C $Id: the_model_main.F,v 1.2 2006/05/12 22:26:20 ce107 Exp $
      SUBROUTINE THE_MODEL_MAIN
C     /==========================================================\
C     | SUBROUTINE THE_MODEL_MAIN                                |
C     | o Master controlling routine for model using the MITgcm  |
C     |   UV parallel wrapper.                                   |
C     |==========================================================|
C     | THE_MODEL_MAIN is invoked by the MITgcm UV parallel      |
C     | wrapper with a single integer argument "myThid". This    |
C     | variable identifies the thread number of an instance of  |
C     | THE_MODEL_MAIN. Each instance of THE_MODEL_MAIN works    |
C     | on a particular region of the models domain and          |
C     | synchronises with other instances as necessary. The      |
C     | routine has to "understand" the MITgcm parallel          |
C     | environment and the numerical algorithm. Editing this    |
C     | routine is best done with some knowledge of both aspects.|
C     | Notes                                                    |
C     | =====                                                    |
C     | C*P* comments indicating place holders for which code is |
C     |      presently being developed.                          |
C     \==========================================================/

C     == Global variables ===
#include "SIZE.h"
#include "EEPARAMS.h"
#include "PARAMS.h"
#include "CG2D.h"
#ifdef ALLOW_MPI
#include "mpif.h"
#include "MPI_INFO.h"
#endif
#if defined(USE_PAPI_FLOPS) || defined(USE_PAPI_FLIPS)
#include "PAPI.h"
#endif

C     == Routine arguments ==
C     myThid - Thread number for this instance of the routine.
      INTEGER myThid

C     == Local variables ==
      INTEGER I
      REAL*8 wTime1, wTime2
      REAL*8 fCount, fTime, fRate, wSetSize

C--   Set model initial conditions 
      CALL INITIALISE( myThid )

C--   Begin time stepping loop
#ifdef USE_MPI_TIME
      wTime1=MPI_Wtime()
#else
      CALL CLOC(wTime1)
#endif

C initialize PAPI stuff
#if defined(USE_PAPI_FLOPS)
       call PAPIF_flops(real_time, proc_time, flpops, mflops, check)
#else
#if defined(USE_PAPI_FLIPS)
       call PAPIF_flips(real_time, proc_time, flpops, mflops, check)
#endif
#endif

      DO I=1, nTimeSteps
       nIter = nIter0 + I
       CALL CG2D
      ENDDO

#ifdef USE_MPI_TIME
      wTime2=MPI_Wtime()
#else
      CALL CLOC(wTime2)
#endif

C     CALL PLOT_FIELD_XYR8( cg2d_x , 'CG2D_X  AFTER SOLVE')
C     CALL PLOT_FIELD_XYR8( cg2d_Ax, 'CG2D_AX AFTER SOLVE')
C     CALL PLOT_FIELD_XYR8( cg2d_b,  'CG2D_B  AFTER SOLVE')
C     CALL PLOT_FIELD_XYR8( cg2d_r,  'CG2D_R  AFTER SOLVE')

#if defined(USE_PAPI_FLOPS)
#if !defined(PAPI_PER_ITERATION) && !defined(PAPI_PER_TIMESTEP)
       call PAPIF_flops(real_time, proc_time, flpops, mflops, check)
       WRITE(6,'(F10.3,A7,F10.3,A36)') mflops, ' user ',
     $      mflops*proc_time/real_time,
     $      ' wallclock Mflop/s during execution' 
#endif
#else
#if defined(USE_PAPI_FLIPS)
#if !defined(PAPI_PER_ITERATION) && !defined(PAPI_PER)
       call PAPIF_flips(real_time, proc_time, flpops, mflops, check)
       WRITE(6,'(F10.3,A7,F10.3,A36)') mflops, ' user ',
     $      mflops*proc_time/real_time,
     $      ' wallclock Mflip/s during execution' 
#endif
#endif
#endif
      WRITE(6,*) 'Wall clock time = ', wTime2-wTime1
      fTime = wTime2-wTime1
      fCount = DBLE(nTImeSteps)*DBLE(cg2dMaxIters)*34.D0*DBLE(sNx)
     $     *DBLE(sNy) 
      fRate  = fCount/fTime/1.D6
      WRITE(6,*) 'PID ',myProcId+1,' OF ',numberOfProcs,' MFLOP/s = ',
     $     fRate 
      wSetSize=DBLE((sNx+2*OLx)*(sNy+2*OLy)*8*11)/1024./1024.
      WRITE(6,*) 'PID ',myProcId+1,' OF ',numberOfProcs,' MB = ',
     $     wSetSize 
      WRITE(6,'(4(1X,I4),3F10.4)') 
     &  numberOfProcs, myProcId, sNx, sNy,
     &  fTime, fRate, wSetSize

      RETURN
      END