home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!think.com!bromley
- From: bromley@think.com (Mark Bromley)
- Newsgroups: comp.programming
- Subject: Re: first-year programming languages
- Date: 20 Nov 1992 15:33:06 GMT
- Organization: Thinking Machines Corporation, Cambridge MA, USA
- Lines: 20
- Distribution: na
- Message-ID: <1ej0fiINNapi@early-bird.think.com>
- References: <aelman.721693402@Xenon.Stanford.EDU> <dnebing-141192140340@m64-143.bgsu.edu> <1992Nov19.000628.18932@linus.mitre.org>
- NNTP-Posting-Host: luna.think.com
-
- In article <1992Nov19.000628.18932@linus.mitre.org> crawford@boole.mitre.org (Randy Crawford) writes:
- >I'm not so ready to predict that computers in 2020 will so closely resemble
- >today's crop of assembly machines. After all, hasn't assembly programming
- >pretty much disappeared from most software development in the past decade?
- >Ever try to outguess optimizing compilers on architectures like SPARC or
- >PA-RISC? Think you could better their scheduling/pipelining/register-window
- >use by programming in assembly? If so, you're a far better assembly programmer
- >than I.
-
- I wouldn't say that assembly has disappeared in the last decade. It is
- rare to find large programes totally implemented in assembler, but some
- speed critical sections of code are still implemented in assembler.
- Scheduling, pipelining, caches etc are not mysterious. Given any
- reasonably short sequence of code, say on the order of a page, I would
- expect humans to be better than compilers when they are willing to expend
- the effort.
-
- From first hand experience I can tell you that the SPARC compilers are not
- exactly omniscient.
-
-