Open sangallidavide opened 1 month ago
The PL tests in the test-suite are tagged as BROKEN
This was fixed in this branch: https://github.com/sangallidavide/yambo/tree/fixes-photoluminescence
@sangallidavide Looking at the changes, I see that K_diago_residuals are changed. Could you please tell me what changed compared to the previous version. I am asking this because, I am about the complete the Ydiago interface and rewrote some of these functions to take care of parallel distribution of eigen-vectors . Here is my new implementation : https://github.com/muralidhar-nalabothula/yambo/blob/master/src/bse/K_residuals.F
Yeah, I recently merge in tech master with changes done some time ago. Sorry it overlapped with your development on distributed residuals.
In the previous version:
In the new version:
Bse run (same inputs) with
BSEprop= "photolum"
crashed with array mismatch betweenBS_T_grp(i_T_g)%dipoles_opt(:,i_T,1)
andPL_weights(:)
in file https://github.com/yambo-code/yambo/blob/master/src/bse/K_IP.F#L188 This seems like a valid bug as printing their sizes gave me different values (1 and 3) respectively. also in https://github.com/yambo-code/yambo/blob/master/src/bse/K_components_folded_in_serial_arrays.F#L111One more In https://github.com/yambo-code/yambo/blob/master/src/bse/PL_diago_residual.F ,the shape of BSS_dipoles_PL is assumed to be
(BS_H_dim,3)
, but its allocated array shape is different(3,BS_H_dim)
in https://github.com/yambo-code/yambo/blob/master/src/modules/mod_PHOTOLUM.F#L47 Again in https://github.com/yambo-code/yambo/blob/master/src/bse/K_components_folded_in_serial_arrays.F#L111 the shape is assumed to be(3,BS_H_dim)
. This is an undefined behaviour, not sure how the test cases passed in this caseSize mismatch error : https://github.com/yambo-code/yambo/blob/master/src/bse/PL_diago_residual.F#L126
BSE inputfile : bse.txt
Originally posted by @muralidhar-nalabothula in https://github.com/yambo-code/yambo/issues/111#issuecomment-2249351535