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
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
|
/*
* -----------------------------------------------------------------
* $Revision$
* $Date$
* -----------------------------------------------------------------
* Programmer(s): Slaven Peles @ LLNL
* -----------------------------------------------------------------
* LLNS Copyright Start
* Copyright (c) 2014, Lawrence Livermore National Security
* This work was performed under the auspices of the U.S. Department
* of Energy by Lawrence Livermore National Laboratory in part under
* Contract W-7405-Eng-48 and in part under Contract DE-AC52-07NA27344.
* Produced at the Lawrence Livermore National Laboratory.
* All rights reserved.
* For details, see the LICENSE file.
* LLNS Copyright End
* -----------------------------------------------------------------
*/
/**
* Vector class
*
* Manages vector data layout for CUDA implementation of N_Vector.
*
*/
#ifndef _NVECTOR_HPP_
#define _NVECTOR_HPP_
#include <cstdlib>
#include <iostream>
#include <cuda_runtime.h>
#include "ThreadPartitioning.hpp"
#include <nvector/nvector_cuda.h>
namespace suncudavec
{
template <typename T, typename I>
class Vector : public _N_VectorContent_Cuda
{
public:
Vector(I N)
: size_(N),
mem_size_(N*sizeof(T)),
ownPartitioning_(true)
{
// Set partitioning
partStream_ = new StreamPartitioning<T, I>(N, 256);
partReduce_ = new ReducePartitioning<T, I>(N, 256);
allocate();
}
/// Copy constructor does not copy values
explicit Vector(const Vector& v)
: size_(v.size()),
mem_size_(size_*sizeof(T)),
partStream_(v.partStream_),
partReduce_(v.partReduce_),
ownPartitioning_(false)
{
allocate();
}
~Vector()
{
if (ownPartitioning_)
{
delete partReduce_;
delete partStream_;
}
clear();
}
void allocate()
{
cudaError_t err;
h_vec_ = static_cast<T*>(malloc(mem_size_));
if(h_vec_ == NULL)
std::cerr << "Failed to allocate host vector!\n";
err = cudaMalloc((void**) &d_vec_, mem_size_);
if(err != cudaSuccess)
std::cerr << "Failed to allocate device vector (error code " << err << ")!\n";
}
void clear()
{
free(h_vec_);
cudaError_t err = cudaFree(d_vec_);
if(err != cudaSuccess)
std::cerr << "Failed to free device vector (error code " << err << ")!\n";
}
int size() const
{
return size_;
}
T* host()
{
return h_vec_;
}
const T* host() const
{
return h_vec_;
}
T* device()
{
return d_vec_;
}
const T* device() const
{
return d_vec_;
}
void copyToDev()
{
cudaError_t err = cudaMemcpy(d_vec_, h_vec_, mem_size_, cudaMemcpyHostToDevice);
if(err != cudaSuccess)
std::cerr << "Failed to copy vector from host to device (error code " << err << ")!\n";
}
void copyFromDev()
{
cudaError_t err = cudaMemcpy(h_vec_, d_vec_, mem_size_, cudaMemcpyDeviceToHost);
if(err != cudaSuccess)
std::cerr << "Failed to copy vector from device to host (error code " << err << ")!\n";
}
StreamPartitioning<T, I>& partStream() const
{
return *partStream_;
}
ReducePartitioning<T, I>& partReduce() const
{
return *partReduce_;
}
private:
I size_;
I mem_size_;
T* h_vec_;
T* d_vec_;
StreamPartitioning<T, I>* partStream_;
ReducePartitioning<T, I>* partReduce_;
bool ownPartitioning_;
};
// Vector extractor
template <typename T, typename I>
inline Vector<T, I> *extract(N_Vector v)
{
return static_cast<Vector<T, I>*>(v->content);
}
// Get Vector device data
template <typename T, typename I>
inline T *getDevData(N_Vector v)
{
Vector<T,I> *vp = static_cast<Vector<T, I>*>(v->content);
return vp->device();
}
// Get Vector length
template <typename T, typename I>
inline I getSize(N_Vector v)
{
Vector<T,I> *vp = static_cast<Vector<T, I>*>(v->content);
return vp->size();
}
} // namespace suncudavec
#endif // _NVECTOR_HPP_
|
