An idea on multithreading implementation

[Geoff Canyon Rev]

On Sat, Jan 29, 2011 at 3:31 PM, Jeffrey Massung <massung at gmail.com> wrote:

> Geoff (btw, Jeff here ;-)),
>
> Okay, I think I completely understand where the disconnect lies - and it's
> with my understanding of the LC internals. I put together a very simple
> stack that's nothing more than a field "Test" and a button. The button
> script looks like this:
>
...

>
> Now, if LC essentially already had coroutines, when done, the output in the
> field should be something like "1a2b3c..." (with newlines of course).
> However, that's not what's in the field. Instead the output is
> "123456...abcdef..." This is likely what you were alluding to and I wasn't
> getting, and this puts an entirely different face on the conversation.
>

Yep, that's what I was talking about. Sorry I wasn't clearer.


>
> The issue isn't coroutines vs. whatever so much as LC doesn't actually
> allow (from what I can tell) for multiple execution contexts. Coroutines -
> obviously - is one method of achieving this goal, and I would agree that it
> is a preferred solution.


> I hope the test above puts this discussion to rest and is a hint to the Rev
> team on a direction they can take.
>

I'd vote for that.


>
> Now, not to be pedantic, but your last paragraph didn't really make much
> sense [to me]. Coroutines - since they must yield and don't run in parallel
> - are are still single-threaded. This means that you could run 100 contexts
> on 100 HW threads, and your program will run at exactly the same speed as if
> they were on 1 HW thread (this assumes that the OS isn't hyperthreading your
> application with others).
>
>
In what way do you think threads would speed up performance?