threading.c 30.4 KB
Newer Older
John Koleszar's avatar
John Koleszar committed
1
2
3
/*
 *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
 *
4
 *  Use of this source code is governed by a BSD-style license
5
6
 *  that can be found in the LICENSE file in the root of the source
 *  tree. An additional intellectual property rights grant can be found
7
 *  in the file PATENTS.  All contributing project authors may
8
 *  be found in the AUTHORS file in the root of the source tree.
John Koleszar's avatar
John Koleszar committed
9
10
11
12
13
14
 */


#ifndef WIN32
# include <unistd.h>
#endif
15
16
17
#ifdef __APPLE__
#include <mach/mach_init.h>
#endif
John Koleszar's avatar
John Koleszar committed
18
19
20
21
22
23
24
25
#include "onyxd_int.h"
#include "vpx_mem/vpx_mem.h"
#include "threading.h"

#include "loopfilter.h"
#include "extend.h"
#include "vpx_ports/vpx_timer.h"

26
27
#define MAX_ROWS 256

John Koleszar's avatar
John Koleszar committed
28
29
30
31
32
33
34
35
extern void vp8_decode_mb_row(VP8D_COMP *pbi,
                              VP8_COMMON *pc,
                              int mb_row,
                              MACROBLOCKD *xd);

extern void vp8_build_uvmvs(MACROBLOCKD *x, int fullpixel);
extern void vp8_decode_macroblock(VP8D_COMP *pbi, MACROBLOCKD *xd);

36
37
void vp8_thread_loop_filter(VP8D_COMP *pbi, MB_ROW_DEC *mbrd, int ithread);

John Koleszar's avatar
John Koleszar committed
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
void vp8_setup_decoding_thread_data(VP8D_COMP *pbi, MACROBLOCKD *xd, MB_ROW_DEC *mbrd, int count)
{
#if CONFIG_MULTITHREAD
    VP8_COMMON *const pc = & pbi->common;
    int i, j;

    for (i = 0; i < count; i++)
    {
        MACROBLOCKD *mbd = &mbrd[i].mbd;
#if CONFIG_RUNTIME_CPU_DETECT
        mbd->rtcd = xd->rtcd;
#endif
        mbd->subpixel_predict        = xd->subpixel_predict;
        mbd->subpixel_predict8x4     = xd->subpixel_predict8x4;
        mbd->subpixel_predict8x8     = xd->subpixel_predict8x8;
        mbd->subpixel_predict16x16   = xd->subpixel_predict16x16;
        mbd->gf_active_ptr            = xd->gf_active_ptr;

        mbd->mode_info        = pc->mi - 1;
        mbd->mode_info_context = pc->mi   + pc->mode_info_stride * (i + 1);
        mbd->mode_info_stride  = pc->mode_info_stride;

        mbd->frame_type = pc->frame_type;
        mbd->frames_since_golden      = pc->frames_since_golden;
        mbd->frames_till_alt_ref_frame  = pc->frames_till_alt_ref_frame;

64
65
        mbd->pre = pc->yv12_fb[pc->lst_fb_idx];
        mbd->dst = pc->yv12_fb[pc->new_fb_idx];
John Koleszar's avatar
John Koleszar committed
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83

        vp8_setup_block_dptrs(mbd);
        vp8_build_block_doffsets(mbd);
        mbd->segmentation_enabled    = xd->segmentation_enabled;
        mbd->mb_segement_abs_delta     = xd->mb_segement_abs_delta;
        vpx_memcpy(mbd->segment_feature_data, xd->segment_feature_data, sizeof(xd->segment_feature_data));

        mbd->mbmi.mode = DC_PRED;
        mbd->mbmi.uv_mode = DC_PRED;

        mbd->current_bc = &pbi->bc2;

        for (j = 0; j < 25; j++)
        {
            mbd->block[j].dequant = xd->block[j].dequant;
        }
    }

84
85
    for (i=0; i< pc->mb_rows; i++)
        pbi->current_mb_col[i]=-1;
John Koleszar's avatar
John Koleszar committed
86
87
88
89
90
91
92
93
#else
    (void) pbi;
    (void) xd;
    (void) mbrd;
    (void) count;
#endif
}

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
void vp8_setup_loop_filter_thread_data(VP8D_COMP *pbi, MACROBLOCKD *xd, MB_ROW_DEC *mbrd, int count)
{
#if CONFIG_MULTITHREAD
    VP8_COMMON *const pc = & pbi->common;
    int i, j;

    for (i = 0; i < count; i++)
    {
        MACROBLOCKD *mbd = &mbrd[i].mbd;
//#if CONFIG_RUNTIME_CPU_DETECT
//        mbd->rtcd = xd->rtcd;
//#endif

        //mbd->subpixel_predict        = xd->subpixel_predict;
        //mbd->subpixel_predict8x4     = xd->subpixel_predict8x4;
        //mbd->subpixel_predict8x8     = xd->subpixel_predict8x8;
        //mbd->subpixel_predict16x16   = xd->subpixel_predict16x16;
        //mbd->gf_active_ptr            = xd->gf_active_ptr;

        mbd->mode_info        = pc->mi - 1;
        mbd->mode_info_context = pc->mi   + pc->mode_info_stride * (i + 1);
        mbd->mode_info_stride  = pc->mode_info_stride;

