Closed lastephey closed 1 year ago
When the gpu version of gpu_specter is fully ready, profile the code to understand where our major gpu bottlenecks are.
At the moment our best tools are nsight systems and nvvp (for multiple mpi ranks).
Save/post these results somewhere they won't be lost. They will be useful for NESAP before/after speedup calculation.
Start recording GPU profiling data
If possible:
Track first GPU port
Track addition of MPS
Track batched eigh
Thanks for your help. Closing in the spirit of Closember.
When the gpu version of gpu_specter is fully ready, profile the code to understand where our major gpu bottlenecks are.
At the moment our best tools are nsight systems and nvvp (for multiple mpi ranks).
Save/post these results somewhere they won't be lost. They will be useful for NESAP before/after speedup calculation.