FW: Hanging QS PDB 6CGV

[kristen.browne@…]

Sorry sent this to the admin email by accident:

PDB QS 6CGV

This one looks like it's hanging after:
Using preset: Initial Styles / Original Look
Preset implemented in Python; no expansion to individual ChimeraX commands available.

Complete log is attached.  This is similar to the other hang for PDB QS 6B1T.
Kristen Browne, MSc, MscBMC
Contractor - MSC, Inc. | A Guidehouse Company
3D Modeling and Biovisualization Specialist
Bioinformatics and Computational Biosciences Branch (BCBB)<https://www.niaid.nih.gov/research/bioinformatics-computational-biosciences-branch>
OCICB/OSMO/OD/NIAID/NIH

5601 Fishers Lane, Room 4A60
Rockville, MD 20852
Office 202-253-5228

logs (7).txt​

[kristen.browne@…]

Added by email2trac

[pett]

Also 6 million atoms

[pett]

Tom,

The "color zone #1 near #1 distance 20" command on this structure takes a tremendous amount of memory, >20GB, whereas opening the structure is 3.7GB and adding a blobby surface (surface #1 enclose #1 resolution 18 grid 6) adds another 1.1GB. Is this memory use something you feel is to be expected, or does it seem like a bug?

--Eric

[Tom Goddard]

In my test the color zone increased memory use about 0.5 GB, reaching total of 6 GB. Tested with recent development build on Mac M1. I would not expect the color zone to increase memory use much.

open /Users/goddard/Downloads/6cgv-assembly1.cif
color bypolymer
surface #1 enclose #1 resolution 18 gridSpacing 6
color zone #1 near #1 distance 20

[pett]

Okay, I didn't provide the correct "surface" command to replicate the behavior with "color zone". The command is "surface #1 resolution 18 grid 6" which produces 1260 gaussian surfaces. Whereas "surface #1 enclose #1 resolution 18 grid 6" sidesteps the behavior. So is large memory usage from "color zone" something you would expect from the former command? Regardless, I will probably switch the NIH presets to the latter command.

[pett]

Put "enclose #1" into the blob presets' surface command. Will send updated bundle to Michal separately.

[Tom Goddard]

Regarding "color zone #1 near #1 distance 20" taking a lot of memory if 1260 chain surfaces are made with "surface #1 res 18 grid 6", I tried it and it reached 100 Gbytes and then my Mac asked to kill it. A look at the code shows why. By default color zone is set to automatically update the colors if the surface changes. That means it needs to keep a copy of the 6 million atom positions, 24 bytes per atom, so about 150 Mbytes, and each of the 1260 colored surfaces holds its own copy, so over 150 Gbytes. The easy fix is to tell color zone not to auto-update when the surface changes "color zone #1 near #1 distance 20 update false". That makes it use less than 0.5 Gbytes. The code could be improved so that all surfaces don't get their own copy of the points. The reason that happens is because the atom coordinates are transformed in the the surface model coordinate system which makes a copy. But all the surface models in this case have the same coordinate system so only one copy of the points would be needed. This is a pretty rare problem but I put in that optimization (daily build, not in 1.5).