This is the mail archive of the
mailing list for the GCC project.
Re: gcc compile-time performance
- From: dewar at gnat dot com (Robert Dewar)
- To: Richard dot Earnshaw at arm dot com, dewar at gnat dot com
- Cc: gcc at gcc dot gnu dot org
- Date: Mon, 20 May 2002 10:41:01 -0400 (EDT)
- Subject: Re: gcc compile-time performance
<<The only *possible* reason I can think of (note: possible, not good) is
that it would help to ensure that cross-compilers would behave the
identically regardless of native FP format/accuracy etc. A simple
fixed-point library would fix that.
That's a good point, probably decisive. It is indeed unfortunate to have
different cross-compilers generating different code, or for that matter
a cross compiler behaving indentically to the native compiler (I know
that at least Richard Kenner counts on that, since he prefers to debug
on a cross-compiler wherever possible to avoid jumping around on
multiple systems, and that makes good sense).
So simple scaled integer arithmetic is indeed preferable.