Re: Gempak build with PGI compilers

Jeff Wolfe wrote:

Jeff,

GEMPAK was compiled with straight -O on both boxes.  Both were running Solaris
8 and patched.  I am not sure about compiler patches (I don't administer
the 6500) but it is FORTE 6 (just plain 6, not the updates).  The 6500 has
like 10G of memory..over 8G free..the U10 512MB.  I was comparing the time it 
took
to run gdradr which compites every NEXRAD site.  It took 15 seconds longer
on the 6500 AND the data was local there while it was NFS mounted on the U10.


> In message <3B886369.5E11697D@xxxxxxxxxxxx>, Robert Mullenax writes:
> > s
> >
> > Yes this is what we saw.  Maybe you can answer another question as I am new 
> > to multi-
> > CPU
> > machines.  Even with the non-SMP option compile a single GEMPAK script will 
> > run slowe
> > r on the E6500
> > than it will on an ordinary Ultra 10 (440Mhz vs. 400Mhz).  If you look at 
> > top while t
> > he script is
> > running on the E6500 you will see it "hopping" between CPUs..which of 
> > course slows it
> >  down.
> > Is there a way to prevent this?
>
> While the memory bus has a little more latentcy on the 6500 than the single
> CPU boxes, there is no way running on different CPUs should cause a noticeable
> performance impact on somthing like GEMPAK.
>
> What compiler options did you use to build gempack?
> Are you up-to-date with FORTE and Solaris patches?
> Are you using the same binary on the U10 and the 6500?
> Are the Solaris versions the same on both machines?
> How's the memory usage on both machines?
>
> -JEff


  • 2001 messages navigation, sorted by:
    1. Thread
    2. Subject
    3. Author
    4. Date
    5. ↑ Table Of Contents
  • Search the ldm-users archives: