[Hadstruct] Bug fix in redstar_corr_graph

[Robert Edwards]

All,
   Sadly, in   redstar_corr_graph,  there was a bug that affected nucleon 3pt functions when wrapping around the lattice. It is manifested as a spurious minus sign change when the t-slice steps around the edge of the lattice. If this is nucleons at source/sink, and you have a t_origin+t0 combo such that stepping up to the t_sep, then this bug impacts the output.

   To be clear, nucleon & mesons 2pts are fine. Only 3pts for nucleons were affected (mesons are fine). Doesn’t matter if the insertion is disconnected or gluonic.

   Also, this bug was present day-one for redstar_corr_graph. If you have data generated using redstar_gen_graph/redstar_npt, then it is fine.

   The fix was very simple, and is applied to redstar “devel” and “devel-pdf”. No changes to any other library needed.

   For Chris Chamness, this spurious sign flip will mess up your 3pts. Simplest thing is to rerun the contractions on a config if the  t_origin+t0+t_sep is greater than Lt=64.

Robert



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/hadstruct/attachments/20221221/43213470/attachment.html>