[gmx-users] Using the CHARMM TIP3P water model in Gromacs

Erik Lindahl lindahl at cbr.su.se
Wed Oct 11 09:42:44 CEST 2006

Hi Mark,

On Oct 11, 2006, at 7:57 AM, Mark Abraham wrote:

> Mostly correct - non-CHARMM TIP3P has LJ on the oxygen. However,  
> the decision about which code to use is made on-the-fly by mdrun.  
> If the molecule matches the pattern of a water molecule and only  
> has LJ parameters on the oxygen then the optimized inner loops are  
> used for water-nonwater and water-water interactions. The  
> optimization, obviously, omits calculating LJ interactions with  
> hydrogens.
> You can see what is happening by looking at the accounting at the  
> end of .log files... lines like "Coul(T) + LJ [W3-W3]" indicate the  
> time spent calling the innerloop routine that does coulomb lookups  
> on a table, computes LJ and does so for water-water interactions.  
> If the description of CHARMM TIP3P water is correct, you should see  
> no lines with the [W3] indicators, and thus the LJ terms are being  
> treated correctly.
> As an aside, I did develop assembly innerloops for CHARMM TIP3P  
> water for x86 some time back and posted here. My main compute  
> platform isn't x86 though, so I've not used them in anger. See  
> discussions back in Feb and March this year on gmx-developers.

I've thought about that too, and will enable some sort of similarly  
optimized assembly loops on all architectures when I have time. To  
tell the truth, my main reason isn't the charm "mutation" of  
Jorgensen's water but the abundance of CH2/CH3 groups in lipid  
chains. The main problem is that there are quite a few more  
combinations (CH2-Water, CH2-CH3, CH3-water, etc), so I likely won't  
optimize them quite as hard for a specific case (when we do strict  
water-water interactions we don't even look up charges or LJ  
parameters), but it should help charmm TIP3p too.



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://maillist.sys.kth.se/pipermail/gromacs.org_gmx-users/attachments/20061011/6ad190a1/attachment.html>

More information about the gromacs.org_gmx-users mailing list