[squeak-dev] Prepare for Thousands of Cores --- oh my Chip -
it's full of cores!
Joshua Gargus
schwa at fastmail.us
Sun Jul 6 19:21:00 UTC 2008
On Jul 6, 2008, at 11:41 AM, Peter William Lount wrote:
> Hi,
>
> Oh, clearly I misunderstood about Hydra. I thought it could handle
> more than one native thread per image in memory.
Nope, only one.
> My mistake if it can't. Unfortunately I've not yet had time to take
> a look at hydro up close.
>
> The one native thread per image model is an ok idea idea however it
> fails to cover many situations for which multiple threads provides
> the better or even simpler solution.
>
> Besides as soon as you have two threads of either kind you
> essentially have many of the same issues and the same level of
> complexity of issues involved with regards to concurrency and data
> coherency of objects within the image. As soon as you have two or
> more images you have data coherency issues across those images.
> These problems don't magically go away by limiting the native
> threads per image to one.
That's true. However, it provides a simple way to conceptualize which
threads are accessing which memory; you can then design your
application accordingly. This simple model doesn't exist with
multiple-threads-per-image (not that *something* couldn't be designed
to play the same role, but it's not immediately clear to me how that
would work).
Best,
Josh
>
>
> All systems regardless of design have scalability issues. It's the
> nature of the beast.
>
> Erlang certainly has a lot going for it's model even though it's not
> fully message based nor object based.
>
> Thinking to see further than simplistic solutions and doing the hard
> work to get there are both important.
>
> All the best,
>
> Peter
>
>
>
More information about the Squeak-dev
mailing list
|