        //mbd->frame_type = pc->frame_type;
        //mbd->frames_since_golden      = pc->frames_since_golden;
        //mbd->frames_till_alt_ref_frame  = pc->frames_till_alt_ref_frame;

        //mbd->pre = pc->yv12_fb[pc->lst_fb_idx];
        //mbd->dst = pc->yv12_fb[pc->new_fb_idx];

        //vp8_setup_block_dptrs(mbd);
        //vp8_build_block_doffsets(mbd);
        mbd->segmentation_enabled    = xd->segmentation_enabled;  //
        mbd->mb_segement_abs_delta     = xd->mb_segement_abs_delta;  //
        vpx_memcpy(mbd->segment_feature_data, xd->segment_feature_data, sizeof(xd->segment_feature_data));   //

        //signed char ref_lf_deltas[MAX_REF_LF_DELTAS];
        vpx_memcpy(mbd->ref_lf_deltas, xd->ref_lf_deltas, sizeof(xd->ref_lf_deltas));
        //signed char mode_lf_deltas[MAX_MODE_LF_DELTAS];
        vpx_memcpy(mbd->mode_lf_deltas, xd->mode_lf_deltas, sizeof(xd->mode_lf_deltas));
        //unsigned char mode_ref_lf_delta_enabled;
        //unsigned char mode_ref_lf_delta_update;
        mbd->mode_ref_lf_delta_enabled    = xd->mode_ref_lf_delta_enabled;
        mbd->mode_ref_lf_delta_update    = xd->mode_ref_lf_delta_update;

        //mbd->mbmi.mode = DC_PRED;
        //mbd->mbmi.uv_mode = DC_PRED;
        //mbd->current_bc = &pbi->bc2;

        //for (j = 0; j < 25; j++)
        //{
        //    mbd->block[j].dequant = xd->block[j].dequant;
        //}
    }

    for (i=0; i< pc->mb_rows; i++)
        pbi->current_mb_col[i]=-1;
#else
    (void) pbi;
    (void) xd;
    (void) mbrd;
    (void) count;
#endif
}
John Koleszar's avatar
John Koleszar committed
158
159
160
161
162
163
164
165
166
167
168
169
170
171

THREAD_FUNCTION vp8_thread_decoding_proc(void *p_data)
{
#if CONFIG_MULTITHREAD
    int ithread = ((DECODETHREAD_DATA *)p_data)->ithread;
    VP8D_COMP *pbi = (VP8D_COMP *)(((DECODETHREAD_DATA *)p_data)->ptr1);
    MB_ROW_DEC *mbrd = (MB_ROW_DEC *)(((DECODETHREAD_DATA *)p_data)->ptr2);
    ENTROPY_CONTEXT mb_row_left_context[4][4];

    while (1)
    {
        if (pbi->b_multithreaded_rd == 0)
            break;

172
173
        //if(WaitForSingleObject(pbi->h_event_start_decoding[ithread], INFINITE) == WAIT_OBJECT_0)
        if (sem_wait(&pbi->h_event_start_decoding[ithread]) == 0)
John Koleszar's avatar
John Koleszar committed
174
175
176
177
178
179
180
181
        {
            if (pbi->b_multithreaded_rd == 0)
                break;
            else
            {
                VP8_COMMON *pc = &pbi->common;
                MACROBLOCKD *xd = &mbrd->mbd;

182
183
                int mb_row;
                int num_part = 1 << pbi->common.multi_token_partition;
John Koleszar's avatar
John Koleszar committed
184
185
                volatile int *last_row_current_mb_col;

186
                for (mb_row = ithread+1; mb_row < pc->mb_rows; mb_row += (pbi->decoding_thread_count + 1))
John Koleszar's avatar
John Koleszar committed
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
                    int i;
                    int recon_yoffset, recon_uvoffset;
                    int mb_col;
                    int ref_fb_idx = pc->lst_fb_idx;
                    int dst_fb_idx = pc->new_fb_idx;
                    int recon_y_stride = pc->yv12_fb[ref_fb_idx].y_stride;
                    int recon_uv_stride = pc->yv12_fb[ref_fb_idx].uv_stride;

                    pbi->mb_row_di[ithread].mb_row = mb_row;
                    pbi->mb_row_di[ithread].mbd.current_bc =  &pbi->mbc[mb_row%num_part];

                    last_row_current_mb_col = &pbi->current_mb_col[mb_row -1];

                    recon_yoffset = mb_row * recon_y_stride * 16;
                    recon_uvoffset = mb_row * recon_uv_stride * 8;
                    // reset above block coeffs

                    xd->above_context[Y1CONTEXT] = pc->above_context[Y1CONTEXT];
                    xd->above_context[UCONTEXT ] = pc->above_context[UCONTEXT];
                    xd->above_context[VCONTEXT ] = pc->above_context[VCONTEXT];
                    xd->above_context[Y2CONTEXT] = pc->above_context[Y2CONTEXT];
                    xd->left_context = mb_row_left_context;
                    vpx_memset(mb_row_left_context, 0, sizeof(mb_row_left_context));
                    xd->up_available = (mb_row != 0);

                    xd->mb_to_top_edge = -((mb_row * 16)) << 3;
                    xd->mb_to_bottom_edge = ((pc->mb_rows - 1 - mb_row) * 16) << 3;

                    for (mb_col = 0; mb_col < pc->mb_cols; mb_col++)
John Koleszar's avatar
John Koleszar committed
217
                    {
218
219
220
221
222
223
224
225
                        if ((mb_col & 7) == 0)
                        {
                            while (mb_col > (*last_row_current_mb_col - 8) && *last_row_current_mb_col != pc->mb_cols - 1)
                            {
                                x86_pause_hint();
                                thread_sleep(0);
                            }
                        }
John Koleszar's avatar
John Koleszar committed
226

227
228
229
230
                        // Take a copy of the mode and Mv information for this macroblock into the xd->mbmi
                        // the partition_bmi array is unused in the decoder, so don't copy it.
                        vpx_memcpy(&xd->mbmi, &xd->mode_info_context->mbmi,
                                   sizeof(MB_MODE_INFO) - sizeof(xd->mbmi.partition_bmi));
John Koleszar's avatar
John Koleszar committed
231

232
                        if (xd->mbmi.mode == SPLITMV || xd->mbmi.mode == B_PRED)
John Koleszar's avatar
John Koleszar committed
233
                        {
234
235
236
237
238
                            for (i = 0; i < 16; i++)
                            {
                                BLOCKD *d = &xd->block[i];
                                vpx_memcpy(&d->bmi, &xd->mode_info_context->bmi[i], sizeof(B_MODE_INFO));
                            }
John Koleszar's avatar
John Koleszar committed
239
240
                        }

241
242
243
244
                        // Distance of Mb to the various image edges.
                        // These specified to 8th pel as they are always compared to values that are in 1/8th pel units
                        xd->mb_to_left_edge = -((mb_col * 16) << 3);
                        xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3;
John Koleszar's avatar
John Koleszar committed
245

246
247
248
                        xd->dst.y_buffer = pc->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
                        xd->dst.u_buffer = pc->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
                        xd->dst.v_buffer = pc->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
John Koleszar's avatar
John Koleszar committed
249

250
                        xd->left_available = (mb_col != 0);
251

252
253
254
255
256
257
258
                        // Select the appropriate reference frame for this MB
                        if (xd->mbmi.ref_frame == LAST_FRAME)
                            ref_fb_idx = pc->lst_fb_idx;
                        else if (xd->mbmi.ref_frame == GOLDEN_FRAME)
                            ref_fb_idx = pc->gld_fb_idx;
                        else
                            ref_fb_idx = pc->alt_fb_idx;
John Koleszar's avatar
John Koleszar committed
259

260
261
262
                        xd->pre.y_buffer = pc->yv12_fb[ref_fb_idx].y_buffer + recon_yoffset;
                        xd->pre.u_buffer = pc->yv12_fb[ref_fb_idx].u_buffer + recon_uvoffset;
                        xd->pre.v_buffer = pc->yv12_fb[ref_fb_idx].v_buffer + recon_uvoffset;
John Koleszar's avatar
John Koleszar committed
263

264
                        vp8_build_uvmvs(xd, pc->full_pixel);
John Koleszar's avatar
John Koleszar committed
265

266
                        vp8_decode_macroblock(pbi, xd);
John Koleszar's avatar
John Koleszar committed
267

268
269
                        recon_yoffset += 16;
                        recon_uvoffset += 8;
John Koleszar's avatar
John Koleszar committed
270

271
                        ++xd->mode_info_context;  /* next mb */
John Koleszar's avatar
John Koleszar committed
272

273
                        xd->gf_active_ptr++;      // GF useage flag for next MB
John Koleszar's avatar
John Koleszar committed
274

275
276
277
278
                        xd->above_context[Y1CONTEXT] += 4;
                        xd->above_context[UCONTEXT ] += 2;
                        xd->above_context[VCONTEXT ] += 2;
                        xd->above_context[Y2CONTEXT] ++;
John Koleszar's avatar
John Koleszar committed
279

280
281
282
                        //pbi->mb_row_di[ithread].current_mb_col = mb_col;
                        pbi->current_mb_col[mb_row] = mb_col;
                    }
John Koleszar's avatar
John Koleszar committed
283

284
285
                    // adjust to the next row of mbs
                    vp8_extend_mb_row(
286
                    &pc->yv12_fb[dst_fb_idx],
John Koleszar's avatar
John Koleszar committed
287
                    xd->dst.y_buffer + 16, xd->dst.u_buffer + 8, xd->dst.v_buffer + 8
288
                    );
John Koleszar's avatar
John Koleszar committed
289

290
                    ++xd->mode_info_context;      /* skip prediction column */
John Koleszar's avatar
John Koleszar committed
291

292
293
                    // since we have multithread
                    xd->mode_info_context += xd->mode_info_stride * pbi->decoding_thread_count;
John Koleszar's avatar
John Koleszar committed
294
295
296
297
298
299

                    pbi->last_mb_row_decoded = mb_row;

                }
            }
        }
300
301
302
303
304
305
306
307
308
        //  add this to each frame
        if ((mbrd->mb_row == pbi->common.mb_rows-1) || ((mbrd->mb_row == pbi->common.mb_rows-2) && (pbi->common.mb_rows % (pbi->decoding_thread_count+1))==1))
        {
            //SetEvent(pbi->h_event_end_decoding);
            sem_post(&pbi->h_event_end_decoding);
        }

        if ((pbi->b_multithreaded_lf) &&(pbi->common.filter_level))
            vp8_thread_loop_filter(pbi, mbrd, ithread);
John Koleszar's avatar
John Koleszar committed
309

310
    }
John Koleszar's avatar
John Koleszar committed
311
312
313
314
315
316
317
#else
    (void) p_data;
#endif

    return 0 ;
}

318
319

void vp8_thread_loop_filter(VP8D_COMP *pbi, MB_ROW_DEC *mbrd, int ithread)
John Koleszar's avatar
John Koleszar committed
320
321
322
{
#if CONFIG_MULTITHREAD

323
        if (sem_wait(&pbi->h_event_start_lpf[ithread]) == 0)
John Koleszar's avatar
John Koleszar committed
324
        {
325
326
327
           // if (pbi->b_multithreaded_lf == 0) // we're shutting down      ????
           //     break;
           // else
John Koleszar's avatar
John Koleszar committed
328
329
            {
                VP8_COMMON *cm  = &pbi->common;
330
                MACROBLOCKD *mbd = &mbrd->mbd;
John Koleszar's avatar
John Koleszar committed
331
332
                int default_filt_lvl = pbi->common.filter_level;

333
                YV12_BUFFER_CONFIG *post = cm->frame_to_show;
John Koleszar's avatar
John Koleszar committed
334
                loop_filter_info *lfi = cm->lf_info;
335
                //int frame_type = cm->frame_type;
John Koleszar's avatar
John Koleszar committed
336
337
338
339
340
341
342
343
344
345

                int mb_row;
                int mb_col;

                int filter_level;
                int alt_flt_enabled = mbd->segmentation_enabled;

                int i;
                unsigned char *y_ptr, *u_ptr, *v_ptr;

346
                volatile int *last_row_current_mb_col;
John Koleszar's avatar
John Koleszar committed
347
348

                // Set up the buffer pointers
349
350
351
                y_ptr = post->y_buffer + post->y_stride  * 16 * (ithread +1);
                u_ptr = post->u_buffer + post->uv_stride *  8 * (ithread +1);
                v_ptr = post->v_buffer + post->uv_stride *  8 * (ithread +1);
John Koleszar's avatar
John Koleszar committed
352
353

                // vp8_filter each macro block
354
                for (mb_row = ithread+1; mb_row < cm->mb_rows; mb_row+= (pbi->decoding_thread_count + 1))
John Koleszar's avatar
John Koleszar committed
355
                {
356
                    last_row_current_mb_col = &pbi->current_mb_col[mb_row -1];
John Koleszar's avatar
John Koleszar committed
357
358
359
360
361

                    for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
                    {
                        int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;

362
363
364
365
366
367
368
369
370
371
                        if ((mb_col & 7) == 0)
                        {
                            while (mb_col > (*last_row_current_mb_col-8) && *last_row_current_mb_col != cm->mb_cols - 1)
                            {
                                x86_pause_hint();
                                thread_sleep(0);
                            }
                        }

                        filter_level = pbi->mt_baseline_filter_level[Segment];
John Koleszar's avatar
John Koleszar committed
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396

                        // Apply any context driven MB level adjustment
                        vp8_adjust_mb_lf_value(mbd, &filter_level);

                        if (filter_level)
                        {
                            if (mb_col > 0)
                                cm->lf_mbv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);

                            if (mbd->mode_info_context->mbmi.dc_diff > 0)
                                cm->lf_bv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);

                            // don't apply across umv border
                            if (mb_row > 0)
                                cm->lf_mbh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);

                            if (mbd->mode_info_context->mbmi.dc_diff > 0)
                                cm->lf_bh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
                        }

                        y_ptr += 16;
                        u_ptr += 8;
                        v_ptr += 8;

                        mbd->mode_info_context++;     // step to next MB
397
                        pbi->current_mb_col[mb_row] = mb_col;
John Koleszar's avatar
John Koleszar committed
398
399
400
401
                    }

                    mbd->mode_info_context++;         // Skip border mb

402
403
404
                    y_ptr += post->y_stride  * 16 * (pbi->decoding_thread_count + 1) - post->y_width;
                    u_ptr += post->uv_stride *  8 * (pbi->decoding_thread_count + 1) - post->uv_width;
                    v_ptr += post->uv_stride *  8 * (pbi->decoding_thread_count + 1) - post->uv_width;
John Koleszar's avatar
John Koleszar committed
405

