Skip to content

strong-scaling issues #17

@ShashankSubramanian

Description

@ShashankSubramanian

[] Only powers of 2 for number of cores - else pnetcdf hangs at ncmpi_open for 448 cores and beyond
[] fftw null plan returned (and related PetscLayout segmentation fault) at 4096 cores and beyond - possible due to fftw version change which might not be incorporated in accfft
[] data-driven gaussians does not scale from 2048 cores and beyond - solver hangs here (probably some communication issue; never been tested for scaling)

Metadata

Metadata

Labels

No labels
No labels

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions