22 #include "libavutil/log.h"
23 #include "libavutil/opt.h"
42 const uint8_t *obmc1= obmc + y*obmc_stride;
43 const uint8_t *obmc2= obmc1+ (obmc_stride>>1);
44 const uint8_t *obmc3= obmc1+ obmc_stride*(obmc_stride>>1);
48 int v= obmc1[x] * block[3][x + y*src_stride]
49 +obmc2[x] * block[2][x + y*src_stride]
50 +obmc3[x] * block[1][x + y*src_stride]
51 +obmc4[x] * block[0][x + y*src_stride];
60 if(v&(~255)) v= ~(v>>31);
61 dst8[x + y*src_stride] =
v;
70 int plane_index,
level, orientation;
72 for(plane_index=0; plane_index<3; plane_index++){
74 for(orientation=level ? 1:0; orientation<4; orientation++){
102 for(i=0; i<
QROOT; i++){
104 v *= pow(2, 1.0 / QROOT);
119 static const uint8_t brane[256]={
120 0x00,0x01,0x01,0x01,0x01,0x01,0x01,0x01,0x11,0x12,0x12,0x12,0x12,0x12,0x12,0x12,
121 0x04,0x05,0xcc,0xcc,0xcc,0xcc,0xcc,0x41,0x15,0x16,0xcc,0xcc,0xcc,0xcc,0xcc,0x52,
122 0x04,0xcc,0x05,0xcc,0xcc,0xcc,0x41,0xcc,0x15,0xcc,0x16,0xcc,0xcc,0xcc,0x52,0xcc,
123 0x04,0xcc,0xcc,0x05,0xcc,0x41,0xcc,0xcc,0x15,0xcc,0xcc,0x16,0xcc,0x52,0xcc,0xcc,
124 0x04,0xcc,0xcc,0xcc,0x41,0xcc,0xcc,0xcc,0x15,0xcc,0xcc,0xcc,0x16,0xcc,0xcc,0xcc,
125 0x04,0xcc,0xcc,0x41,0xcc,0x05,0xcc,0xcc,0x15,0xcc,0xcc,0x52,0xcc,0x16,0xcc,0xcc,
126 0x04,0xcc,0x41,0xcc,0xcc,0xcc,0x05,0xcc,0x15,0xcc,0x52,0xcc,0xcc,0xcc,0x16,0xcc,
127 0x04,0x41,0xcc,0xcc,0xcc,0xcc,0xcc,0x05,0x15,0x52,0xcc,0xcc,0xcc,0xcc,0xcc,0x16,
128 0x44,0x45,0x45,0x45,0x45,0x45,0x45,0x45,0x55,0x56,0x56,0x56,0x56,0x56,0x56,0x56,
129 0x48,0x49,0xcc,0xcc,0xcc,0xcc,0xcc,0x85,0x59,0x5A,0xcc,0xcc,0xcc,0xcc,0xcc,0x96,
130 0x48,0xcc,0x49,0xcc,0xcc,0xcc,0x85,0xcc,0x59,0xcc,0x5A,0xcc,0xcc,0xcc,0x96,0xcc,
131 0x48,0xcc,0xcc,0x49,0xcc,0x85,0xcc,0xcc,0x59,0xcc,0xcc,0x5A,0xcc,0x96,0xcc,0xcc,
132 0x48,0xcc,0xcc,0xcc,0x49,0xcc,0xcc,0xcc,0x59,0xcc,0xcc,0xcc,0x96,0xcc,0xcc,0xcc,
133 0x48,0xcc,0xcc,0x85,0xcc,0x49,0xcc,0xcc,0x59,0xcc,0xcc,0x96,0xcc,0x5A,0xcc,0xcc,
134 0x48,0xcc,0x85,0xcc,0xcc,0xcc,0x49,0xcc,0x59,0xcc,0x96,0xcc,0xcc,0xcc,0x5A,0xcc,
135 0x48,0x85,0xcc,0xcc,0xcc,0xcc,0xcc,0x49,0x59,0x96,0xcc,0xcc,0xcc,0xcc,0xcc,0x5A,
138 static const uint8_t needs[16]={
148 int16_t *tmpI= tmpIt;
152 r= brane[dx + 16*dy]&15;
153 l= brane[dx + 16*dy]>>4;
155 b= needs[l] | needs[
r];
161 for(x=0; x < b_w; x++){
162 int a_1=src[x + HTAPS_MAX/2-4];
163 int a0= src[x + HTAPS_MAX/2-3];
164 int a1= src[x + HTAPS_MAX/2-2];
165 int a2= src[x + HTAPS_MAX/2-1];
166 int a3= src[x + HTAPS_MAX/2+0];
167 int a4= src[x + HTAPS_MAX/2+1];
