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
|
/* SPDX-License-Identifier: Apache-2.0
* Copyright 2011-2022 Blender Foundation */
#ifdef WITH_OPTIX
# include "device/optix/queue.h"
# include "device/optix/device_impl.h"
# include "util/time.h"
# define __KERNEL_OPTIX__
# include "kernel/device/optix/globals.h"
CCL_NAMESPACE_BEGIN
/* CUDADeviceQueue */
OptiXDeviceQueue::OptiXDeviceQueue(OptiXDevice *device) : CUDADeviceQueue(device)
{
}
void OptiXDeviceQueue::init_execution()
{
CUDADeviceQueue::init_execution();
}
static bool is_optix_specific_kernel(DeviceKernel kernel)
{
return (kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE ||
kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_MNEE ||
kernel == DEVICE_KERNEL_INTEGRATOR_INTERSECT_CLOSEST ||
kernel == DEVICE_KERNEL_INTEGRATOR_INTERSECT_SHADOW ||
kernel == DEVICE_KERNEL_INTEGRATOR_INTERSECT_SUBSURFACE ||
kernel == DEVICE_KERNEL_INTEGRATOR_INTERSECT_VOLUME_STACK);
}
bool OptiXDeviceQueue::enqueue(DeviceKernel kernel,
const int work_size,
DeviceKernelArguments const &args)
{
if (!is_optix_specific_kernel(kernel)) {
return CUDADeviceQueue::enqueue(kernel, work_size, args);
}
if (cuda_device_->have_error()) {
return false;
}
debug_enqueue_begin(kernel, work_size);
const CUDAContextScope scope(cuda_device_);
OptiXDevice *const optix_device = static_cast<OptiXDevice *>(cuda_device_);
const device_ptr sbt_data_ptr = optix_device->sbt_data.device_pointer;
const device_ptr launch_params_ptr = optix_device->launch_params.device_pointer;
cuda_device_assert(
cuda_device_,
cuMemcpyHtoDAsync(launch_params_ptr + offsetof(KernelParamsOptiX, path_index_array),
args.values[0], // &d_path_index
sizeof(device_ptr),
cuda_stream_));
if (kernel == DEVICE_KERNEL_INTEGRATOR_INTERSECT_CLOSEST ||
kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE ||
kernel == DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_MNEE) {
cuda_device_assert(
cuda_device_,
cuMemcpyHtoDAsync(launch_params_ptr + offsetof(KernelParamsOptiX, render_buffer),
args.values[1], // &d_render_buffer
sizeof(device_ptr),
cuda_stream_));
}
cuda_device_assert(cuda_device_, cuStreamSynchronize(cuda_stream_));
OptixPipeline pipeline = nullptr;
OptixShaderBindingTable sbt_params = {};
switch (kernel) {
case DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_RAYTRACE:
pipeline = optix_device->pipelines[PIP_SHADE_RAYTRACE];
sbt_params.raygenRecord = sbt_data_ptr + PG_RGEN_SHADE_SURFACE_RAYTRACE * sizeof(SbtRecord);
break;
case DEVICE_KERNEL_INTEGRATOR_SHADE_SURFACE_MNEE:
pipeline = optix_device->pipelines[PIP_SHADE_MNEE];
sbt_params.raygenRecord = sbt_data_ptr + PG_RGEN_SHADE_SURFACE_MNEE * sizeof(SbtRecord);
break;
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_CLOSEST:
pipeline = optix_device->pipelines[PIP_INTERSECT];
sbt_params.raygenRecord = sbt_data_ptr + PG_RGEN_INTERSECT_CLOSEST * sizeof(SbtRecord);
break;
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_SHADOW:
pipeline = optix_device->pipelines[PIP_INTERSECT];
sbt_params.raygenRecord = sbt_data_ptr + PG_RGEN_INTERSECT_SHADOW * sizeof(SbtRecord);
break;
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_SUBSURFACE:
pipeline = optix_device->pipelines[PIP_INTERSECT];
sbt_params.raygenRecord = sbt_data_ptr + PG_RGEN_INTERSECT_SUBSURFACE * sizeof(SbtRecord);
break;
case DEVICE_KERNEL_INTEGRATOR_INTERSECT_VOLUME_STACK:
pipeline = optix_device->pipelines[PIP_INTERSECT];
sbt_params.raygenRecord = sbt_data_ptr + PG_RGEN_INTERSECT_VOLUME_STACK * sizeof(SbtRecord);
break;
default:
LOG(ERROR) << "Invalid kernel " << device_kernel_as_string(kernel)
<< " is attempted to be enqueued.";
return false;
}
sbt_params.missRecordBase = sbt_data_ptr + MISS_PROGRAM_GROUP_OFFSET * sizeof(SbtRecord);
sbt_params.missRecordStrideInBytes = sizeof(SbtRecord);
sbt_params.missRecordCount = NUM_MIS_PROGRAM_GROUPS;
sbt_params.hitgroupRecordBase = sbt_data_ptr + HIT_PROGAM_GROUP_OFFSET * sizeof(SbtRecord);
sbt_params.hitgroupRecordStrideInBytes = sizeof(SbtRecord);
sbt_params.hitgroupRecordCount = NUM_HIT_PROGRAM_GROUPS;
sbt_params.callablesRecordBase = sbt_data_ptr + CALLABLE_PROGRAM_GROUPS_BASE * sizeof(SbtRecord);
sbt_params.callablesRecordCount = NUM_CALLABLE_PROGRAM_GROUPS;
sbt_params.callablesRecordStrideInBytes = sizeof(SbtRecord);
/* Launch the ray generation program. */
optix_device_assert(optix_device,
optixLaunch(pipeline,
cuda_stream_,
launch_params_ptr,
optix_device->launch_params.data_elements,
&sbt_params,
work_size,
1,
1));
debug_enqueue_end();
return !(optix_device->have_error());
}
CCL_NAMESPACE_END
#endif /* WITH_OPTIX */
|
