[Vm-dev] Multi-core system Squeak (was Re: Sista alternate bytecodes + Java Bytecode ?)

David T. Lewis lewis at mail.msen.com
Fri Oct 20 13:59:24 UTC 2017


On Fri, Oct 20, 2017 at 02:39:51PM +0800, Ben Coman wrote:
>  
> On Fri, Oct 20, 2017 at 9:05 AM, David T. Lewis <lewis at mail.msen.com> wrote:
> 
> >
> > The Spur image format allows object memory to be segmented such that it will
> > support fast and efficient forking of images even for extremely large images.
> >
> 
> Would it be efficient enough to benefit forking off the display rendering
> half of the world loop?

No, it is much too slow for that. Reading the results back into the parent image
takes time. But it might be good enough for things like background generation of
Sista bytecodes to be used in future method calls (I think Tim had that in mind).

I think that forkSqueak opens some interesting possibilities for multiprocessing,
but I would expect it to be useful only for coarse-grained problems, possibly such
as a numeric method that can be partitioned easily. I doubt that it will prove
useful for small or short processing where the complexity would outweigh any
possible benefits.

It does make for a good parlor trick though. Try doing this for a completely
useless but otherwise impressive example:

   (RemoteTask do: [2 + 2]) inspect

Dave



More information about the Vm-dev mailing list