CUDA Thread communication and synchronization mechanisms

Asked Dec 09 '11 at 22:23

Active Feb 21 '16 at 07:13

Viewed 997 times

I am a bit confused about the thread communication and synchronization mechanisms provided by CUDA. Can someone help me to verify my assumptions below ?

Threads within a warp communicate using shared or global memory and synchronize using implicit synchronization
Warps within a thread block communicate using shared or global memory and synchronize using barrier synchronization
Thread blocks within a given grid or kernel communicate using global memory and synchronize using atomic memory operations
Thread blocks from different grids or kernels communicate using global memory and synchronize using implicit synchronization

edited Feb 21 '16 at 07:13

talonmies

70,661
34
192
269

asked Dec 09 '11 at 22:23

nurabha

1,152
3
18
42

Yes, your understanding is correct. – necromancer Dec 09 '11 at 22:44
2

If you can reword so that this is not a yes/no question, we can reopen and possibly get some better answers. – Dec 10 '11 at 20:41
2

Being a yes/no question doesn't mean it isn't a real question. The poster put some work in here, it seems a shame it was closed. – Adrian Mouat Feb 29 '12 at 11:53

CUDA Thread communication and synchronization mechanisms

0 Answers0