[R-sig-ME] Generalized least squares program

[Gang Chen]

I'm using function gls in nlme to run some extended linear model with  
generalized least squares many many times in a loop with the same  
design matrix X and same residual covariance structure but different  
values for the response variable. A lot of time is wasted in  
reformulating the same model structures and overhead matrix  
calculations. Such a waste makes the runtime range from one day to  
one month. Are there some slots available in gls that would allow me  
to avoid the unnecessary waste? If not, is there a counterpart of gls  
for extended linear model with generalized least squares available in  
lme4 that would give such slots?

Thanks,
Gang