168 int a5= src[x + HTAPS_MAX/2+2];
169 int a6= src[x + HTAPS_MAX/2+3];
172 am= 20*(a2+
a3) - 5*(a1+a4) + (a0+
a5);
181 if(am&(~255)) am= ~(am>>31);
194 for(y=0; y < b_h; y++){
195 for(x=0; x < b_w+1; x++){
206 am= (20*(a2+
a3) - 5*(a1+a4) + (a0+
a5) + 16)>>5;
210 if(am&(~255)) am= ~(am>>31);
222 for(y=0; y < b_h; y++){
223 for(x=0; x < b_w; x++){
234 am= (20*(a2+
a3) - 5*(a1+a4) + (a0+
a5) + 512)>>10;
237 if(am&(~255)) am= ~(am>>31);
251 hpel[ 6]= tmp2t[1] + 1;
254 hpel[ 9]= hpel[1] + 64;
255 hpel[10]= hpel[8] + 1;
257 #define MC_STRIDE(x) (needs[x] ? 64 : stride)
260 int dxy = dx / 8 + dy / 8 * 4;
261 const uint8_t *src1 = hpel[dxy ];
262 const uint8_t *src2 = hpel[dxy + 1];
263 const uint8_t *src3 = hpel[dxy + 4];
264 const uint8_t *src4 = hpel[dxy + 5];
271 for(y=0; y < b_h; y++){
272 for(x=0; x < b_w; x++){
273 dst[x]= ((8-dx)*(8-dy)*src1[x] + dx*(8-dy)*src2[x]+
274 (8-dx)* dy *src3[x] + dx* dy *src4[x]+32)>>6;
287 int a= weight[((dx&7) + (8*(dy&7)))];
289 for(y=0; y < b_h; y++){
290 for(x=0; x < b_w; x++){
291 dst[x]= (a*src1[x] + b*src2[x] + 4)>>3;
300 void ff_snow_pred_block(
SnowContext *
s,
uint8_t *dst,
uint8_t *tmp, ptrdiff_t
stride,
int sx,
int sy,
int b_w,
int b_h,
BlockNode *
block,
int plane_index,
int w,
int h){
303 const unsigned color = block->
color[plane_index];
304 const unsigned color4 = color*0x01010101;
306 for(y=0; y < b_h; y++){
307 *(uint32_t*)&dst[0 + y*stride]= color4;
308 *(uint32_t*)&dst[4 + y*stride]= color4;
309 *(uint32_t*)&dst[8 + y*stride]= color4;
310 *(uint32_t*)&dst[12+ y*stride]= color4;
311 *(uint32_t*)&dst[16+ y*stride]= color4;
312 *(uint32_t*)&dst[20+ y*stride]= color4;
313 *(uint32_t*)&dst[24+ y*stride]= color4;
314 *(uint32_t*)&dst[28+ y*stride]= color4;
317 for(y=0; y < b_h; y++){
318 *(uint32_t*)&dst[0 + y*stride]= color4;
319 *(uint32_t*)&dst[4 + y*stride]= color4;
320 *(uint32_t*)&dst[8 + y*stride]= color4;
321 *(uint32_t*)&dst[12+ y*stride]= color4;
324 for(y=0; y < b_h; y++){
325 *(uint32_t*)&dst[0 + y*stride]= color4;
326 *(uint32_t*)&dst[4 + y*stride]= color4;
329 for(y=0; y < b_h; y++){
330 *(uint32_t*)&dst[0 + y*stride]= color4;
333 for(y=0; y < b_h; y++){
334 for(x=0; x < b_w; x++){
342 int mx= block->
mx*scale;
343 int my= block->
my*scale;
346 const int tab_index= 3 - (b_w>>2) + (b_w>>4);
361 av_assert2((tab_index>=0 && tab_index<4) || b_w==32);
362 if((dx&3) || (dy&3) || !(b_w == b_h || 2*b_w == b_h || b_w == 2*b_h) || (b_w&(b_w-1)) || !s->
plane[plane_index].
