[Pcamal-dev] Combining file
Philippe P. Pebay
pppebay at sandia.gov
Fri Dec 8 14:10:32 MST 2006
OpenMP: that is an interesting point of view. In my knowledge, shared
memory architectures are famous for their inability to scale, especially
within reasonable costs.
Regarding file recombination: if some applications can readily deal with
distributed files, this is good. But I would agree with Darryl's
opinion, regarding the fact that it is good to have at least the ability
(for the sake of genericity) to recombine all files into a single one.
If the number of generated nodes is known, then we're good: it suffices
to add the cumulative per-block offsets to the local node IDs. BTW, this
is *not* what we are doing with the StreamingTessellator (because the
number of generated nodes is not known upfront).
P.
Owen, Steven James wrote:
> I had an interesting discussion with Kevin Copps yesterday regarding our
> parallel hex project.
>
> He mentioned that he thinks OpenMP will be the preferred method for
> parallel computing using Dual/multi-core technology. Not sure if this
> is something we should even be concerned about right now, but he seemed
> to be pretty convinced that distributed computing was not the best way
> to do things.
>
> I asked about recombining files. He indicated that it would be fine to
> leave them separate. Sierra could handle the multiple files, although
> it was not the typical use case. Perhaps we provide the option to
> recombine or not? He mentioned a SEACAS tool for doing the
> recombination. Perhaps we just use that rather than reinventing our
> own?
>
> I still think it would be good to get Greg's opinion on the recombining
> of files however.
>
> Another point he brought up - we may have already considered. Are we
> ensuring that the node numbers will be unique across the multiple files.
> Seems like it would be pretty easy to do since we know up front the
> number of nodes we will generate (ie. N sweep-layers times M nodes per
> layer)
>
> -steve
>
> _______________________________________________
> Pcamal-dev mailing list
> Pcamal-dev at software.sandia.gov
> http://software.sandia.gov/mailman/listinfo/pcamal-dev
--
Philippe P. Pebay | pppebay at ca.sandia.gov
Sandia National Laboratories | Phone: +1 925 294 2024
PO Box 969, MS 9051 | Cell : +1 925 784 3255
Livermore, CA 94550 U.S.A. | Fax : +1 925 294 2595
More information about the Pcamal-dev
mailing list