Skip to content

GitLab

Switch branch/tag
Find file Normal viewHistoryPermalink
GpuResources_c.h 2.48 KB
Newer Older
huchen's avatar
init the faiss for rocm  
huchen committed
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 
/**
 * Copyright (c) Facebook, Inc. and its affiliates.
 *
 * This source code is licensed under the MIT license found in the
 * LICENSE file in the root directory of this source tree.
 */

// Copyright 2004-present Facebook. All Rights Reserved.
// -*- c -*-

#ifndef FAISS_GPU_RESOURCES_C_H
#define FAISS_GPU_RESOURCES_C_H

#include <cublas_v2.h>
#include <cuda_runtime_api.h>
#include "../faiss_c.h"

#ifdef __cplusplus
extern "C" {
#endif

/// Base class of GPU-side resource provider; hides provision of
/// cuBLAS handles, CUDA streams and a temporary memory manager
FAISS_DECLARE_CLASS(GpuResources)

FAISS_DECLARE_DESTRUCTOR(GpuResources)

/// Call to pre-allocate resources for a particular device. If this is
/// not called, then resources will be allocated at the first time
/// of demand
int faiss_GpuResources_initializeForDevice(FaissGpuResources*, int);

/// Returns the cuBLAS handle that we use for the given device
int faiss_GpuResources_getBlasHandle(FaissGpuResources*, int, cublasHandle_t*);

/// Returns the stream that we order all computation on for the
/// given device
int faiss_GpuResources_getDefaultStream(FaissGpuResources*, int, cudaStream_t*);

/// Returns the available CPU pinned memory buffer
int faiss_GpuResources_getPinnedMemory(FaissGpuResources*, void**, size_t*);

/// Returns the stream on which we perform async CPU <-> GPU copies
int faiss_GpuResources_getAsyncCopyStream(
        FaissGpuResources*,
        int,
        cudaStream_t*);

/// Calls getBlasHandle with the current device
int faiss_GpuResources_getBlasHandleCurrentDevice(
        FaissGpuResources*,
        cublasHandle_t*);

/// Calls getDefaultStream with the current device
int faiss_GpuResources_getDefaultStreamCurrentDevice(
        FaissGpuResources*,
        cudaStream_t*);

/// Synchronizes the CPU with respect to the default stream for the
/// given device
// equivalent to cudaDeviceSynchronize(getDefaultStream(device))
int faiss_GpuResources_syncDefaultStream(FaissGpuResources*, int);

/// Calls syncDefaultStream for the current device
int faiss_GpuResources_syncDefaultStreamCurrentDevice(FaissGpuResources*);

/// Calls getAsyncCopyStream for the current device
int faiss_GpuResources_getAsyncCopyStreamCurrentDevice(
        FaissGpuResources*,
        cudaStream_t*);

FAISS_DECLARE_CLASS(GpuResourcesProvider)

FAISS_DECLARE_DESTRUCTOR(GpuResourcesProvider)

int faiss_GpuResourcesProvider_getResources(
        FaissGpuResourcesProvider*,
        FaissGpuResources**);

#ifdef __cplusplus
}
#endif
#endif