#6558: Be more selective in patching ATLAS on Solaris
-------------------------+--------------------------------------------------
Reporter: drkirkby | Owner: tbd
Type: enhancement | Status: new
Priority: minor | Milestone: sage-4.1.1
Component: solaris | Keywords: solaris atlas sun4v
Reviewer: | Author:
Merged: |
-------------------------+--------------------------------------------------
Trac ticket #6276 was a patch I added to the ATLAS code, as the tuning
process had dumped core on a Sun T5240 machine called ('t2'). This machine
is based on the latest T2+ processor from Sun. The patch was suggested by
Clint Whaley - the main ATLAS developer.
Despite no known problems with ATLAS dumping core on any of the older
systems, or Solaris systems based on the x86 processor, I had applied this
patch to all Solaris systems. However, it is really a hack more than a
patch, as it allows ATLAS to build by returning a reasonable value for a
parameter that the system could not tune properly.
This patch is an improvement, which simply checks if the system is Solaris
and the architecture is 'sun4v' before applying the hack. So only Solaris
systems with Sun T1, T2 or T2+ processors will be patched. Those would
form only a small fraction of the Solaris machines. On the vast majority
of systems in use today, the patch will no longer be applied, so the
tuning process will be more accurate.
A patch for this will be very simple. Although I have not checked it yet,
I believe changing:
{{{
if os.uname()[0] == 'SunOS' :
shutil.copy2('patches/mmsearch-with-temp-Solaris-
fix.c','src/tune/blas/gemm/mmsearch.c')
}}}
to
{{{
if os.uname()[0] == 'SunOS' and os.uname()[4] == 'sun4v':
shutil.copy2('patches/mmsearch-with-temp-Solaris-
fix.c','src/tune/blas/gemm/mmsearch.c')
}}}
will fix this.
I'll attach a patch for review, once its operation has been carefully
checked both on a Sun T5240 (a sun4v machine called 't2') and a Sun Blade
2000 (a sun4u machine called 'kestrel').
--
Ticket URL: <http://trac.sagemath.org/sage_trac/ticket/6558>
Sage <http://sagemath.org/>
Sage: Creating a Viable Open Source Alternative to Magma, Maple, Mathematica,
and MATLAB
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"sage-trac" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/sage-trac?hl=en
-~----------~----~----~----~------~----~------~--~---
