chapter2 GPU VS CPU runtime

[packtpublishing]

chapter2 GPU VS CPU runtime

[deepesh-packt]

The code for measuring performance is shown in the book. The code to measure time using clock_t should be wrapped around the call to the kernel or CPU function like shown below:
clock_t start_d = clock();
gpuAdd << <N, 1 >> >(d_a, d_b, d_c);
cudaThreadSynchronize();
clock_t end_d = clock();
double time_d = (double)(end_d - start_d) / CLOCKS_PER_SEC;
The maximum value of N will be dependent on memory available on the system.

[KansaiUser]

The code for measuring performance is shown in the book. The code to measure time using clock_t should be wrapped around the call to the kernel or CPU function like shown below:
clock_t start_d = clock();
gpuAdd << <N, 1 >> >(d_a, d_b, d_c);
cudaThreadSynchronize();
clock_t end_d = clock();
double time_d = (double)(end_d - start_d) / CLOCKS_PER_SEC;
The maximum value of N will be dependent on memory available on the system.

Yes, but in the book there is also time_h which appears nowhere.
It would be nice to have the code in the repo but now in Chapter 2 we have the same code for matrix addition 3 times without any change between them.