|
Discontinuous Galerkin Library
#include "dg/algorithm.h"
|
Simple tool for performance measuring. More...
Public Member Functions | |
| void | tic () |
| Start timer. | |
| void | toc () |
| Stop timer. | |
| double | diff () const |
| Return time in seconds elapsed between tic and toc. | |
Simple tool for performance measuring.
The implementation of this class is chosen with compile-time MACROS THRUST_DEVICE_SYSTEM and MPI_VERSION.
| double dg::Timer::diff | ( | ) | const |
Return time in seconds elapsed between tic and toc.
| void dg::Timer::tic | ( | ) |
Start timer.
uses omp_get_wtime() if available, else gettimeofday. If compiled with nvcc we place cudaEvent_t in the gpu stream. The mpi version places an MPI_Barrier(MPI_COMM_WORLD) and then uses MPI_Wtime. MPI + Cuda adds an additional cudaDeviceSynchronize
| void dg::Timer::toc | ( | ) |
Stop timer.
uses omp_get_wtime() if available, else gettimeofday. If compiled with nvcc we place cudaEvent_t in the gpu stream. The mpi version places an MPI_Barrier(MPI_COMM_WORLD) and then uses MPI_Wtime. MPI + Cuda adds an additional cudaDeviceSynchronize