VMD-L Mailing List

From: Matthew Kenney (matthew.kenney_at_cellsignal.com)
Date: Mon Jul 16 2018 - 15:49:49 CDT

Forgot to address one test you suggested--
Vmd's startup info, both within and locally on my machine has the same
startup info, line-by-line. Let me know if there are any other tests I
should run. Thanks, and no rush.

Best,
Matt

On Mon, Jul 16, 2018 at 4:37 PM Matthew Kenney <
matthew.kenney_at_cellsignal.com> wrote:

> Hi John,
>
> Sorry about the binaries issue. I have now removed them from the image and
> asked that users download them directly from the vmd site and place them
> alongside the Dockerfile that builds the binaries into the container. I
> have also made my Dockerhub repository private, in case this is a violation
> as well.
>
> Thank you for your advice on the GLX rendering. To make sure VMD was
> working well with these containers, I ran the same "test scenes" that you
> included in your VMD image hosted on nvcr.io and compared the performance
> of the containers I just created to yours, and found that they achieved the
> quicksurf calculation on the HIV capsid molecule at the approximately the
> same rate. I ran these tests a few different times, using both
> openglpbuffer and opengl as --dispdev options, and found that both the
> nvcr.io container and the containers I just published were
> comparable...sometimes the one I built won out by a second or two, and
> sometimes the nvcr.io container won out by a second or two.
>
> I had not thought of using the glxinfo command, but I just installed
> mesa-utils in both of the new containers and ran glxinfo, and found that
> they had exactly the same output as running glxinfo from my local machine,
> and loaded all the same glx extensions as my local machine. Crucially, each
> container displayed "direct rendering: Yes" and recognized my cuda
> accelerated GPU.
>
> If you have any other concerns, please let me know. Also, if you have any
> other advice about the glx performance, feel free to share as I would
> definitely like to make sure that these containers are performing optimally.
>
> Best,
> Matt
>
>
> On Mon, Jul 16, 2018 at 4:06 PM John Stone <johns_at_ks.uiuc.edu> wrote:
>
>> Hi,
>> I'm still out of the office, so I have to keep this short:
>> Before you assume that this is working well, you need to address
>> whether or not this docker image is using direct or indirect GLX.
>>
>> I mentioned this in the initial discussions but I haven't seen any
>> indication that adding the missing X server pieces is actually
>> allowing GLX passthrough (with a direct context) vs. using indirect
>> GLX via network protocol. Using indirect GLX will be substantially
>> lower performance than direct GLX. The EGL containers I have built
>> use a hardware-direct context, so they run at full performance.
>>
>> If you use GLX, you have to check on direct vs. indirect otherwise
>> VMD may perform quite slowly vs. what it is normally capable of.
>> You can determine which way the container is running by looking
>> at the VMD startup text and compare with the output of 'glxinfo'.
>>
>> I should also note that the VMD license doesn't allow redistribution
>> of the binaries without a special arrangement. We are allowing the
>> containers we built to be obtained via NGC, but that is not generally
>> the case for just anyone. This is necessary because we have to report
>> user statistics to our funding agency.
>>
>> Best,
>> John Stone
>> vmd_at_ks.uiuc.edu
>>
>> On Mon, Jul 16, 2018 at 11:00:16AM -0400, Matthew Kenney wrote:
>> > Hi Everyone,
>> > If anyone has the need to use a fully visual version of VMD in
>> > containerized form, feel free to pull from my Dockerhub or clone
>> from my
>> > Github. I've provided both a Centos7 and Ubuntu16.04 option for the
>> two
>> > VMD containers, and each can be run in either text mode or visual
>> mode.
>> > Images can be easily modified to fit your system's CUDA version (or
>> lack
>> > of CUDA support). Further, each are below 3GB, so you shouldn't have
>> to
>> > worry too much about storage space. Hope these can be of use to you!
>> > Dockerhub:Â [1]https://hub.docker.com/r/mkenney1/vmd-cudagl/
>> > GIthub:Â [2]https://github.com/kenneym/vmd-cudagl
>> > Best,
>> > Matt
>> > This message contains information which may be confidential and/or
>> > protected by attorney-client privilege. Â Unless you are the
>> addressee,
>> > you may not use, copy or disclose to anyone this message or any
>> > information contained in this message. Â If you have received this
>> message
>> > in error, please send me an email and delete this message. Â Thank
>> you.
>> >
>> > References
>> >
>> > Visible links
>> > 1. https://hub.docker.com/r/mkenney1/vmd-cudagl/
>> > 2. https://github.com/kenneym/vmd-cudagl
>>
>> --
>> NIH Center for Macromolecular Modeling and Bioinformatics
>> Beckman Institute for Advanced Science and Technology
>> University of Illinois, 405 N. Mathews Ave, Urbana, IL 61801
>> http://www.ks.uiuc.edu/~johns/ Phone: 217-244-3349
>> http://www.ks.uiuc.edu/Research/vmd/
>>
>
>
> --
> ___________________
> Matt Kenney
> Summer Intern for Ivan Gregoretti
> Bioinformatics Dept.
> 

-- 
___________________
Matt Kenney
Summer Intern for Ivan Gregoretti
Bioinformatics Dept.
-- 
This message contains information which may be confidential and/or 
protected by attorney-client privilege.  Unless you are the addressee, you 
may not use, copy or disclose to anyone this message or any information 
contained in this message.  If you have received this message in error, 
please send me an email and delete this message.   Thank you.