Skip to content

Grimme-D3 phonon problem in parallel

giannozz requested to merge giannozz/q-e:d3-ph-fix into develop

Grimme-D3 dynamical matrix contribution was computed only on processor zero, but the dynamical matrix is diagonalized on all processors. Quick and dirty fix: broadcast the Grimme-D3 contribution. A better fix would be to diagonalize on a single processor and broadcast.

Merge request reports