Victor Eijkhout <[email protected]> writes: > Since multiple iterations of NSF reviewers & conference referees have > failed to understand my theory (the fact that I'm still converging on > the best explanation may have something to do with it) I've decided to > put on my coding hat and show that IMP can actually work.
How would you propose to handle irregular communication patterns such as you would get with a small number of dense rows in a sparse matrix, or the row/column communicators for dense linear algebra? How are reductions expressed and implemented? I think that the setup costs are actually critical. They are not always amortized in either dynamic execution models or methods with frequent adaptivity, particle transport, self-contact, etc. The MPI model may already have two-sided knowledge about the new communication pattern or know a good way to use collectives. It's more work for the programmer to specify both sides of the communication, but dynamic exchange is significantly more expensive.
pgpPpurQV0MTg.pgp
Description: PGP signature
