Lecture 11 Flashcards
1
Q
cudaDeviceSynchronize()
A
is used from the host.
Wait until all current kernels finish
2
Q
cudaStreamSynchronize()
A
waits until all kernels
in a stream finish.
3
Q
__syncthreads()
A
is used inside a kernel.
Stop thread until all threads reach the location!
4
Q
cudaEventCreate
A
initialize an event variable
5
Q
cudaEventRecord
A
place a marker in the queue
6
Q
cudaEventSynchronize
A
wait until all markers
have received values
7
Q
cudaEventElapsedTime
A
get the time difference
between two events
8
Q
Coalescing memory
A
Always access global memory ”in order”
If threads access global memory in order of thread
numbers, performance will be improved!
9
Q
CUDA can be coupled closer to ________
A
OpenGL