[mesa-users] 1M_pre_ms_to_wd
Francis Timmes
fxt44 at mac.com
Mon Oct 19 13:37:40 EDT 2015
thanks bill.
> mesa-web is a great resource but not a magic solution that suddenly makes all difficult things easy!
hear, hear!
with mesa-web we are providing a free and easy-to-use infrastructure for educational purposes.
the human and computational resources are not free on our side, imposing a limit on
what can be expected to run to the desired completion point within the time window.
fxt
> On Oct 19, 2015, at 10:11 AM, Bill Paxton <paxton at kitp.ucsb.edu> wrote:
>
> Hi Frank,
>
> mesa-web is awesome! a great new resource -- congratulations and many thanks to Carl Fields and you for putting it together.
>
> but however many cores you add for it to use, there will of course still be ways for users to specify jobs that will take too long. For example, Noah's inlist had winds turned off, so the model was going to be stuck at the tip of the AGB doing computationally expensive flashes until it burned off the entire H/He envelope to make a 1M WD the hard way. Brutal, but it should eventually finish if you are willing to burn a lot of CPU hours burning H and He!
>
> I know you and Carl are clear on this, but I wanted to remind the community that mesa-web is a great resource but not a magic solution that suddenly makes all difficult things easy! It is wonderful to have mesa-web to try out new ideas, but everyone needs to know how easy it is to turn a reasonable run into a very expensive one (e.g., turn off winds in the 1M_pre_ms_to_wd test!). So there will be times when mesa-web comes back and says "this is taking too long. i give up." That isn't a fault of mesa-web (or mesa/star!) --- it is the nature of the beast that some problems are just too expensive to solve with current hardware or are just beyond the scope of the code.
>
> Cheers,
> Bill
>
>
>
> On Oct 19, 2015, at 9:50 AM, Francis Timmes wrote:
>
>> thanks willie. i agree the 1M_pre_ms_to_wd case can be run in under
>> an hour with sufficient horsepower.
>>
>> we’re mid-steam on contracting additional cores for mesa-web.asu.edu
>> as jobs currently run on just one core. all it takes is $ ;)
>>
>> fxt
>>
>>
>>
>>
>>> On Oct 19, 2015, at 2:26 AM, Willie Strickland <cwskas at gmail.com> wrote:
>>>
>>> I have run the test case to which Bill refers (1M_pre_ms_to_wd) many times to various stopping points including all the way to WD.
>>>
>>> I primarily use this test and the work directory (15M star on which you built your inlist) anytime I install a new version of MacOS or of MESA. I compare the results to previous runs to make sure the final profile and final model are in agreement and I also pay attention to the time taken to run.
>>>
>>> I cannot attest to the astrophysical accuracy of my results, but would be happy to share specifics with the mailing list for evaluation if that would be helpful.
>>>
>>> After installing El Capitan, I started with a pre-main sequence model and ran to WD in 58 minutes and 29 seconds on my 4 year old, 2 core i7 MacBook Pro specifying 4 threads. Attached is the Grid8 plot from the final model.
>>>
>>> My inlist does have some modifications which I have found helpful in running my tests, but may not be as accurate.
>>>
>>> Willie
>>>
>>>
>>> <grid8_000170.png>
>>>
>>>
>>>> On Oct 18, 2015, at 3:15 PM, Bill Paxton <paxton at kitp.ucsb.edu> wrote:
>>>>
>>>> Hi Noah,
>>>>
>>>> There is a test case that runs Z=0.02, 1Msun from pre-main-sequence to cool white dwarf. You should run that test to get an idea of where the runtime is spent --- e.g., climbing the RGB. Then let us know what you find. Perhaps you'll find ways to save models along the way so that students can use them to run different interesting parts of the evolution (e.g., he core flash) without the computational cost of running the entire evolution. I think that's probably a better approach than trying to make a "brain dead" version of mesa that gives up lots of accuracy in an attempt to run fast. (for the fast but loose approach, you might want to run my old "EZ" code -- but I don't recommend it.)
>>>>
>>>> bill
>>>>
>>>> btw: it is okay to say "thank you" when I volunteer to look at something for you. ;D
>>>>
>>>>
>>>>
>>>>
>>>> On Oct 18, 2015, at 1:00 PM, Glasman, Noah B wrote:
>>>>
>>>>> Here is my inlist_project for that run. I am just trying to run the entire life of a 1Msun star with Z = 0.001. Does anyone know if MESA works on OSX El Capitan yet?
