Diagonalizing Hessian: gettting_approximate_run_time

[karolk]

Hi,

I am running all-atom NMA (NO rtb, NO 4-bead) for an exceptionally large structure, which consists of ~ 5000 of Calpha atoms and total number of atoms is around 38k. It is diagonalizing hessian for more than 10 days...and I can't see the end. It needs like 250-300 GB of hard drive to store the matrix.

Is there any way of deducing an approximate run time for all atom NMA, for example, knowing how much time it took, using one of the ENM force-fields, or knowing the size of the hessian matrix? Yes, I know, it is not the way to go with big structures, but I have resources and some time, so I want to try.

thanks a lot

[Xinqiu Yao]

It is hard to tell the time to take, but for hessian matrix the size is 3N-by-3N, where N is the total number of atoms.

If the system is large, you might want to try C-alpha based ENM instead...

[karolk]

Hey,

I have already tried different ENM approaches, but I also want to try this one. Based on this what I read, all-atom still gives best results, so I want to know whether it pays of using it for my systems. All very similar and all very large.

My suggestion would be to implement a counter, kind of clock, which would display an approximate time, which is required to diagonalize hessian. For standard systems you don't need it, but for large ones, it would be useful

You can close this ticket. Thanks for your input.

[Lars Skjærven]

eigen value decomposition is performed with eigen() and I don't think its possible to implement a progress bar for this.