On Fri, Mar 31, 2023 at 2:39 PM LawsonEnglish <LEnglish5 at cox.net> wrote:

> The Shared Memory facility in most modern OS’s (Mac, Linux and Windows)
> will allow a pointer between processes. The problem is to get it working
> with Squeak. Eliot outlined the issues in his response. They are more
> convoluted than I was hoping for, but presumably it's doable.
>
> The point of doing it is to create a shared memory IPC option for Squeak,
> as, in principle, that is the fastest way of doing IPC, and if you’re
> sharing large chunks of data (e.g. the pixels of an image), the savings can
> be significant.
>
> The Mandelbrot Set is simply a simple (and potentially entertaining) way
> of providing visual feedback that its actually working. Its actually not a
> good use case for shared memory IPC because, as you drill down into the
> Set, the time spent for calculations is far more than the time saved using
> the shared memory, but it would visually demonstrate that something is
> happening, and far less boring than simply sending arbitrary chunks of
> memory during testing of the protocol.
>

And note that with image segments one could exchange arbitrary object
graphs above a memory-mapped word array with relative ease and speed.

Lawson
>
>
> On Mar 31, 2023, at 00:49, Marcel Taeumel via Squeak-dev  wrote:
>
> Hi L --
>
> > The real question is about setting a pointer to the memory used by a
> Bitmap object.
>
> I would not do that, I don't think it is possible as each project has its
> own address space. Virtual memory and so on... You cannot simply share
> pointers between processes. Thus, allocate some inter-process memory and
> copy initial data (from any Squeak object memory) to there. Then read/write
> it there.
>
>
>
>

-- 
_,,,^..^,,,_
best, Eliot
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.squeakfoundation.org/pipermail/squeak-dev/attachments/20230331/c90bb033/attachment.html>