This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Complex arithmetic improvements?
- To: David Edelsohn <dje at watson dot ibm dot com>
- Subject: Re: Complex arithmetic improvements?
- From: Gabriel Dos Reis <Gabriel dot Dos-Reis at dptmaths dot ens-cachan dot fr>
- Date: Sat, 29 Aug 1998 12:11:39 +0200 (MET DST)
- Cc: etzpc at unix dot ccc dot nottingham dot ac dot uk, Martin Kahlert <martin dot kahlert at mchp dot siemens dot de>, N8TM at aol dot com, Sol Foster <colomon at ralf dot org>, egcs at cygnus dot com
- References: <peter@eexpc.eee.nott.ac.uk><Pine.LNX.3.95.980821142325.21246B-100000@eexpc.eee.nott.ac.uk><9808212324.AA32562@marc.watson.ibm.com>
>>>>> «David», David Edelsohn <dje@watson.ibm.com> wrote:
>>>>> Peter Chang writes:
Peter> So it would seem that these changes would be architecture-dependent. This
Peter> doesn't surprise me in the least. I'm certainly not up-to-date on the
Peter> latest fp performances so does anyone have any references or figures for,
Peter> say the clock cycles for fp operations on different architectures?
David> Could the RTX_COSTS macro allow the algorithm to be chosen in a
David> machine-dependent manner? The macro mostly is setup for INTs, though.
David> David
I don't recommand it: the proposed scheme (well known in the numerics
are) is numerically *unstable*.
-- Gaby
"One reason that life is complex is that it has a
real part and imaginary part." -- Andrew Koenig