Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!elroy.jpl.nasa.gov!decwrl!sgi!shinobu!odin!soul.esd.sgi.com!jeffs
From: jeffs@soul.esd.sgi.com (Jeff Smith)
Newsgroups: comp.unix.aix
Subject: Re: BASH and GCC
Message-ID: <1991Mar6.190227.14110@odin.corp.sgi.com>
Date: 6 Mar 91 19:02:27 GMT
References: <1274@dkunix9.dk.oracle.com> <2547@sirius.ucs.adelaide.edu.au>
Sender: news@odin.corp.sgi.com (Net News)
Reply-To: jeffs@sgi.com
Distribution: comp.unix.aix
Organization: Silicon Graphics, Mt View, Ca
Lines: 27

|> Well that is rather amusing in its own right. We have been told (in
Australia)
|> that the optimiser for the compiler does not work (hence the OS is
not optimised
|> code), so the beauty of gcc is made even better as it would most probably
|> be the only optimising compiler for the RS6000.

I doubt this.  The RS/6000 architecture replies heavily on optimizing compilers
to keep it's multiple instruction units pipelining well.  I'm also pretty sure
most of the OS is build -O (although I haven't worked on AIX v3.1 since before
GA).

The IBM compiler does do all the major optimizations, and some very fancy stuff
for instruction scheduling.

The major problem with the IBM compilers is that some of these techniques take
alot of RAM, and the compiler process can get quite large.  Hopefully this
will get cleaned up over a few releases.

I think it will be some time before gcc can produce code as good as the IBM
compiler since the instruction scheduling is pretty complex (as compared to
other RISC architectures).

jeffs@sgi.com

"just another ex-compiler guy"
"Look Mom, I didn't even slam AIX"