QuEST/quest/src/gpu/gpu_config.hpp at main · InvictusWingsSRL/QuEST · GitHub

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
/** @file
 * Utility signatures for querying GPU hardware,
 * and allocating and copying GPU VRAM data.
 *
 * Note that this header is included by /core/ and
 * parsed by non-CUDA compilers, so must never contain
 * any CUDA-specific signatures
 *
 * @author Tyson Jones
 */

#ifndef GPU_CONFIG_HPP
#define GPU_CONFIG_HPP

#include "quest/include/config.h"
#include "quest/include/types.h"
#include "quest/include/qureg.h"
#include "quest/include/matrices.h"
#include "quest/include/channels.h"


/*
 * CUDA ERROR HANDLING
 */

#if COMPILE_CUDA

#define CUDA_CHECK(cmd) \
    assertCudaCallSucceeded((int) (cmd), #cmd, __func__, __FILE__, __LINE__)

void assertCudaCallSucceeded(int code, const char* call, const char* caller, const char* file, int line);

#endif


/*
 * HARDWARE AVAILABILITY
 */

bool gpu_isGpuCompiled();

bool gpu_isCuQuantumCompiled();

bool gpu_isGpuAvailable();

bool gpu_isDirectGpuCommPossible();

int gpu_getNumberOfLocalGpus();

int gpu_getComputeCapability();

size_t gpu_getCurrentAvailableMemoryInBytes();

size_t gpu_getTotalMemoryInBytes();

bool gpu_doesGpuSupportMemPools();

qindex gpu_getMaxNumConcurrentThreads();


/*
 * ENVIRONMENT MANAGEMENT
 */

void gpu_bindLocalGPUsToNodes();

bool gpu_areAnyNodesBoundToSameGpu();

void gpu_sync();

void gpu_initCuQuantum();

void gpu_finalizeCuQuantum();


/*
 * MEMORY MANAGEMENT
 */

qcomp* gpu_allocArray(qindex numLocalAmps);
void gpu_deallocArray(qcomp* amps);

void gpu_copyArray(qcomp* dest, qcomp* src, qindex dim);

void gpu_copyCpuToGpu(qcomp* cpuArr, qcomp* gpuArr, qindex numElems);
void gpu_copyGpuToCpu(qcomp* gpuArr, qcomp* cpuArr, qindex numElems);

void gpu_copyGpuToCpu(Qureg qureg, qcomp* gpuArr, qcomp* cpuArr, qindex numElems);
void gpu_copyGpuToCpu(Qureg qureg);

void gpu_copyCpuToGpu(Qureg qureg, qcomp* cpuArr, qcomp* gpuArr, qindex numElems);
void gpu_copyCpuToGpu(Qureg qureg);

void gpu_copyCpuToGpu(CompMatr matr);
void gpu_copyGpuToCpu(CompMatr matr);

void gpu_copyCpuToGpu(DiagMatr matr);
void gpu_copyGpuToCpu(DiagMatr matr);

void gpu_copyCpuToGpu(SuperOp op);
void gpu_copyGpuToCpu(SuperOp op);

// funnily, there is no need for GPU-to-CPU or FullStateDiagMatr;
// the invoking printer.cpp function uses localiser_fullstatediagamtr_...
// copying to handle the distributed nuisance, which spoofs a Qureg
void gpu_copyCpuToGpu(FullStateDiagMatr matr);


/*
 * CACHE MANAGEMENT
 */

qcomp* gpu_getCacheOfSize(qindex numElemsPerThread, qindex numThreads);

void gpu_clearCache();

size_t gpu_getCacheMemoryInBytes();


#endif // GPU_CONFIG_HPP