Closed acmnpv closed 6 years ago
I need to update the fix because I didn't do the correct division by the number of coordinates.
The original code had ncoords defined as
ncoords = 3*mask%included (line 153 in trj.f90), each atom gets assigned three raw floating point numbers
while the new code has
ncoords = mask%included (line 169 in trj.90), where each atom gets one coordinate vector.
Thanks for the notification!
On 30/01/18 13:54, Mauricio Esguerra wrote:
This is just reverting to original.
https://github.com/qusers/qsource/blob/testing/akesandgren/src/avetr.f90
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/qusers/Q6/pull/9#issuecomment-361584813, or mute the thread https://github.com/notifications/unsubscribe-auth/AGKLaCAIv-SEjDzrcuk6yIduc9Q8TRCMks5tPxEWgaJpZM4RyL1F.
Hej,
OK, so, I checked both locally in the mac, and at kebnekeise using and intel compiled Qprep. Both created the same exact pdb. Also checked against vmd and there are minor differences, which are most likely due to differences in the implementations. The VMD and Q6 average files for David's example are attached with .txt extension since github doesn't like the .pdb extension because it thinks it a python debugger file.
Hej,
thanks again for reviewing and testing! I'll delete the bugfix branch so we can use it again later.
Cheers!
The coordinate array in avetr was still allocated as if we had the old coordinate handling and not the vector based one. This lead to the array being too small and further errors.
Fixes #8