[mesa-users] 40M_z1m4_to_si_burn test case
jhshiode at berkeley.edu
Wed Oct 24 20:29:30 EDT 2012
Updating to 4607 does cut the # time steps and total time ~ in half (3160 steps in ~ 54 minutes).
The evolution in terms of the HR diagram and is much closer to what I quoted in my email than what appears on mesastar.org, so perhaps we should update that. Regarding that, on a stellar evolution level.. the evolutionary track is kind of odd for this test case (something I only recently appreciated). It does one of these weird roller coaster loops in the HR diagram after core Hydrogen exhaustion and before moving to the red. Does it make sense to do a resolution/mixing study and update the test case to something that looks more physical? (Yes, I'm offering my help with this undertaking, as it fits pretty well into my general research direction at the moment..)
On Oct 24, 2012, at 3:31 PM, Bill Paxton wrote:
> Hi Josh,
> The "expected" time is for my 12 core monster-mac, so you'll need to upgrade your hardware to match that.
> However, even on my machine, the 40M test took about 30 minutes for 4589
> (so your 2 core takes about 3 times longer than my 12 core --- that sounds about right).
> Back at release 4576 it was taking about 10 minutes, so the change happened after that.
> With a recent development version (4607), the time is 12.5 minutes.
> So 4589 has a performance bug. Sorry I missed it.
> 4607 was okay for the full test_suite, so you might want to jump to it and see how you do.
> let us know. if it looks okay, I'll do another release.
> On Oct 24, 2012, at 2:50 PM, Josh Shiode wrote:
>> In the newest installation of MESA 4589, I'm finding significantly different evolution in the star/test_suite/40M_z1m4_to_si_burn test case as compared to the documentation on mesastar.org
>> I'm seeing, to name a few things:
>> Fewer retries ~ 18 total
>> nearly double the number of timesteps to get to the fe54 limit
>> Significantly longer runtime ~ 118 minutes (not 10 as expected)
>> My setup:
>> Mac OS X 10.6.8
>> (tried w/ and w/o SDK)
>> OMP_NUM_THREADS = 2
>> gfortran version 4.6.0 (and 4.7.2)
>> Can anyone else confirm these results? Or are people not having any issue running this test case?
>> Everyone hates slow websites. So do we.
>> Make your web apps faster with AppDynamics
>> Download AppDynamics Lite for free today:
>> mesa-users mailing list
>> mesa-users at lists.sourceforge.net
More information about the Mesa-users