406
407
408
                    mbd->mode_info_context += pbi->decoding_thread_count * mbd->mode_info_stride;         // Skip border mb
                }
            }
John Koleszar's avatar
John Koleszar committed
409
410
        }

411
412
413
414
415
        //  add this to each frame
        if ((mbrd->mb_row == pbi->common.mb_rows-1) || ((mbrd->mb_row == pbi->common.mb_rows-2) && (pbi->common.mb_rows % (pbi->decoding_thread_count+1))==1))
        {
          sem_post(&pbi->h_event_end_lpf);
        }
John Koleszar's avatar
John Koleszar committed
416
#else
417
    (void) pbi;
John Koleszar's avatar
John Koleszar committed
418
419
420
421
422
423
424
425
426
427
428
429
#endif
}

void vp8_decoder_create_threads(VP8D_COMP *pbi)
{
#if CONFIG_MULTITHREAD
    int core_count = 0;
    int ithread;

    pbi->b_multithreaded_rd = 0;
    pbi->b_multithreaded_lf = 0;
    pbi->allocated_decoding_thread_count = 0;
430
    core_count = (pbi->max_threads > 16) ? 16 : pbi->max_threads;
John Koleszar's avatar
John Koleszar committed
431
432
433
434

    if (core_count > 1)
    {
        pbi->b_multithreaded_rd = 1;
435
436
        pbi->b_multithreaded_lf = 1;  // this can be merged with pbi->b_multithreaded_rd ?
        pbi->decoding_thread_count = core_count -1;
John Koleszar's avatar
John Koleszar committed
437
438

        CHECK_MEM_ERROR(pbi->h_decoding_thread, vpx_malloc(sizeof(pthread_t) * pbi->decoding_thread_count));
439
        CHECK_MEM_ERROR(pbi->h_event_start_decoding, vpx_malloc(sizeof(sem_t) * pbi->decoding_thread_count));
John Koleszar's avatar
John Koleszar committed
440
441
442
443
        CHECK_MEM_ERROR(pbi->mb_row_di, vpx_memalign(32, sizeof(MB_ROW_DEC) * pbi->decoding_thread_count));
        vpx_memset(pbi->mb_row_di, 0, sizeof(MB_ROW_DEC) * pbi->decoding_thread_count);
        CHECK_MEM_ERROR(pbi->de_thread_data, vpx_malloc(sizeof(DECODETHREAD_DATA) * pbi->decoding_thread_count));

444
445
446
        CHECK_MEM_ERROR(pbi->current_mb_col, vpx_malloc(sizeof(int) * MAX_ROWS));  // pc->mb_rows));
        CHECK_MEM_ERROR(pbi->h_event_start_lpf, vpx_malloc(sizeof(sem_t) * pbi->decoding_thread_count));

John Koleszar's avatar
John Koleszar committed
447
448
        for (ithread = 0; ithread < pbi->decoding_thread_count; ithread++)
        {
449
450
            sem_init(&pbi->h_event_start_decoding[ithread], 0, 0);
            sem_init(&pbi->h_event_start_lpf[ithread], 0, 0);
John Koleszar's avatar
John Koleszar committed
451
452
453
454
455
456
457
458

            pbi->de_thread_data[ithread].ithread  = ithread;
            pbi->de_thread_data[ithread].ptr1     = (void *)pbi;
            pbi->de_thread_data[ithread].ptr2     = (void *) &pbi->mb_row_di[ithread];

            pthread_create(&pbi->h_decoding_thread[ithread], 0, vp8_thread_decoding_proc, (&pbi->de_thread_data[ithread]));
        }

459
460
461
        sem_init(&pbi->h_event_end_decoding, 0, 0);
        sem_init(&pbi->h_event_end_lpf, 0, 0);

John Koleszar's avatar
John Koleszar committed
462
463
464
465
466
467
468
469
470
471
472
473
474
475
        pbi->allocated_decoding_thread_count = pbi->decoding_thread_count;
    }

#else
    (void) pbi;
#endif
}

void vp8_decoder_remove_threads(VP8D_COMP *pbi)
{
#if CONFIG_MULTITHREAD

    if (pbi->b_multithreaded_lf)
    {
476
        int i;
John Koleszar's avatar
John Koleszar committed
477
        pbi->b_multithreaded_lf = 0;
478
479
480
481
482

        for (i = 0; i < pbi->allocated_decoding_thread_count; i++)
            sem_destroy(&pbi->h_event_start_lpf[i]);

        sem_destroy(&pbi->h_event_end_lpf);
John Koleszar's avatar
John Koleszar committed
483
484
485
486
487
    }

    //shutdown MB Decoding thread;
    if (pbi->b_multithreaded_rd)
    {
488
489
        int i;

John Koleszar's avatar
John Koleszar committed
490
        pbi->b_multithreaded_rd = 0;
491

John Koleszar's avatar
John Koleszar committed
492
        // allow all threads to exit
493
        for (i = 0; i < pbi->allocated_decoding_thread_count; i++)
John Koleszar's avatar
John Koleszar committed
494
        {
495
496
            sem_post(&pbi->h_event_start_decoding[i]);
            pthread_join(pbi->h_decoding_thread[i], NULL);
John Koleszar's avatar
John Koleszar committed
497
498
        }

