...
The are the commands for ASE/python mode and the "native" (no ASE/python) mode:
Code Block |
---|
esp-ver-bsub <version> myscript.py
esp-ver-bsub-native <version> -q suncat-test -o my.log -n 8 pw.x -in pw.inp
|
Dealing With Memory Issues (e.g. Swapping Jobs)
For a "typical" espresso job (default planewave parallelization):
...
Reducing Memory Usage for Large Systems
From http://www.democritos.it/pipermail/pw_forum/2008-January/008101.html
Excerpt (relevant for the "native" (non-ASE) mode):
- consider reducing the planewave cutoff, IF it won't affect your results too much
is your system an isolated system? then use this keyword:
Code Block K_POINTS gamma
it will use k=0 only (which is all you need for an isolated system) and exploit various tricks to reduce memory usage
- setting option "diago_david_ndim" to the minimum (2) and "mixing_ndim" to a smaller value (4) reduces memory usage, but may increase CPU time
- using diagonalization='cg' will also reduce memory usage, but it will increase CPU time by a sizable amount
- do not calculate stress if you do not need to: it is expensive
In ASE-mode we are currently unable to set the "K_POINTS gamma" field and "diago_david_ndim". The other two can be specified with the convergence keyword. "mixing_ndim" is "mix", and "diagonalization" is "diag". For example:
Code Block |
---|
convergence = {'energy':1e-6,
'mixing':0.7,
'maxsteps':100,
'mix':4,
'diag':'cg'},
|
...
A simple optimization: esp.py
Calculate density-of-states: espdos.py
Plot density-of-states: espdosplot.py
NEB: espneb.py
Running in "native" mode (see esp-ver-bsub-native
command above): pw.inp
Versions
Version | Date | Comment |
1 | 12/3/2012 | initial version |
2 | 12/5/2012 | use mkl fftw |
3 | 12/7/2012 | UNSTABLE version: developers allowed to change espresso.py. Users can overwride espresso.py by putting their own espresso.py in directory $HOME/espresso |
4,4a | 12/10/2012 | update to the latest svn espresso-src and espresso python |
5 | 2/14/2013 | Entropy corrections added and default parameters changed (smearing type and width) |
6,6a | 3/7/2013 | Many changes: move to combination of dacapo/espresso pseudo potentials (previously just dacapo), add spin polarized BEEF |
7,7a | 4/5/2013 | Update the python interface for bug fixes. Numbers shouldn't change from v6 |
8,8a | 4/5/2013 | Important bug fixes: no need for calc.stop(), support for kpoint parallelization with ASE, fix for rhel5 nfs auto mount problem. Numbers shouldn't change from v6/v7. |
9,9a | 5/28/2013 | Add PDOS/NEB calculations. new libbeef interface allows for adding additional beef functionals in future. |
10,10a | 6/19/2013 | Fix problem with pipe buffering that crashed NEB. Dump more information about python/fortran executables to output. |
11 | 7/2/2013 | Bug fix for end of job race condition giving "broken pipe" error |
12 | 7/11/2013 | Chuan adds new 'diskio' option to allow get_work_function to succeed (was incompatible with avoidio=True). |
13,13a | 9/4/2013 | BROKEN (setupenv renamed). band structure calculations and ase constraints being passed down to espresso's internal relaxation routines. calculation results should not change |
14,14a | 9/10/2013 | fix bug in espresso.py forces, fixed bug with LDA+U PDOS in espresso fortran |
15,15a | 9/10/2013 | fix bug where espresso.py would crash if it was given a not-understand ASE constraint, even using an ASE optimization |
SUNCAT Quantum Espresso Talks
...
Copy this script, and then edit the appropriate lines at the top:
Code Block |
---|
/afs/slac/g/suncat/share/scripts/privesp.csh |
Avoid syntax error in Python run scripts
There is a simple trick to catch syntax errors even before submitting them to the test queue. Just wrap the ase submit command in the following function (e.g. via your bashrc) and a faulty python script will be exited with an error message without waiting in any queue.
Code Block |
---|
function esp(){ python -m py_compile ${1} if [ $? -ne 0 ] then echo "Not compiling" return 1 fi chmod -v 755 $1 esp-ver-bsub 18 $@ } |
Espresso ASE To-Do List
- merge branch with espresso trunk
- become part of ASE svn (need to follow new ASE guidelines)
- dry-run mode to get memory estimate
- understand failing espresso tests
- record uspp and executable directory in output (and/or svn version, somehow?)
- neb (done)
- constraints interface to pass ASE constraints to espresso
- dos (done)
- bandgaps (done)
- separation of site-specific code from ASE code (including site-specific "scratch") (done)
- make beef errors accessible from ASE
- beef self-tests integrated with espresso self-tests
- support kpoint parallelization (done)
- look into other parallelization (openmp, scalapack)
- documentation/examples (including on ASE website)
- fix popen warnings on suncat3
- eliminate difference between batch/non-batch running
- how to get automatic python recompilation with setup.py build approach?
- can we eliminate os-dependent stuff, like grep/egrep/sed?
- eliminate need for calc.stop() with multiple calculations (done)
- get work function without dumping out the electrostatic cube file? (chuan has tools for this)
- dipole correction goes in the middle of unit cell by default (in python, chuan makes sure it goes in the biggest gap) (done)