On Sat, 30 Dec 2000 Geoff Canyon <[EMAIL PROTECTED]> wrote:
> At 3:18 PM -0700 12/29/00, Scott Raney wrote:
> >No but not because we're even considering adding thread support to the
> >language. It's because an event-driven application (e.g., using "read
> >to socket with message" and "write to socket with message") is more
> >efficient and usually easier to write and debug than one that uses
> >"wait with message" or any other multi-threading system. In fact,
> >multithreading is such a pain in the ass that we've gone to
> >considerable effort to *remove* all of the existing multithreading
> >code from the engines for the next release and will rely totally on
> >event-driven architectures instead.
>
> At 3:18 PM -0700 12/29/00, Scott Raney wrote:
> >No: this feature is only for programming convenience and would offer
> >nothing in terms of performance (in fact, it will most likely slow
> >things down, just as multithreading in other languages does).
>
> Just out of curiosity, Scott, would the one exception to this
> statement (that threading hurts performance) be in the case of
> multi-processor systems? Under those circumstances, doesn't threading
> simplify the passing out of work to multiple CPUs?
Depending on the OS, maybe. But MacOS doesn't support this and even
support on OS X is not all it could be (e.g., you have to use a
different, incompatible, API). Threading is pretty good NT in this
respect, but none of the UNIXs are all that great so in most cases
where maximum throughput is required, you end up running multiple
processes instead of threads within a process. Not sure which is the
cause and which is the effect, here, but there aren't enough
multi-processor systems out there (large servers excepted) to make
supporting them worth the trouble except for extremely specialized
applications in which case library-based multithreading works as well
as OS-based. For example, the library approach is what the image
editing tools on MacOS take, where the other processor(s) sit idle
until the application specifically calls on them to do something.
Regards,
Scott
> Regards,
>
> Geoff
********************************************************
Scott Raney [EMAIL PROTECTED] http://www.metacard.com
MetaCard: You know, there's an easier way to do that...
Archives: http://www.mail-archive.com/[email protected]/
Info: http://www.xworlds.com/metacard/mailinglist.htm
Please send bug reports to <[EMAIL PROTECTED]>, not this list.
