Page 1 of 1
Projects 7610 & 7611 Adv -> Full FAH
Posted: Mon Jun 13, 2011 11:41 pm
by tjlane
Projects 7610 and 7611, shooting of the FiP35 WW domain from the ultra-long trajectories of Shaw et al (Science, 2010), are now in open beta (adv). These projects compare FAH-style simulation methodology to more traditional techniques, recently implemented in an optimized fashion by DE Shaw Research.
Both projects have the same points/core info:
Points: 788
Preferred deadline: 11.71 days
Final deadline: 19.52 days
k-factor: 0.75
Core: A4
Operating system(s): Linux, Win32
Please contact with any issues!
Edit: A4 core not compatible with OSX as noted in open beta. Updated to reflect this.
Re: Projects 7610 & 7611 on Adv
Posted: Wed Jun 15, 2011 7:45 pm
by tjlane
These projects have been rolled back to closed beta pending issues with the A4 core. Updates forthcoming.
Re: Projects 7610 & 7611 on Adv
Posted: Fri Jun 17, 2011 5:26 pm
by tjlane
P7600 has returned to Advanced Methods. During testing we discovered and crushed a number of bugs, listed below.
This project runs the A4 core, which should be compatible with both SMP and uniprocessor clients. It should use all available cores a machine has allocated to folding. There were some issues with older clients not recognizing this, and running an SMP job on a single core.
Known/Fixed Issues:
-Project was assigning to OSX clients, though the A4 core is not compatible with OSX v6. A4 support is included with OSX v7. P76** are no longer assigning to OSX.
-Clients below v6.34 were unstable with the A4 core. Project is no longer being assigned to sub-6.34 clients.
-There may be a problem with the '-smp' flag. It seems for some clients the project wanted to run as a uniprocessor job unless the specified the number of threads to run, (e.g. '-smp' vs. '-smp 8'). This is hopefully resolved by restricting the min client version to 6.34, but at this point we are unsure.
-Confusion over whether A4 is smp or uniprocessor. Answer: it should be able to run as *either*, depending on a user's setup/argument specifications.
-Some simulations are known to be unstable. This is an unavoidable aspect of simulation methodology in general. Some experimental aspects of the P76** series mean this may occur more often than not early on in simulations. This may be conflating other issues. I will do my best to aggressively shut down bad units if they occur. Some of these have been dealt with.
Thanks for your patience & help FAH! I will continue to post here with relevant information.