VMD-L Mailing List

From: Karl Decker (kaelenmitharos_at_gmail.com)
Date: Fri Dec 16 2016 - 11:27:49 CST

One of my compatriots on this project reminds me volmap communicates with
the rest of our code by a file on disk and a disk operation always takes
times comparable to a second, which is too long for our preference. Is it
too presumptuous to ask for the source code for the volmap distance
function so we can keep the result of the volmap function in memory rather
than writing it to disk immediately? That would be amazingly helpful.

Thanks,
Karl

On Fri, Dec 16, 2016 at 10:36 AM, Karl Decker <kaelenmitharos_at_gmail.com>
wrote:

> Thanks for the quick reply! The distance map command really is the main
> one as far as Volmap goes. We have some other custom calculations that
> we're hoping to implement in C++ somehow, as the inefficiency of doing them
> in Tcl is another show-stopper, but it's not clear how we'll do that at the
> moment.
>
> Thanks,
> Karl Decker
>
> On Thu, Dec 15, 2016 at 11:27 AM, John Stone <johns_at_ks.uiuc.edu> wrote:
>
>> Hi,
>> The 'volmap' commands don't exploit GPU acceleration yet, but some of
>> them
>> have been superceded or augmented by other methods of doing similar
>> things.
>> In the longer term, the volmap commands are getting updated/replaced by
>> new
>> implementations that do equivalent calculations but that support GPUs and
>> also some of the new many-core CPUs that also require a very different
>> algorithmic approach. If you can tell me what operations you need to be
>> very fast (other than the "distance" calculation below, I can likely give
>> some suggestions on how we could accomplish that soon.
>>
>> Cheers,
>> John Stone
>> vmd_at_ks.uiuc.edu
>>
>> On Thu, Dec 15, 2016 at 11:13:26AM -0600, Karl Decker wrote:
>> > Hello all! I'm running a script that uses the Volmap distance mapping
>> > command. The line looks like:
>> > volmap distance $sel -res 1 -cutoff 7 -minmax "{-40 -40 -40} {40 40
>> 40} -o
>> > foo.dx
>> > where $sel is an atomselect selection, as in [atomselect top
>> protein].
>> > It does not seem, based on performance, that this Volmap operation is
>> > being accelerated with CUDA. How can we make it so?
>> >
>> > Thanks,
>> > Karl
>>
>> --
>> 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/
>>
>
>