fast_mc )
363 mc_block(&s->
plane[plane_index], dst, src, stride, b_w, b_h, dx, dy);
366 for(y=0; y<b_h; y+=16){
383 #define mca(dx,dy,b_w)\
384 static void mc_block_hpel ## dx ## dy ## b_w(uint8_t *dst, const uint8_t *src, ptrdiff_t stride, int h){\
386 mc_block(NULL, dst, src-(HTAPS_MAX/2-1)-(HTAPS_MAX/2-1)*stride, stride, b_w, b_w, dx, dy);\
413 s->dsp.put_qpel_pixels_tab [0][dy+dx/4]=\
414 s->dsp.put_no_rnd_qpel_pixels_tab[0][dy+dx/4]=\
415 s->h264qpel.put_h264_qpel_pixels_tab[0][dy+dx/4];\
416 s->dsp.put_qpel_pixels_tab [1][dy+dx/4]=\
417 s->dsp.put_no_rnd_qpel_pixels_tab[1][dy+dx/4]=\
418 s->h264qpel.put_h264_qpel_pixels_tab[1][dy+dx/4];
438 s->hdsp.put_pixels_tab [0][dy/4+dx/8]=\
439 s->hdsp.put_no_rnd_pixels_tab[0][dy/4+dx/8]=\
440 mc_block_hpel ## dx ## dy ## 16;\
441 s->hdsp.put_pixels_tab [1][dy/4+dx/8]=\
442 s->hdsp.put_no_rnd_pixels_tab[1][dy/4+dx/8]=\
443 mc_block_hpel ## dx ## dy ## 8;
483 int plane_index,
level, orientation;
484 int ret, emu_buf_size;
500 for(plane_index=0; plane_index < s->
nb_planes; plane_index++){
512 for(orientation=level ? 1 : 0; orientation<4; orientation++){
518 b->
width = (w + !(orientation&1))>>1;
519 b->
height= (h + !(orientation>1))>>1;
553 #define USE_HALFPEL_PLANE 0
568 if (!halfpel[1][p] || !halfpel[2][p] || !halfpel[3][p])
576 halfpel[1][p][i]= (20*(src[i] + src[i+1]) - 5*(src[i-1] + src[i+2]) + (src[i-2] + src[i+3]) + 16 )>>5;
583 halfpel[2][p][i]= (20*(src[i] + src[i+ls]) - 5*(src[i-ls] + src[i+2*ls]) + (src[i-2*ls] + src[i+3*ls]) + 16 )>>5;
591 halfpel[3][p][i]= (20*(src[i] + src[i+ls]) - 5*(src[i-ls] + src[i+2*ls]) + (src[i-2*ls] + src[i+3*ls]) + 16 )>>5;
670 int plane_index,
level, orientation, i;
697 for(plane_index=0; plane_index < s->
nb_planes; plane_index++){
699 for(orientation=level ? 1 : 0; orientation<4; orientation++){
uint8_t * scratchpad
data area for the ME algo, so that the ME does not need to malloc/free
av_cold int ff_snow_common_init(AVCodecContext *avctx)
int ff_snow_frame_start(SnowContext *s)
av_cold void ff_dsputil_init(DSPContext *c, AVCodecContext *avctx)
This structure describes decoded (raw) audio or video data.
void ff_snow_inner_add_yblock(const uint8_t *obmc, const int obmc_stride, uint8_t **block, int b_w, int b_h, int src_x, int src_y, int src_stride, slice_buffer *sb, int add, uint8_t *dst8)
void(* emulated_edge_mc)(uint8_t *dst, ptrdiff_t dst_stride, const uint8_t *src, ptrdiff_t src_stride, int block_w, int block_h, int src_x, int src_y, int w, int h)
Copy a rectangular area of samples to a temporary buffer and replicate the border samples...
void(* draw_edges)(uint8_t *buf, int wrap, int width, int height, int w, int h, int sides)
void av_log(void *avcl, int level, const char *fmt,...) av_printf_format(3
Send the specified message to the log if the level is less than or equal to the current av_log_level...
enum AVPixelFormat pix_fmt
Pixel format, see AV_PIX_FMT_xxx.
uint32_t * score_map
map to store the scores
static const uint8_t obmc4[16]
int ff_snow_common_init_after_header(AVCodecContext *avctx)
void ff_snow_reset_contexts(SnowContext *s)
void av_freep(void *ptr)
Free a memory block which has been allocated with av_malloc(z)() or av_realloc() and set the pointer ...
