From: Thomas Evangelidis (tevang3_at_gmail.com)
Date: Sat Dec 01 2012 - 12:09:03 CST
I was only talking about was the CPU version. For the GPU
version, the problem
seems to be worse --- the VDW energy part doesn't match either.
Now, I really can't say anything with absolute certainty as the simulation
you're doing involves two modifications of the NAMD code that I'm not familiar
with -- one is GBIS and the other is GPU. My guess is that the energy reduction
is where the problem is, meaning that aMD is not getting the correct ELECT
energy when implicit solvent is used (and the correct VDW when GPU is used). So
aMD total boost or dual boost cannot be trusted with implicit solvent. But the
bonded calculations shouldn't be affected, so aMDdihedral boost should be fine.
However, like I said, this is my guess only.
On Nov 30, 2012, at 4:29 AM, Thomas Evangelidis wrote:
> Hi Yi,> > Many thanks again for your thorough explanation. I don't use the boost on> the total energy, just the boost on the dihedral energy. I started a> production phase with implicit solvent and aMD on the GPUs of the> supercomputer. Do you mean that the dihedral aMD does not work correctly> in implicit solvent with NAMD-CUDA? Please let me know to stop it now. I> don't want to waste my limited GPU time.> > many thanks again,> Thomas
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:22:19 CST