| Bezoar 2007-06-18, 10:11 pm |
| On Jun 18, 5:24 pm, Alexandre Ferrieux <alexandre.ferri...@gmail.com>
wrote:
> On Jun 18, 7:07 pm, gavino <gavcom...@gmail.com> wrote:
>
>
>
>
> No magic nor technology will parallelize a single-threaded, event-
> driven task without considerable human help (assuming it makes sense,
> otherwise the human won't cooperate either ;-).
> A "hunk of tcl code" spawning no threads nor child processes will not
> gain much from your grid of 128 transputers...
> Now maybe you could describe the task you are trying to "optimize"
> this way.
>
> -Alex
I don't agree with that , I once wrote a Tcl script to stress test an
application that used a socket based api
to accept jobs and return the result. My script would fork off
children to run their jobs and the idea was to
simulate a number of simultaneous users. When the parent script
issued the stop command the children would
cleanup and report back their results. The parent would database the
results and quit. I could easily see this
type of paradigm being used to process huge files where each child
would get a file or a segment of a file and process
the data and return back the results. Distcc is exactly that where it
uses cpp on other machines to preprocess files
before compiling on the main compile machine. This would not be
difficult as you say at all particularily if no ordering
of execution is necessary.
|