/*
* Copyright (c) 2010 The WebM project authors. All Rights Reserved.
*
* Use of this source code is governed by a BSD-style license
* that can be found in the LICENSE file in the root of the source
* tree. An additional intellectual property rights grant can be found
* in the file PATENTS. All contributing project authors may
* be found in the AUTHORS file in the root of the source tree.
*/
#include "onyx_int.h"
#include "vp8/common/threading.h"
#include "vp8/common/common.h"
#include "vp8/common/extend.h"
#include "bitstream.h"
#include "encodeframe.h"
#if CONFIG_MULTITHREAD
extern void vp8cx_mb_init_quantizer(VP8_COMP *cpi, MACROBLOCK *x, int ok_to_skip);
extern void vp8_loopfilter_frame(VP8_COMP *cpi, VP8_COMMON *cm);
static THREAD_FUNCTION thread_loopfilter(void *p_data)
{
VP8_COMP *cpi = (VP8_COMP *)(((LPFTHREAD_DATA *)p_data)->ptr1);
VP8_COMMON *cm = &cpi->common;
while (1)
{
if (cpi->b_multi_threaded == 0)
break;
if (sem_wait(&cpi->h_event_start_lpf) == 0)
{
if (cpi->b_multi_threaded == 0) /* we're shutting down */
break;
vp8_loopfilter_frame(cpi, cm);
sem_post(&cpi->h_event_end_lpf);
}
}
return 0;
}
static
THREAD_FUNCTION thread_encoding_proc(void *p_data)
{
int ithread = ((ENCODETHREAD_DATA *)p_data)->ithread;
VP8_COMP *cpi = (VP8_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr1);
MB_ROW_COMP *mbri = (MB_ROW_COMP *)(((ENCODETHREAD_DATA *)p_data)->ptr2);
ENTROPY_CONTEXT_PLANES mb_row_left_context;
while (1)
{
if (cpi->b_multi_threaded == 0)
break;
if (sem_wait(&cpi->h_event_start_encoding[ithread]) == 0)
{
const int nsync = cpi->mt_sync_range;
VP8_COMMON *cm = &cpi->common;
int mb_row;
MACROBLOCK *x = &mbri->mb;
MACROBLOCKD *xd = &x->e_mbd;
TOKENEXTRA *tp ;
#if CONFIG_REALTIME_ONLY & CONFIG_ONTHEFLY_BITPACKING
TOKENEXTRA *tp_start = cpi->tok + (1 + ithread) * (16 * 24);
const int num_part = (1 << cm->multi_token_partition);
#endif
int *segment_counts = mbri->segment_counts;
int *totalrate = &mbri->totalrate;
if (cpi->b_multi_threaded == 0) /* we're shutting down */
break;
for (mb_row = ithread + 1; mb_row < cm->mb_rows; mb_row += (cpi->encoding_thread_count + 1))
{
int recon_yoffset, recon_uvoffset;
int mb_col;
int ref_fb_idx = cm->lst_fb_idx;
int dst_fb_idx = cm->new_fb_idx;
int recon_y_stride = cm->yv12_fb[ref_fb_idx].y_stride;
int recon_uv_stride = cm->yv12_fb[ref_fb_idx].uv_stride;
int map_index = (mb_row * cm->mb_cols);
volatile const int *last_row_current_mb_col;
volatile int *current_mb_col = &cpi->mt_current_mb_col[mb_row];
#if (CONFIG_REALTIME_ONLY & CONFIG_ONTHEFLY_BITPACKING)
vp8_writer *w = &cpi->bc[1 + (mb_row % num_part)];
#else
tp = cpi->tok + (mb_row * (cm->mb_cols * 16 * 24));
cpi->tplist[mb_row].start = tp;
#endif
last_row_current_mb_col = &cpi->mt_current_mb_col[mb_row - 1];
/* reset above block coeffs */
xd->above_context = cm->above_context;
xd->left_context = &mb_row_left_context;
vp8_zero(mb_row_left_context);
xd->up_available = (mb_row != 0);
recon_yoffset = (mb_row * recon_y_stride * 16);
recon_uvoffset = (mb_row * recon_uv_stride * 8);
/* Set the mb activity pointer to the start of the row. */
x->mb_activity_ptr = &cpi->mb_activity_map[map_index];
/* for each macroblock col in image */
for (mb_col = 0; mb_col < cm->mb_cols; mb_col++)
{
*current_mb_col = mb_col - 1;
if ((mb_col & (nsync - 1)) == 0)
{
while (mb_col > (*last_row_current_mb_col - nsync))
{
x86_pause_hint();
thread_sleep(0);
}
}
#if CONFIG_REALTIME_ONLY & CONFIG_ONTHEFLY_BITPACKING
tp = tp_start;
#endif
/* 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 = ((cm->mb_cols - 1 - mb_col) * 16) << 3;
xd->mb_to_top_edge = -((mb_row * 16) << 3);
xd->mb_to_bottom_edge = ((cm->mb_rows - 1 - mb_row) * 16) << 3;
/* Set up limit values for motion vectors used to prevent
* them extending outside the UMV borders
*/
x->mv_col_min = -((mb_col * 16) + (VP8BORDERINPIXELS - 16));
x->mv_col_max = ((cm->mb_cols - 1 - mb_col) * 16) + (VP8BORDERINPIXELS - 16);
x->mv_row_min = -((mb_row * 16) + (VP8BORDERINPIXELS - 16));
x->mv_row_max = ((cm->mb_rows - 1 - mb_row) * 16) + (VP8BORDERINPIXELS - 16);
xd->dst.y_buffer = cm->yv12_fb[dst_fb_idx].y_buffer + recon_yoffset;
xd->dst.u_buffer = cm->yv12_fb[dst_fb_idx].u_buffer + recon_uvoffset;
xd->dst.v_buffer = cm->yv12_fb[dst_fb_idx].v_buffer + recon_uvoffset;
xd->left_available = (mb_col != 0);
x->rddiv = cpi->RDDIV;
x->rdmult = cpi->RDMULT;
/* Copy current mb to a buffer */
vp8_copy_mem16x16(x->src.y_buffer, x->src.y_stride, x->thismb, 16);
if (cpi->oxcf.tuning == VP8_TUNE_SSIM)
vp8_activity_masking(cpi, x);
/* Is segmentation enabled */
/* MB level adjustment to quantizer */
if (xd->segmentation_enabled)
{
/* Code to set segment id in xd->mbmi.segment_id for
* current MB (with range checking)
*/
if (cpi->segmentation_map[map_index + mb_col] <= 3)
xd->mode_info_context->mbmi.segment_id = cpi->segmentation_map[map_index + mb_col];
else
xd->mode_info_context->mbmi.segment_id = 0;
vp8cx_mb_init_quantizer(cpi, x, 1);
}
else
/* Set to Segment 0 by default */
xd->mode_info_context->mbmi.segment_id = 0;
x->active_ptr = cpi->active_map + map_index + mb_col;
if (cm->frame_type == KEY_FRAME)
{
*totalrate += vp8cx_encode_intra_macroblock(cpi, x, &tp);
#ifdef MODE_STATS
y_modes[xd->mbmi.mode] ++;
#endif
}
else
{
*totalrate += vp8cx_encode_inter_macroblock(cpi, x, &tp, recon_yoffset, recon_uvoffset, mb_row, mb_col);
#ifdef MODE_STATS
inter_y_modes[xd->mbmi.mode] ++;
if (xd->mbmi.mode == SPLITMV)
{
int b;
for (b = 0; b < xd->mbmi.partition_count; b++)
{
inter_b_modes[x->partition->bmi[b].mode] ++;
}