This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: gcc & g4 processor?
- To: David Edelsohn <dje at watson dot ibm dot com>
- Subject: Re: gcc & g4 processor?
- From: adam morley <axm135 at po dot cwru dot edu>
- Date: Mon, 7 May 2001 12:59:44 -0400
- Cc: gcc at gcc dot gnu dot org
- References: <axm135@po.cwru.edu> <200105071651.MAA22172@makai.watson.ibm.com>
On Mon, May 07, 2001 at 12:51:58PM -0400, David Edelsohn wrote:
> >>>>> adam morley writes:
>
> adam> so i was looking @ apples website, specifically what compiler they
> include with the now shipping Mac OS X. its gcc 2.95.2. it has a special
> option -faltivec to compile with altivec enhancements.
> adam> is this included in regular download of gcc from one of the mirrors?
> ie, can i put it on linux ppc, openbsd, whatever, compile it and get that
> action?
>
> Motorola has not assigned the changes to the FSF, so they cannot
> be included in the FSF sources. In the interim, the GCC infrastructure
> for supporting SIMD instructions has been improved by an Intel contract
> with Red Hat to support MMX/SSE instructions. Apple has stated that it is
> investigating rewriting the AltiVec support from scratch without any
> Motorola IP using the new GCC SIMD infrastructure.
interesting. this brings up another point: will there be support for things like VIS from sun? i know sun already has a compiler which does a very very good job of optimizing, so i doubt if sun would build support into gcc.
is there a timetable for apple's altivec support?
>
> adam> plus, how good is PPC optimization under gcc?
>
> Good, not great. It depends a lot on the type of algorithm.
> Informal performance comparisons seem to show that GCC is anywhere from
> 30% worse to 5% better than IBM's Visual Age compiler on specific
> workloads.
and im guessing this doesn't run under openbsd.
>
> David
--
thanks
adam
any and all ideas herein are the sole property of the author, with no implied warranties or guarantees. unless its somebody else's already.