search

harrymchugh / pinns

Other
0 stars 0 forks source link

Performance compared to OpenFOAM; scalability (e.g MPI vs finer mesh). #23

Closed harrymchugh closed 1 year ago

harrymchugh commented 1 year ago

For the 220x220 mesh case, with a given reynolds number (or bigger mesh if needed to avoid overdecomposition), explore how well the simulation scales with increasing number of processors.

compare to inference

harrymchugh commented 1 year ago

Would be nice to show profiling of the GPU during inference

https://dev-discuss.pytorch.org/t/using-nsight-systems-to-profile-gpu-workload/59

harrymchugh commented 1 year ago

no impact using MPI as 20x20 and even 100x100 too overdecomposed