*-- Author : S.Burke / J.V. Morris
      SUBROUTINE FKMXM(X,G,Q)
**********************************************************************                                        
*                                                                    *                                        
* Optimised matrix transformation.                                   *                                        
* Used to calculate the smoothed covariance matrix; the gain matrix  *                                        
* (G) is known to have G(3,x)=delta(3,x).                            *                                        
*                                                                    *                                        
**********************************************************************                                        
**********************************************************************                                        
* First work out (G*X)j1 k2 for each k2 (contracting over k1).                                                
* Then do: (Q)j1 j2 = Sum(k2) (G*Q)j1 k2 * (G)j2 k2                                                           
*