From: mullinwm AT picard DOT ml DOT wpafb DOT af DOT mil ("William M. Mullins")
Subject: numerical errors in f2c -> gcc
20 Feb 1997 07:42:17 -0800
Approved: cygnus DOT gnu-win32 AT cygnus DOT com
Distribution: cygnus
Message-ID: <330C6A28.63A7.cygnus.gnu-win32@picard.ml.wpafb.af.mil>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
X-Mailer: Mozilla 3.01Gold (WinNT; I)
Original-To: gnu-win32 AT cygnus DOT com
Original-Sender: owner-gnu-win32 AT cygnus DOT com

b17.1 on NT4

f2c builds with no problems...installs without a hitch

Numerical codes ported from sun and sgi type machines compile, link and
run without a hitch.

Almost all numerical results are slightly wrong (4-9th sig fig) when
compared to the "test cases".  Occasionally, when numerical second
derivatives are used, the differences are very large (second sig fig).

No fixes or workarounds have been found...very troubling

No, this is not an old pentium machine.  It is a micron P6-200 system,
but all of our NT machines have shown the same problems.

Some kind of answer is really needed or we will have to move back to
the old suns.


mullinwm AT picard DOT ml DOT wpafb DOT af DOT mil
-
For help on using this list, send a message to
"gnu-win32-request AT cygnus DOT com" with one line of text: "help".