[FLASH-USERS] Runtime error when trying to test Sod or Sedov

[Garishvili, Irakli]

Hi Chris,

Thanks a lot for your suggestion. It was right on money.

I simply added "-fdefault-real-8" to FFLAGS and it compiled and run just
fine.

Thanks again,
Cheers,
Irakli.

  


On 3/16/11 2:01 PM, "Chris Daley" <cdaley at flash.uchicago.edu> wrote:

> Hi Irakli,
> 
> I recommend checking that the FFLAGS macros in your Makefile.h promote
> reals to double precision.  After that I recommend checking the
> consistency of your software stack:
> 
> 1). Are you using the same MPI installation to build FLASH (mpif90) as
> you are to launch FLASH (mpirun)?
> 2). Are you building FLASH with the same compiler that was used to
> build HDF5?
> 3). If you have a shared library installation of HDF5 you should
> check that the expected version of HDF5 is loaded (see output of ldd
> ./flash3).
> 
> For testing purposes you can build FLASH without I/O using the +noio
> setup shortcut.  This may help you to isolate the problem.
> 
> You mention that you have run FLASH under valgrind and that you see
> memory leaks.  For the purposes of finding the actual error I would
> ignore the leaks and focus on "invalid read" and "invalid write"
> valgrind messages that indicate memory problems.  For usage with
> valgrind, I recommend FLASH applications with pm4dev (+pm4dev setup
> shortcut) because the output is less noisy than FLASH applications
> with pm40 (default). I spent some time last year removing false
> positives, but only in FLASH applications with pm4dev mesh package.
> 
> If you are still having trouble you should send us your setup line and
> Makefile.h.
> 
> Chris
> 
> 
> Garishvili, Irakli wrote:
>> Dear Flash Experts,
>> 
>> I am trying to test newest FLASH (3.3) version on a ³chaos_4_x86_64²
>> linux machine.
>> 
>> I separately tried to built/run Sod and Sedov problems and I got pretty
>> much the same errors in both
>> cases. Setup and build (gmake) processes give me absolutely no trouble.
>>  However, I get ³hdf5 I/O²-related runtime errors
>>  after executing Œmpirun ­np 1 flash3¹ command.
>> 
>> The strange thing is that runtime error messages totally change if I
>> modify ³lrefine_max² (default was set to 6) value inside
>>  flash.par file. Below I copied last parts of 2 different error
>> printouts I get when running Sod problem.
>> First when lrefine_max = 6,  and second when lrefine_max = 3.
>> 
>> I also tested flash3 executable using valgrind which seemed to indicate
>> some memory leak issues.
>> 
>> I¹d really appreciate any suggestion on this matter.
>> Cheers,
>> Irakli.  
>> 
>> 
>> 1.)
>> 
>>       84 1.9390E-01 1.5864E-03 |  1.58640E-03
>>       85 1.9707E-01 1.5865E-03 |  1.58646E-03
>>   iteration, no. not moved =            0           0
>>  refined: total leaf blocks =          541
>>  refined: total blocks =          721
>>       86 2.0024E-01 1.5865E-03 |  1.58647E-03
>> HDF5-DIAG: Error detected in HDF5 library version: 1.6.5 thread 0.  Back
>> trace follows.
>>   #000: H5Dio.c line 587 in H5Dwrite(): can't write data
>>     major(15): Dataset interface
>>     minor(25): Write failed
>>   #001: H5Dio.c line 953 in H5D_write(): can't write data
>>     major(15): Dataset interface
>>     minor(25): Write failed
>>   #002: H5Dio.c line 1324 in H5D_contig_write(): optimized write failed
>>     major(15): Dataset interface
>>     minor(25): Write failed
>>   #003: H5Dselect.c line 654 in H5D_select_write(): write error
>>     major(14): Dataspace interface
>>     minor(25): Write failed
>>   #004: H5Dcontig.c line 739 in H5D_contig_writevv(): block write failed
>>     major(05): Low-level I/O layer
>>     minor(25): Write failed
>>   #005: H5F.c line 3338 in H5F_block_write(): file write failed
>>     major(05): Low-level I/O layer
>>     minor(25): Write failed
>>   #006: H5FD.c line 3486 in H5FD_write(): driver write request failed
>>     major(22): Virtual File Layer
>>     minor(25): Write failed
>>   #007: H5FDsec2.c line 804 in H5FD_sec2_write(): file write failed
>>     major(05): Low-level I/O layer
>>     minor(25): Write failed
>> Driver_abortC called
>> Error: Unable to write unknowns
>> 
>> Calling MPI_Abort for immediate shutdown
>> [0] [MPI Abort by user] Aborting Program!
>> mpirun_rsh: Abort signaled from [1]
>> 
>> 
>> 2.)
>> 
>>       24 3.3554E-03 1.6777E-03 |  1.92661E-02
>>       25 6.7109E-03 3.3554E-03 |  1.73809E-02
>>       26 1.3422E-02 6.7109E-03 |  1.52721E-02
>>       27 2.6844E-02 1.3422E-02 |  1.36666E-02
>>       28 5.3687E-02 1.2506E-02 |  1.25062E-02
>>       29 7.8699E-02 1.2473E-02 |  1.24730E-02
>>       30 1.0365E-01 1.2544E-02 |  1.25439E-02
>>       31 1.2873E-01 1.2581E-02 |  1.25813E-02
>>       32 1.5390E-01 1.2615E-02 |  1.26146E-02
>>       33 1.7913E-01 1.2648E-02 |  1.26484E-02
>>       34 2.0442E-01 1.2663E-02 |  1.26628E-02
>>  *** Wrote checkpoint file to sod_hdf5_chk_0001 ****
>>  exiting: reached max SimTime
>>  *** Wrote plotfile to sod_forced_hdf5_plt_cnt_0000 ****
>> Segmentation fault (core dumped)
>> 
>> ------ End of Forwarded Message
>