From: Toon Moene <toon@moene.indiv.nluug.nl>
To: egcs@cygnus.com
Subject: Re: FWD: FLOATING-POINT CONSISTENCY, -FFLOAT-STORE, AND X86
Date: Tue, 15 Dec 1998 12:24:00 -0000 [thread overview]
Message-ID: <3676C529.FE8214F@moene.indiv.nluug.nl> (raw)
> Using 80-bit spills is a quick approximation to
> extending the FP stack into memory, and it should give > some of the benefit with very little (hopefully)
> hassle. Of course, 80 bits is wider than the normal
> spill, so it eats more memory bandwidth, cache space,
> etc. Anyone who's that concerned will bend over
> backwards to avoid spills anyways.
Am I the only one - apart from Harvey J. Stein and Tim Prince - who
finds this whole discussion unreal ? Surely, 80 bit temporaries might
seem a neat hack to a numerical analyst like Dr. Kahan, but the ordinary
computational physicist or chemist knows better than to choose "poorly
conditioned" algorithms.
My main concern is that there is a grid spacing that will render the
basic equation of geostrophy badly approximated in 32-bit arithmetic:
1 dp
--- -- = f v
rho dx
p pressure
rho mass of air per unit volume (1 kg / m^3)
x distance
f Coriolis parameter (10^-4 s^-1)
v wind speed
You can do the math (p ~ 10^5 kg m^-1 s^-2, v ~ 10 m/s, what dx will
make dp < 10^-3 p ?)
If that's the case we have to rethink our finite difference code for the
first time in 13 years and use a trick like subtracting a basic state
from the equations - big deal.
The last thing I need is to have egcs slowed down to a crawl by having
it spill unaligned 80-bit temporaries for something that shouldn't be
larger than 32 bits in the first place.
Please make this and other "accuracy" options a "-pedantic-numerics"
one.
--
Toon Moene (toon@moene.indiv.nluug.nl)
Saturnushof 14, 3738 XG Maartensdijk, The Netherlands
Phone: +31 346 214290; Fax: +31 346 214286
g77 Support: fortran@gnu.org; egcs: egcs-bugs@cygnus.com
next reply other threads:[~1998-12-15 12:24 UTC|newest]
Thread overview: 65+ messages / expand[flat|nested] mbox.gz Atom feed top
1998-12-15 12:24 Toon Moene [this message]
1998-12-15 12:55 ` Joe Buck
1998-12-15 15:05 ` Edward Jason Riedy
1998-12-16 10:05 ` Craig Burley
-- strict thread matches above, loose matches on Subject: below --
1998-12-17 14:38 Toon Moene
1998-12-17 15:30 ` Harvey J. Stein
1998-12-18 1:54 ` Toon Moene
1998-12-18 3:05 ` Harvey J. Stein
1998-12-18 9:01 ` Toon Moene
1998-12-18 15:59 ` Richard Henderson
1998-12-18 13:26 ` Marc Lehmann
1998-12-18 12:50 ` Dave Love
1998-12-17 14:37 tprince
1998-12-17 15:15 ` Stephen L Moshier
1998-12-17 11:27 Brad Lucier
1998-12-17 14:51 ` Marc Lehmann
1998-12-19 0:17 ` Craig Burley
1998-12-19 6:42 ` Emil Hallin
1998-12-19 14:26 ` Dave Love
1998-12-16 13:52 Toon Moene
1998-12-17 10:06 ` Craig Burley
1998-12-17 12:16 ` Harvey J. Stein
1998-12-19 0:29 ` Craig Burley
1998-12-17 11:20 ` Dave Love
1998-12-15 12:10 Geert Bosch
1998-12-15 13:09 ` Jeffrey A Law
1998-12-15 1:45 Geert Bosch
1998-12-15 3:34 ` Harvey J. Stein
1998-12-16 10:36 ` Craig Burley
1998-12-16 12:47 ` Harvey J. Stein
1998-12-17 10:22 ` Craig Burley
1998-12-17 14:54 ` Marc Lehmann
1998-12-19 0:27 ` Craig Burley
1998-12-19 5:06 ` Stephen L Moshier
1998-12-15 6:43 ` Stephen L Moshier
1998-12-16 10:14 ` Craig Burley
1998-12-15 9:29 ` Joe Buck
1998-12-15 10:14 ` Jeffrey A Law
1998-12-16 8:32 ` Sylvain Pion
1998-12-16 9:20 ` Craig Burley
1998-12-13 18:23 Stephen L Moshier
1998-12-14 1:52 ` Harvey J. Stein
1998-12-14 14:56 ` Edward Jason Riedy
1998-12-14 17:20 ` Joe Buck
1998-12-14 18:51 ` Edward Jason Riedy
1998-12-14 21:54 ` Craig Burley
1998-12-15 14:31 ` Edward Jason Riedy
1998-12-15 17:11 ` Jamie Lokier
1998-12-16 0:26 ` Harvey J. Stein
1998-12-16 9:33 ` Craig Burley
1998-12-16 12:18 ` Harvey J. Stein
1998-12-16 9:38 ` Craig Burley
1998-12-16 12:25 ` Marc Lehmann
1998-12-16 12:50 ` Tim Hollebeek
1998-12-16 13:04 ` Harvey J. Stein
1998-12-16 14:01 ` Marc Lehmann
1998-12-17 11:26 ` Dave Love
1998-12-17 15:06 ` Marc Lehmann
1998-12-18 12:50 ` Dave Love
1998-12-19 14:09 ` Marc Lehmann
1998-12-20 11:28 ` Dave Love
1998-12-20 11:24 ` Dave Love
1998-12-16 23:11 ` Joern Rennecke
1998-12-17 6:07 ` Jamie Lokier
1998-12-14 22:54 ` Craig Burley
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=3676C529.FE8214F@moene.indiv.nluug.nl \
--to=toon@moene.indiv.nluug.nl \
--cc=egcs@cygnus.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).