This would be a useful feature. There have been hints that someone will work on it in time, but it seems to have been lost in the crowd of other to-do's.
#1 Updated by Matthew Zwier over 7 years ago
Joe Kaus (firstname.lastname@example.org, all credit to him) implemented this for GROMACS 4.5.3 while in our group at U. Pittsburgh, and sent this patch to Per Larsson. It at least works in single-precision, and seems to permit stability with 1 - 1.5 nm cutoffs and scaling far beyond the 1-2 core limit Justin has mentioned several times in the mailing list. It also compares favorably in both absolute energies and energy conservation compared to AMBER 11 (though I'm not sure where that data wound up, so I can't offer any hard data to back up that claim.) We don't have the time to update the patch to the latest version, or even to test it properly, but the patch is attached if it is at all useful.
The salt term is activated by specifying userint1=1, with concentration set with userreal1.