swarm-support
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Drone (was Re: ObjectLoader & batch processing)


From: Theodore C. Belding
Subject: Re: Drone (was Re: ObjectLoader & batch processing)
Date: Tue, 29 Oct 1996 22:29:23 -0500

So the problem was with using libtclobjc in particular, and not Tcl in
general?  It's actually possible to just use Tcl as the control language
for Obj C without libtclobjc, although it's obviously not as nice or
object-oriented.  Ousterhout gives some examples of this in his book.  I
can dig up an example that I wrote, if anyone's interested.

As far as using Drone with a built-in scripting language, as far as I can
see, Drone would simply treat the script as just another input file for the
program.  Then the script could tell the program to load in any additional
parameter files that were necessary.  Any parameters that you sweep would
still be specified on the command line.  And instead of sweeping
"parameters", you could instead sweep over a set of messages to an object:
-D<object>=<message> -- this is all handled by the program, and the script
doesn't care.

One thing that intrigues me is the idea of writing the control file that
controls Drone's operation (how many runs to execute, etc.) in Tcl, then
have Drone interpret that as a script, instead of having an arcane control
file syntax for this.  This is how you write a .rc file for some Tcl/Tk
newsreaders, etc. The problem with this is that it seems too powerful --
the user can alter any variable, and maybe cause a lot of damage.  I'd like
to look into this more though -- maybe I missed an easy solution.
-Ted

At 4:48 PM -0700 10/29/96, Manor Askenazi wrote:
>Y'all,
>
>> On a related topic, I think there was some discussion a while ago about
>> more general scripting.  If you'd like to build a completely general
>> scripting language into your Swarm application, so that you can control
>> every aspect of the program using a script, I'd recommend taking a look at
>> Tcl...
>
>This was the original plan... However, some pilot tests conducted by
>Ginger showed that dynamic calls involving doubles and floats weren't
>working correctly, i.e. the following was not possible:
>
>InitScript.tcl:
>
>  [MyMainSwarm setHeat: 0.789] ;
>
>This was not surprising because __builtin_apply() wasn't working properly
>and libtclObjc depended on it.
>
>Now, there has been an announcement of an upgrade to gcc which may have
>solved this problem so that we could perhaps get back on track for using
>tcl as the scripting approach for Swarm. A really cool example of this
>can be found at:
>
>  http://www-sfb288.math.tu-berlin.de/oorange/Oorange.html
>
>As far as I understood the Oorange group were able to proceed completely
>with interpreted Objective-C because they were running on 1-2 architectures.
>But due to the announcement of gcc-2.7.2.1-objc-960906 there may be a way to
>proceed on this and do the double stuff and do it on any architecture.
>
>In the meanwhile I made the horrendous blunder of proposing an interim
>solution which stinks because y'all want to do much more than just the
>equivalent of typing into a probe: you want to set structs arrays etc.
>save arrays etc. etc.
>
>For the record, if we do have a cool scripting language, how would that
>mesh in with the DRONE stuff?
>
>Regards,
>
>Manor.
>
>PS> We may be losing an opportunity to design all of this cleanly into
>    Swarm, but the prime imperative here is to provide solutions today
>    to whoever is doing batch-stuff (that includes me as a user :-)
>    So, I for one will take whatever works!!!


--
Ted Belding                      <mailto:address@hidden>
University of Michigan Program for the Study of Complex Systems
<http://www-personal.engin.umich.edu/~streak/>




reply via email to

[Prev in Thread] Current Thread [Next in Thread]