "movd %2, %%mm4 \n\t"
198 "movd %3, %%mm5 \n\t"
199 "movd %4, %%mm6 \n\t"
200 "packssdw %%mm4, %%mm4 \n\t"
201 "packssdw %%mm5, %%mm5 \n\t"
202 "packssdw %%mm6, %%mm6 \n\t"
203 "packssdw %%mm4, %%mm4 \n\t"
204 "packssdw %%mm5, %%mm5 \n\t"
205 "packssdw %%mm6, %%mm6 \n\t"
206 REQUANT_CORE( (%1), 8(%1), 16(%1), 24(%1), (%0), 8(%0), 64(%0), 72(%0))
207 REQUANT_CORE(32(%1), 40(%1), 48(%1), 56(%1),16(%0),24(%0), 48(%0), 56(%0))
208 REQUANT_CORE(64(%1), 72(%1), 80(%1), 88(%1),32(%0),40(%0), 96(%0),104(%0))
209 REQUANT_CORE(96(%1),104(%1),112(%1),120(%1),80(%0),88(%0),112(%0),120(%0))
210 : : "r" (src), "r" (dst), "g" (threshold1+1), "g" (threshold1+5), "g" (threshold1-4) //FIXME maybe more accurate then needed?
211 );
212 dst[0]= (src[0] + 4)>>3;
213 }
214
215 static void softthresh_mmx(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation){
216 int bias= 0; //FIXME
217 unsigned int threshold1;
218
219 threshold1= qp*((1<<4) - bias) - 1;
220
221 __asm__ volatile(
222 #undef REQUANT_CORE
223 #define REQUANT_CORE(dst0, dst1, dst2, dst3, src0, src1, src2, src3) \
224 "movq " #src0 ", %%mm0 \n\t"\
225 "movq " #src1 ", %%mm1 \n\t"\
226 "pxor %%mm6, %%mm6 \n\t"\
227 "pxor %%mm7, %%mm7 \n\t"\
228 "pcmpgtw %%mm0, %%mm6 \n\t"\
229 "pcmpgtw %%mm1, %%mm7 \n\t"\
230 "pxor %%mm6, %%mm0 \n\t"\
231 "pxor %%mm7, %%mm1 \n\t"\
232 "psubusw %%mm4, %%mm0 \n\t"\
233 "psubusw %%mm4, %%mm1 \n\t"\
234 "pxor %%mm6, %%mm0 \n\t"\
235 "pxor %%mm7, %%mm1 \n\t"\
236 "movq " #src2 ", %%mm2 \n\t"\
237 "movq " #src3 ", %%mm3 \n\t"\
238 "pxor %%mm6, %%mm6 \n\t"\
239 "pxor %%mm7, %%mm7 \n\t"\
240 "pcmpgtw %%mm2, %%mm6 \n\t"\
241 "pcmpgtw %%mm3, %%mm7 \n\t"\
242 "pxor %%mm6, %%mm2 \n\t"\
243 "pxor %%mm7, %%mm3 \n\t"\
244 "psubusw %%mm4, %%mm2 \n\t"\
245 "psubusw %%mm4, %%mm3 \n\t"\
246 "pxor %%mm6, %%mm2 \n\t"\
247 "pxor %%mm7, %%mm3 \n\t"\
248 \
249 "paddsw %%mm5, %%mm0 \n\t"\
250 "paddsw %%mm5, %%mm1 \n\t"\
251 "paddsw %%mm5, %%mm2 \n\t"\
252 "paddsw %%mm5, %%mm3 \n\t"\
253 "psraw $3, %%mm0 \n\t"\
254 "psraw $3, %%mm1 \n\t"\
255 "psraw $3, %%mm2 \n\t"\
256 "psraw $3, %%mm3 \n\t"\
257 \
258 "movq %%mm0, %%mm7 \n\t"\
259 "punpcklwd %%mm2, %%mm0 \n\t" /*A*/\
260 "punpckhwd %%mm2, %%mm7 \n\t" /*C*/\
261 "movq %%mm1, %%mm2 \n\t"\
262 "punpcklwd %%mm3, %%mm1 \n\t" /*B*/\
263 "punpckhwd %%mm3, %%mm2 \n\t" /*D*/\
264 "movq %%mm0, %%mm3 \n\t"\
265 "punpcklwd %%mm1, %%mm0 \n\t" /*A*/\
266 "punpckhwd %%mm7, %%mm3 \n\t" /*C*/\
267 "punpcklwd %%mm2, %%mm7 \n\t" /*B*/\
268 "punpckhwd %%mm2, %%mm1 \n\t" /*D*/\
269 \
270 "movq %%mm0, " #dst0 " \n\t"\
271 "movq %%mm7, " #dst1 " \n\t"\
272 "movq %%mm3, " #dst2 " \n\t"\
273 "movq %%mm1, " #dst3 " \n\t"
274
275 "movd %2, %%mm4 \n\t"
276 "movd %3, %%mm5 \n\t"
