--- Begin Message ---
Guys,
I have released 3.7.36. Please give it a testing, as we are now ready to
start iterating this guy to stable. I have started the dev2stable page at:
http://www.cs.utsa.edu/~whaley/atlas/dev2stab_38.html
As you can see, I've run the full tester on a few archs with a BETA of 36.
36 is better at using the windows compilers, but I've decided to support
only gcc for arch defaults. The main reason is that when I use icc and
ask for any kind of respect for IEEE, icc runs **muuuuch* slower than gcc.
The other factor is that I don't have a contact at intel, so my icc license
is always expiring after 30 days, and it's not clear I can use the free
license even if I want to install every 30 days, so I'm just not going to
support. MSVC++ can build the kernel, but can't handle the search (some of
the routines ATLAS tries exceed its ability to do jumps). However, gcc is
as fast as anything (when using safe flags), so it's what I'm supporting
for arch defaults. The prototype of the new errata file contains instructions
on making the library with MSVC++/ifort interfaces, so that should provide
what people really need:
http://www.cs.utsa.edu/~whaley/atlas/errata3.8.0.html#WinComp
I have not tried to build a dynamic lib on windows, and frankly doubt that
works at present. Probably I'll look at that for the first patch of the
stable after release . . .
The big news for this release is that, at least on the Core2Duo and Hammer,
ATLAS now passes the LAPACK testers roughly as well as the reference BLAS.
I had to fix a couple of ATLAS errors for this to happen, here's the ChangeLog:
ATLAS 3.7.36 released 08/09/07, changes from 3.7.35:
* Some smoothing ops allowing easier use of windows compilers
* Fixed error in mmsearch causing PPC searches to die wt latency problems
* Fixed error where wrong flags caused snrm2 to be incorrect on Core2Duo
* Changed GER to heavily favor applying alpha to X, in order to keep LAPACK
from barfing up a lung on those tiny matrix test cases
* Fixed error in complex syreflect causing wrong answers in [c,z]gemm when
gemm is used to do a syrk
Since we are now in stable mode, please let me know of any results you get
so they can be added to the dev2stable page. If you have a platform that
is not supported with architectural defaults, let me know ASAP, so we can
add them (assuming I can get access to a machine). I still plan to add
G4 defaults, but I'm waiting to see if I can get access to a Linux/G4 with
gcc 4.2.
Regards,
Clint
**************************************************************************
** R. Clint Whaley, PhD ** Assist Prof, UTSA ** www.cs.utsa.edu/~whaley **
**************************************************************************
-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
Still grepping through log files to find problems? Stop.
Now Search log events and configuration files using AJAX and a browser.
Download your FREE copy of Splunk now >> http://get.splunk.com/
_______________________________________________
Math-atlas-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/math-atlas-devel
--- End Message ---
