Hi, are you using intel compilers? If so, did you try adding '-heap-arrays'
flag at 'FFLAGS'? It seems that intel compilers try to put allocatable
arrays at stack memory, which can cause problems when running large systems
simulations with multi-nodes.

Cheers,


Pedro

On Sat, Jul 18, 2015 at 12:07 AM, Nick Papior <[email protected]> wrote:

> APPLICATION terminated looks like an error issued in your intel library,
> there is not much we can do about those without more information as to why
> :(
>
> Maybe you could try the later version?
>
>
> 2015-07-17 20:29 GMT+00:00 Oleksandr Voznyy <[email protected]>:
>
>> Hi all,
>> I have problems with the trunk-424 version: I think it is the the new
>> grid balancing system. It works for a few SCF cycles and then crashes.
>>
> So did it work on previous versions?
>
>>
>> The unit cell is 26x26x50 Ang, 1000 atoms.
>> Is my system too large? It crashes on anything from 64 to 512 CPUs.
>>
>> I don't think updating to the latest version would help since load
>> balancing was untouched there.
>>
> Alex.
>>
>> -------------------------------------------------------------
>>
>> InitMesh: MESH =   288 x   288 x   540 =    44789760
>> InitMesh: (bp) =   144 x   144 x   270 =     5598720
>> InitMesh: Mesh cutoff (required, used) =   300.000   316.798 Ry
>> ExtMesh (bp) on 0 =   244 x   118 x   134 =     3858128
>> New grid distribution:   2
>>            1      71:  108   38:   74  199:  235
>>            2       1:   36    1:   37   30:  198
>> ...
>>           63       1:   37  110:  144  199:  237
>>           64       1:   36  111:  144  238:  270
>> New grid distribution:   3
>>            1      37:   72   73:  108  181:  225
>>            2      37:   72    1:   36  181:  225
>> ...
>>           63      73:  108   37:   72  181:  225
>>           64       1:   36  109:  144  226:  270
>> APPLICATION TERMINATED WITH THE EXIT STRING: Hangup (signal 1)
>>
>
>
>
> --
> Kind regards Nick
>

Responder a