Files prepared and density convergence / ELF calculation put into the queue on 32 nodes with 8 ppn and a walltime of 36 hours.

Density successfully converged after 26 SCF steps and ELF printed.
Final etot = -1139.9019901145 Hartrees.
Final diff in etot = -3.430E-08.
nkpt = 258.
natom = 180.
nband = 317.
ngfft = 240 45 100.
ecut = 25 Hartrees.
Run on 32 nodes, 8 ppn.
It took 118821.0 seconds per processor = 33 hours 0 minutes
Size of abinit ELF output file: 9.3 mb
Size of ELF vesta file: 24.8 mb

Next step is to change the input file to read in the density, converge the wavefunction, and print the PDOS. I'm going to let iatsph and ratsph be set to default. Default iatsph goes from 1 to natsph, exactly as we want. Default ratsph sets all r to 2.0 Bohr for norm conserving pseudopotentials, which I believe we have. If the calculation is successful, we can tweak the ratsph and see what happens. The calculation is now the queue.

When run on 32 nodes with 8 ppn = 256 processors for 36 hours, job did not finish. Furthermore, calculation didn't even get to the point of being out of memory, as mz84 and mz133 did, however yvte1 has a much larger unit cell. I suspect that it would have gotten to the "out of memory" stage if it had been able to run for longer. This presents a problem, as I think I'm going to need to use 4 ppn in order to deal with the memory issue, however at 64 nodes, 4 ppn, 24 hours, I'll have less computational time than I did here.

Talk to Josh about this.

Note: Unit cell said to not be primitive in first calculation. If this is true, memory needs could be dramatically reduced.

This compound has been in a bit of a holding pattern for a while because its unit cell is so damn big. However, in light of recent successes with laqnoh, I'm going to try and calculate the PDOS using a 3x3x3 k-point grid, which gives 40 k-points. The calculation is in the queue.

The calculation ran, and it got further than it has previously, but it did not complete. Previously, it died here:

vtorho : nnsclo_now=250, note that nnsclo,dbl_nnsclo,istep= 0 0 1
mpirun: killing job…

This time:

vtorho : nnsclo_now=250, note that nnsclo,dbl_nnsclo,istep= 0 0 1
-P-0000 leave_test : synchronization done…
vtorho: loop on k-points and spins done in parallel
ioarr: writing density data
ioarr: file name is yvte1o_DEN
ioarr: data written to disk file yvte1o_DEN
partial_dos_fractions : spaceComm = 0
partial_dos_fractions : rmax = 2.0000000000000000
Unit cell volume ucvol= 1.0903275E+04 bohr^3
Angles (23,13,12)= 9.00000000E+01 9.00600000E+01 9.00000000E+01 degrees

mpirun noticed that process rank 4 with PID 10675 on node c0232-ib exited on signal 9 (Killed).

NOTE: log for this run is named "furtherlog"

In successful calculations, the Angles line is followed by "get dens in sphere for i band ", however clearly here we did not get to that step, after which the entire angular analysis occurs (which I expect to take a good amount of time too). The only other thing I can think to do is to run it on the regular x_long queue. Let's say that we expect this calculation to take 2x as long as it did = 96 hours on 10 nodes

However in the x_long queue we're limited to 4 nodes. The only multiple available to us is therefore 2 nodes, which is a 5th of what we're using now, and should therefore take 5x as long = 480 hours. The maximum available wallclock is 504 hours, therefore I think I'll just set it to that. I'm putting the calculation in the queue now. Also, if the job does finish it will most likely require "filling in" as the others have, therefore I'm putting an identical job in the queue but holding it such that it can be run without waiting as long in the event filling in is required.

Alex took a look at the abinit generated ELFs and the whole induced inert pair stuff is not at all present. Because we trust the ELFs from abinit a lot more than those from LMTO this means that this part of the story is simply gone. Kinda sucks, but it's better to be right.

This calculation currently shows a walltime of 700 hours, however I noticed that I forgot to remove the commas in iatsph. I don't think this will be a problem, but it's worth noting.

Damnit… the calculation ran out of memory at 780:50:1. Guess that's the end of that.

Te coordination examined with Shape. This compound is listed under ajiqeq. See Shape Calculations for more information.

Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License