I now work with lisp and/or scheme most of the time
29 Sep 2006
Update to below. ECL is running on OpenBSD/Sparc64. The key was to install the Boehm garbage collector from the OpenBSD ports collection, and then use the --enable-boehm=system flag with configure. This also worked on my iBook when I just edited the ports makefile such that it installed on powerpc. It seems ok.
I also switched over to a Ultra 10 for now. The SPARCstation 4/110 was just too slow.
21 Sep 2006
Freely downloadable books from Niklaus Wirth
Got ECL to build on my Sparc64 systems. On first pass it was slower (by about 3x) then CLISP, which is bad, especially on the main SPARCstation 4. Things got better when I switched from CLOS classes to structures (ie defstruct). I think, for this use, I like defstruct better. They do print nicer. Then I changed the bit that reads in the rest of the data as an array to just skip the data (because the pretty printer was making a mess of this) and the time went down by a factor of 10 (from 643 secs to 65 secs). I had already halved this time by switching to structures but I was quite impressed. Ergo, remember that the combo of (make-array) followed by (read-sequence) isn't fast when you do it 10000 times. On the MicroSparc II/110 it takes 0.06 secs to do that sequence.
The overall good news is that everything runs on all of my systems.
12 Sep 2006
StrongTalk with open source for the VM
Free 64-bit single-core OpenSPARC processor core emerges
6 Sep 2006
Panic time. My proposal was accepted by EuroBSDCon so that means I have to have it working well in time to write the paper by 15 Oct. This means, for now, I've turned off the OpenWRT systems and am back to using lisp.
Because both the home SparcStation 20 and the carry around iBook are now running OpenBSD I had to switch from ECL. ECL works fine, it seems, on OpenBSD/i386, but I have some adversion to x86 systems. Ergo we are now running CLISP. CLISP is quite nice.
So, why back to lisp rather than scheme? My problem is that I've worked with lisp a bit in the past, and, combined with good docs it turns out I can work very quickly in Lisp. I wrote a fairly complete libpcap reader (reading the bits and bytes directly) starting in the evening of 31 Aug and finishing on 6 Sep in a few hours a day. I was quite pleased. Of course one could do it in scheme as well, but, as shown above I don't have spare time.
So I now have a month and 10 days or so to get pckSwarms working, produce results, and write a reasonable paper. Guess I won't have much spare time.
I've always wanted to go to Siggraph. Leo went and has a nice report
In particular there was some interesting city generation work done by Pascal Mueller.
Last modified: Tue Sep 12 13:18:54 MEST 2006