I want to employ the code introduced in the paper “On the Robust Mapping of Dynamic Programming onto a Graphics Processing Unit” (http://synergy.cs.vt.edu/pubs/papers/xiao-icpads2009-gpu.pdf). However, I am not quite sure about the function of the __threadfence(), since the other paper “Inter-Block GPU Communication via Fast Barrier Synchronization” (http://eprints.cs.vt.edu/archive/00001087/01/TR_GPU_synchronization.pdf) doesn’t use the __threadfence() function.
Thanks a lot!
I want to employ the code introduced in the paper “On the Robust Mapping of Dynamic Programming onto a Graphics Processing Unit” (http://synergy.cs.vt.edu/pubs/papers/xiao-icpads2009-gpu.pdf). However, I am not quite sure about the function of the __threadfence(), since the other paper “Inter-Block GPU Communication via Fast Barrier Synchronization” (http://eprints.cs.vt.edu/archive/00001087/01/TR_GPU_synchronization.pdf) doesn’t use the __threadfence() function.
Thanks a lot!
You really shouldn’t be doing inter-block synchronization. It’s technically a violation of the execution model.
You really shouldn’t be doing inter-block synchronization. It’s technically a violation of the execution model.