Hi,
Backprojection of particles happens in the E step of Refine3D, although
mathematically backprojection should be called the M step. Thus, comparison of
runtime with the maximization step alone is not fair.
> I run out of memory if I use more than 3 mpi processes.
By rebuilding RELION in CPU single precision (cmake -DDoblePrec_CPU OFF),
you can reduce the memory consumption to almost half.
This is almost always safe but we know some cases where CtfRefine produces NaN.
Best regards,
Takanori Nakane
> I am trying Ewald sphere correction on my reconstruction, but the relion_reconstruct_mpi command line program is much slower than maximization done by
> relion_refine_mpi called through the gui, at least with the options I'm using. (mostly defaults except the ewald correction options)
> In particular, I run out of memory if I use more than 3 mpi processes.
> With 3, each half map takes several hours to generate, and I need 2 with the --reverse_curvature and 2 without, correct?
> Any tips to speed up reconstruction would be appreciated!
> thanks,
> Michael Woodson,
> UTMB Galveston
>
>
> ------------------------------------------------------------------------------------------------------------------------------------------------------
>
> To unsubscribe from the CCPEM list, click the following link:
> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1
>
########################################################################
To unsubscribe from the CCPEM list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCPEM&A=1
This message was issued to members of www.jiscmail.ac.uk/CCPEM, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
|