Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!uunet!ficc!peter From: peter@ficc.ferranti.com (Peter da Silva) Newsgroups: comp.arch Subject: Re: Benchmarking interactive software (Re: It looks like he's at it again!) Message-ID: Date: 1 Aug 90 11:19:56 GMT References: <1990Jul12.012730.4248@Stardent.COM> <64044@sgi.sgi.com> <13392@cbmvax.commodore.com> <_YV4Y9A@ggpc2.ferranti.com> <784@sibyl.eleceng.ua.OZ> <13533@cbmvax.commodore.com> Reply-To: peter@ficc.ferranti.com (Peter da Silva) Organization: Xenix Support, FICC Lines: 20 In article <13533@cbmvax.commodore.com> jesup@cbmvax (Randell Jesup) writes: > Don't forget the other reason to write in assembler, something that > many Unix programmers seem to have forgotten (amusing, since it was once > one of the advantages of Unix): space efficiency. If you *really* want space efficiency the obvious thing to do is write in Forth, or some other threaded language. Writing in assembler to save memory is like writing in COBOL to improve readability. That way you retain quite a high level of portability, you can still implement the critical code in machine dependent fashion, and you save a *lot* more space than an assembly implementation. Alternatively: write the critical code in assembly, include a small Scheme or other tiny interpreted language, and put most of the code in that. It'd still be fast enough for humans and you'd save simply gobs of RAM. Assembly should be the last resort, not the first. (slams on the size of SVR4 deleted... funny, I was just complaining about that in comp.sys.amiga.tech) -- Peter da Silva. `-_-' +1 713 274 5180. 'U`