Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: notesfiles Path: utzoo!linus!decvax!tektronix!hplabs!hp-pcd!orstcs!richardt From: richardt@orstcs.UUCP (richardt) Newsgroups: net.arch Subject: RE: RISC (Actually 68K densities) Message-ID: <12200008@orstcs.UUCP> Date: Wed, 10-Jul-85 01:48:00 EDT Article-I.D.: orstcs.12200008 Posted: Wed Jul 10 01:48:00 1985 Date-Received: Mon, 15-Jul-85 01:43:17 EDT Organization: Oregon State University - Corvallis, OR Lines: 24 Nf-ID: #N:orstcs:12200008:000:1351 Nf-From: orstcs!richardt Jul 9 21:48:00 1985 As I read, I see a *lot* of comments about how "most 68xxx & 32xxx instructions take 32 bits anyway!" I would like to suggest that a large factor in this may be due to *sloppy compilers*! I have written 68000 code as a hobbyist for about the last three years. I have started writing BASIC interpreters several times and a FORTH-relative once. In all of those cases, I found that the average instruction width # of instructions / # of bits for code = ~18 bits That's a rough estimate; It may be a little high. When you add in data, then things get interesting. A program that uses a lot of predetermined data runs up that average a lot more than you think. When I sat down and started to write op system routines, the average instruction was about 20 bits. (Note: I am using a mathematical average here.) My point is, If most of your programs have larger average instruction widths, maybe you've got a sloppily-written compiler. The average instruction width on a 'thousand needn't be anywhere near 32 bits. So look at your compiler before you go running of screaming 'Give me a RISC.' I'll step down off my soapbox now, and someone else can defend the National chips. I don't have the info (yet.) ---------------------------------------- Is there an assembly-language programmer in the house? orstcs!richardt /* ---------- */