Path: utzoo!utgpu!water!watmath!clyde!rutgers!sri-spam!ames!oliveb!sun!gorodish!guy
From: guy@gorodish.Sun.COM (Guy Harris)
Newsgroups: comp.arch
Subject: Re: Performance increase - a suggestion
Message-ID: <40627@sun.uucp>
Date: 1 Feb 88 21:19:01 GMT
References: <235@unicom.UUCP> <28200089@ccvaxa> <3127@phri.UUCP> <408@micropen>
Sender: news@sun.uucp
Lines: 17

> Alternately, perhaps hardware/arch. people can convince C compiler writers
> that (as I often find) single precision is more than sufficient for all but
> the most numerically nasty problems and should be used for float arithmetic
> unless code directed otherwise.

They already have: see the latest ANSI C draft standard.  The reason why double
precision was specified in K&R was, in part (as was, I think, stated here
earlier) that it was easier to implement on the PDP-11 that way, as the
PDP-11's floating point instructions (FP11-style) didn't come in
single-precision and double-precision flavors; instead, there was a mode bit
that you had to set to single-precision or double-precision mode with special
instructions.  (It was certainly *possible* to support both single-precision
and double-precision arithmetic in compiled code - FORTRAN-IV Plus offers an
existence proof - but I suspect it was deemed unnecessary in the context of C.)
	Guy Harris
	{ihnp4, decvax, seismo, decwrl, ...}!sun!guy
	guy@sun.com