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
|
// Copyright 2024 Intel Corporation
// SPDX-License-Identifier: Apache-2.0
#include "DeviceRTImpl_sycl.h"
namespace ospray {
namespace devicert {
/////////////////////////////////////////////////////////////////////
// AsyncEvent implementation
void AsyncEventImpl::wait() const
{
event.wait();
}
bool AsyncEventImpl::finished() const
{
return (event.get_info<sycl::info::event::command_execution_status>()
== sycl::info::event_command_status::complete);
}
float AsyncEventImpl::getDuration() const
{
try {
const auto t0 =
event.get_profiling_info<sycl::info::event_profiling::command_start>();
const auto t1 =
event.get_profiling_info<sycl::info::event_profiling::command_end>();
return (t1 - t0) * 1.0e-9f;
} catch (...) {
// In case no profiling data is available
return 0.f;
}
}
void *AsyncEventImpl::getSyclEventPtr()
{
return &event;
}
/////////////////////////////////////////////////////////////////////
// Device implementation
DeviceImpl::DeviceImpl(bool debug)
: Device(debug),
device(sycl::gpu_selector_v),
context(device),
queue(context,
device,
{sycl::property::queue::enable_profiling(),
sycl::property::queue::in_order()})
{
postStatusMsg(OSP_LOG_INFO)
<< "Using SYCL GPU device on "
<< device.get_info<sycl::info::device::name>() << " device (default)";
}
namespace {
sycl::device getDeviceByIndex(uint32_t id)
{
auto gpuDevices = sycl::device::get_devices(sycl::info::device_type::gpu);
return gpuDevices[id % gpuDevices.size()];
}
} // namespace
DeviceImpl::DeviceImpl(uint32_t deviceId, bool debug)
: Device(debug),
device(getDeviceByIndex(deviceId)),
context(device),
queue(context,
device,
{sycl::property::queue::enable_profiling(),
sycl::property::queue::in_order()})
{
postStatusMsg(OSP_LOG_INFO) << "Using SYCL GPU device on "
<< device.get_info<sycl::info::device::name>()
<< " device (number: " << deviceId << ")";
}
DeviceImpl::DeviceImpl(void *devicePtr, void *contextPtr, bool debug)
: Device(debug),
device(*static_cast<sycl::device *>(devicePtr)),
context(*static_cast<sycl::context *>(contextPtr)),
queue(context,
device,
{sycl::property::queue::enable_profiling(),
sycl::property::queue::in_order()})
{
postStatusMsg(OSP_LOG_INFO) << "Using SYCL GPU device on "
<< device.get_info<sycl::info::device::name>()
<< " device (provided externally)";
}
void *DeviceImpl::deviceMalloc(std::size_t size)
{
return sycl::malloc_device(size, queue);
}
void *DeviceImpl::sharedMalloc(std::size_t size)
{
return sycl::malloc_shared(
size, queue, sycl::ext::oneapi::property::usm::device_read_only());
}
void *DeviceImpl::hostMalloc(std::size_t size)
{
return sycl::malloc_host(size, queue);
}
void DeviceImpl::free(void *ptr)
{
sycl::free(ptr, queue);
}
Alloc DeviceImpl::getPointerType(void *ptr) const
{
sycl::usm::alloc type = sycl::get_pointer_type(ptr, queue.get_context());
switch (type) {
case sycl::usm::alloc::host:
return Alloc::Host;
case sycl::usm::alloc::device:
return Alloc::Device;
case sycl::usm::alloc::shared:
return Alloc::Shared;
case sycl::usm::alloc::unknown:
return Alloc::Unknown;
}
return Alloc::Unknown;
}
void DeviceImpl::wait()
{
queue.wait();
}
AsyncEvent DeviceImpl::createAsyncEvent()
{
std::shared_ptr<AsyncEventImpl> eventImpl =
std::make_shared<AsyncEventImpl>();
return AsyncEvent(eventImpl);
}
AsyncEvent DeviceImpl::memcpy(void *dest, const void *src, std::size_t size)
{
// Create async event and do memory copying
std::shared_ptr<AsyncEventImpl> eventImpl =
std::make_shared<AsyncEventImpl>();
eventImpl->event = queue.memcpy(dest, src, size);
return AsyncEvent(eventImpl);
}
AsyncEvent DeviceImpl::launchRendererKernel(const vec3ui &itemDims,
RendererKernel kernel,
ispc::Renderer *renderer,
ispc::FrameBuffer *fb,
ispc::Camera *camera,
ispc::World *world,
const uint32_t *taskIDs,
const FeatureFlags &ff)
{
// Create async event and launch a kernel
std::shared_ptr<AsyncEventImpl> eventImpl =
std::make_shared<AsyncEventImpl>();
kernel(&queue,
&eventImpl->event,
itemDims,
renderer,
fb,
camera,
world,
taskIDs,
ff);
return AsyncEvent(eventImpl);
}
AsyncEvent DeviceImpl::launchFrameOpKernel(const vec2ui &itemDims,
FrameOpKernel kernel,
const ispc::FrameBufferView *fbv)
{
// Create async event and launch a kernel
std::shared_ptr<AsyncEventImpl> eventImpl =
std::make_shared<AsyncEventImpl>();
kernel(&queue, &eventImpl->event, itemDims, fbv);
return AsyncEvent(eventImpl);
}
AsyncEvent DeviceImpl::launchHostTask(const std::function<void()> &task)
{
// Create async event and launch a task
std::shared_ptr<AsyncEventImpl> eventImpl =
std::make_shared<AsyncEventImpl>();
eventImpl->event = queue.submit(
[&](sycl::handler &cgh) { cgh.host_task([=]() { task(); }); });
return AsyncEvent(eventImpl);
}
void *DeviceImpl::getSyclDevicePtr()
{
// Return SYCL device pointer
return &device;
}
void *DeviceImpl::getSyclContextPtr()
{
// Return SYCL context pointer
return &context;
}
void *DeviceImpl::getSyclQueuePtr()
{
// Return SYCL command queue pointer
return &queue;
}
} // namespace devicert
} // namespace ospray
|
