[FLASH-USERS] Maclaurin test problem
Philip Chang
chang65 at uwm.edu
Thu Jul 18 15:16:33 EDT 2013
Hi Petros,
Thanks for your suggestions. It worked. I diagnose the problem and it is due to the -O3 optimization flag in my compilation. -O2 works fine. The compiler used was ifort version 12.1.3.
I'm putting this on the list so that others may benefit in the future.
Thanks,
Phil
On Jul 18, 2013, at 11:48 AM, Petros Tzeferacos wrote:
> Hi Philip,
>
> I apologize, I was not explicit enough. What I was aiming for was to resetup without setting maxblocks=400. Another thing that can be at fault here is the heavy optimization done. Could you resetup with this exact line
>
> ./setup Maclaurin -auto -3d +newMpole -site=scinetgpc
>
> and run with 4 MPI ranks? If it still crashes can you try with the following line?
>
> ./setup Maclaurin -auto -3d +newMpole -site=scinetgpc -debug
>
> Thanks and sorry for the mixup.
> -- Petros
>
>
>
> On Thu, Jul 18, 2013 at 12:14 PM, Philip Chang <chang65 at uwm.edu> wrote:
> Hi Petros,
>
> I have compiled it appropriately and attached the log file. It looks the same.
>
> Cheers,
> Phil
>
>
> On Jul 18, 2013, at 8:50 AM, Petros Tzeferacos wrote:
>
>> Hey Philip, thanks for the log file. Just so you know, when FLASH performs multiple runs in the same directory, it appends the log of each run in the same file making it harder to go through. It is good practice to get rid/store elsewhere/rename the .log file.
>>
>> Could you try out a quick test on your machine for me?
>>
>> Compile with ./setup Maclaurin -auto -3d +newMpole -site=scinetgpc, and run with 4 MPI ranks,
>>
>> $ > mpirun -np 4 ./flash4
>>
>> This will be exactly the run I did on my laptop yesterday and completed without problems. Should you get an error on the number of blocks, retry to run with 8 MPI ranks. Do not use a different flash.par file so we are on the same page. Thanks!
>>
>> Best,
>> -- Petros
>>
>>
>>
>>
>>
>> On Wed, Jul 17, 2013 at 7:22 PM, Philip Chang <chang65 at uwm.edu> wrote:
>> Yes,
>>
>> I am using FLASH 4.0.1 and attached is my maclaurin.log file. I did not use the unsplit solver.
>>
>> I can recompile with the unsplit hydro solver, smaller initial timestep and different Riemann solver. Which one did you use?
>>
>> Cheers,
>> Phil
>>
>>
>>
>>
>>
>> On Jul 17, 2013, at 4:22 PM, Petros Tzeferacos wrote:
>>
>>> Thanks for the info Phil. Two more pieces of information: which version of FLASH 4 are you using and, if possible, could you share your maclaurin.log file?
>>>
>>> I was able to run the problem with unsplit hydro albeit I had to use a more diffusive Riemann solver and decrease the initial timestep which was set too high. I have not yet seen the error you are reporting on my end. I'll look into it but the log file will tell us more.
>>>
>>> Best,
>>> -- Petros
>>>
>>>
>>> On Wed, Jul 17, 2013 at 5:39 PM, Philip Chang <chang65 at uwm.edu> wrote:
>>> Hi Petros,
>>>
>>> The test that I ran was the straight up Maclaurin test problem:
>>>
>>> ./setup Maclaurin -auto -3d +uhd +newMpole -site=scinetgpc -maxblocks=400
>>>
>>> on execution, it creates the level 1 block and then proceeds to fail.
>>>
>>> The last few line are
>>> [GRID amr_refine_derefine] min blks 1 max blks 1 tot blks 1
>>> [GRID amr_refine_derefine] min leaf blks 1 max leaf blks 1 tot leaf blks
>>> 1
>>> [ 07-17-2013 14:13:06.043 ] [GRID amr_refine_derefine]: refinement complete
>>> [ 07-17-2013 14:13:06.044 ] [GRID gr_expandDomain]: iteration=1, create level=2
>>> [ 07-17-2013 14:13:06.047 ] [DRIVER_ABORT]: Driver_abort() called by PE 0
>>> [ 07-17-2013 14:13:06.047 ] abort_message: [Eos] unrecognized eos_type.
