I had run the previous version of Hydrolight on a serial computer with
about 800 processors. We had the challenge of performing thousands of
short simulations (each taking about a minute to execute). There was no
parralelization, though.

We simply compiled the HL code to run it under Linux. We did not modify
the code. Every simulation was being separately assigned to a different
processor. There was a homemade batch queuing software that was taking
care of an optimal assignment of executions to different processors.

servet


erehm wrote:
>
>
> Has anyone parallelized Hydrolight and/or Ecolight to run on either
> multicore processors on a single machine (requiring rework of code to
> be reentrant) or on multiple machines (e.g., via MPI)?
>
> Let's assume for the moment that we're not including any inelastic
> scattering, so that one could compute all requested wavelengths in
> parallel...
>
> .
> Web Bug from
>
http://geo.yahoo.com/serv?s=97359714/grpId=14196913/grpspId=1705083247/msgId=66/\
stime=1243584225/nc1=1/nc2=2/nc3=3
>
>