[mesa-users] retry

[Aaron Dotter]

Hi Jing,

The problem with your implementation is that the routine run1_star, which
run_star.f calls, only checks to see if return = keep_going after
extras_finish_step.  If not, it exits the evolve loop and the run ends.

If you want to force a retry, you ought to use extras_check_model instead of
extras_finish_step.  The former is able to force a retry.  You can see how
all of this is implemented if you read the source code in
mesa/star/public/run_star_support.f, look for evolve_loop in run1_star.

Aaron



On Fri, Jul 1, 2011 at 8:13 PM, Jing Luan <jingluan at caltech.edu> wrote:

> Hello :-)
>
> I force mesa to do a retry at some model but after retrying that model
> mesa stops without any report;
>
> Here is the content added to my extras_finish_step in run_star_extras.f, I
> just set extras_finish_step to be retry when kn is even, and set
> extras_finish_step to be keep_going when kn is odd, where kn is a save
> integer and increases by 1 every time extras_finish_step is called:
>
> integer, save :: kn
>
>         write (*,*) 'call extras_finish_step'
>         if (s% model_number == 3582) kn=0 ! model 3582 is the first model
> because I start the run with model 3581
>
>         if (mod(kn,2)==0) then
>           write (*,*) 'retry,kn',kn
>           kn=kn+1
>           extras_finish_step = retry
>         else
>           write (*,*) 'keep_going,kn',kn
>           kn=kn+1
>           extras_finish_step = keep_going
>         end if
>
>
> I tell mesa to print out 'call extras_finish_step' if extras_finish_step
> is executed. I find that extras_finish_step is executed after the first
> model with model number=3582, but NOT executed after the second model with
> model number=3582; That is to say mesa terminates after the retried model
> 3582 is calculated and before extras_finish_step is called.
>
> Because mesa does not report any information when it terminates, I really
> have no idea where to trace the problem. Any help or suggestion is
> appreciated!
>
> Thank you very much :-)
>
> Sincerely,
> Jing
>
>
>
>   Hi,
>>
>>  Mesa/star doesn't retrain complete information about previous states,
>>  so you should save what you need yourself in your run_star_extras.
>>
>>  Good luck,
>>  Bill
>>
>>
>>
>>
>>  On Jun 30, 2011, at 9:17 PM, jingluan at caltech.edu wrote:
>>
>>   Dear Bill,
>>>
>>>  May I ask a question about the pointer s in mesa? Because I am not
>>>  familiar with pointer in fortran and failed to find direct-relevant
>>>  example by google. Thank you very much :-)
>>>
>>>  For example, the current model is model2, the information contained in
>>>  its
>>>  s is different with the previous s in model1. After model2 is
>>>  calculated,
>>>  mesa goes on to extras_finish_step. In this extras_finish_step, is there
>>>  some way to use the information in the previous s in model1?
>>>
>>>  I define s_old like this : type(star_info), pointer, save :: s_old
>>>  and then set: s_old=s in model1's extras_finish_step.But mesa reports
>>>  segmentation err. I think it should be because s and s_old are both
>>>  pointers, but did not yet find the right way to refer the previous s.
>>>
>>>  I need to do it because I need some kind of mid-point integration, where
>>>  the mid-point is some linear combination of s and s_old.
>>>
>>>  Thank you very much :-)
>>>
>>>  Sincerely,
>>>  Jing
>>>
>>>
>>
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20110701/fbf95d05/attachment.html>