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
|
#if GOOGLE_CUDA
#define EIGEN_USE_GPU
#include "tensorflow/core/common_runtime/gpu/gpu_device.h"
#include "tensorflow/core/common_runtime/gpu/process_state.h"
namespace tensorflow {
void RequireGPUDevice() {}
class GPUDevice : public BaseGPUDevice {
public:
GPUDevice(const SessionOptions& options, const string& name,
Bytes memory_limit, BusAdjacency bus_adjacency, int gpu_id,
const string& physical_device_desc, Allocator* gpu_allocator,
Allocator* cpu_allocator)
: BaseGPUDevice(options, name, memory_limit, bus_adjacency, gpu_id,
physical_device_desc, gpu_allocator, cpu_allocator) {}
Allocator* GetAllocator(AllocatorAttributes attr) override {
if (attr.on_host()) {
ProcessState* ps = ProcessState::singleton();
if (attr.gpu_compatible()) {
return ps->GetCUDAHostAllocator(0);
} else {
return cpu_allocator_;
}
} else {
return gpu_allocator_;
}
}
};
class GPUDeviceFactory : public BaseGPUDeviceFactory {
private:
LocalDevice* CreateGPUDevice(const SessionOptions& options,
const string& name, Bytes memory_limit,
BusAdjacency bus_adjacency, int gpu_id,
const string& physical_device_desc,
Allocator* gpu_allocator,
Allocator* cpu_allocator) override {
return new GPUDevice(options, name, memory_limit, bus_adjacency, gpu_id,
physical_device_desc, gpu_allocator, cpu_allocator);
}
};
REGISTER_LOCAL_DEVICE_FACTORY("GPU", GPUDeviceFactory);
} // namespace tensorflow
#endif // GOOGLE_CUDA
|
