What about overlapping run periods by an appropriate
number of days to help match the ends?
--- In EnergyPlus_Support@yahoogroups.com,
Fred Buhl <wfbuhl@...> wrote:
>
> Nice job, Phylroy. I'll echo Edwin's comments. Generally speaking,
> EnergyPlus is not particularly
> well-suited to parallelization. However some folks have used multiple
> cpus successfully when doing
> optimization.
>
> Fred Buhl
>
> On 12/7/2009 9:06 AM, Edwin Lee
wrote:
> >
> >
> > Phylroy
> >
> > One issue about this is that there is a history effect carried from
> > one time step to another in EnergyPlus. If you just re-initialize to
> > a different day of the year, there are a number of transient
> > components in EnergyPlus which will start out with different values
> > than if you ran an annual simulation straight through. May be fine
> > for whatever case you are wanting, but in many cases, I could see
this
> > adding inaccuracy to the simulation.
> >
> > On the other hand, an 8x boost sure is appealing though! If there is
> > little transient effect in the simulation, that is a nice tool.
> >
> > Edwin
> >
> > On Mon, Dec 7, 2009 at 11:00, Phylroy <phylroy.lopez@...
> > <mailto:phylroy.lopez@...>> wrote:
> >
> >
> >
> >
> > Hey,
> >
> > I just posted a file that will divide the annual simulation period
> > into the number of cpus present on your system and run the slices
> > on separate threads.
> >
> > All that remains is for you to write a script to sum or append the
> > results together for what you wish to examine.
> >
> > This has deceased that simulation time by roughly 8x on our 8 core
> > systems.
> >
> > Here is the script the code is rather straight-foward. You will
> > need ruby installed on your system.
> >
> > http://f1.grp.yahoofs.com/v1/gCYdS218IhkWpX26bNjiFFvdlaOuFZatdpHi03pCWl_31hFtjHZ3tzRw5Gb3iM-NE73gylgX3i3EtfhMZXp4h3f9F-4GUHJotn8L/Utilities/AnnualSimulationSplitter.rb
> >
> > Phylroy
> >
> >
> >
> >
> >
>