Article: Adaptive boundaries in multiscale simulations

Moderators: Site Moderators, FAHC Science Team

Post Reply
wuffy68
Posts: 169
Joined: Wed Jun 04, 2014 11:06 pm
Hardware configuration: 1x nVidia 1080Ti, 1x nVidia 1070, 1x nVidia 1060, 1x nVidia 750Ti, AMD Radeon R7 M460
Location: Roxborough, Colorado USA
Contact:

Article: Adaptive boundaries in multiscale simulations

Post by wuffy68 »

Published in the The Journal of Chemical Physics Communication: Adaptive boundaries in multiscale simulations by Jason A. Wagoner and Vijay Pande

Will Folding@home molecular simulations benefit from the Adaptive Boundaries outlined in the article? If so, will this affect Work Unit size in terms of; number of atoms, TPF, download and upload file sizes?

Which FAH projects could benefit most from the implementation of Adaptive Boundaries? CPU or GPU projects? Which disease research could benefit most from a modified modeling technique? Cancer, Neural Cognitive or Motor Neuron Disorders, or Infectious Disease?

Any additional information would be greatly appreciated - Thank you.

Edits: minor formatting changes, added article authors.
1x nVidia 1070, 1x nVidia 1060 3g,
1x nVidia 970, 2x nVidia 960,
1x nVidia 555, 1x AMD R7, 2x AMD 295,
6x i5 CPU-only rigs
gbowman
Posts: 208
Joined: Fri Nov 30, 2007 9:51 pm

Re: Article: Adaptive boundaries in multiscale simulations

Post by gbowman »

Great question. The paper you posted about is a really nice advance. There is ongoing effort to integrate it into OpenMM and it would be great to use on FAH someday. We don't plan a new core in the immediate future though. Changing the number of particles in a simulation on the fly requires some significant advances in book-keeping as existing molecular dynamics codes typically been able to safely assume the number of particles is constant. Deployment on FAH is further complicated by the fact that the current points system would also need to be altered. With adaptive boundaries, two users who simulate the same system for the same amount of simulation time could end up simulating systems with dramatically different numbers of atoms, resulting in large variations in how much wall-clock time running a WU requires. As the science advances, we'll be able to solve these problems for simulations run on dedicated clusters and then adapt FAH based on the lessons we learn.
wuffy68
Posts: 169
Joined: Wed Jun 04, 2014 11:06 pm
Hardware configuration: 1x nVidia 1080Ti, 1x nVidia 1070, 1x nVidia 1060, 1x nVidia 750Ti, AMD Radeon R7 M460
Location: Roxborough, Colorado USA
Contact:

Re: Article: Adaptive boundaries in multiscale simulations

Post by wuffy68 »

gbowman wrote:As the science advances, we'll be able to solve these problems for simulations run on dedicated clusters and then adapt FAH based on the lessons we learn.
Great explanation! I would suggest Pande Labs issue a similar statement with every published paper to the community either here, or on reddit. Most experienced folders have a crude idea what the implications are, but for new folks who may only have high-school level chemistry and physics, presenting these in more "bite-size" chunks would go a long way in attracting, and retaining new community members. Like you say, it's a great advancement ... now we need to telegraph that to the masses in a way average news and social media readers can digest.

Thank you again Dr. Bowman
Last edited by wuffy68 on Wed Apr 25, 2018 4:20 pm, edited 1 time in total.
1x nVidia 1070, 1x nVidia 1060 3g,
1x nVidia 970, 2x nVidia 960,
1x nVidia 555, 1x AMD R7, 2x AMD 295,
6x i5 CPU-only rigs
gbowman
Posts: 208
Joined: Fri Nov 30, 2007 9:51 pm

Re: Article: Adaptive boundaries in multiscale simulations

Post by gbowman »

Thanks, and good idea. I'll try to make a habit of that for our publications.
Post Reply