commands: playback speed: adjust video timing after change
[mplayer.git] / libmpcodecs / vf_spp.c
blob2402d7255431840f6b7cbbedbd81cbb3297af0fc
1 /*
2 * Copyright (C) 2003 Michael Niedermayer <michaelni@gmx.at>
4 * This file is part of MPlayer.
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.
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.
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.
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)
29 #include <stdio.h>
30 #include <stdlib.h>
31 #include <string.h>
32 #include <inttypes.h>
33 #include <math.h>
35 #include "config.h"
37 #include "mp_msg.h"
38 #include "cpudetect.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavcodec/avcodec.h"
42 #include "libavcodec/dsputil.h"
44 #include "img_format.h"
45 #include "mp_image.h"
46 #include "vf.h"
47 #include "libvo/fastmemcpy.h"
49 #define XMIN(a,b) ((a) < (b) ? (a) : (b))
51 //===========================================================================//
52 static const uint8_t __attribute__((aligned(8))) dither[8][8]={
53 { 0, 48, 12, 60, 3, 51, 15, 63, },
54 { 32, 16, 44, 28, 35, 19, 47, 31, },
55 { 8, 56, 4, 52, 11, 59, 7, 55, },
56 { 40, 24, 36, 20, 43, 27, 39, 23, },
57 { 2, 50, 14, 62, 1, 49, 13, 61, },
58 { 34, 18, 46, 30, 33, 17, 45, 29, },
59 { 10, 58, 6, 54, 9, 57, 5, 53, },
60 { 42, 26, 38, 22, 41, 25, 37, 21, },
63 static const uint8_t offset[127][2]= {
64 {0,0},
65 {0,0}, {4,4},
66 {0,0}, {2,2}, {6,4}, {4,6},
67 {0,0}, {5,1}, {2,2}, {7,3}, {4,4}, {1,5}, {6,6}, {3,7},
69 {0,0}, {4,0}, {1,1}, {5,1}, {3,2}, {7,2}, {2,3}, {6,3},
70 {0,4}, {4,4}, {1,5}, {5,5}, {3,6}, {7,6}, {2,7}, {6,7},
72 {0,0}, {0,2}, {0,4}, {0,6}, {1,1}, {1,3}, {1,5}, {1,7},
73 {2,0}, {2,2}, {2,4}, {2,6}, {3,1}, {3,3}, {3,5}, {3,7},
74 {4,0}, {4,2}, {4,4}, {4,6}, {5,1}, {5,3}, {5,5}, {5,7},
75 {6,0}, {6,2}, {6,4}, {6,6}, {7,1}, {7,3}, {7,5}, {7,7},
77 {0,0}, {4,4}, {0,4}, {4,0}, {2,2}, {6,6}, {2,6}, {6,2},
78 {0,2}, {4,6}, {0,6}, {4,2}, {2,0}, {6,4}, {2,4}, {6,0},
79 {1,1}, {5,5}, {1,5}, {5,1}, {3,3}, {7,7}, {3,7}, {7,3},
80 {1,3}, {5,7}, {1,7}, {5,3}, {3,1}, {7,5}, {3,5}, {7,1},
81 {0,1}, {4,5}, {0,5}, {4,1}, {2,3}, {6,7}, {2,7}, {6,3},
82 {0,3}, {4,7}, {0,7}, {4,3}, {2,1}, {6,5}, {2,5}, {6,1},
83 {1,0}, {5,4}, {1,4}, {5,0}, {3,2}, {7,6}, {3,6}, {7,2},
84 {1,2}, {5,6}, {1,6}, {5,2}, {3,0}, {7,4}, {3,4}, {7,0},
87 struct vf_priv_s {
88 int log2_count;
89 int qp;
90 int mode;
91 int mpeg2;
92 int temp_stride;
93 uint8_t *src;
94 int16_t *temp;
95 AVCodecContext *avctx;
96 DSPContext dsp;
97 char *non_b_qp;
100 #define SHIFT 22
102 static void hardthresh_c(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation){
103 int i;
104 int bias= 0; //FIXME
105 unsigned int threshold1, threshold2;
107 threshold1= qp*((1<<4) - bias) - 1;
108 threshold2= (threshold1<<1);
110 memset(dst, 0, 64*sizeof(DCTELEM));
111 dst[0]= (src[0] + 4)>>3;
113 for(i=1; i<64; i++){
114 int level= src[i];
115 if(((unsigned)(level+threshold1))>threshold2){