499
500
501
        for (i = 0; i < pbi->allocated_decoding_thread_count; i++)
        {
            sem_destroy(&pbi->h_event_start_decoding[i]);
John Koleszar's avatar
John Koleszar committed
502
503
        }

504
        sem_destroy(&pbi->h_event_end_decoding);
John Koleszar's avatar
John Koleszar committed
505
506
507
508
509
510
511

        if (pbi->h_decoding_thread)
        {
            vpx_free(pbi->h_decoding_thread);
            pbi->h_decoding_thread = NULL;
        }

512
        if (pbi->h_event_start_decoding)
John Koleszar's avatar
John Koleszar committed
513
        {
514
515
516
517
518
519
520
521
            vpx_free(pbi->h_event_start_decoding);
            pbi->h_event_start_decoding = NULL;
        }

        if (pbi->h_event_start_lpf)
        {
            vpx_free(pbi->h_event_start_lpf);
            pbi->h_event_start_lpf = NULL;
John Koleszar's avatar
John Koleszar committed
522
523
524
525
526
527
528
529
530
531
532
533
534
535
        }

        if (pbi->mb_row_di)
        {
            vpx_free(pbi->mb_row_di);
            pbi->mb_row_di = NULL ;
        }

        if (pbi->de_thread_data)
        {
            vpx_free(pbi->de_thread_data);
            pbi->de_thread_data = NULL;
        }

536
537
538
539
540
541
        if (pbi->current_mb_col)
        {
            vpx_free(pbi->current_mb_col);
            pbi->current_mb_col = NULL ;
        }
    }
John Koleszar's avatar
John Koleszar committed
542
543
544
545
546
547
548
549
550
#else
    (void) pbi;
#endif
}


void vp8_start_lfthread(VP8D_COMP *pbi)
{
#if CONFIG_MULTITHREAD
551
  /*
John Koleszar's avatar
John Koleszar committed
552
553
554
    memcpy(&pbi->lpfmb, &pbi->mb, sizeof(pbi->mb));
    pbi->last_mb_row_decoded = 0;
    sem_post(&pbi->h_event_start_lpf);
555
556
    */
    (void) pbi;
John Koleszar's avatar
John Koleszar committed
557
558
559
560
561
562
563
564
#else
    (void) pbi;
#endif
}

void vp8_stop_lfthread(VP8D_COMP *pbi)
{
#if CONFIG_MULTITHREAD
565
  /*
John Koleszar's avatar
John Koleszar committed
566
567
568
569
    struct vpx_usec_timer timer;

    vpx_usec_timer_start(&timer);

570
    sem_wait(&pbi->h_event_end_lpf);
John Koleszar's avatar
John Koleszar committed
571
572
573

    vpx_usec_timer_mark(&timer);
    pbi->time_loop_filtering += vpx_usec_timer_elapsed(&timer);
574
575
    */
    (void) pbi;
John Koleszar's avatar
John Koleszar committed
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
#else
    (void) pbi;
#endif
}


void vp8_mtdecode_mb_rows(VP8D_COMP *pbi,
                          MACROBLOCKD *xd)
{
#if CONFIG_MULTITHREAD
    int mb_row;
    VP8_COMMON *pc = &pbi->common;

    int ibc = 0;
    int num_part = 1 << pbi->common.multi_token_partition;
591
592
    int i;
    volatile int *last_row_current_mb_col = NULL;
John Koleszar's avatar
John Koleszar committed
593
594
595

    vp8_setup_decoding_thread_data(pbi, xd, pbi->mb_row_di, pbi->decoding_thread_count);

596
597
598
    for (i = 0; i < pbi->decoding_thread_count; i++)
        sem_post(&pbi->h_event_start_decoding[i]);

John Koleszar's avatar
John Koleszar committed
599
600
601
602
    for (mb_row = 0; mb_row < pc->mb_rows; mb_row += (pbi->decoding_thread_count + 1))
    {
        int i;

603
        xd->current_bc = &pbi->mbc[mb_row%num_part];
John Koleszar's avatar
John Koleszar committed
604

605
        //vp8_decode_mb_row(pbi, pc, mb_row, xd);
John Koleszar's avatar
John Koleszar committed
606
        {
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
            int i;
            int recon_yoffset, recon_uvoffset;
            int mb_col;
            int ref_fb_idx = pc->lst_fb_idx;
            int dst_fb_idx = pc->new_fb_idx;
            int recon_y_stride = pc->yv12_fb[ref_fb_idx].y_stride;
            int recon_uv_stride = pc->yv12_fb[ref_fb_idx].uv_stride;

