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
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
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
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
|
/*
* Copyright 2011-2013 Blender Foundation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <stdlib.h>
#include <string.h>
#include "bvh/bvh2.h"
#include "device/device.h"
#include "device/queue.h"
#include "device/cpu/device.h"
#include "device/cuda/device.h"
#include "device/dummy/device.h"
#include "device/hip/device.h"
#include "device/multi/device.h"
#include "device/optix/device.h"
#include "util/foreach.h"
#include "util/half.h"
#include "util/log.h"
#include "util/math.h"
#include "util/string.h"
#include "util/system.h"
#include "util/time.h"
#include "util/types.h"
#include "util/vector.h"
CCL_NAMESPACE_BEGIN
bool Device::need_types_update = true;
bool Device::need_devices_update = true;
thread_mutex Device::device_mutex;
vector<DeviceInfo> Device::cuda_devices;
vector<DeviceInfo> Device::optix_devices;
vector<DeviceInfo> Device::cpu_devices;
vector<DeviceInfo> Device::hip_devices;
uint Device::devices_initialized_mask = 0;
/* Device */
Device::~Device() noexcept(false)
{
}
void Device::build_bvh(BVH *bvh, Progress &progress, bool refit)
{
assert(bvh->params.bvh_layout == BVH_LAYOUT_BVH2);
BVH2 *const bvh2 = static_cast<BVH2 *>(bvh);
if (refit) {
bvh2->refit(progress);
}
else {
bvh2->build(progress, &stats);
}
}
Device *Device::create(const DeviceInfo &info, Stats &stats, Profiler &profiler)
{
if (!info.multi_devices.empty()) {
/* Always create a multi device when info contains multiple devices.
* This is done so that the type can still be e.g. DEVICE_CPU to indicate
* that it is a homogeneous collection of devices, which simplifies checks. */
return device_multi_create(info, stats, profiler);
}
Device *device = NULL;
switch (info.type) {
case DEVICE_CPU:
device = device_cpu_create(info, stats, profiler);
break;
#ifdef WITH_CUDA
case DEVICE_CUDA:
if (device_cuda_init())
device = device_cuda_create(info, stats, profiler);
break;
#endif
#ifdef WITH_OPTIX
case DEVICE_OPTIX:
if (device_optix_init())
device = device_optix_create(info, stats, profiler);
break;
#endif
#ifdef WITH_HIP
case DEVICE_HIP:
if (device_hip_init())
device = device_hip_create(info, stats, profiler);
break;
#endif
default:
break;
}
if (device == NULL) {
device = device_dummy_create(info, stats, profiler);
}
return device;
}
DeviceType Device::type_from_string(const char *name)
{
if (strcmp(name, "CPU") == 0)
return DEVICE_CPU;
else if (strcmp(name, "CUDA") == 0)
return DEVICE_CUDA;
else if (strcmp(name, "OPTIX") == 0)
return DEVICE_OPTIX;
else if (strcmp(name, "MULTI") == 0)
return DEVICE_MULTI;
else if (strcmp(name, "HIP") == 0)
return DEVICE_HIP;
return DEVICE_NONE;
}
string Device::string_from_type(DeviceType type)
{
if (type == DEVICE_CPU)
return "CPU";
else if (type == DEVICE_CUDA)
return "CUDA";
else if (type == DEVICE_OPTIX)
return "OPTIX";
else if (type == DEVICE_MULTI)
return "MULTI";
else if (type == DEVICE_HIP)
return "HIP";
return "";
}
vector<DeviceType> Device::available_types()
{
vector<DeviceType> types;
types.push_back(DEVICE_CPU);
#ifdef WITH_CUDA
types.push_back(DEVICE_CUDA);
#endif
#ifdef WITH_OPTIX
types.push_back(DEVICE_OPTIX);
#endif
#ifdef WITH_HIP
types.push_back(DEVICE_HIP);
#endif
return types;
}
vector<DeviceInfo> Device::available_devices(uint mask)
{
/* Lazy initialize devices. On some platforms OpenCL or CUDA drivers can
* be broken and cause crashes when only trying to get device info, so
* we don't want to do any initialization until the user chooses to. */
thread_scoped_lock lock(device_mutex);
vector<DeviceInfo> devices;
#if defined(WITH_CUDA) || defined(WITH_OPTIX)
if (mask & (DEVICE_MASK_CUDA | DEVICE_MASK_OPTIX)) {
if (!(devices_initialized_mask & DEVICE_MASK_CUDA)) {
if (device_cuda_init()) {
device_cuda_info(cuda_devices);
}
devices_initialized_mask |= DEVICE_MASK_CUDA;
}
if (mask & DEVICE_MASK_CUDA) {
foreach (DeviceInfo &info, cuda_devices) {
