ceph/src/boost/libs/compute/perf/perf_thrust_partition.cu

   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 <algorithm>
  12 #include <cstdlib>
  13 #include <iostream>
  14
  15 #include <thrust/copy.h>
  16 #include <thrust/device_vector.h>
  17 #include <thrust/generate.h>
  18 #include <thrust/host_vector.h>
  19 #include <thrust/partition.h>
  20
  21 #include "perf.hpp"
  22
  23 int rand_int()
  24 {
  25     return static_cast<int>((rand() / double(RAND_MAX)) * 25.0);
  26 }
  27
  28 struct less_than_ten : public thrust::unary_function<bool, int>
  29 {
  30     __device__ bool operator()(int x) const
  31     {
  32         return x < 10;
  33     }
  34 };
  35
  36 int main(int argc, char *argv[])
  37 {
  38     perf_parse_args(argc, argv);
  39
  40     std::cout << "size: " << PERF_N << std::endl;
  41     thrust::host_vector<int> h_vec(PERF_N);
  42     std::generate(h_vec.begin(), h_vec.end(), rand_int);
  43
  44     thrust::device_vector<int> d_vec(PERF_N);
  45
  46     perf_timer t;
  47     for(size_t trial = 0; trial < PERF_TRIALS; trial++){
  48         d_vec = h_vec;
  49
  50         t.start();
  51         thrust::partition(
  52             d_vec.begin(), d_vec.end(), less_than_ten()
  53         );
  54         cudaDeviceSynchronize();
  55         t.stop();
  56     }
  57     std::cout << "time: " << t.min_time() / 1e6 << " ms" << std::endl;
  58
  59     return 0;
  60 }