>>>
>>> It appears that the call the Eos is failing, which at some level is unclear why this is happening as the internal energy, pressure and density are populated in Simulation_initBlock.F90
>>>
>>> Cheers,
>>> Phil
>>>
>>>
>>> On Jul 17, 2013, at 1:10 PM, Petros Tzeferacos wrote:
>>>
>>>> With all that being said, Phil could you let us know more about the problem you are having? That is: setup line, any modifications and such. I just ran the test locally with -auto -3d and it concluded smoothly.
>>>>
>>>> Best,
>>>> -- Petros
>>>>
>>>>
>>>> On Wed, Jul 17, 2013 at 3:05 PM, Norbert Flocke <flocke at flash.uchicago.edu> wrote:
>>>> Hello Philip,
>>>>
>>>> The MacLaurin test problem as it is currently in FLASH4 is outdated. It has not been updated with the many changes that occured during the last 3 years. The MacLaurin unit test is the one that has been designed and used for checking all new multipole ideas. It's main purpose for creation was to get rid of all the other developments in eos, hydro, etc...that happened in parallel and which could cause interference with the sole testing of the multipole routines. The MacLaurin unit test uses a customized driver routine that has been stripped off all the unnecessary parts (like eos and hydro for example).
>>>>
>>>> Hope this helps,
>>>> Norbert
>>>>
>>>>
>>>> On Wed, 17 Jul 2013, Philip Chang wrote:
>>>>
>>>>
>>>> Hello FLASH users,
>>>>
>>>> I'm trying to set up the Maclaurin test problem and it doesn't appear to work. In particular, it appears that it is failing on calling the equation of state, which is strange because it is just an adiabatic gamma-law EOS. On the other hand, the unitTest/Multipole seems to work just fine. Because they are so similar, it is confusing why this is happening, but it appears that the difference is that the evolution step in the unitTest is empty, i.e., no hydro step, while the Maclaurin uses the hydro step. This suggests to me that the included Maclaurin test problem no longer works, i.e., these Maclaurin test problem does not evolve. Is this correct?
>>>>
>>>> Cheers,
>>>> Phil
>>>>
>>>>
>>>>
>>>>
>>>> --
>>>> /* ************************************************* *
>>>> Petros Tzeferacos, PhD
>>>>
>>>> Research Scientist & Code Group leader,
>>>> Flash Center for Computational Science,
>>>> Department of Astronomy & Astrophysics,
>>>> University of Chicago.
>>>> 5747 S. Ellis Ave, Jones Lab. room 317
>>>> Chicago, IL 60637
>>>> Tel. 773-834-3236
>>>> petros.tzeferacos at flash.uchicago.edu
>>>> * ************************************************* */
>>>
>>>
>>>
>>>
>>> --
>>> /* ************************************************* *
>>> Petros Tzeferacos, PhD
>>>
>>> Research Scientist & Code Group leader,
>>> Flash Center for Computational Science,
>>> Department of Astronomy & Astrophysics,
>>> University of Chicago.
>>> 5747 S. Ellis Ave, Jones Lab. room 317
>>> Chicago, IL 60637
>>> Tel. 773-834-3236
>>> petros.tzeferacos at flash.uchicago.edu
>>> * ************************************************* */
>>
>>
>>
>>
>>
>> --
>> /* ************************************************* *
>> Petros Tzeferacos, PhD
>>
>> Research Scientist & Code Group leader,
>> Flash Center for Computational Science,
>> Department of Astronomy & Astrophysics,
>> University of Chicago.
>> 5747 S. Ellis Ave, Jones Lab. room 317
>> Chicago, IL 60637
>> Tel. 773-834-3236
>> petros.tzeferacos at flash.uchicago.edu
>> * ************************************************* */
>
>
>
>
>
> --
> /* ************************************************* *
> Petros Tzeferacos, PhD
>
> Research Scientist & Code Group leader,
> Flash Center for Computational Science,
> Department of Astronomy & Astrophysics,
> University of Chicago.
> 5747 S. Ellis Ave, Jones Lab. room 317
> Chicago, IL 60637
> Tel. 773-834-3236
> petros.tzeferacos at flash.uchicago.edu
> * ************************************************* */
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20130718/dd835b7b/attachment-0001.htm>
More information about the flash-users
mailing list