Timing policy for the measurement of the GPU time.
The DeviceSynchronizePolicy class represents the timing policy for GPU time measurements that can be used in combination with the Timer class template. This combination is realized with the DeviceSynchronizePolicy type definition. This class uses device synchronization internally and is therefore not suited for CUDA applications with overlapping kernels.
#include <DeviceSynchronizePolicy.h>
Static Public Member Functions | |
Timing functions | |
static double | getTimestamp () |
Returns a timestamp of the current GPU time in seconds. More... | |
|
inlinestatic |
Returns a timestamp of the current GPU time in seconds.
Uses wall clock time and device synchronization internally.