Slater determinant

[Lucas Wagner]

FYI, everyone, I plan to make some changes in the Slater determinant
object over the next week or so, which should dramatically improve the
multideterminant speed in the default object. I will try my best not
to commit bugs to mainline (that is, I will test thoroughly before
committing!), but be aware, and please let me know if you find a
problem.

Cheers!

Lucas



##################################

Dear Paul,

Thanks for the quick response. I do need to do some coding one way or
another, since there are several new features that I need fast
multideterminants to support (analytic derivatives of the laplacian
wrt parameters, fast 1-RDM evaluation, complex orbitals). I may as
well clean up while I do it.

Given what you said, my current plan is to implement Bryan's algorithm
and depreciate FSlat_wf. This makes me feel a bit sad since it was
nice work, but I think it's the correct decision code-wise. Let me
know if you have some objections.

Best,

Lucas

On Thursday, August 9, 2012 at 9:53 PM, Kent, Paul R. wrote:

Hi Lucas

Well, my algorithm is already implemented and works :-)

However, Bryan's algorithm is faster. Not by much, but it is faster.
The low order polynomial terms will be smaller in Bryan's algorithm
for most runs. Unfortunately the Urbana folks never properly
implemented my algorithm (that is what Jeongnim says - they didn't
solve the graph problem properly) so I can't give you a figure for a
real calculation. If you are motivated to write some code, definitely
implement Bryan's stuff. The code is easy to write and easy to test…

Cheers, Paul

On 9 Aug 2012, at 6:30 PM, Lucas Wagner wrote:

Dear Paul,

I hope you're doing well! I have a quick question: I'm currently
devoting a little time to clean up the Slater determinant stuff in
QWalk and part of that is implementing some sort of fast
multideterminant evaluation in the standard object. I'm currently
weighing between pulling your implementation in and using Bryan
Clark's implementation (JCP 135, 244105). His seems more simple to me
and he claims that it's faster. Is this true? Do you have any
thoughts?

Thanks, best regards,

Lucas