1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
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
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
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
|
/* SPDX-License-Identifier: Apache-2.0
* Copyright 2011-2022 Blender Foundation */
#include "integrator/work_tile_scheduler.h"
#include "device/queue.h"
#include "integrator/tile.h"
#include "session/buffers.h"
#include "util/atomic.h"
#include "util/log.h"
CCL_NAMESPACE_BEGIN
WorkTileScheduler::WorkTileScheduler()
{
}
void WorkTileScheduler::set_accelerated_rt(bool accelerated_rt)
{
accelerated_rt_ = accelerated_rt;
}
void WorkTileScheduler::set_max_num_path_states(int max_num_path_states)
{
max_num_path_states_ = max_num_path_states;
}
void WorkTileScheduler::reset(const BufferParams &buffer_params,
int sample_start,
int samples_num,
int sample_offset,
float scrambling_distance)
{
/* Image buffer parameters. */
image_full_offset_px_.x = buffer_params.full_x;
image_full_offset_px_.y = buffer_params.full_y;
image_size_px_ = make_int2(buffer_params.width, buffer_params.height);
scrambling_distance_ = scrambling_distance;
offset_ = buffer_params.offset;
stride_ = buffer_params.stride;
/* Samples parameters. */
sample_start_ = sample_start;
samples_num_ = samples_num;
sample_offset_ = sample_offset;
/* Initialize new scheduling. */
reset_scheduler_state();
}
void WorkTileScheduler::reset_scheduler_state()
{
tile_size_ = tile_calculate_best_size(
accelerated_rt_, image_size_px_, samples_num_, max_num_path_states_, scrambling_distance_);
VLOG(3) << "Will schedule tiles of size " << tile_size_;
if (VLOG_IS_ON(3)) {
/* The logging is based on multiple tiles scheduled, ignoring overhead of multi-tile scheduling
* and purely focusing on the number of used path states. */
const int num_path_states_in_tile = tile_size_.width * tile_size_.height *
tile_size_.num_samples;
const int num_tiles = max_num_path_states_ / num_path_states_in_tile;
VLOG(3) << "Number of unused path states: "
<< max_num_path_states_ - num_tiles * num_path_states_in_tile;
}
num_tiles_x_ = divide_up(image_size_px_.x, tile_size_.width);
num_tiles_y_ = divide_up(image_size_px_.y, tile_size_.height);
total_tiles_num_ = num_tiles_x_ * num_tiles_y_;
num_tiles_per_sample_range_ = divide_up(samples_num_, tile_size_.num_samples);
next_work_index_ = 0;
total_work_size_ = total_tiles_num_ * num_tiles_per_sample_range_;
}
bool WorkTileScheduler::get_work(KernelWorkTile *work_tile_, const int max_work_size)
{
/* Note that the `max_work_size` can be higher than the `max_num_path_states_`: this is because
* the path trace work can decide to use smaller tile sizes and greedily schedule multiple tiles,
* improving overall device occupancy.
* So the `max_num_path_states_` is a "scheduling unit", and the `max_work_size` is a "scheduling
* limit". */
DCHECK_NE(max_num_path_states_, 0);
const int work_index = next_work_index_++;
if (work_index >= total_work_size_) {
return false;
}
const int sample_range_index = work_index % num_tiles_per_sample_range_;
const int start_sample = sample_range_index * tile_size_.num_samples;
const int tile_index = work_index / num_tiles_per_sample_range_;
const int tile_y = tile_index / num_tiles_x_;
const int tile_x = tile_index - tile_y * num_tiles_x_;
KernelWorkTile work_tile;
work_tile.x = tile_x * tile_size_.width;
work_tile.y = tile_y * tile_size_.height;
work_tile.w = tile_size_.width;
work_tile.h = tile_size_.height;
work_tile.start_sample = sample_start_ + start_sample;
work_tile.num_samples = min(tile_size_.num_samples, samples_num_ - start_sample);
work_tile.sample_offset = sample_offset_;
work_tile.offset = offset_;
work_tile.stride = stride_;
work_tile.w = min(work_tile.w, image_size_px_.x - work_tile.x);
work_tile.h = min(work_tile.h, image_size_px_.y - work_tile.y);
work_tile.x += image_full_offset_px_.x;
work_tile.y += image_full_offset_px_.y;
const int tile_work_size = work_tile.w * work_tile.h * work_tile.num_samples;
DCHECK_GT(tile_work_size, 0);
if (max_work_size && tile_work_size > max_work_size) {
/* The work did not fit into the requested limit of the work size. Unschedule the tile,
* so it can be picked up again later. */
next_work_index_--;
return false;
}
*work_tile_ = work_tile;
return true;
}
CCL_NAMESPACE_END
|