116 const int j= permutation[i];
117 dst[j]= (level + 4)>>3;
122 static void softthresh_c(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation){
123 int i;
124 int bias= 0; //FIXME
125 unsigned int threshold1, threshold2;
127 threshold1= qp*((1<<4) - bias) - 1;
128 threshold2= (threshold1<<1);
130 memset(dst, 0, 64*sizeof(DCTELEM));
131 dst[0]= (src[0] + 4)>>3;
133 for(i=1; i<64; i++){
134 int level= src[i];
135 if(((unsigned)(level+threshold1))>threshold2){
136 const int j= permutation[i];
137 if(level>0)
138 dst[j]= (level - threshold1 + 4)>>3;
139 else
140 dst[j]= (level + threshold1 + 4)>>3;
145 #if HAVE_MMX
146 static void hardthresh_mmx(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation){
147 int bias= 0; //FIXME
148 unsigned int threshold1;
150 threshold1= qp*((1<<4) - bias) - 1;
152 __asm__ volatile(
153 #define REQUANT_CORE(dst0, dst1, dst2, dst3, src0, src1, src2, src3) \
154 "movq " #src0 ", %%mm0 \n\t"\
155 "movq " #src1 ", %%mm1 \n\t"\
156 "movq " #src2 ", %%mm2 \n\t"\
157 "movq " #src3 ", %%mm3 \n\t"\
158 "psubw %%mm4, %%mm0 \n\t"\
159 "psubw %%mm4, %%mm1 \n\t"\
160 "psubw %%mm4, %%mm2 \n\t"\
161 "psubw %%mm4, %%mm3 \n\t"\
162 "paddusw %%mm5, %%mm0 \n\t"\
163 "paddusw %%mm5, %%mm1 \n\t"\
164 "paddusw %%mm5, %%mm2 \n\t"\
165 "paddusw %%mm5, %%mm3 \n\t"\
166 "paddw %%mm6, %%mm0 \n\t"\
167 "paddw %%mm6, %%mm1 \n\t"\
168 "paddw %%mm6, %%mm2 \n\t"\
169 "paddw %%mm6, %%mm3 \n\t"\
170 "psubusw %%mm6, %%mm0 \n\t"\
171 "psubusw %%mm6, %%mm1 \n\t"\
172 "psubusw %%mm6, %%mm2 \n\t"\
173 "psubusw %%mm6, %%mm3 \n\t"\
174 "psraw $3, %%mm0 \n\t"\
175 "psraw $3, %%mm1 \n\t"\
176 "psraw $3, %%mm2 \n\t"\
177 "psraw $3, %%mm3 \n\t"\
179 "movq %%mm0, %%mm7 \n\t"\
180 "punpcklwd %%mm2, %%mm0 \n\t" /*A*/\
181 "punpckhwd %%mm2, %%mm7 \n\t" /*C*/\
182 "movq %%mm1, %%mm2 \n\t"\
183 "punpcklwd %%mm3, %%mm1 \n\t" /*B*/\
184 "punpckhwd %%mm3, %%mm2 \n\t" /*D*/\
185 "movq %%mm0, %%mm3 \n\t"\
186 "punpcklwd %%mm1, %%mm0 \n\t" /*A*/\
187 "punpckhwd %%mm7, %%mm3 \n\t" /*C*/\
188 "punpcklwd %%mm2, %%mm7 \n\t" /*B*/\
189 "punpckhwd %%mm2, %%mm1 \n\t" /*D*/\
191 "movq %%mm0, " #dst0 " \n\t"\
192 "movq %%mm7, " #dst1 " \n\t"\
193 "movq %%mm3, " #dst2 " \n\t"\
194 "movq %%mm1, " #dst3 " \n\t"
196 "movd %2, %%mm4 \n\t"
197 "movd %3, %%mm5 \n\t"
198 "movd %4, %%mm6 \n\t"
199 "packssdw %%mm4, %%mm4 \n\t"
200 "packssdw %%mm5, %%mm5 \n\t"
201 "packssdw %%mm6, %%mm6 \n\t"
202 "packssdw %%mm4, %%mm4 \n\t"
203 "packssdw %%mm5, %%mm5 \n\t"
204 "packssdw %%mm6, %%mm6 \n\t"
205 REQUANT_CORE( (%1), 8(%1), 16(%1), 24(%1), (%0), 8(%0), 64(%0), 72(%0))
206 REQUANT_CORE(32(%1), 40(%1), 48(%1), 56(%1),16(%0),24(%0), 48(%0), 56(%0))
207 REQUANT_CORE(64(%1), 72(%1), 80(%1), 88(%1),32(%0),40(%0), 96(%0),104(%0))
208 REQUANT_CORE(96(%1),104(%1),112(%1),120(%1),80(%0),88(%0),112(%0),120(%0))
209 : : "r" (src), "r" (dst), "g" (threshold1+1), "g" (threshold1+5), "g" (threshold1-4) //FIXME maybe more accurate then needed?
