QuantumATK Forum

QuantumATK => General Questions and Answers => Topic started by: Jahangiri on July 7, 2010, 08:26

Title: ATKError
Post by: Jahangiri on July 7, 2010, 08:26
Dear users and Developers

I have a question. Maximum for how many atoms for two probe system atk can work ?

I am asking because I faced an error which I think is due to memory allocation .
I am having this error:


======================================================================
 /opt/mpich2/bin/mpiexec -n 8 /opt/vnl-2008.10.0/atk/bin/atk twoprobe_configuration_script.py > OUT1.out
Traceback (most recent call last):
  File "<string>", line 1139, in ?
ATKError: St9bad_alloc
=======================================================================

I should mention the job gets over after one day ( by getting the error ) , and the number of atoms are 749.


Could any one give any suggestion?
Thank you.
Best regards
Akbar

Title: Re: ATKError
Post by: Anders Blom on July 7, 2010, 09:31
Over 700 atoms is certainly pushing the limit. There is no strict limit, since it depends on your parameters. The memory consumption depends critically on the k-point sampling, the basis set size (and of course the number of atoms), and the grid size (total simulation volume, taking the mesh cut-off into account).

Also the parallelization strategy is important. I see you use -n 8, but do you have 8 computers, or one machine with two quad-cores? Each MPI processes uses the same amount of memory, so if you run 8 processes on one single machine, each process can max use 1/8 of the machine RAM. In this case, you can only run small systems - or use fewer MPI processes.
Title: Re: ATKError
Post by: Jahangiri on July 7, 2010, 09:59
Dear Anders Blom

We have one machine with 8 cores. But by referring to the system configuration. I think 16GB RAM would be sufficient !
Am I right?

The system Information is as following :
===============================================================
MemTotal:       16461608 kB
MemFree:        15317608 kB
Buffers:           19184 kB
Cached:           231664 kB
SwapCached:       106636 kB
Active:            28292 kB
Inactive:         714956 kB
Active(anon):      12044 kB
Inactive(anon):   480364 kB
Active(file):      16248 kB
Inactive(file):   234592 kB
Unevictable:           0 kB
Mlocked:               0 kB
SwapTotal:       5119992 kB
SwapFree:        4829652 kB
Dirty:                 0 kB
Writeback:             0 kB
AnonPages:        422284 kB
Mapped:            20444 kB
Slab:              56960 kB
SReclaimable:      25944 kB
SUnreclaim:        31016 kB
PageTables:        15456 kB
NFS_Unstable:          0 kB
Bounce:                0 kB
WritebackTmp:          0 kB
CommitLimit:    13350796 kB
Committed_AS:     930184 kB
VmallocTotal:   34359738367 kB
VmallocUsed:      368876 kB
VmallocChunk:   34359339227 kB
HugePages_Total:       0
HugePages_Free:        0
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:       2048 kB
DirectMap4k:        7680 kB
DirectMap2M:    16760832 kB
=======================================================================

Thank you very much.
Best regards
Akbar
Title: Re: ATKError
Post by: Anders Blom on July 7, 2010, 10:08
That gives you 2 Gb per MPI process. 2 Gb for 700 atoms is tight, esp. if you want something better than SingleZeta basis set (and you do), and a couple of k-points.
Title: Re: ATKError
Post by: Jahangiri on July 7, 2010, 12:20
Dear Anders Blom

Thank you very much.

Best regards
Akbar