If we do not converge soon on porting vectorization to automatic code generation, one alternative could be to try manual vectorization of ggtt(g(g)) in the meantime. In principle all ixx/oxx/... funxtions are vectorized. What is missing is the relevant FFV functions and their calling from alculate_wavefunctions. Maybe some of this can be done by simple query/replace in the text editor.
Low priority, just an idea that was mentioned. Keep this as a placeholder. Of course this will become irrelevant if/when we move to automatic code generation (dedicated meeting scheduled for Sep 16 https://indico.cern.ch/event/1061524)
This can be cancelled. Vectorization is now fully implemented in code generating code in epochX, see issue #244 (epochX3 PR #253 and epochX4 PR #267 and #270)
As discussed in the June 14 meeting, https://indico.cern.ch/event/1047119
If we do not converge soon on porting vectorization to automatic code generation, one alternative could be to try manual vectorization of ggtt(g(g)) in the meantime. In principle all ixx/oxx/... funxtions are vectorized. What is missing is the relevant FFV functions and their calling from alculate_wavefunctions. Maybe some of this can be done by simple query/replace in the text editor.
Low priority, just an idea that was mentioned. Keep this as a placeholder. Of course this will become irrelevant if/when we move to automatic code generation (dedicated meeting scheduled for Sep 16 https://indico.cern.ch/event/1061524)