Cuda 4 inter-GPU synchronization ?

Did anyone successfully test this feature on Cuda4 ? It seems I can’t have it worked correctly, not sure if it’s a bug in my code or the feature is not fully implemented yet.

For example, I want to do a kernel call after copying a memory from 1 gpu to another. Stream 0 is created on gpu0 stream 1 is created on gpu1. Both are Fermi

cudaMemcpyAsync( mem1, mem0, size, cudaMemcpyDefault, stream0 );
cudaEventRecord(P2Pevent, stream0);

cudaStreamWaitEvent(stream1, P2Pevent, 0);

Sometime it seems the memory is not ready for the kernel yet so the result is incorrect. If I add an cudaDeviceSynchronize on Gpu0 then it works fine.


any idea ?

Shouldn’t you use cudaMemcpyDeviceToDevice?

Not sure if cudaStreamWaitEvent works for event on a different device

default should be working. can you post full code? this might be a known bug that is resolved in RC2, but I’d like to find out for sure.

I guess it’s a bug in my code, changed something and it works… Thanks guy. This is a cool feature, hopefully there is no bug

so what did you change to make it work?