277 "packssdw %%mm4, %%mm4 \n\t"
278 "packssdw %%mm5, %%mm5 \n\t"
279 "packssdw %%mm4, %%mm4 \n\t"
280 "packssdw %%mm5, %%mm5 \n\t"
281 REQUANT_CORE( (%1), 8(%1), 16(%1), 24(%1), (%0), 8(%0), 64(%0), 72(%0))
282 REQUANT_CORE(32(%1), 40(%1), 48(%1), 56(%1),16(%0),24(%0), 48(%0), 56(%0))
283 REQUANT_CORE(64(%1), 72(%1), 80(%1), 88(%1),32(%0),40(%0), 96(%0),104(%0))
284 REQUANT_CORE(96(%1),104(%1),112(%1),120(%1),80(%0),88(%0),112(%0),120(%0))
285 : : "r" (src), "r" (dst), "g" (threshold1), "rm" (4) //FIXME maybe more accurate then needed?
286 );
287
288 dst[0]= (src[0] + 4)>>3;
289 }
290 #endif
291
292 static inline void add_block(int16_t *dst, int stride, DCTELEM block[64]){
293 int y;
294
295 for(y=0; y<8; y++){
296 *(uint32_t*)&dst[0 + y*stride]+= *(uint32_t*)&block[0 + y*8];
297 *(uint32_t*)&dst[2 + y*stride]+= *(uint32_t*)&block[2 + y*8];
298 *(uint32_t*)&dst[4 + y*stride]+= *(uint32_t*)&block[4 + y*8];
299 *(uint32_t*)&dst[6 + y*stride]+= *(uint32_t*)&block[6 + y*8];
300 }
301 }
302
303 static void store_slice_c(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int height, int log2_scale){
304 int y, x;
305
306 #define STORE(pos) \
307 temp= ((src[x + y*src_stride + pos]<<log2_scale) + d[pos])>>6;\
308 if(temp & 0x100) temp= ~(temp>>31);\
309 dst[x + y*dst_stride + pos]= temp;
310
311 for(y=0; y<height; y++){
312 const uint8_t *d= dither[y];
313 for(x=0; x<width; x+=8){
314 int temp;
315 STORE(0);
316 STORE(1);
317 STORE(2);
318 STORE(3);
319 STORE(4);
320 STORE(5);
321 STORE(6);
322 STORE(7);
323 }
324 }
325 }
326
327 #if HAVE_MMX
328 static void store_slice_mmx(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int height, int log2_scale){
329 int y;
330
331 for(y=0; y<height; y++){
332 uint8_t *dst1= dst;
333 int16_t *src1= src;
334 __asm__ volatile(
335 "movq (%3), %%mm3 \n\t"
336 "movq (%3), %%mm4 \n\t"
337 "movd %4, %%mm2 \n\t"
338 "pxor %%mm0, %%mm0 \n\t"
339 "punpcklbw %%mm0, %%mm3 \n\t"
340 "punpckhbw %%mm0, %%mm4 \n\t"
341 "psraw %%mm2, %%mm3 \n\t"
342 "psraw %%mm2, %%mm4 \n\t"
343 "movd %5, %%mm2 \n\t"
344 "1: \n\t"
345 "movq (%0), %%mm0 \n\t"
346 "movq 8(%0), %%mm1 \n\t"
347 "paddw %%mm3, %%mm0 \n\t"
348 "paddw %%mm4, %%mm1 \n\t"
349 "psraw %%mm2, %%mm0 \n\t"
350 "psraw %%mm2, %%mm1 \n\t"
351 "packuswb %%mm1, %%mm0 \n\t"
352 "movq %%mm0, (%1) \n\t"
353 "add $16, %0 \n\t"
354 "add $8, %1 \n\t"
355 "cmp %2, %1 \n\t"
356 " jb 1b \n\t"
357 : "+r" (src1), "+r"(dst1)
358 : "r"(dst + width), "r"(dither[y]), "g"(log2_scale), "g"(6-log2_scale)
359 );
360 src += src_stride;
361 dst += dst_stride;
362 }
363 // if(width != mmxw)
364 // store_slice_c(dst + mmxw, src + mmxw, dst_stride, src_stride, width - mmxw, log2_scale);
365 }
366 #endif
367
