14.09.2014 Views

CASINO manual - Theory of Condensed Matter

CASINO manual - Theory of Condensed Matter

CASINO manual - Theory of Condensed Matter

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

undesirable circumstance for a particular example, system/run. You know that this did not happen<br />

a few hundred commits ago when you last ran the autotest at commit 123456, say.<br />

In that case you may want to use git-bisect. Open a new terminal and change into ~/<strong>CASINO</strong>, so<br />

that you have one terminal to run the autotest (the ‘autotest terminal’) and another to do version<br />

switching and compiling (the ‘git terminal’). In the ‘git terminal’ type<br />

git bisect start<br />

git bisect bad # the current commit has the bug<br />

git bisect good 123456 # an old commit (123456) does not have the bug<br />

git will then jump to an intermediate commit. Type make to recompile casino at that point, then<br />

switch to the ‘autotest terminal’ and type, for example,<br />

./autotest --only system/run<br />

Then switch to the ‘git terminal’ and, if the problem is not present in this run, type<br />

git bisect good<br />

else type<br />

git bisect bad<br />

and type make again to rebuild the code at the next commit git decides to test, and so<br />

on. If at any point you land on a commit which you would rather skip (it won’t compile, or it<br />

has a known problem which won’t let you see whether the bug you are looking for is there or not), type<br />

git bisect skip<br />

and git will pretend the commit is not there.<br />

After a few bisects git should be able to tell you which commit introduced the regression. To return<br />

to the initial state (and this can be done at any point during the bisect operation), type<br />

git bisect reset<br />

For more information see the git-bisect manpage.<br />

C Appendix 3: Converting <strong>CASINO</strong> v1.x input files to<br />

<strong>CASINO</strong> v2.x format<br />

There is a utility called update input included with casino which should be capable <strong>of</strong> updating<br />

most input files from version prior to 2.0 into ones suitable for the current code. As the changeover<br />

took place ages ago (2003?) hardly anyone will want to do this, and this appendix is in serious danger<br />

<strong>of</strong> being deleted. However, if you really do want to do such conversions, and have any problems using<br />

this script, then here are some useful tips.<br />

• The formats <strong>of</strong> the Jastrow and backflow sets in the correlation.data file are exactly the same<br />

as the formats <strong>of</strong> the old jastrow.data and backflow.data files: therefore you simply need<br />

to copy the contents <strong>of</strong> the old files into a file called correlation.data. If you don’t have a<br />

backflow.data file then you can simply rename jastrow.data as correlation.data.<br />

• If your calculation uses the archaic jasfun.data file <strong>of</strong> casino’s old Jastrow factor or the<br />

swfn.data file used to store ‘spline’ orbitals then it is not possible to continue your calculation<br />

using version 2, as both the old Jastrow factor and the spline representation <strong>of</strong> orbitals have<br />

been dropped from casino. You must therefore use an archived version <strong>of</strong> casino version 1.<br />

• Any blwfn.data files should be regenerated. The file format can also be updated by hand—ask<br />

Neil to do this.<br />

218

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!