[gmx-developers] Water innerloop

[Mark Abraham]

Yang Ye wrote:
> Thanks, Eric and Mark.
> 
> Erik Lindahl wrote:
> 
>> Hi,
>>
>>>
>>> Presumably the nonbonded list generation will take care of itself.
>>>
>>
>> Currently the detection of 'water' molecules is performed when setting 
>> up the forcerecord. If you just modify that you can have any type of 
>> molecule end up the water or water-water lists.
> 
> I have identified this code in gromacs source. The criteria for water 
> innerloop is stated in manual B.2.1. I see that CHARMM TIP3P does not 
> fulfill the requirements. Get the TIP3P-SPC water inner loop to run for 
> CHARMM TIP3P will not produce desired results. This is what I have 
> understood.

Yes. One would need to tweak code so that H atoms contribute to the vdW 
interactions.

>>> Would that be all that's necessary?
>>
>> Otherwise, modifying the general-case algorithms (for various $x, $y, 
>> $z) in gmxlib/nonbonded/nb_kernel_$x/nb_kernel$y$z0_$x.s along the 
>> lines of the existing specialisations nb_kernel$y$z[1-4]_$x.s for the 
>> specific case of TIP3P in CHARMM is the way forward. Obviously the 
>> mechanism that chooses which nb_kernel function to call would need to 
>> be expanded.
>>
>> Yes, in principle. However, if you only do it in C or Fortran the 
>> non-water Gromacs assembly loops will probably still be faster.
>>
>> You won't gain quite as much as for normal TIP3P/SPC/TIP4P either, 
>> since you're performing more calculations per coordinate load/force 
>> store.
> 
> A general question towards GROMACS source. May I know which kernel is 
> used for a system contains CHARMM TIP3P?
> For example, for SSE, what does "No Coul" in 010 (No Coul, vdw=L-J, no 
> water optimization) mean? Does it imply that it is for system without 
> coulomb interactions?

Yes.

> And why there is no 110, or 200, etc?

Documentation is spread out a bit... 
src/gmxlib/nonbonded/nb_kernel_*/README has a bit of stuff, as does 
include/types/nrnb.h and src/mdlib/ns.c - they all overlap a bit.

Mark