On Thu, Jun 20, 2002 at 05:15:44PM -0400, Dave McGuire wrote:

 >   While I agree in principle with your points, I cannot and will not
 > accept the notion that modern GCC "gives back" enough additional
 > functionality to justify its increased size and decreased speed over,
 > say, v1.38, the first release GCC I ran.
 > 
 >   Shouldn't that relationship be a bit closer to linear?  Sure,
 > there's a lot of, say, Pentium optimization crap in there...but is
 > that even being executed when I run it on a VAX?

It's not a lot of "Pentium optimization", and if you think so, you're
not very familiar with the internals of GCC (which, for your own sanity,
is probably a good thing :-)

 >   Good, clean, high-performance programming isn't a standard of
 > "yesterday".

Yes, and the GCC people have realized that their compiler has become
needlessly slow.  They are actively working on the problem.  If you
are interested in seeing the NetBSD system compiler not become glacier-
slow, then I suggest you assist them in this effort.

Face it: There is no reasonable alternative to GCC.  We require an
open-source compiler which is activeley maintained and supports the
range of architectures that we support.

There simple is no other compiler which does what we want.  Therefore,
I would suggest that the CORRECT course of action is to stop bickering
about how awful GCC is and spend some effort improving it.

Some people (including myself) have put an awful lot of effort into GCC
recently getting it in shape to be the NetBSD system compiler so that
for the FIRST TIME SINCE THE NETBSD PROJECT STARTED, the stock FSF GCC
will work out-of-box on NetBSD (which means that we'll be able to put
effort into improving other parts of the OS rather than maintaining a
pile of local changes to the toolchain).

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>