368 static void (*store_slice)(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int height, int log2_scale)= store_slice_c;
369
370 static void (*requantize)(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation)= hardthresh_c;
371
372 static void filter(struct vf_priv_s *p, uint8_t *dst, uint8_t *src, int dst_stride, int src_stride, int width, int height, uint8_t *qp_store, int qp_stride, int is_luma){
373 int x, y, i;
374 const int count= 1<<p->log2_count;
375 const int stride= is_luma ? p->temp_stride : ((width+16+15)&(~15));
376 uint64_t __attribute__((aligned(16))) block_align[32];
377 DCTELEM *block = (DCTELEM *)block_align;
378 DCTELEM *block2= (DCTELEM *)(block_align+16);
379
380 if (!src || !dst) return; // HACK avoid crash for Y8 colourspace
381 for(y=0; y<height; y++){
382 int index= 8 + 8*stride + y*stride;
383 fast_memcpy(p->src + index, src + y*src_stride, width);
384 for(x=0; x<8; x++){
385 p->src[index - x - 1]= p->src[index + x ];
386 p->src[index + width + x ]= p->src[index + width - x - 1];
387 }
388 }
389 for(y=0; y<8; y++){
390 fast_memcpy(p->src + ( 7-y)*stride, p->src + ( y+8)*stride, stride);
391 fast_memcpy(p->src + (height+8+y)*stride, p->src + (height-y+7)*stride, stride);
392 }
393 //FIXME (try edge emu)
394
395 for(y=0; y<height+8; y+=8){
396 memset(p->temp + (8+y)*stride, 0, 8*stride*sizeof(int16_t));
397 for(x=0; x<width+8; x+=8){
398 const int qps= 3 + is_luma;
399 int qp;
400
401 if(p->qp)
402 qp= p->qp;
403 else{
404 qp= qp_store[ (XMIN(x, width-1)>>qps) + (XMIN(y, height-1)>>qps) * qp_stride];
405 qp = FFMAX(1, norm_qscale(qp, p->mpeg2));
406 }
407 for(i=0; i<count; i++){
408 const int x1= x + offset[i+count-1][0];
409 const int y1= y + offset[i+count-1][1];
410 const int index= x1 + y1*stride;
411 p->dsp.get_pixels(block, p->src + index, stride);
412 p->dsp.fdct(block);
413 requantize(block2, block, qp, p->dsp.idct_permutation);
414 p->dsp.idct(block2);
415 add_block(p->temp + index, stride, block2);
416 }
417 }
418 if(y)
419 store_slice(dst + (y-8)*dst_stride, p->temp + 8 + y*stride, dst_stride, stride, width, XMIN(8, height+8-y), 6-p->log2_count);
420 }
421 #if 0
422 for(y=0; y<height; y++){
423 for(x=0; x<width; x++){
424 if((((x>>6) ^ (y>>6)) & 1) == 0)
425 dst[x + y*dst_stride]= p->src[8 + 8*stride + x + y*stride];
426 if((x&63) == 0 || (y&63)==0)
427 dst[x + y*dst_stride] += 128;
428 }
429 }
430 #endif
431 //FIXME reorder for better caching
432 }
433
434 static int config(struct vf_instance *vf,
435 int width, int height, int d_width, int d_height,
436 unsigned int flags, unsigned int outfmt){
437 int h= (height+16+15)&(~15);
438
439 vf->priv->temp_stride= (width+16+15)&(~15);
440 vf->priv->temp= malloc(vf->priv->temp_stride*h*sizeof(int16_t));
