[HPGMG Forum] [EXTERNAL] Re: Acceptable rounding errors

Jeff Hammond jeff.science at gmail.com
Mon Aug 3 18:08:01 UTC 2015


>
>
> The summation order during restriction might not be reproducible
> (depends on implementation details including whether you use
> MPI_Waitsome versus MPI_Waitall), but can be made reproducible and
> parallel-invariant if desired.
>
> Implementing unnecessary determinism at the expense of parallelism is
absolutely the worst way to handle numerical problems.  The best solutions
are (1) math that doesn't care and (2) floating-point that does care,
meaning extended precisions.

Jeff
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://hpgmg.org/lists/archives/hpgmg-forum/attachments/20150803/2d6d426d/attachment.html>


More information about the HPGMG-Forum mailing list