May 05, 2004
Cumulative Disarray #1


I have quite a bit of Lisp-related news to catch up on. Some of it is weeks old, but given that this language has been around for over 40 years a few weeks probably won't make much difference one way or the other (anyway, doesn't it feel a little relentless sometimes, the way that weblogs seem so focused on the news of the minute? lemonodor's jumping off that treadmill, ha!). There won't be anything you haven't already seen at Planet Lisp, but I find it's handy to use lemonodor as a sort of personal notebook of record, and the very act of writing a post about something helps me to remember that thing. I suppose it's possible not everyone reads Planet Lisp, too.

So here's the first in a series of posts to get me caught up to my archival desires:

Ivan Boldyrev announced a set of PVM bindings (using UFFI) [via markus fix].

Back when I started grad school, I was pretty into ray tracing. I was also flush with excitement over my newly gained access to 20 or 30 Sun workstations (ray tracing on my own 16 MHz 386sx had already taught me as much patience as I was willing to learn), so when I found a patch for the ray tracing software I preferred that used PVM to do distributed ray tracing across multiple machines, I had to try it.

Unfortunately the patch as published was buggy and only accomplished the feat of turning the ray tracer into a very effective fork bomb. The first time I ran it on the main shared departmental server I thought it was just coincidence that an instant after I hit Enter my dialup connection froze. So a few minutes later when I was able to dial back in and login to my account, I ran it again. Oops, the machine crashed. Again. I just thought it was being really flaky. The next time I reconnected, before I could run the ray tracer a third time a message scrolled onto my terminal:

John Wiseman: please call techstaff immediately!  773-555-9054.

Man, I felt awful.

That reminds me: Ivan's code is “pre-alpha”. Be careful.

For more background on PVM and parallel computing with Lisp, see this post by Bill Clementson.

Posted by jjwiseman at May 05, 2004 08:38 AM

I remember that! I was lucky that the machines I tried it on were under my control, and also that I had the patience (and the stamina... those were the days :-) to work on it until 3 in the morning until it worked.

As I remember, the neat thing about PVM (over more industrially-accepted message-passing systems such as MPI) is the automatic heterogeneous adaptibility — when I got pvmpov working it ran on six sparc Ultra Is (one of which I still use, seven years later, to build sbcl :-), a really early powermac, a linux box or two and an SGI running IRIX. Ah, the good old days...

Posted by: Christophe Rhodes on May 5, 2004 03:33 PM
Post a comment

Email Address:


Unless you answer this question, your comment will be classified as spam and will not be posted.
(I'll give you a hint: the answer is “lisp”.)


Remember info?