ABINIT Discussion Forums

Hi all,

In my calculations, there is a comment as follows:

symanal : COMMENT -
The Bravais lattice determined only from the primitive
vectors, bravais(1)= 4, is more symmetric
than the real one, iholohedry= 1, obtained by taking into
account the atomic positions.
symspgr : the symmetry operation no. 1 is the identity
symspgr : spgroup= 1 P1 (=C1^1)
getkgrid : length of smallest supercell vector (bohr)= 8.946400E+01
Simple Lattice Grid
npfft, npband and npkpt 1 1 8
mpi_enreg%sizecart(1),np_fft 1 1
mpi_enreg%sizecart(2),np_band 1 1
mpi_enreg%sizecart(3),np_kpt 8 8
in initmpi_grid : me_fft, me_band, me_kpt are 0 0 0
invars1: mkmem undefined in the input file. Use default mkmem = nkpt
invars1: With nkpt_me= 32 and mkmem = 256, ground state wf handled in core.
Resetting mkmem to nkpt_me to save memory space.
invars1: mkqmem undefined in the input file. Use default mkqmem = nkpt
invars1: With nkpt_me= 32 and mkqmem = 256, ground state wf handled in core.
Resetting mkqmem to nkpt_me to save memory space.
invars1: mk1mem undefined in the input file. Use default mk1mem = nkpt
invars1: With nkpt_me= 32 and mk1mem = 256, ground state wf handled in core.
Resetting mk1mem to nkpt_me to save memory space.


As we know, the default of mkqmem is nkpt. Why nkpt_me is different from mkqmem. what is the difference between nkpt and nkpt_me? How to change the setup of nkpt_me or mkqmem?

Thank you very much!

According to the comment of invars1.F90, nkpt_me means " the maximal number of k points treated per processor" and "nkpt_me may differ from processor to processor".

Sincerely,
Guangfu Luo

At this stage, during the treatment of input, mkmem will be equal to either
* 0 if you are using mkmem 0 for disk paging, or
* nkpt (256) which is your case.

as Guangfu says, nkpt_me is the number of kpt per processor

Matthieu

Just a small comment: mkmem, mkqmen and nkpt_me are not used in the GW code.
One should use gwmem to control the amount of memory allocated for the wavefunctions and
the dielectric matrix.