Closed smparker closed 5 months ago
Shane, this is likely a logic error in the Engine
... I will try to reproduce. Unlike 1e 2c and 2e 4c derivatives which have been stressed hard 2e 2c and 2e 3c have not been tested.
So could this be a bug? Or maybe it is time for me to learn the C interface...
NOOOOOOOOOOOOOOOOOOOOOOOO ...
I learned a couple more details about this that I wanted to share just in case they'd be helpful:
Okay, I pretty much know what's going on now. It seems to be caused by using different max angular momentum values for the eri2 and eri3 options. Originally, I used --with-max-am=4 --with-eri-max-am=4 --with-eri2-max-am=6 --with-eri3-max-am=6
when generating the library. When I replace all that with just --with-max-am=6
then the above code seems to work just fine. I can live with that as a solution.
@smparker this is addressed by #324 , will try to push out 2.8.2 today/tomorrow
Hi there,
I've run into a crash caused by what looks like an internal assertion failing. I'm still not sure if it's a bug or if I'm just misunderstanding the interface, so I'm hoping you can help figure it out. I'm trying to compute the first derivative of the 3-center ERIs using the c++11 interface. The basic error comes from line 1745 of my
engine.impl.h
(I realize the line number might change with configure options) and the failing assertion readsbuildfnptrs_[buildfnidx] && "null build function ptr"
.I have been able to make a minimum example that shows the same behavior that I'll paste below. But first, details.
Braket::xs_xx
Braket::xs_xs
engine.compute()
orengine.compute2()
interfacesmax_l
doesn't seem to change this at allThe test code that triggers this is below.
The full output I see from this is:
So could this be a bug? Or maybe it is time for me to learn the C interface...
update: I updated the code snippet and results to correspond to 2.8.1