[Sidefx-houdini-list] simulation multithreaded?
joe at sidefx.com
Mon Oct 29 23:47:25 EDT 2007
Abdul-Kareem Abo-Namous wrote:
> funny thing happening here (9.741 on w32): I run a p-fluid simulation with a
> low number of particles, and it uses - as expected - only one core. I
> increase the amount of particles to something maybe a bit extreme (0.1 for
> particle separation on a sphere with radius 2), and suddenly all 8 cores are
> busy, and houdini is eating up 100% cpu. bug or sudden multithreading? I'm
> guessing the first since the simulation isn't actually proceeding much.. :)
Most likely what can be considered "sudden multithreading."
Multithreading has a certain amount of overhead, and so it's not always
worthwhile to engage it (e.g. when using only a few particles in
particle fluids -- this is only an example, as I'm not sure what the
conditions for multithreading actually are on the solver).
When those conditions -- which it seems your "extreme" simulation
fulfills -- are met, multithreading engages (as you've noticed).
More information about the Sidefx-houdini-list