211 dst[0]= (src[0] + 4)>>3;
214 static void softthresh_mmx(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation){
215 int bias= 0; //FIXME
216 unsigned int threshold1;
218 threshold1= qp*((1<<4) - bias) - 1;
220 __asm__ volatile(
221 #undef REQUANT_CORE
222 #define REQUANT_CORE(dst0, dst1, dst2, dst3, src0, src1, src2, src3) \
223 "movq " #src0 ", %%mm0 \n\t"\
224 "movq " #src1 ", %%mm1 \n\t"\
225 "pxor %%mm6, %%mm6 \n\t"\
226 "pxor %%mm7, %%mm7 \n\t"\
227 "pcmpgtw %%mm0, %%mm6 \n\t"\
228 "pcmpgtw %%mm1, %%mm7 \n\t"\
229 "pxor %%mm6, %%mm0 \n\t"\
230 "pxor %%mm7, %%mm1 \n\t"\
231 "psubusw %%mm4, %%mm0 \n\t"\
232 "psubusw %%mm4, %%mm1 \n\t"\
233 "pxor %%mm6, %%mm0 \n\t"\
234 "pxor %%mm7, %%mm1 \n\t"\
235 "movq " #src2 ", %%mm2 \n\t"\
236 "movq " #src3 ", %%mm3 \n\t"\
237 "pxor %%mm6, %%mm6 \n\t"\
238 "pxor %%mm7, %%mm7 \n\t"\
239 "pcmpgtw %%mm2, %%mm6 \n\t"\
240 "pcmpgtw %%mm3, %%mm7 \n\t"\
241 "pxor %%mm6, %%mm2 \n\t"\
242 "pxor %%mm7, %%mm3 \n\t"\
243 "psubusw %%mm4, %%mm2 \n\t"\
244 "psubusw %%mm4, %%mm3 \n\t"\
245 "pxor %%mm6, %%mm2 \n\t"\
246 "pxor %%mm7, %%mm3 \n\t"\
248 "paddsw %%mm5, %%mm0 \n\t"\
249 "paddsw %%mm5, %%mm1 \n\t"\
250 "paddsw %%mm5, %%mm2 \n\t"\
251 "paddsw %%mm5, %%mm3 \n\t"\
252 "psraw $3, %%mm0 \n\t"\
253 "psraw $3, %%mm1 \n\t"\
254 "psraw $3, %%mm2 \n\t"\
255 "psraw $3, %%mm3 \n\t"\
257 "movq %%mm0, %%mm7 \n\t"\
258 "punpcklwd %%mm2, %%mm0 \n\t" /*A*/\
259 "punpckhwd %%mm2, %%mm7 \n\t" /*C*/\
260 "movq %%mm1, %%mm2 \n\t"\
261 "punpcklwd %%mm3, %%mm1 \n\t" /*B*/\
262 "punpckhwd %%mm3, %%mm2 \n\t" /*D*/\
263 "movq %%mm0, %%mm3 \n\t"\
264 "punpcklwd %%mm1, %%mm0 \n\t" /*A*/\
265 "punpckhwd %%mm7, %%mm3 \n\t" /*C*/\
266 "punpcklwd %%mm2, %%mm7 \n\t" /*B*/\
267 "punpckhwd %%mm2, %%mm1 \n\t" /*D*/\
269 "movq %%mm0, " #dst0 " \n\t"\
270 "movq %%mm7, " #dst1 " \n\t"\
271 "movq %%mm3, " #dst2 " \n\t"\
272 "movq %%mm1, " #dst3 " \n\t"
274 "movd %2, %%mm4 \n\t"
275 "movd %3, %%mm5 \n\t"
276 "packssdw %%mm4, %%mm4 \n\t"
277 "packssdw %%mm5, %%mm5 \n\t"
278 "packssdw %%mm4, %%mm4 \n\t"
279 "packssdw %%mm5, %%mm5 \n\t"
280 REQUANT_CORE( (%1), 8(%1), 16(%1), 24(%1), (%0), 8(%0), 64(%0), 72(%0))
281 REQUANT_CORE(32(%1), 40(%1), 48(%1), 56(%1),16(%0),24(%0), 48(%0), 56(%0))
282 REQUANT_CORE(64(%1), 72(%1), 80(%1), 88(%1),32(%0),40(%0), 96(%0),104(%0))
283 REQUANT_CORE(96(%1),104(%1),112(%1),120(%1),80(%0),88(%0),112(%0),120(%0))
284 : : "r" (src), "r" (dst), "g" (threshold1), "rm" (4) //FIXME maybe more accurate then needed?
