TCE(CCSD(T)) crash during integral transformation


Clicked A Few Times
I have experienced different kinds of errors when I use different 2emet settings on the latest checked out source code.

If I use default 2emet setting (1) , it runs out of memory
If I use 2emet 2 or 3, it crashed with segmentation fault...
Anything greater than 3, will be considered invalid 2emet (seems someone taken out 2emet=4 recently, must I set 2eorb to use 2emet option greater than 3 ?)

 2-e (intermediate) file size =     99291110400
2-e (intermediate) file name = ./benzene.v2i.001
forrtl: severe (174): SIGSEGV, segmentation fault occurred

Here is my input.

echo
memory stack 5000 mb heap 100 mb global 27000 mb noverify
start benzene
scratch_dir /home/chiensh/scratch/junk
print "task time" "ga stats" "ma stats"
geometry
SYMMETRY c1
C 0.00000 1.40272 0.00000
H 0.00000 2.49029 0.00000
C -1.21479 0.70136 0.00000
H -2.15666 1.24515 0.00000
C -1.21479 -0.70136 0.00000
H -2.15666 -1.24515 0.00000
C 0.00000 -1.40272 0.00000
H 0.00000 -2.49029 0.00000
C 1.21479 -0.70136 0.00000
H 2.15666 -1.24515 0.00000
C 1.21479 0.70136 0.00000
H 2.15666 1.24515 0.00000
end

basis cartesian
* library aug-cc-pvtz
end

driver
clear
end

scf
singlet
#VECTORS HCORE
THRESH 1.0e-9
maxiter 90
end

ccsd
maxiter 100
TOL2E 10e-12
FREEZE atomic
end

TCE
FREEZE atomic
ccsd(t)
idiskx 1
2emet 3
thresh 1.0e-12
io eaf
tilesize 30
attilesize 40
END
set tce:read_integrals F F F F
set ccsd:use_trpdrv_nb T
set ccsd:use_ccsd_omp T
set ccsd:use_trpdrv_omp T

#task ccsd(t) energy
task tce energy

Forum Vet
I always recommend user to keep the input as simple as possible.
The following input works when using 10 nodes with 8 processes each (for a total of 80 processes)

echo
memory stack 5000 mb heap 100 mb global 27000 mb noverify

start benzene

geometry
SYMMETRY c1
  C        0.00000        1.40272        0.00000
  H        0.00000        2.49029        0.00000
  C       -1.21479        0.70136        0.00000
  H       -2.15666        1.24515        0.00000
  C       -1.21479       -0.70136        0.00000
  H       -2.15666       -1.24515        0.00000
  C        0.00000       -1.40272        0.00000
  H        0.00000       -2.49029        0.00000
  C        1.21479       -0.70136        0.00000
  H        2.15666       -1.24515        0.00000
  C        1.21479        0.70136        0.00000
  H        2.15666        1.24515        0.00000
end

basis cartesian
  * library aug-cc-pvtz
end

scf
singlet
THRESH  1.0e-9
maxiter 90
end

TCE
   FREEZE atomic
   ccsd(t)
   2eorb
   2emet 15
   tilesize 24
   attilesize 40
END

set ccsd:use_ccsd_omp T
set ccsd:use_trpdrv_omp T

task tce  energy

Output snippet
 4-electron integrals stored in orbital form
 
 v2    file size   =       6043878300
 4-index algorithm nr.  15 is used
 imaxsize =       40
 imaxsize ichop =        0
 starting step 1 at                87.24 secs 
 starting step 2 at               176.46 secs 
 starting step 3 at               199.18 secs 
 starting step 4 at               232.93 secs 
 done step 4 at               319.40 secs 
 Cpu & wall time / sec          232.9          232.2
 T1-number-of-tasks                    19
 
 t1 file size   =             6630
 t1 file name   = ./benzene.t1
 t1 file handle =       -998
 T2-number-of-boxes                   551
 
 t2 file size   =         67092525
 t2 file name   = ./benzene.t2
 t2 file handle =       -996
CCSD iterations
 ---------------------------------------------------------
 Iter          Residuum       Correlation     Cpu    Wall 
 ---------------------------------------------------------
NEW TASK SCHEDULING
CCSD_T1_NTS --- OK
CCSD_T2_NTS --- OK
    1   0.2553083450496  -0.9715759672782   125.6   121.1
    2   0.0651398915140  -0.9674552396587   126.7   121.9
    3   0.0242216017050  -0.9897865752363   128.3   123.6
    4   0.0104249540394  -0.9899069468234   126.0   121.4
    5   0.0050915558657  -0.9919464561545   124.7   120.1
 MICROCYCLE DIIS UPDATE:                     5                     5
    6   0.0019074082373  -0.9924283406677   125.5   121.0

Clicked A Few Times
Thank you, Edo.

BTW, is there a way to estimate how much memory and disk space is required before the job is submitted to the computing node?

~Dominic

Forum Vet
Dominic
1) My suggestion is to avoid any disk-based algorithm
2) the memory requirement should scale as the sixth power of the tilesize


Forum >> NWChem's corner >> Running NWChem