devices.push_back(info);
}
}
}
#endif
#ifdef WITH_OPTIX
if (mask & DEVICE_MASK_OPTIX) {
if (!(devices_initialized_mask & DEVICE_MASK_OPTIX)) {
if (device_optix_init()) {
device_optix_info(cuda_devices, optix_devices);
}
devices_initialized_mask |= DEVICE_MASK_OPTIX;
}
foreach (DeviceInfo &info, optix_devices) {
devices.push_back(info);
}
}
#endif
#ifdef WITH_HIP
if (mask & DEVICE_MASK_HIP) {
if (!(devices_initialized_mask & DEVICE_MASK_HIP)) {
if (device_hip_init()) {
device_hip_info(hip_devices);
}
devices_initialized_mask |= DEVICE_MASK_HIP;
}
foreach (DeviceInfo &info, hip_devices) {
devices.push_back(info);
}
}
#endif
if (mask & DEVICE_MASK_CPU) {
if (!(devices_initialized_mask & DEVICE_MASK_CPU)) {
device_cpu_info(cpu_devices);
devices_initialized_mask |= DEVICE_MASK_CPU;
}
foreach (DeviceInfo &info, cpu_devices) {
devices.push_back(info);
}
}
return devices;
}
DeviceInfo Device::dummy_device(const string &error_msg)
{
DeviceInfo info;
info.type = DEVICE_DUMMY;
info.error_msg = error_msg;
return info;
}
string Device::device_capabilities(uint mask)
{
thread_scoped_lock lock(device_mutex);
string capabilities = "";
if (mask & DEVICE_MASK_CPU) {
capabilities += "\nCPU device capabilities: ";
capabilities += device_cpu_capabilities() + "\n";
}
#ifdef WITH_CUDA
if (mask & DEVICE_MASK_CUDA) {
if (device_cuda_init()) {
capabilities += "\nCUDA device capabilities:\n";
capabilities += device_cuda_capabilities();
}
}
#endif
#ifdef WITH_HIP
if (mask & DEVICE_MASK_HIP) {
if (device_hip_init()) {
capabilities += "\nHIP device capabilities:\n";
capabilities += device_hip_capabilities();
}
}
#endif
return capabilities;
}
DeviceInfo Device::get_multi_device(const vector<DeviceInfo> &subdevices,
int threads,
bool background)
{
assert(subdevices.size() > 0);
if (subdevices.size() == 1) {
/* No multi device needed. */
return subdevices.front();
}
DeviceInfo info;
info.type = DEVICE_NONE;
info.id = "MULTI";
info.description = "Multi Device";
info.num = 0;
info.has_nanovdb = true;
info.has_osl = true;
info.has_profiling = true;
info.has_peer_memory = false;
info.denoisers = DENOISER_ALL;
foreach (const DeviceInfo &device, subdevices) {
/* Ensure CPU device does not slow down GPU. */
if (device.type == DEVICE_CPU && subdevices.size() > 1) {
if (background) {
int orig_cpu_threads = (threads) ? threads : system_cpu_thread_count();
int cpu_threads = max(orig_cpu_threads - (subdevices.size() - 1), 0);
VLOG(1) << "CPU render threads reduced from " << orig_cpu_threads << " to " << cpu_threads
<< ", to dedicate to GPU.";
if (cpu_threads >= 1) {
DeviceInfo cpu_device = device;
cpu_device.cpu_threads = cpu_threads;
info.multi_devices.push_back(cpu_device);
}
else {
continue;
}
}
else {
VLOG(1) << "CPU render threads disabled for interactive render.";
continue;
}
}
else {
info.multi_devices.push_back(device);
}
/* Create unique ID for this combination of devices. */
info.id += device.id;
/* Set device type to MULTI if subdevices are not of a common type. */
if (info.type == DEVICE_NONE) {
info.type = device.type;
}
else if (device.type != info.type) {
info.type = DEVICE_MULTI;
}
/* Accumulate device info. */
info.has_nanovdb &= device.has_nanovdb;
info.has_osl &= device.has_osl;
info.has_profiling &= device.has_profiling;
info.has_peer_memory |= device.has_peer_memory;
info.denoisers &= device.denoisers;
}
return info;
}
void Device::tag_update()
{
free_memory();
}
void Device::free_memory()
{
devices_initialized_mask = 0;
cuda_devices.free_memory();
optix_devices.free_memory();
hip_devices.free_memory();
cpu_devices.free_memory();
}
unique_ptr<DeviceQueue> Device::gpu_queue_create()
{
LOG(FATAL) << "Device does not support queues.";
return nullptr;
}
const CPUKernels *Device::get_cpu_kernels() const
{
LOG(FATAL) << "Device does not support CPU kernels.";
return nullptr;
}
void Device::get_cpu_kernel_thread_globals(
vector<CPUKernelThreadGlobals> & /*kernel_thread_globals*/)
{
LOG(FATAL) << "Device does not support CPU kernels.";
}
void *Device::get_cpu_osl_memory()
{
return nullptr;
}
/* DeviceInfo */
CCL_NAMESPACE_END
|