>>>>>
>>>>> ! inlist to evolve a 15 solar mass star
>>>>>
>>>>> ! For the sake of future readers of this file (yourself included),
>>>>> ! ONLY include the controls you are actually using. DO NOT include
>>>>> ! all of the other controls that simply have their default values.
>>>>>
>>>>> &star_job
>>>>>
>>>>> ! begin with a pre-main sequence model
>>>>> create_pre_main_sequence_model = .true.
>>>>>
>>>>> ! save a model at the end of the run
>>>>> save_model_when_terminate = .false.
>>>>> save_model_filename = '15M_at_TAMS.mod'
>>>>>
>>>>> ! display on-screen plots
>>>>> pgstar_flag = .true.
>>>>>
>>>>> / !end of star_job namelist
>>>>>
>>>>>
>>>>> &controls
>>>>>
>>>>> ! starting specifications
>>>>> initial_mass = 1 ! in Msun units
>>>>> initial_Z = 0.001
>>>>> !max_model_number = -1
>>>>>
>>>>> ! stop when the star nears ZAMS (Lnuc/L > 0.99)
>>>>> !Lnuc_div_L_zams_limit = 0.99d0
>>>>> !stop_near_zams = .false.
>>>>>
>>>>> ! stop when the center mass fraction of h1 drops below this limit
>>>>> !xa_central_lower_limit_species(1) = 'h1'
>>>>> !xa_central_lower_limit(1) = 1d-3
>>>>>
>>>>> / ! end of controls namelist
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> From: Bill Paxton <paxton at kitp.ucsb.edu>
>>>>> Sent: Sunday, October 18, 2015 12:23 PM
>>>>> To: mesa-users at lists.sourceforge.net group group
>>>>> Subject: [mesa-users] MESA in education -- where's the website for this?
>>>>>
>>>>> Dear MESA-USERS who use mesa for teaching,
>>>>>
>>>>> How do you manage with the issues Noah raises in the attached email?
>>>>>
>>>>> Have you found problems that are good for combining academic interest and short runtimes?
>>>>>
>>>>> Wouldn't it be great if there were a website/forum/social-media-something-or-other specifically to serve "mesa in education"?
>>>>>
>>>>> -Bill
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On Oct 17, 2015, at 6:05 PM, Glasman, Noah B wrote:
>>>>>
>>>>>> I am wondering if there is a way to speed up how fast MESA runs on my computer? At the moment it takes an hour plus to run it for a 1Msun star with Z = 0.001 and after an hour it's only at 6.349e9 years. Is there something I can do to speed up the program or maybe allocate how much ram it uses? At the pace it is going it is almost impractical to use for any research or classwork for my Astronomy class. I have a 2012 Macbook Pro 13 inch, 4gb DDR3 RAM, 2.5 GHz Intel Core i5 cpu, with OSX 10.10.5. I am running MESA-r7624, with Version 20150908 of the MESA-sdk.
>>>>>>
>>>>>>
>>>>>> Sincerely,
>>>>>> Noah Glasman
>>>>>> ------------------------------------------------------------------------------
>>>>>> _______________________________________________
>>>>>> mesa-users mailing list
>>>>>> mesa-users at lists.sourceforge.net
>>>>>> https://lists.sourceforge.net/lists/listinfo/mesa-users
>>>>>
>>>>>
>>>>
>>>> ------------------------------------------------------------------------------
>>>> _______________________________________________
>>>> mesa-users mailing list
>>>> mesa-users at lists.sourceforge.net
>>>> https://lists.sourceforge.net/lists/listinfo/mesa-users
>>>
>>> ------------------------------------------------------------------------------
>>> _______________________________________________
>>> mesa-users mailing list
>>> mesa-users at lists.sourceforge.net
>>> https://lists.sourceforge.net/lists/listinfo/mesa-users
>>
>> ------------------------------------------------------------------------------
>> _______________________________________________
>> mesa-users mailing list
>> mesa-users at lists.sourceforge.net
>> https://lists.sourceforge.net/lists/listinfo/mesa-users
>
More information about the Mesa-users
mailing list