287 dst[0]= (src[0] + 4)>>3;
289 #endif
291 static inline void add_block(int16_t *dst, int stride, DCTELEM block[64]){
292 int y;
294 for(y=0; y<8; y++){
295 *(uint32_t*)&dst[0 + y*stride]+= *(uint32_t*)&block[0 + y*8];
296 *(uint32_t*)&dst[2 + y*stride]+= *(uint32_t*)&block[2 + y*8];
297 *(uint32_t*)&dst[4 + y*stride]+= *(uint32_t*)&block[4 + y*8];
298 *(uint32_t*)&dst[6 + y*stride]+= *(uint32_t*)&block[6 + y*8];
302 static void store_slice_c(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int height, int log2_scale){
303 int y, x;
305 #define STORE(pos) \
306 temp= ((src[x + y*src_stride + pos]<<log2_scale) + d[pos])>>6;\
307 if(temp & 0x100) temp= ~(temp>>31);\
308 dst[x + y*dst_stride + pos]= temp;
310 for(y=0; y<height; y++){
311 const uint8_t *d= dither[y];
312 for(x=0; x<width; x+=8){
313 int temp;
314 STORE(0);
315 STORE(1);
316 STORE(2);
317 STORE(3);
318 STORE(4);
319 STORE(5);
320 STORE(6);
321 STORE(7);
326 #if HAVE_MMX
327 static void store_slice_mmx(uint8_t *dst, int16_t *src, int dst_stride, int src_stride, int width, int height, int log2_scale){
328 int y;
330 for(y=0; y<height; y++){
331 uint8_t *dst1= dst;
332 int16_t *src1= src;
333 __asm__ volatile(
334 "movq (%3), %%mm3 \n\t"
335 "movq (%3), %%mm4 \n\t"
336 "movd %4, %%mm2 \n\t"
337 "pxor %%mm0, %%mm0 \n\t"
338 "punpcklbw %%mm0, %%mm3 \n\t"
339 "punpckhbw %%mm0, %%mm4 \n\t"
340 "psraw %%mm2, %%mm3 \n\t"
341 "psraw %%mm2, %%mm4 \n\t"
342 "movd %5, %%mm2 \n\t"
343 "1: \n\t"
344 "movq (%0), %%mm0 \n\t"
345 "movq 8(%0), %%mm1 \n\t"
346 "paddw %%mm3, %%mm0 \n\t"
347 "paddw %%mm4, %%mm1 \n\t"
348 "psraw %%mm2, %%mm0 \n\t"
349 "psraw %%mm2, %%mm1 \n\t"
350 "packuswb %%mm1, %%mm0 \n\t"
351 "movq %%mm0, (%1) \n\t"
352 "add $16, %0 \n\t"
353 "add $8, %1 \n\t"
354 "cmp %2, %1 \n\t"
355 " jb 1b \n\t"
356 : "+r" (src1), "+r"(dst1)
357 : "r"(dst + width), "r"(dither[y]), "g"(log2_scale), "g"(6-log2_scale)
359 src += src_stride;
360 dst += dst_stride;
362 // if(width != mmxw)
363 // store_slice_c(dst + mmxw, src + mmxw, dst_stride, src_stride, width - mmxw, log2_scale);
365 #endif
367 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 static void (*requantize)(DCTELEM dst[64], DCTELEM src[64], int qp, uint8_t *permutation)= hardthresh_c;
371 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){
372 int x, y, i;
373 const int count= 1<<p->log2_count;
374 const int stride= is_luma ? p->temp_stride : ((width+16+15)&(~15));
375 uint64_t __attribute__((aligned(16))) block_align[32];
376 DCTELEM *block = (DCTELEM *)block_align;
377 DCTELEM *block2= (DCTELEM *)(block_align+16);
379 if (!src || !dst) return; // HACK avoid crash for Y8 colourspace
380 for(y=0; y<height; y++){
381 int index= 8 + 8*stride + y*stride;
382 fast_memcpy(p->src + index, src + y*src_stride, width);
383 for(x=0; x<8; x++){
384 p->src[index - x - 1]= p->src[index + x ];
385 p->src[index + width + x ]= p->src[index + width - x - 1];
388 for(y=0; y<8; y++){
389 fast_memcpy(p->src + ( 7-y)*stride, p->src + ( y+8)*stride, stride);
390 fast_memcpy(p->src + (height+8+y)*stride, p->src + (height-y+7)*stride, stride);
392 //FIXME (try edge emu)
394 for(y=0; y<height+8; y+=8){
395 memset(p->temp + (8+y)*stride, 0, 8*stride*sizeof(int16_t));
396 for(x=0; x<width+8; x+=8){
397 const int qps= 3 + is_luma;
398 int qp;
400 if(p->qp)
401 qp= p->qp;
402 else{
403 qp= qp_store[ (XMIN(x, width-1)>>qps) + (XMIN(y, height-1)>>qps) * qp_stride];
404 qp = FFMAX(1, norm_qscale(qp, p->mpeg2));
406 for(i=0; i<count; i++){
407 const int x1= x + offset[i+count-1][0];
408 const int y1= y + offset[i+count-1][1];
409 const int index= x1 + y1*stride;
410 p->dsp.get_pixels(block, p->src + index, stride);
411 p->dsp.fdct(block);
412 requantize(block2, block, qp, p->dsp.idct_permutation);
413 p->dsp.idct(block2);
414 add_block(p->temp + index, stride, block2);
417 if(y)
418 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 #if 0
421 for(y=0; y<height; y++){
422 for(x=0; x<width; x++){
423 if((((x>>6) ^ (y>>6)) & 1) == 0)
424 dst[x + y*dst_stride]= p->src[8 + 8*stride + x + y*stride];
425 if((x&63) == 0 || (y&63)==0)
426 dst[x + y*dst_stride] += 128;
429 #endif
430 //FIXME reorder for better caching
433 static int config(struct vf_instance *vf,
434 int width, int height, int d_width, int d_height,
435 unsigned int flags, unsigned int outfmt){
436 int h= (height+16+15)&(~15);
438 vf->priv->temp_stride= (width+16+15)&(~15);
439 vf->priv->temp= malloc(vf->priv->temp_stride*h*sizeof(int16_t));
440 vf->priv->src = malloc(vf->priv->temp_stride*h*sizeof(uint8_t));
442 return vf_next_config(vf,width,height,d_width,d_height,flags,outfmt);
445 static void get_image(struct vf_instance *vf, mp_image_t *mpi){
446 if(mpi->flags&MP_IMGFLAG_PRESERVE) return; // don't change
447 // ok, we can do pp in-place (or pp disabled):
448 vf->dmpi=vf_get_image(vf->next,mpi->imgfmt,
449 mpi->type, mpi->flags | MP_IMGFLAG_READABLE, mpi->width, mpi->height);
450 mpi->planes[0]=vf->dmpi->planes[0];
451 mpi->stride[0]=vf->dmpi->stride[0];
452 mpi->width=vf->dmpi->width;
453 if(mpi->flags&MP_IMGFLAG_PLANAR){
454 mpi->planes[1]=vf->dmpi->planes[1];
455 mpi->planes[2]=vf->dmpi->planes[2];
456 mpi->stride[1]=vf->dmpi->stride[1];
457 mpi->stride[2]=vf->dmpi->stride[2];
459 mpi->flags|=MP_IMGFLAG_DIRECT;
462 static int put_image(struct vf_instance *vf, mp_image_t *mpi, double pts){
463 mp_image_t *dmpi;
465 if(!(mpi->flags&MP_IMGFLAG_DIRECT)){
466 // no DR, so get a new image! hope we'll get DR buffer:
467 dmpi=vf_get_image(vf->next,mpi->imgfmt,
468 MP_IMGTYPE_TEMP,
469 MP_IMGFLAG_ACCEPT_STRIDE|MP_IMGFLAG_PREFER_ALIGNED_STRIDE,
470 mpi->width,mpi->height);
471 vf_clone_mpi_attributes(dmpi, mpi);
472 }else{
473 dmpi=vf->dmpi;
476 vf->priv->mpeg2= mpi->qscale_type;
477 if(mpi->pict_type != 3 && mpi->qscale && !vf->priv->qp){
478 int w = mpi->qstride;
479 int h = (mpi->h + 15) >> 4;
480 if (!w) {
481 w = (mpi->w + 15) >> 4;
482 h = 1;
484 if(!vf->priv->non_b_qp)
485 vf->priv->non_b_qp= malloc(w*h);
486 fast_memcpy(vf->priv->non_b_qp, mpi->qscale, w*h);
488 if(vf->priv->log2_count || !(mpi->flags&MP_IMGFLAG_DIRECT)){