           // volatile int *last_row_current_mb_col = NULL;
            if (mb_row > 0)
                last_row_current_mb_col = &pbi->current_mb_col[mb_row -1];

            vpx_memset(pc->left_context, 0, sizeof(pc->left_context));
            recon_yoffset = mb_row * recon_y_stride * 16;
            recon_uvoffset = mb_row * recon_uv_stride * 8;
            // reset above block coeffs

            xd->above_context[Y1CONTEXT] = pc->above_context[Y1CONTEXT];
            xd->above_context[UCONTEXT ] = pc->above_context[UCONTEXT];
            xd->above_context[VCONTEXT ] = pc->above_context[VCONTEXT];
            xd->above_context[Y2CONTEXT] = pc->above_context[Y2CONTEXT];
            xd->up_available = (mb_row != 0);

            xd->mb_to_top_edge = -((mb_row * 16)) << 3;
            xd->mb_to_bottom_edge = ((pc->mb_rows - 1 - mb_row) * 16) << 3;

            for (mb_col = 0; mb_col < pc->mb_cols; mb_col++)
            {
John Koleszar's avatar
John Koleszar committed
635

636
637
638
639
640
641
642
                if ( mb_row > 0 && (mb_col & 7) == 0){
               //if ( mb_row > 0 ){
                  while (mb_col > (*last_row_current_mb_col - 8) && *last_row_current_mb_col != pc->mb_cols - 1)
                  {
                      x86_pause_hint();
                      thread_sleep(0);
                  }
John Koleszar's avatar
John Koleszar committed
643

644
                }
John Koleszar's avatar
John Koleszar committed
645

646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
                // Take a copy of the mode and Mv information for this macroblock into the xd->mbmi
                // the partition_bmi array is unused in the decoder, so don't copy it.
                vpx_memcpy(&xd->mbmi, &xd->mode_info_context->mbmi,
                           sizeof(MB_MODE_INFO) - sizeof(xd->mbmi.partition_bmi));

                if (xd->mbmi.mode == SPLITMV || xd->mbmi.mode == B_PRED)
                {
                    for (i = 0; i < 16; i++)
                    {
                        BLOCKD *d = &xd->block[i];
                        vpx_memcpy(&d->bmi, &xd->mode_info_context->bmi[i], sizeof(B_MODE_INFO));
                    }
                }

                // Distance of Mb to the various image edges.
                // These specified to 8th pel as they are always compared to values that are in 1/8th pel units
                xd->mb_to_left_edge = -((mb_col * 16) << 3);
                xd->mb_to_right_edge = ((pc->mb_cols - 1 - mb_col) * 16) << 3;

                xd->dst.y_buffer = pc->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
                xd->dst.u_buffer = pc->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
                xd->dst.v_buffer = pc->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;

                xd->left_available = (mb_col != 0);

                // Select the appropriate reference frame for this MB
                if (xd->mbmi.ref_frame == LAST_FRAME)
                    ref_fb_idx = pc->lst_fb_idx;
                else if (xd->mbmi.ref_frame == GOLDEN_FRAME)
                    ref_fb_idx = pc->gld_fb_idx;
                else
                    ref_fb_idx = pc->alt_fb_idx;

                xd->pre.y_buffer = pc->yv12_fb[ref_fb_idx].y_buffer + recon_yoffset;
                xd->pre.u_buffer = pc->yv12_fb[ref_fb_idx].u_buffer + recon_uvoffset;
                xd->pre.v_buffer = pc->yv12_fb[ref_fb_idx].v_buffer + recon_uvoffset;

                vp8_build_uvmvs(xd, pc->full_pixel);

                vp8_decode_macroblock(pbi, xd);

                recon_yoffset += 16;
                recon_uvoffset += 8;

                ++xd->mode_info_context;  /* next mb */

                xd->gf_active_ptr++;      // GF useage flag for next MB

                xd->above_context[Y1CONTEXT] += 4;
                xd->above_context[UCONTEXT ] += 2;
                xd->above_context[VCONTEXT ] += 2;
                xd->above_context[Y2CONTEXT] ++;

                //pbi->current_mb_col_main = mb_col;
                pbi->current_mb_col[mb_row] = mb_col;
            }

            // adjust to the next row of mbs
            vp8_extend_mb_row(
                &pc->yv12_fb[dst_fb_idx],
                xd->dst.y_buffer + 16, xd->dst.u_buffer + 8, xd->dst.v_buffer + 8
            );
John Koleszar's avatar
John Koleszar committed
708

709
            ++xd->mode_info_context;      /* skip prediction column */
John Koleszar's avatar
John Koleszar committed
710

711
712
            pbi->last_mb_row_decoded = mb_row;
        }
John Koleszar's avatar
John Koleszar committed
713
        xd->mode_info_context += xd->mode_info_stride * pbi->decoding_thread_count;
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
    }

    sem_wait(&pbi->h_event_end_decoding);   // add back for each frame
#else
    (void) pbi;
    (void) xd;
#endif
}


void vp8_mt_loop_filter_frame( VP8D_COMP *pbi)
{
#if CONFIG_MULTITHREAD
    VP8_COMMON *cm  = &pbi->common;
    MACROBLOCKD *mbd = &pbi->mb;
    int default_filt_lvl = pbi->common.filter_level;

