The ABINIT Users Mailing List ( CLOSED )

[<kuroki@vos.nagaokaut.ac.jp>]

Thank you for your replies. (Dr./Mr. Alain Jacques and Damien Caliste)

I understood that the error massage occur at different situation for 
indivisual configurations.

In my case, the system is ubuntu8.04 which may cause the error because 
of 32 bit version (i686).
===========================================================================
kuroki@iodine-t2:~$ uname -a
Linux iodine-t2 2.6.24-23-server #1 SMP Wed Apr 1 22:22:14 UTC 2009 i686 
GNU/Linux
===========================================================================

Then I checked user memory limit using "ulimit -a" resulting that 
maximum memory size was unlimited.

However, I'm now confusing.

When the abinit was running with 64 atoms using (mkmem = 0) parameter, 
used memory was about 14GB which is much larger than the estimation of 
memana (about 1GB). Total of the momory need for DATASET1 listed below 
was about 2GB. How to comprehend?

===========================================================================
kuroki@iodine-t2:~$ free
             total       used       free     shared    buffers    cached
Mem:      24937664   14216456   10721208          0      52504   7868852
-/+ buffers/cache:    6295100   18642564
Swap:      6000236          0    6000236
===========================================================================
 memory : analysis of memory needs
===========================================================================
 Values of the parameters that define the memory need for DATASET  1.
intxc =         0  ionmov =         0    iscf =         7 xclevel =    2
lmnmax=         4   lnmax =         4   mband =       256  mffmem =    1
mgfft =       125   mkmem =         0 mpssoang=         4     mpw =114503
  mqgrid =      3001   natom =        64    nfft =   1875000    nkpt = 
       10
  nloalg =         4  nspden =         1 nspinor =         1  nsppol = 
        1
    nsym =        16  n1xccc =         0  ntypat =         3  occopt = 
        1
===========================================================================
P This job should need less than                    1008.820 Mbytes of 
memory.
===========================================================================

Thank for your kind help!
Yuichiro KUROKI



> Hello,
> 
> Le 17/06/2009, <kuroki@vos.nagaokaut.ac.jp> a écrit :
>> How to increase the limit of memory?
>> Is there the parameter to control the memory limitation?
> Yes, and if I understand correctly your mail, this may not come from
> ABINIT.
> 
> I've never used GW calculations so my answer may not be accurate.
> 
>> ==================================================================
>> memana : ERROR -
>>   Test failed to allocate    5060.225 Mbytes
>>   It is not worth to continue
>>   Action : modify input variable to fit the available memory.
>>   or increase limit on available memory.
> Here, ABINIT fails to allocate 5Gb of RAM. Of course it does not know
> the reason, just that the Fortran line "allocate(bigarray)" fails.
> 
>> ==================================================================
>> kuroki@iodine-t2:~/abinit-5.7.4/CuAlS2/Cu16Al16S32$ free
>>           total       used       free     shared
>> buffers         cached Mem:    24937664    5626760
>> 19310904       0      83156    5308000 -/+
>> buffers/cache:     235604    24702060 Swap:
>> 6000236          0    6000236
>> ==================================================================
> It seems that you have enough memory on your machine. So the failure
> may comes from :
> - a user memory limit. If you use Bash, type "ulimit -a" and look at
> the result, specifically the "max memory size". If you use another
> shell, try to find the equivalent command. Read the man also on how to
> change the value if you don't have unlimited in the "max meomry size".
> - what is your OS ? It may be a 32bit system with indirect memory
> addressing. In that case no single process can allocate more than 4Gb
> of memory. Type "uname -a" and see the architecture.
> 
> Otherwise, no idea...
> 
> May be give us some clue on your compiler, system...
> 
> Damien.
> 
>