blob: f94367cc987b44c53986467e8b2ac9122a98fe97 (
plain)
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
|
/* Copyright 2015 Google Inc. All Rights Reserved.
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.
==============================================================================*/
#ifndef TENSORFLOW_COMMON_RUNTIME_GPU_GPU_BFC_ALLOCATOR_H_
#define TENSORFLOW_COMMON_RUNTIME_GPU_GPU_BFC_ALLOCATOR_H_
#include <memory>
#include <string>
#include <unordered_map>
#include <vector>
#include "tensorflow/core/common_runtime/allocator_retry.h"
#include "tensorflow/core/common_runtime/bfc_allocator.h"
#include "tensorflow/core/platform/stream_executor.h"
#include "tensorflow/core/platform/thread_annotations.h"
#include "tensorflow/core/platform/types.h"
#include "tensorflow/core/protobuf/config.pb.h"
namespace gpu = ::perftools::gputools;
namespace tensorflow {
// A GPU memory allocator that implements a 'best-fit with coalescing'
// algorithm.
class GPUBFCAllocator : public BFCAllocator {
public:
// 'device_id' refers to the StreamExecutor ID of the device within
// the process and must reference a valid ID in the process.
GPUBFCAllocator(int device_id, size_t total_memory);
GPUBFCAllocator(int device_id, size_t total_memory,
const GPUOptions& gpu_options);
virtual ~GPUBFCAllocator() {}
TF_DISALLOW_COPY_AND_ASSIGN(GPUBFCAllocator);
};
// Suballocator for GPU memory.
class GPUMemAllocator : public SubAllocator {
public:
// Note: stream_exec cannot be null.
explicit GPUMemAllocator(perftools::gputools::StreamExecutor* stream_exec)
: stream_exec_(stream_exec) {
CHECK(stream_exec_ != nullptr);
}
~GPUMemAllocator() override {}
void* Alloc(size_t alignment, size_t num_bytes) override {
void* ptr = nullptr;
if (num_bytes > 0) {
ptr = stream_exec_->AllocateArray<char>(num_bytes).opaque();
}
return ptr;
}
void Free(void* ptr, size_t num_bytes) override {
if (ptr != nullptr) {
gpu::DeviceMemoryBase gpu_ptr(ptr);
stream_exec_->Deallocate(&gpu_ptr);
}
}
private:
perftools::gputools::StreamExecutor* stream_exec_; // not owned, non-null
TF_DISALLOW_COPY_AND_ASSIGN(GPUMemAllocator);
};
} // namespace tensorflow
#endif // TENSORFLOW_COMMON_RUNTIME_GPU_GPU_BFC_ALLOCATOR_H_
|
