search

MRChemSoft / mrchem

MultiResolution Chemistry
GNU Lesser General Public License v3.0
27 stars 21 forks source link

Nuclear potential prec scaled with box size #386

Closed gitpeterwind closed 3 years ago

gitpeterwind commented 3 years ago

Scale the relative prec for Nuclear potential with box size. V.norm increase with 6th root of box volume = additional scaling factor.

Note that the prec for the " allreducePotential" should be rethinked too. It seems to use absolute prec, not relative prec, but probably it is so small, that there is no neglect of nodes anyway, which is probably what we want.

The correct treatment would be to not copy the entire nuclear potential everywhere, but to treat it "locally", i.e. save the nodes in bank, and when multiplied by an orbitalvector, loop over orbitals for fixed node, (instead of loop over nodes for fixed orbitals). This can easily(?) be done transparently.

codecov[bot] commented 3 years ago

Codecov Report

Merging #386 (c9c16e5) into master (5ec4c2b) will increase coverage by 0.01%. The diff coverage is 100.00%.

Impacted file tree graph

@@            Coverage Diff             @@
##           master     #386      +/-   ##
==========================================
+ Coverage   69.95%   69.96%   +0.01%     
==========================================
  Files         177      177              
  Lines       14110    14115       +5     
==========================================
+ Hits         9871     9876       +5     
  Misses       4239     4239
Impacted Files Coverage Δ
src/driver.cpp 70.16% <100.00%> (ø)
src/qmoperators/one_electron/NuclearOperator.cpp 92.20% <100.00%> (+0.54%) :arrow_up:

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 5ec4c2b...c9c16e5. Read the comment docs.