IR-MP2 Calculation stops with error ival=10


Click here for full thread
Just Got Here
4:06:16 PM PDT - Thu, Jul 19th 2012
Dear Bert,

I have decreased the memory setting in the input file,
The calculation is running perfectly now.
thank you very much for your help!!!

kind regars
johannes


Quote:Bert Jul 18th 1:32 pm
The memory keyword in the input deck is per core. So, in your input deck you are asking for 16*8 GByte per node, which far exceeds the memory you have on the system. You should adjust accordingly, i.e. heap + stack + global should be less then 1500-1700mb .

Bert



Quote:Johannes Jul 18th 10:54 am
Dear Bert,

Each node has 16 cores and a total memory of 32GB. The configuration is set up in a way that 30GB can be used for the calculation and the remaining 2GB are reserved for the operating sytem.

I am not sure where the error comes from. Is the calculation running out of memory or is there a problem with the install configuration?

kind regards
johannes

32GB
Quote:Bert Jul 16th 10:59 am
Johannes,

You are asking for 8 gbyte per processor. How much memory do you have available per core (i.e. memory per node/#cores per node).

Bert


Quote:Johannes Jul 12th 11:02 am
Dear NWChem community,

I am trying to perform a RI-MP2 calculation on a rather large molecule with aug-cc-dz basis set.
However, the calculation ends with the error message 0: error ival=4
Does anyone has an idea what is going wrong here?

here is my input deck

title "zr4_m2_td rimp2"
restart zr4_m2_rimp2
memory stack 2000 heap 2000 global 4000 mb
SCRATCH_DIR /global/lv70279/jkreutze/scr/admin
charge 1
scf
print high
maxiter 100
direct
SINGLET
end

mp2
freeze 0
end

geometry


Geometry in xyz

end

basis

basis sets from EMSL

end

task rimp2 energy

and the last part of the output:

         RI integral approximation: 'V'

   Threshold for Schwarz screening:  1.000E-09

Smallest eigenvalue retained in fit  1.000E-08

    Base file name for MO integrals 'mo3cint'

   Type of storage for MO integrals 'EAF library'



Transformation block size  203, requiring     1 integral passes.

XForm_3cInts: Completed     1:  203 pass.             @Thu Jul 12 11:17:59 2012

Energy evaluation using block sizes   203   29
0: error ival=10
(rank:0 hostname:r08n35.gb pid:23493):ARMCI DASSERT fail. ../../ga-5-1/armci/src/devices/openib/openib.c:armci_call_data_server():2193 cond:(pdscr->status==IBV_WC_SUCCESS)
32: error ival=4
(rank:32 hostname:r01n21.gb pid:25604):ARMCI DASSERT fail. ../../ga-5-1/armci/src/devices/openib/openib.c:armci_call_data_server():2193 cond:(pdscr->status==IBV_WC_SUCCESS)
8: error ival=10
(rank:8 hostname:r08n41.gb pid:2670):ARMCI DASSERT fail. ../../ga-5-1/armci/src/devices/openib/openib.c:armci_call_data_server():2193 cond:(pdscr->status==IBV_WC_SUCCESS)
APPLICATION TERMINATED WITH THE EXIT STRING: Hangup (signal 1)
"zr4_m2_rimp2.out" [readonly] 27163L, 1590822C

Since the SCF finishes normally and the error occures after starting the MP2 part i assumer an error related to the provided memory for the MP2 calculation.
Hace i set the memory in the input deck (stack, heap, global etc.) correctly?

I would really appreciate your help in this matter,
kind regards
johannes