1 //---------------------------------------------------------------------------//
2 // Copyright (c) 2015 Jakub Szuppe <j.szuppe@gmail.com>
3 //
4 // Distributed under the Boost Software License, Version 1.0
5 // See accompanying file LICENSE_1_0.txt or copy at
6 // http://www.boost.org/LICENSE_1_0.txt
7 //
8 // See http://boostorg.github.com/compute for more information.
9 //---------------------------------------------------------------------------//
10
11 #include <iostream>
12 #include <algorithm>
13 #include <vector>
14
15 #include <bolt/cl/copy.h>
16 #include <bolt/cl/device_vector.h>
17 #include <bolt/cl/transform.h>
18
19 #include "perf.hpp"
20
BOLT_FUNCTOR(saxpy_functor,struct saxpy_functor{ float _a; saxpy_functor(float a) : _a(a) {}; float operator() (const float &x, const float &y) const { return _a * x + y; }; };)21 BOLT_FUNCTOR(saxpy_functor,
22 struct saxpy_functor
23 {
24 float _a;
25 saxpy_functor(float a) : _a(a) {};
26
27 float operator() (const float &x, const float &y) const
28 {
29 return _a * x + y;
30 };
31 };
32 )
33
34 int main(int argc, char *argv[])
35 {
36 perf_parse_args(argc, argv);
37
38 std::cout << "size: " << PERF_N << std::endl;
39
40 bolt::cl::control ctrl = bolt::cl::control::getDefault();
41 ::cl::Device device = ctrl.getDevice();
42 std::cout << "device: " << device.getInfo<CL_DEVICE_NAME>() << std::endl;
43
44 // create host vectors
45 std::vector<float> host_x(PERF_N);
46 std::vector<float> host_y(PERF_N);
47 std::generate(host_x.begin(), host_x.end(), rand);
48 std::generate(host_y.begin(), host_y.end(), rand);
49
50 // create device vectors
51 bolt::cl::device_vector<float> device_x(PERF_N);
52 bolt::cl::device_vector<float> device_y(PERF_N);
53
54 // transfer data to the device
55 bolt::cl::copy(host_x.begin(), host_x.end(), device_x.begin());
56 bolt::cl::copy(host_y.begin(), host_y.end(), device_y.begin());
57
58 perf_timer t;
59 for(size_t trial = 0; trial < PERF_TRIALS; trial++){
60 t.start();
61 bolt::cl::transform(
62 device_x.begin(), device_x.end(),
63 device_y.begin(),
64 device_y.begin(),
65 saxpy_functor(2.5f)
66 );
67 t.stop();
68 }
69 std::cout << "time: " << t.min_time() / 1e6 << " ms" << std::endl;
70
71 // transfer data back to host
72 bolt::cl::copy(device_x.begin(), device_x.end(), host_x.begin());
73 bolt::cl::copy(device_y.begin(), device_y.end(), host_y.begin());
74
75 return 0;
76 }
77