search

LLNL / libROM

Model reduction library with an emphasis on large scale parallelism and linear subspace methods
https://www.librom.net
Other
201 stars 36 forks source link

Add init samples to S_OPT for Laghos #117

Closed kevinhkhuynh closed 2 years ago

kevinhkhuynh commented 2 years ago

It wasn't entirely clear to me what the GNAT version of this was doing, so I tried to re-implement it here.

Here, I assume each process has its own unique init_samples vector using the local row indices (i.e., the indices for each init_samples vector start from zero on each processor) and that num_samples <= init_samples.

  1. Sample all init_samples from each processor in a random order.
  2. Continue on with the normal S_OPT algorithm assuming we still have some samples left over to sample, skipping the special case for the very first sample.
kevinhkhuynh commented 2 years ago

Added init_samples to S_OPT, similarly to GNAT to fit the Laghos template, and added a simple test.

kevinhkhuynh commented 2 years ago

Once this is reviewed, I will merge this into PR 112 and merge PR 112, along with the corresponding Laghos PR.