Langou, Julien wrote:
I agree with Sven.
We have for a long time waited to solve the problem and enumerated the
solutions. I think it's time to move on and release a full thread safe
Sven's solution is fine with me. It's a starting point (does not work on non
IEEE machine). I am not against using FORTRAN 90 INTRINSIC as is done in
This latter solution would impose on the user to have a F90 compiler.
I can't believe that many people are running on non-IEEE machinees.
And to cater for them, all you need to do is decide at install time
whether you are IEEE or not (by calling the current xLAMCH).
If you are, use the hard coded IEEE xLAMCH. Otherwise, use the current
xLAMCH, and no-one is worse off.
From: lapack-bounces@Domain.Removed on behalf of Sven Hammarling
Sent: Sat 9/1/2007 1:45 PM
Cc: Mick Pont
Subject: [Lapack] xLAMCH
Following the helpful messages from Jeremy Barnes, I would like to press
the request, that I have made in the past, that we supply an IEEE
version of xLAMCH. My colleague, Mick Pont, has pointed out that we
have already moved xLAMCH into the INSTALL directory, and so we could
test for an IEEE machine at install time using the current xLAMCH and
move a hard coded version into SRC if the machine parameters are IEEE.
Lapack mailing list
This e-mail has been scanned for all viruses by Star.
The Numerical Algorithms Group Ltd is a company registered in England
and Wales with company number 1249803. The registered office is:
Wilkinson House, Jordan Hill Road, Oxford OX2 8DR, United Kingdom.
This e-mail has been scanned for all viruses by Star. The service is
powered by MessageLabs.