av_cold void ff_h264qpel_init(H264QpelContext *c, int bit_depth)
static const uint32_t color[16+AV_CLASS_CATEGORY_NB]
uint32_t * ref_scores[MAX_REF_FRAMES]
#define FF_CEIL_RSHIFT(a, b)
void ff_snow_release_buffer(AVCodecContext *avctx)
#define AV_LOG_ERROR
Something went wrong and cannot losslessly be recovered.
void av_free(void *ptr)
Free a memory block which has been allocated with av_malloc(z)() or av_realloc(). ...
int16_t(*[MAX_REF_FRAMES] ref_mvs)[2]
static int halfpel_interpol(SnowContext *s, uint8_t *halfpel[4][4], AVFrame *frame)
AVFrame * av_frame_alloc(void)
Allocate an AVFrame and set its fields to default values.
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Libavcodec external API header.
qpel_mc_func put_h264_qpel_pixels_tab[4][16]
uint8_t * emu_edge_buffer
av_cold void ff_hpeldsp_init(HpelDSPContext *c, int flags)
#define USE_HALFPEL_PLANE
av_cold void ff_videodsp_init(VideoDSPContext *ctx, int bpc)
void av_frame_unref(AVFrame *frame)
Unreference all the buffers referenced by frame and reset the frame fields.
static void mc_block(Plane *p, uint8_t *dst, const uint8_t *src, int stride, int b_w, int b_h, int dx, int dy)
static av_cold void init_qexp(void)
int width
picture width / height.
void * av_malloc(size_t size) av_malloc_attrib 1(1)
Allocate a block of size bytes with alignment suitable for all memory accesses (including vectors if ...
uint8_t * halfpel_plane[MAX_REF_FRAMES][4][4]
SubBand band[MAX_DWT_LEVELS][4]
int format
format of the frame, -1 if unknown or unset Values correspond to enum AVPixelFormat for video frames...
uint8_t block_state[128+32 *128]
int ff_scale_mv_ref[MAX_REF_FRAMES][MAX_REF_FRAMES]
int spatial_decomposition_count
main external API structure.
int8_t hcoeff[HTAPS_MAX/2]
int ff_snow_alloc_blocks(SnowContext *s)
int ff_get_buffer(AVCodecContext *avctx, AVFrame *frame, int flags)
Get a buffer for a frame.
BYTE int const BYTE int int int height
int linesize[AV_NUM_DATA_POINTERS]
For video, size in bytes of each picture line.
DWTELEM * temp_dwt_buffer
static int weight(int i, int blen, int offset)
static const uint16_t scale[4]
AVFrame * current_picture
#define MAX_DECOMPOSITIONS
#define CODEC_FLAG_EMU_EDGE
Don't draw edges.
common internal api header.
IDWTELEM * temp_idwt_buffer
#define FF_ALLOC_OR_GOTO(ctx, p, size, label)
#define slice_buffer_get_line(slice_buf, line_num)
uint8_t * obmc_scratchpad
DWTELEM * spatial_dwt_buffer
void ff_snow_pred_block(SnowContext *s, uint8_t *dst, uint8_t *tmp, ptrdiff_t stride, int sx, int sy, int b_w, int b_h, BlockNode *block, int plane_index, int w, int h)
uint32_t * map
map to avoid duplicate evaluations
IDWTELEM * spatial_idwt_buffer
#define AVERROR_INVALIDDATA
int key_frame
1 -> keyframe, 0-> not
av_cold void ff_snow_common_end(SnowContext *s)
#define av_assert0(cond)
assert() equivalent, that is always enabled.
Used to minimize the amount of memory used in order to optimize cache performance.
AVFrame * last_picture[MAX_REF_FRAMES]
#define AV_GET_BUFFER_FLAG_REF
The decoder will keep a reference to the frame and may reuse it later.
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
#define FF_ALLOCZ_OR_GOTO(ctx, p, size, label)
void ff_dwt_init(SnowDWTContext *c)
void * av_mallocz(size_t size) av_malloc_attrib 1(1)
Allocate a block of size bytes with alignment suitable for all memory accesses (including vectors if ...
#define av_assert2(cond)
assert() equivalent, that does lie in speed critical code.
int stride_line
Stride measured in lines, not pixels.