
***** do mapping in scalar_sum when fftw is used. 
use the ready-to-go routine in fftw3_pme_init.f90 

***** Multiple file reading. 
Read_setup and read_solute changed already. Multiple reading works
   only in MPI: each trajectories reads starting coordinates 
from a file specified in kconf unit or kccord_slt. 

1) In read_parameters for the multiple topologies one should do the following
(ony when MPI)

PROTEIN
ala-o ala -arg etc etec
END 

JOIN SYSTEM 1
drug 1 protein
END 

JOIN SYSTEM 2
drug 2 protein
END 

...

One the topology is knwon, it must be guessed which solute must
   construct the trajectory based on the IPROC variable (as done 
in read_setup). 


 2) In dumprs when save_all_restart is specified, then the routine
   should write normally the restart file for restarting the REM while
   save  restart must be accumulated in a RESTART_DIR (specified in
   input mandatorly) with index runing from 1 to n for each system.  
   taking the conf only form para0=1 replicas from all batteries.

3) see also the multiple-system restart file for interrupted
   alchemical run
