// definition CUDA : typedef struct CUevent_st* cudaEvent_t;

cudaEvent_t  start, stop;
cudaEventCreate(&start);
cudaEventCreate(&stop);

cudaEventRecord(start, 0); // current stream

/* ....appel kernel.... */

cudaEventRecord(stop, 0);
cudaEventSynchronize(stop);

float elapsedTime; // temp écoulé
cudaEventElapsedTime(&elapsedTime, start, stop);
printf("elapsed: %3.1f ms\n", elapsedTime);

cudaEventDestroy(start);
cudaEventDestroy(stop);