    YV12_BUFFER_CONFIG *post = cm->frame_to_show;
    loop_filter_info *lfi = cm->lf_info;
    int frame_type = cm->frame_type;

    int mb_row;
    int mb_col;

    int filter_level;
    int alt_flt_enabled = mbd->segmentation_enabled;

    int i;
    unsigned char *y_ptr, *u_ptr, *v_ptr;

    volatile int *last_row_current_mb_col=NULL;

    vp8_setup_loop_filter_thread_data(pbi, mbd, pbi->mb_row_di, pbi->decoding_thread_count);

    mbd->mode_info_context = cm->mi;          // Point at base of Mb MODE_INFO list

    // Note the baseline filter values for each segment
    if (alt_flt_enabled)
    {
        for (i = 0; i < MAX_MB_SEGMENTS; i++)
        {
            // Abs value
            if (mbd->mb_segement_abs_delta == SEGMENT_ABSDATA)
                pbi->mt_baseline_filter_level[i] = mbd->segment_feature_data[MB_LVL_ALT_LF][i];
            // Delta Value
            else
            {
                pbi->mt_baseline_filter_level[i] = default_filt_lvl + mbd->segment_feature_data[MB_LVL_ALT_LF][i];
                pbi->mt_baseline_filter_level[i] = (pbi->mt_baseline_filter_level[i] >= 0) ? ((pbi->mt_baseline_filter_level[i] <= MAX_LOOP_FILTER) ? pbi->mt_baseline_filter_level[i] : MAX_LOOP_FILTER) : 0;  // Clamp to valid range
            }
        }
    }
    else
    {
        for (i = 0; i < MAX_MB_SEGMENTS; i++)
            pbi->mt_baseline_filter_level[i] = default_filt_lvl;
    }

    // Initialize the loop filter for this frame.
    if ((cm->last_filter_type != cm->filter_type) || (cm->last_sharpness_level != cm->sharpness_level))
        vp8_init_loop_filter(cm);
    else if (frame_type != cm->last_frame_type)
        vp8_frame_init_loop_filter(lfi, frame_type);

    for (i = 0; i < pbi->decoding_thread_count; i++)
        sem_post(&pbi->h_event_start_lpf[i]);
        // sem_post(&pbi->h_event_start_lpf);

    // Set up the buffer pointers
    y_ptr = post->y_buffer;
    u_ptr = post->u_buffer;
    v_ptr = post->v_buffer;
John Koleszar's avatar
John Koleszar committed
786

787
788
789
790
791
792
793
    // vp8_filter each macro block
    for (mb_row = 0; mb_row < cm->mb_rows; mb_row+= (pbi->decoding_thread_count + 1))
    {
        if (mb_row > 0)
            last_row_current_mb_col = &pbi->current_mb_col[mb_row -1];

        for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
John Koleszar's avatar
John Koleszar committed
794
        {
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
            int Segment = (alt_flt_enabled) ? mbd->mode_info_context->mbmi.segment_id : 0;


            if ( mb_row > 0 && (mb_col & 7) == 0){
            // if ( mb_row > 0 ){
                while (mb_col > (*last_row_current_mb_col-8) && *last_row_current_mb_col != cm->mb_cols - 1)
                {
                    x86_pause_hint();
                    thread_sleep(0);
                }
            }

            filter_level = pbi->mt_baseline_filter_level[Segment];

            // Distance of Mb to the various image edges.
            // These specified to 8th pel as they are always compared to values that are in 1/8th pel units
            // Apply any context driven MB level adjustment
            vp8_adjust_mb_lf_value(mbd, &filter_level);

            if (filter_level)
            {
                if (mb_col > 0)
                    cm->lf_mbv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);

                if (mbd->mode_info_context->mbmi.dc_diff > 0)
                    cm->lf_bv(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);

                // don't apply across umv border
                if (mb_row > 0)
                    cm->lf_mbh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);

                if (mbd->mode_info_context->mbmi.dc_diff > 0)
                    cm->lf_bh(y_ptr, u_ptr, v_ptr, post->y_stride, post->uv_stride, &lfi[filter_level], cm->simpler_lpf);
            }

            y_ptr += 16;
            u_ptr += 8;
            v_ptr += 8;

            mbd->mode_info_context++;     // step to next MB
            pbi->current_mb_col[mb_row] = mb_col;
John Koleszar's avatar
John Koleszar committed
836
        }
837
838
839
840
841
842
843
        mbd->mode_info_context++;         // Skip border mb

        //update for multi-thread
        y_ptr += post->y_stride  * 16 * (pbi->decoding_thread_count + 1) - post->y_width;
        u_ptr += post->uv_stride *  8 * (pbi->decoding_thread_count + 1) - post->uv_width;
        v_ptr += post->uv_stride *  8 * (pbi->decoding_thread_count + 1) - post->uv_width;
        mbd->mode_info_context += pbi->decoding_thread_count * mbd->mode_info_stride;
John Koleszar's avatar
John Koleszar committed
844
845
    }

846
    sem_wait(&pbi->h_event_end_lpf);
John Koleszar's avatar
John Koleszar committed
847
848
849
850
#else
    (void) pbi;
#endif
}