MAGMA on distributed GPU Cluster

[Manasi Tiwari]

Hello,

I am interested in using the sparse iterative solvers in MAGMA (specifically, CG).

1. Does MAGMA support CG with multiple GPUs? Within the same node and across multiple nodes? 

2. Are there any hybrid implementations of CG in MAGMA which use both the multicore CPUs using OpenMP and GPU using CUDA? 

3. Are there any optimizations done in MAGMA while implementing CG method?

If you could point me to some useful publications where the above doubts have been addressed, that would be great!

Regards,

Manasi