puzzle1 viennaCL

gistfile1.txt

#include <iostream>
#include <vector>

#include "viennacl/vector.hpp"
#include "benchmark-utils.hpp"

#define BENCHMARK_VECTOR_SIZE   1000
#define BENCHMARK_RUNS          10

int main() {
  //Change this type definition to double if your gpu supports that
  typedef float       ScalarType;
  int v_size = 1000;
  Timer timer;
  double e_time;

  std::vector<ScalarType>       stl_vec(v_size);  
  viennacl::vector<ScalarType>  y1(v_size);
  viennacl::vector<ScalarType>  y2(v_size);
  viennacl::vector<ScalarType>  x1(v_size);
  viennacl::vector<ScalarType>  x2(v_size);

  for(unsigned int i=0; i < v_size; i++) {
	stl_vec[i] = i;
  }
 
  copy(stl_vec.begin(), stl_vec.end(), y1.begin());
  copy(stl_vec.begin(), stl_vec.end(), y2.begin());

  // COMMAND 1
  viennacl::backend::finish();
  timer.start();
  std::cout<<"Executing command 1 : x = y1 + y2"<<std::endl;
  timer.start();

  for(int r=0;r<BENCHMARK_RUNS;r++) {
	  x1 = y1 + y2;
  }

  e_time = timer.get();

  std::cout << "GPU time: " << e_time << std::endl;
  std::cout << "GPU "; printOps(2.0 * static_cast<double>(v_size), static_cast<double>(e_time) / static_cast<double>(BENCHMARK_RUNS));


  // COMMAND 2
  viennacl::backend::finish();
  timer.start();
  std::cout<<"Executing command 2 : x = y1 + y2 + y1 + y2"<<std::endl;
  timer.start();

  for(int r=0;r<BENCHMARK_RUNS;r++) {
	  x2 = y1 + y2 + y1 + y2;
  }

  e_time = timer.get();

  std::cout << "GPU time: " << e_time << std::endl;
  std::cout << "GPU "; printOps(2.0 * static_cast<double>(v_size), static_cast<double>(e_time) / static_cast<double>(BENCHMARK_RUNS));

  return 0;
}