[Bf-committers] Sunday meeting minutes, 25th feb 2007

[Giuseppe Ghibò]

Timothy Baldridge wrote:

>>> So how about improving (or adding)
>>> the parallelization of certain tasks|modules (either trough pthread or
>>> OpenMP, which is scheduled also in next gcc 4.2) in blender, which
>>> might be:
> 
> I agree with this idea. However there is a issue here. Simply going
> multi-threaded is a great idea, but limited in my view. OpenMP and
> pthreads use multiple threads, but the same process, and they are not
> network transparent.
> 
> By using a API like PVM and MPI, we would allow blender to be scalable
> not only across multiple cores/CPUs but also across multiple machines,
> i.e. a render farm.

That sound good, but I would distinguish tasks which scales
over SMP in the current typical multicore archs, from those
which can be scaled over the network (or both). Using MPI or PVM would
require also fast network (indeed this depends on the
tasks), while in SMP the data are always exchanged
in memory which has latency over network of one or two orders
of magnitude smaller. For instance IMHO there are tasks
(probably RT playback) for which maybe you need an extremely low latency,
On the other hand there are other tasks, where the
more you scale across the network the better is, as you can increase
"easily" the number of nodes. So maybe we could start listing this
different kind of cases (up to a client "blender at home" for a huge
gigantic distributed blender network... :-)).

Bye.
Giuseppe.