swarm-support
[Top][All Lists]
Advanced

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

swarm-951209 released


From: nelson
Subject: swarm-951209 released
Date: Sat, 9 Dec 1995 14:21:28 -0700

I'm glad to announce a new version of Swarm, swarm-951209.tar.gz. 
Thanks to everyone who worked with swarm-951014 - we've tried to
integrate your suggestions, fix some of the bugs you found.

The major changes are that gcc 2.7.x is required, there are some
changes to the schedule interface that will require small
modifications of your applications, and there is a new convention for
polling for user events that results in faster execution.

In addition to the new code, this release of Swarm will bring in a few
more people to the beta test group. The plan is to continue to do
incremental beta releases (hopefully more frequently), adding more
people in as we go, until we are confident enough in the solidity and
completeness of Swarm to make a general release.

You will find the code, as well as a more detailed list of changes, on
the Swarm release web page:
  http://www.santafe.edu/projects/swarm/release/
You will need authorization to read these pages:
  username swarm-beta
  password AlanTuring
Please do not give this password out to other people: for the current
beta test period, we want to control the set of people who use Swarm.     


A couple of small changes we said we'd make didn't make it in:

First, I did not change Makefile.appl to require you to list the Swarm
libraries you want to link in. The tradeoff here is between linking in
libraries you don't use (libneuro.a, for instance, if you're not using
neural networks) vs. making authors add an ugly line like this to their
application's Makefile:
  SWARMLIBS=-lga -lneuro -lspace -lsimtools -ltkobjc -lactivity -lrandom 
-lswarmobject -lcollections -ldefobj   

Second, the full Swarm-as-object changes we intended aren't quite
here. The idea is to bundle up SwarmPlan+SwarmActivity+other stuff
into one object called "Swarm". This will be relevant for multi-swarm
simulations. In anticipation of this, the SwarmPlan is now more like a
regular schedule, but some of the other packaging hasn't been done yet.


Several things have been requested by most of our beta users, and are
foremost in our minds. These include:

Better execution of Swarm simulations - batch mode, experiment control
objects, being able to quit the thing cleanly.

Data collection facilities. They're coming, but still
primitive. There's code in swarmlibs/swarmobject (used by Heatbugs,
Mousetrap) that's experimental. The full facility turns in part on the
probe interface, dynamically doing message calls.

More sophisticated space libraries, including coordinate
transformations and the ability to click on objects on the screen.


reply via email to

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