441 vf->priv->src = malloc(vf->priv->temp_stride*h*sizeof(uint8_t));
442
443 return vf_next_config(vf,width,height,d_width,d_height,flags,outfmt);
444 }
445
446 static void get_image(struct vf_instance *vf, mp_image_t *mpi){
447 if(mpi->flags&MP_IMGFLAG_PRESERVE) return; // don't change
448 // ok, we can do pp in-place (or pp disabled):
449 vf->dmpi=vf_get_image(vf->next,mpi->imgfmt,
450 mpi->type, mpi->flags | MP_IMGFLAG_READABLE, mpi->width, mpi->height);
451 mpi->planes[0]=vf->dmpi->planes[0];
452 mpi->stride[0]=vf->dmpi->stride[0];
453 mpi->width=vf->dmpi->width;
454 if(mpi->flags&MP_IMGFLAG_PLANAR){
455 mpi->planes[1]=vf->dmpi->planes[1];
456 mpi->planes[2]=vf->dmpi->planes[2];
457 mpi->stride[1]=vf->dmpi->stride[1];
458 mpi->stride[2]=vf->dmpi->stride[2];
459 }
460 mpi->flags|=MP_IMGFLAG_DIRECT;
461 }
462
463 static int put_image(struct vf_instance *vf, mp_image_t *mpi, double pts){
464 mp_image_t *dmpi;
465
466 if(!(mpi->flags&MP_IMGFLAG_DIRECT)){
467 // no DR, so get a new image! hope we'll get DR buffer:
468 dmpi=vf_get_image(vf->next,mpi->imgfmt,
469 MP_IMGTYPE_TEMP,
470 MP_IMGFLAG_ACCEPT_STRIDE|MP_IMGFLAG_PREFER_ALIGNED_STRIDE,
471 mpi->width,mpi->height);
472 vf_clone_mpi_attributes(dmpi, mpi);
473 }else{
474 dmpi=vf->dmpi;
475 }
476
477 vf->priv->mpeg2= mpi->qscale_type;
478 if(mpi->pict_type != 3 && mpi->qscale && !vf->priv->qp){
479 int w = mpi->qstride;
480 int h = (mpi->h + 15) >> 4;
481 if (!w) {
482 w = (mpi->w + 15) >> 4;
483 h = 1;
484 }
485 if(!vf->priv->non_b_qp)
486 vf->priv->non_b_qp= malloc(w*h);
487 fast_memcpy(vf->priv->non_b_qp, mpi->qscale, w*h);
488 }
489 if(vf->priv->log2_count || !(mpi->flags&MP_IMGFLAG_DIRECT)){
490 char *qp_tab= vf->priv->non_b_qp;
491 if((vf->priv->mode&4) || !qp_tab)
492 qp_tab= mpi->qscale;
493
494 if(qp_tab || vf->priv->qp){
495 filter(vf->priv, dmpi->planes[0], mpi->planes[0], dmpi->stride[0], mpi->stride[0], mpi->w, mpi->h, qp_tab, mpi->qstride, 1);
496 filter(vf->priv, dmpi->planes[1], mpi->planes[1], dmpi->stride[1], mpi->stride[1], mpi->w>>mpi->chroma_x_shift, mpi->h>>mpi->chroma_y_shift, qp_tab, mpi->qstride, 0);
497 filter(vf->priv, dmpi->planes[2], mpi->planes[2], dmpi->stride[2], mpi->stride[2], mpi->w>>mpi->chroma_x_shift, mpi->h>>mpi->chroma_y_shift, qp_tab, mpi->qstride, 0);
498 }else{
499 memcpy_pic(dmpi->planes[0], mpi->planes[0], mpi->w, mpi->h, dmpi->stride[0], mpi->stride[0]);
500 memcpy_pic(dmpi->planes[1], mpi->planes[1], mpi->w>>mpi->chroma_x_shift, mpi->h>>mpi->chroma_y_shift, dmpi->stride[1], mpi->stride[1]);
501 memcpy_pic(dmpi->planes[2], mpi->planes[2], mpi->w>>mpi->chroma_x_shift, mpi->h>>mpi->chroma_y_shift, dmpi->stride[2], mpi->stride[2]);
502 }
503 }
504
