Possible to restart CCSD or CCSDT TCE calculations?
Just Got Here
10:26:52 PM PDT - Sun, Mar 31st 2019
Hi all,

I'm working with a reaction (35 atoms total), which is only marginally within the limits of the hardware I presently have available.

From the documentation I've read, TCE can do ROHF-CCSD in parallel but not CCSD[T]. If there's a way to do ROHF-CCSD[T] in parallel please disabuse me of the misunderstanding and enlighten me as to how.

What I'd like to do (given the above assessment) if possible, is to run the ROHF-CCSD in parallel and restart on a single core to finish the CCSD[T].

Can it be done? How? Would using TCE_CUDA (haven't compiled it yet but could) bear on this at all either way?

In general, I'd also like to know if it's possible to restart CCSD calculations (gamess can from the F70 scratch file it produces)

Thanks in advance!

Forum Vet
7:42:36 PM PDT - Mon, Apr 1st 2019
I use the following input and three cores to do the ROHF calculation of doublet CH free radical with NWCHEM 6.8

echo

start molecule

charge 0
memory ...

geometry units angstroms print xyz
symmetry c1
  C       ...

  H       ...
end

basis
  • library aug-cc-pvdz
end

SCF
   doublet

   ROHF

   END

TCE
SCF

CCSD[T]

io ga
END

TASK TCE ENERGY

and find it can be finished without any problem
Singles contributions



Doubles contributions

CCSD(T)

Using plain CCSD(T) code
...

CCSD[T]  correction energy / hartree =     ...

CCSD[T] correlation energy / hartree =    ...

CCSD[T] total energy / hartree       =       ...

CCSD(T)  correction energy / hartree =      ...

CCSD(T) correlation energy / hartree =       ...

CCSD(T) total energy / hartree       =       ...

Cpu & wall time / sec           ...



Parallel integral file used       3 records with       0 large values

I have also successfully finished ROHF CCSDT for it parallelly.
The initial geometry might need to be reoptimized.

Forum Vet
6:03:52 PM PDT - Tue, Apr 2nd 2019
I recalculate it using the same input and four threads with NWCHEM 6.8 on Ubuntu18.04 and get
...
Iterations converged
CCSD correlation energy / hartree =      ...

CCSD total energy / hartree       =     ...
Singles contributions

Doubles contributions

CCSD(T)

Using plain CCSD(T) code
..
CCSD[T] correction energy / hartree = ...
CCSD[T] correlation energy / hartree =        ...

CCSD[T] total energy / hartree       =       ...

CCSD(T)  correction energy / hartree =        ...

CCSD(T) correlation energy / hartree =        ...

CCSD(T) total energy / hartree       =       ...

Cpu & wall time / sec          ...



Parallel integral file used       4 records with       0 large values

Thus, there is no problem to run this parallelly with NWCHEM 6.8.

Forum Vet
5:48:27 PM PDT - Wed, Apr 3rd 2019
When spherical is used, the result on Ubuntu18.04 becomes
...
Singles contributions

Doubles contributions

CCSD(T)

Using plain CCSD(T) code
...

CCSD[T]  correction energy / hartree =        ...

CCSD[T] correlation energy / hartree =        ...

CCSD[T] total energy / hartree       =       ...

CCSD(T)  correction energy / hartree =        

CCSD(T) correlation energy / hartree =        

CCSD(T) total energy / hartree       =      ...

The repeated one gives
 ...
CCSD[T]  correction energy / hartree =        ...

CCSD[T] correlation energy / hartree =        ...

CCSD[T] total energy / hartree       =      ...

CCSD(T)  correction energy / hartree =       

CCSD(T) correlation energy / hartree =        

CCSD(T) total energy / hartree       =     ..
The two results can be called identical.


Forum >> NWChem's corner >> Running NWChem