observation on stanford

[shadowscotland]

Any one know that the GROMACS33 & GROST cores are?

Originally thought it was a smp or gpu core but there are listed as GRO-SPM and GROGPU respectively.

On a simular note the WU 3301 on my xeon is doing double steps (not seen this before either )

[11:19:59] Project: 3301 (Run 1918, Clone 1, Gen 1)
[11:19:59]
[11:19:59] Assembly optimizations on if available.
[11:19:59] Entering M.D.
[11:20:20] (Starting from checkpoint)
[11:20:20] Protein: p3301_ribcomp
[11:20:20]
[11:20:20] Writing local files
[11:20:22] Extra SSE2 boost OK.
[11:20:24] Writing local files
[11:20:24] Completed 0 out of 25000 steps (0)
[11:28:49] Writing local files
[11:28:49] Completed 500 out of 25000 steps (2)
[11:37:13] Writing local files
[11:37:14] Completed 1000 out of 25000 steps (4)
[11:45:57] Writing local files
[11:45:57] Completed 1500 out of 25000 steps (6)
[11:54:42] Writing local files
[11:54:42] Completed 2000 out of 25000 steps (8)
[12:03:06] Writing local files
[12:03:07] Completed 2500 out of 25000 steps (10)
[12:11:28] Writing local files
[12:11:29] Completed 3000 out of 25000 steps (12)
[12:19:52] Writing local files
[12:19:52] Completed 3500 out of 25000 steps (14)
[12:28:14] Writing local files
[12:28:14] Completed 4000 out of 25000 steps (16)
[12:36:33] Writing local files
[12:36:34] Completed 4500 out of 25000 steps (18)

Edit: I know it's a low pointer but still give above average ppd with SSE2 (AMD's beware)

 

[A.N.Other]

GROST: Gromacs Serial replica exchange with Temperatures

I will release the first project p2903 for Serial Replica Exchange sampling method(SREM) to beta soon. Here is some information about the project:

p2903: Serial Replica Exchange for protein BBA5
credit: 132
timeout: 26 days
deadline: 38 days
upload size: ~3M
download size: <1M

What is SREM?
In the MD simulaitons of large molecules such as proteins, molecules are easy to be trapped in local free energy minima, which slows the sampling of phase space. In the Replica Exchange Method, a few simulations are performed at different temperatures simulatineously, and exchanges of configurations are tried every certain a period of time. Even if a trajectory is temporarily trapped in a local minimum, the simulation can escape by exchanging with a higher temperature configuration.

The usual implementation of REM requires parallel processors where the different replicas must be run synchronously which is not suitable for Folding@home. Serial Replica Exchange(SREM) is a method that is equivalent to standard REM in terms of effciency, yet runs asynchronously on a distributed network of computers.

The gromacs33 core is an update to the original gromacs, although I'm not sure exactly what it does on top. It's a linux-only client.

 

[shadowscotland]

thanks - knew someone would know (and save me looking for myself )

 

[BillytheImpaler]

If you get a Gro33 WU the core will be called FahCore_a0.exe. Every WU I've had under this core is a dog, even on fast machines.