505 #if HAVE_MMX
506 if(gCpuCaps.hasMMX) __asm__ volatile ("emms\n\t");
507 #endif
508 #if HAVE_MMX2
509 if(gCpuCaps.hasMMX2) __asm__ volatile ("sfence\n\t");
510 #endif
511
512 return vf_next_put_image(vf,dmpi, pts);
513 }
514
515 static void uninit(struct vf_instance *vf){
516 if(!vf->priv) return;
517
518 if(vf->priv->temp) free(vf->priv->temp);
519 vf->priv->temp= NULL;
520 if(vf->priv->src) free(vf->priv->src);
521 vf->priv->src= NULL;
522 if(vf->priv->avctx) free(vf->priv->avctx);
523 vf->priv->avctx= NULL;
524 if(vf->priv->non_b_qp) free(vf->priv->non_b_qp);
525 vf->priv->non_b_qp= NULL;
526
527 free(vf->priv);
528 vf->priv=NULL;
529 }
530
531 //===========================================================================//
532 static int query_format(struct vf_instance *vf, unsigned int fmt){
533 switch(fmt){
534 case IMGFMT_YVU9:
535 case IMGFMT_IF09:
536 case IMGFMT_YV12:
537 case IMGFMT_I420:
538 case IMGFMT_IYUV:
539 case IMGFMT_CLPL:
540 case IMGFMT_Y800:
541 case IMGFMT_Y8:
542 case IMGFMT_444P:
543 case IMGFMT_422P:
544 case IMGFMT_411P:
545 return vf_next_query_format(vf,fmt);
546 }
547 return 0;
548 }
549
550 static int control(struct vf_instance *vf, int request, void* data){
551 switch(request){
552 case VFCTRL_QUERY_MAX_PP_LEVEL:
553 return 6;
554 case VFCTRL_SET_PP_LEVEL:
555 vf->priv->log2_count= *((unsigned int*)data);
556 return CONTROL_TRUE;
557 }
558 return vf_next_control(vf,request,data);
559 }
560
561 static int vf_open(vf_instance_t *vf, char *args){
562
563 int log2c=-1;
564
565 vf->config=config;
566 vf->put_image=put_image;
567 vf->get_image=get_image;
568 vf->query_format=query_format;
569 vf->uninit=uninit;
570 vf->control= control;
571 vf->priv=malloc(sizeof(struct vf_priv_s));
572 memset(vf->priv, 0, sizeof(struct vf_priv_s));
573
574 init_avcodec();
575
576 vf->priv->avctx= avcodec_alloc_context();
577 dsputil_init(&vf->priv->dsp, vf->priv->avctx);
578
579 vf->priv->log2_count= 3;
580
581 if (args) sscanf(args, "%d:%d:%d", &log2c, &vf->priv->qp, &vf->priv->mode);
582
583 if( log2c >=0 && log2c <=6 )
584 vf->priv->log2_count = log2c;
585
586 if(vf->priv->qp < 0)
587 vf->priv->qp = 0;
588
589 switch(vf->priv->mode&3){
590 default:
591 case 0: requantize= hardthresh_c; break;
592 case 1: requantize= softthresh_c; break;
593 }
594
595 #if HAVE_MMX
596 if(gCpuCaps.hasMMX){
597 store_slice= store_slice_mmx;
598 switch(vf->priv->mode&3){
599 case 0: requantize= hardthresh_mmx; break;
600 case 1: requantize= softthresh_mmx; break;
601 }
602 }
603 #endif
604
605 return 1;
606 }
607
608 const vf_info_t vf_info_spp = {
609 "simple postprocess",
610 "spp",
611 "Michael Niedermayer",
612 "",
613 vf_open,
614 NULL
615 };
MPlayer development branch