489 char *qp_tab= vf->priv->non_b_qp;
490 if((vf->priv->mode&4) || !qp_tab)
491 qp_tab= mpi->qscale;
493 if(qp_tab || vf->priv->qp){
494 filter(vf->priv, dmpi->planes[0], mpi->planes[0], dmpi->stride[0], mpi->stride[0], mpi->w, mpi->h, qp_tab, mpi->qstride, 1);
495 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);
496 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);
497 }else{
498 memcpy_pic(dmpi->planes[0], mpi->planes[0], mpi->w, mpi->h, dmpi->stride[0], mpi->stride[0]);
499 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]);
500 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]);
504 #if HAVE_MMX
505 if(gCpuCaps.hasMMX) __asm__ volatile ("emms\n\t");
506 #endif
507 #if HAVE_MMX2
508 if(gCpuCaps.hasMMX2) __asm__ volatile ("sfence\n\t");
509 #endif
511 return vf_next_put_image(vf,dmpi, pts);
514 static void uninit(struct vf_instance *vf){
515 if(!vf->priv) return;
517 free(vf->priv->temp);
518 vf->priv->temp= NULL;
519 free(vf->priv->src);
520 vf->priv->src= NULL;
521 free(vf->priv->avctx);
522 vf->priv->avctx= NULL;
523 free(vf->priv->non_b_qp);
524 vf->priv->non_b_qp= NULL;
526 free(vf->priv);
527 vf->priv=NULL;
530 //===========================================================================//
531 static int query_format(struct vf_instance *vf, unsigned int fmt){
532 switch(fmt){
533 case IMGFMT_YVU9:
534 case IMGFMT_IF09:
535 case IMGFMT_YV12:
536 case IMGFMT_I420:
537 case IMGFMT_IYUV:
538 case IMGFMT_CLPL:
539 case IMGFMT_Y800:
540 case IMGFMT_Y8:
541 case IMGFMT_444P:
542 case IMGFMT_422P:
543 case IMGFMT_411P:
544 return vf_next_query_format(vf,fmt);
546 return 0;
549 static int control(struct vf_instance *vf, int request, void* data){
550 switch(request){
551 case VFCTRL_QUERY_MAX_PP_LEVEL:
552 return 6;
553 case VFCTRL_SET_PP_LEVEL:
554 vf->priv->log2_count= *((unsigned int*)data);
555 return CONTROL_TRUE;
557 return vf_next_control(vf,request,data);
560 static int vf_open(vf_instance_t *vf, char *args){
562 int log2c=-1;
564 vf->config=config;
565 vf->put_image=put_image;
566 vf->get_image=get_image;
567 vf->query_format=query_format;
568 vf->uninit=uninit;
569 vf->control= control;
570 vf->priv=malloc(sizeof(struct vf_priv_s));
571 memset(vf->priv, 0, sizeof(struct vf_priv_s));
573 vf->priv->avctx= avcodec_alloc_context();
574 dsputil_init(&vf->priv->dsp, vf->priv->avctx);
576 vf->priv->log2_count= 3;
578 if (args) sscanf(args, "%d:%d:%d", &log2c, &vf->priv->qp, &vf->priv->mode);
580 if( log2c >=0 && log2c <=6 )
581 vf->priv->log2_count = log2c;
583 if(vf->priv->qp < 0)
584 vf->priv->qp = 0;
586 switch(vf->priv->mode&3){
587 default:
588 case 0: requantize= hardthresh_c; break;
589 case 1: requantize= softthresh_c; break;
592 #if HAVE_MMX
593 if(gCpuCaps.hasMMX){
594 store_slice= store_slice_mmx;
595 switch(vf->priv->mode&3){
596 case 0: requantize= hardthresh_mmx; break;
597 case 1: requantize= softthresh_mmx; break;
600 #endif
602 return 1;
605 const vf_info_t vf_info_spp = {
606 "simple postprocess",
607 "spp",
608 "Michael Niedermayer",
610 vf_open,
611 NULL