parker@psuvax1.UUCP (Bruce Parker) (11/29/84)
Can anyone comment on the University of Utah's Portable
Standard Lisp? We've had some poor experiences here with
their USLisp and we fear that we might get the same thing
with PSL -- poor documentation and buggy code.
The only reason for asking is that we are considering getting
Rand's REDUCE, which is written in PSL. We just want to get
an idea of what we're in for.
--
Bruce Parker
Computer Science Department (814) 863-1545
334 Whitmore Lab {allegra|ihnp4}!psuvax1!parker
The Pennsylvania State University parker@penn-state (csnet)
University Park, Pennsylvania 16802 parker@psuvax1 (bitnet)shebs@utah-cs.UUCP (Stanley Shebs) (12/03/84)
As the flamingest person in the PSL group at Utah (:-)), I have been selected to officially respond to the PSL query. Lest I be accused of partiality, let me just say that I used to do a lot of Franz hacking, and have diddled with Zetalisp some also. PSL is *not* buggy. More accurately, it is no more buggy than Franz or Zetalisp (read the franz-friends mailing list sometime!). In over a year of substantial PSL hacking, I have only encountered about two bona-fide bugs. I *have* had problems with various options; PSL provides many more ways to screw yourself than does Franz! But normal users don't do those sorts of things anyway (like messing with *compress, stepping on property lists, and other things that the manual usually warns about...) PSL is *fast*. There were many unhappy people at AAAI-84, when the Gabriel benchmarks showed 68K PSL frequently beating out Symbolics and LMI. VAX PSL was about 1-2 orders of magnitude faster than Franz (depending on the specific benchmark). PSL's weakest showing was in floating-point arithmetic, which was still not too bad. PSL is *portable*. We recently did a demo showing PSL running the same large program on 6 machines - Apollo DN300, Sun, HP 9836, VAX, DEC-20, and Cray. IBM, Univac, and other ports are in the works... PSL is *written in PSL* (about 98-99% to be exact, the remainder being a little LAP code and some machine-dependent routines in C or Pascal or whatever). By contrast, the entire Franz kernel is in C, and legible only to experienced VAX/C hackers (maybe). As a result, improvements in the compiler yields overall gains. The present compiler compares well with C compilers, but we're working on a better one... The PSL distribution *has more tools* (including screen editors and Common Lisp compatibility packages) than Franz does. In summary, PSL is a real live production quality Lisp, and is overall better than Franz for Vaxen (there, now we'll never get anything from UCB again! :-) ) stan shebs (shebs@utah-orion)
holtz@clan.UUCP (Neal Holtz) (12/06/84)
A short note from a happy PSL user -- I use it on a Vax and on Apollos, and am very satisfied. No serious bugs noticed. Haven't used Franz enough to compare, except: - a toy Prolog unifier runs 6 times faster under PSL. - PSL runs on Apollos - Franz doesn't (yet)