1 //---------------------------------------------------------------------------//
2 // Copyright (c) 2013-2014 Kyle Lutz <kyle.r.lutz@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 <iterator>
13 #include <algorithm>
14
15 #include <thrust/device_vector.h>
16 #include <thrust/host_vector.h>
17 #include <thrust/merge.h>
18 #include <thrust/sort.h>
19
20 #include "perf.hpp"
21
main(int argc,char * argv[])22 int main(int argc, char *argv[])
23 {
24 perf_parse_args(argc, argv);
25
26 std::cout << "size: " << PERF_N << std::endl;
27 thrust::host_vector<int> v1(std::floor(PERF_N / 2.0));
28 thrust::host_vector<int> v2(std::ceil(PERF_N / 2.0));
29 std::generate(v1.begin(), v1.end(), rand);
30 std::generate(v2.begin(), v2.end(), rand);
31 std::sort(v1.begin(), v1.end());
32 std::sort(v2.begin(), v2.end());
33
34 // transfer data to the device
35 thrust::device_vector<int> gpu_v1 = v1;
36 thrust::device_vector<int> gpu_v2 = v2;
37 thrust::device_vector<int> gpu_v3(PERF_N);
38
39 perf_timer t;
40 for(size_t trial = 0; trial < PERF_TRIALS; trial++){
41 t.start();
42 thrust::merge(
43 gpu_v1.begin(), gpu_v1.end(),
44 gpu_v2.begin(), gpu_v2.end(),
45 gpu_v3.begin()
46 );
47 cudaDeviceSynchronize();
48 t.stop();
49 }
50 std::cout << "time: " << t.min_time() / 1e6 << " ms" << std::endl;
51
52 thrust::host_vector<int> check_v3 = gpu_v3;
53
54 thrust::host_vector<int> v3(PERF_N);
55 std::merge(v1.begin(), v1.end(), v2.begin(), v2.end(), v3.begin());
56 bool ok = std::equal(check_v3.begin(), check_v3.end(), v3.begin());
57 if(!ok){
58 std::cerr << "ERROR: merged ranges different" << std::endl;
59 return -1;
60 }
61
62 return 0;
63 }
64