post update ibi

[marvinbernhardt]

solves #676
In fact, this could also be used to have non-bonded interactions be updated with IBI.

Aditionally I have simplified how one can exclude interactions from imc. Before you had to provide an extra settings.xml file. Now you can just set the interaction.imc.group to 'none' and it will be excluded.

I prepare a csg-tutorial for propane, where this is used to do imc for nb and ibi for bonded.

One thing that is awkward is that this post-update overwrites potential updates from other methods, but I print a warning in that case.

I have not yet changed the iie interface to use this (see line 111 in update_iie.sh) but will do so later, since I work on iie atm anyway.

[codecov]

Codecov Report

Merging #696 (d5b784c) into master (fd5c7df) will decrease coverage by 0.0%.
The diff coverage is 100.0%.

@@           Coverage Diff            @@
##           master    #696     +/-   ##
========================================
- Coverage    28.1%   28.1%   -0.1%     
========================================
  Files         133     133             
  Lines        8677    8678      +1     
========================================
- Hits         2442    2441      -1     
- Misses       6235    6237      +2     

Impacted Files	Coverage Δ
src/tools/csg_stat_imc.cc	61.4% <100.0%> (-0.4%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update fd5c7df...d5b784c. Read the comment docs.

[junghans]

@votca-bot format

[junghans]

@votca-bot changelog: add ability to run ibi as a postupd method

[junghans]

@marvinbernhardt I am a bit confused that this update doesn't break the spce/hncgn

2021-06-23T17:13:08.8420439Z         Start 209: regression_spce_hncgn
2021-06-23T17:13:26.6095206Z 136/159 Test #209: regression_spce_hncgn ..............................   Passed   17.77 sec

Do we need a PR in csg-tutorials as well? If so can you add the naphtalne examples as well?

[junghans]

@votca-bot copyright

[marvinbernhardt]

@marvinbernhardt I am a bit confused that this update doesn't break the spce/hncgn

2021-06-23T17:13:08.8420439Z         Start 209: regression_spce_hncgn
2021-06-23T17:13:26.6095206Z 136/159 Test #209: regression_spce_hncgn ..............................   Passed   17.77 sec

Do we need a PR in csg-tutorials as well? If so can you add the naphtalne examples as well?

IIE and IBI still do IBI on all bonded interactions by default I belive. Question is if we want to keep that behavior or want the user to explicitly use post_update IBI. I would say it should only be implicit for IBI itself, other methods should use post_update.

[junghans]

@marvinbernhardt I am a bit confused that this update doesn't break the spce/hncgn

2021-06-23T17:13:08.8420439Z         Start 209: regression_spce_hncgn
2021-06-23T17:13:26.6095206Z 136/159 Test #209: regression_spce_hncgn ..............................   Passed   17.77 sec

Do we need a PR in csg-tutorials as well? If so can you add the naphtalne examples as well?

IIE and IBI still do IBI on all bonded interactions by default I belive. Question is if we want to keep that behavior or want the user to explicitly use post_update IBI. I would say it should only be implicit for IBI itself, other methods should use post_update.

Yeah that makes sense!

[junghans]

Merge when ready.

And update the tutorials accordingly.

[junghans]

ping @marvinbernhardt

[marvinbernhardt]

I want to test it again with the propane system from the csg_tutorials

[junghans]

has_bonded isn't used anywhere anymore, so remove.

[marvinbernhardt]

I had to modify a perl script (update_ibi_pot.pl) to get rid of nasty artifacts in my angle potential. Now it looks much better, but the perl code is surely redundant. But I am not sure how to improve it.

[junghans]

I had to modify a perl script (update_ibi_pot.pl) to get rid of nasty artifacts in my angle potential. Now it looks much better, but the perl code is surely redundant. But I am not sure how to improve it.

What were the artifacts and what is redundant?

[marvinbernhardt]

I had to modify a perl script (update_ibi_pot.pl) to get rid of nasty artifacts in my angle potential. Now it looks much better, but the perl code is surely redundant. But I am not sure how to improve it.

What were the artifacts and what is redundant?

Artifacts were building up on the right side of angle potentials.

There is redundancy in these two for loops. Unfortunately I'm not so good at Perl.

csg/share/scripts/inverse/update_ibi_pot.pl

Lines 76 to 110 in 7bc89c0

	my $value=0.0;
	for (my $i=$ndx_max_rdf_cur;$i<=$#r_aim;$i++){
	if (($rdf_aim[$i] > 1e-10) && ($rdf_cur[$i] > 1e-10)) {
	$dpot[$i]=log($rdf_cur[$i]/$rdf_aim[$i])*$pref;
	$flag[$i]="i";
	} else {
	$dpot[$i]=$value;
	$flag[$i]="o";
	}
	if($pot_flags_cur[$i] =~ /[u]/) {
	$dpot[$i]=$value;
	$flag[$i]="o";
	}
	else {
	$value=$dpot[$i];
	}
	}
	# go to beginning
	$value=0.0;
	for (my $i=$ndx_max_rdf_cur-1;$i>=0;$i--){
	if (($rdf_aim[$i] > 1e-10) && ($rdf_cur[$i] > 1e-10)) {
	$dpot[$i]=log($rdf_cur[$i]/$rdf_aim[$i])*$pref;
	$flag[$i]="i";
	} else {
	$dpot[$i]=$value;
	$flag[$i]="o";
	}
	if($pot_flags_cur[$i] =~ /[u]/) {
	$dpot[$i]=$value;
	$flag[$i]="o";
	}
	else {
	$value=$dpot[$i];
	}
	}

[junghans]

There is redundancy in these two for loops. Unfortunately I'm not so good at Perl.

I guess you could make it in a function, but don't bother if it is too complicated.

[marvinbernhardt]

I think this could be merged now.