[313] | 1 | MODULE solsor_e |
---|
| 2 | !!====================================================================== |
---|
| 3 | !! *** MODULE solsor *** |
---|
| 4 | !! Ocean solver : Successive Over-Relaxation solver |
---|
| 5 | !!===================================================================== |
---|
| 6 | |
---|
| 7 | !!---------------------------------------------------------------------- |
---|
| 8 | !! sol_sor : Red-Black Successive Over-Relaxation solver |
---|
| 9 | !!---------------------------------------------------------------------- |
---|
| 10 | !! * Modules used |
---|
| 11 | USE oce ! ocean dynamics and tracers variables |
---|
| 12 | USE dom_oce ! ocean space and time domain variables |
---|
| 13 | USE zdf_oce ! ocean vertical physics variables |
---|
| 14 | USE sol_oce ! solver variables |
---|
| 15 | USE in_out_manager ! I/O manager |
---|
| 16 | USE lib_mpp ! distributed memory computing |
---|
| 17 | USE lbclnk ! ocean lateral boundary conditions (or mpp link) |
---|
| 18 | |
---|
| 19 | IMPLICIT NONE |
---|
| 20 | PRIVATE |
---|
| 21 | |
---|
| 22 | !! * Routine accessibility |
---|
| 23 | PUBLIC sol_sor_e ! ??? |
---|
| 24 | |
---|
| 25 | !!---------------------------------------------------------------------- |
---|
| 26 | !! OPA 9.0 , LOCEAN-IPSL (2005) |
---|
[719] | 27 | !! $Header$ |
---|
[313] | 28 | !! This software is governed by the CeCILL licence see modipsl/doc/NEMO_CeCILL.txt |
---|
| 29 | !!---------------------------------------------------------------------- |
---|
| 30 | |
---|
| 31 | CONTAINS |
---|
| 32 | |
---|
| 33 | SUBROUTINE sol_sor_e( kindic ) |
---|
| 34 | !!---------------------------------------------------------------------- |
---|
| 35 | !! *** ROUTINE sol_sor *** |
---|
| 36 | !! |
---|
| 37 | !! ** Purpose : Solve the ellipic equation for the barotropic stream |
---|
| 38 | !! function system (lk_dynspg_rl=T) or the transport divergence |
---|
[359] | 39 | !! system (lk_dynspg_flt=T) using a red-black successive-over- |
---|
[313] | 40 | !! relaxation method. |
---|
| 41 | !! In the former case, the barotropic stream function trend has a |
---|
| 42 | !! zero boundary condition along all coastlines (i.e. continent |
---|
| 43 | !! as well as islands) while in the latter the boundary condition |
---|
| 44 | !! specification is not required. |
---|
| 45 | !! This routine provides a MPI optimization to the existing solsor |
---|
| 46 | !! by reducing the number of call to lbc. |
---|
| 47 | !! |
---|
| 48 | !! ** Method : Successive-over-relaxation method using the red-black |
---|
| 49 | !! technique. The former technique used was not compatible with |
---|
| 50 | !! the north-fold boundary condition used in orca configurations. |
---|
| 51 | !! Compared to the classical sol_sor, this routine provides a |
---|
| 52 | !! mpp optimization by reducing the number of calls to lnc_lnk |
---|
| 53 | !! The solution is computed on a larger area and the boudary |
---|
| 54 | !! conditions only when the inside domain is reached. |
---|
| 55 | !! |
---|
| 56 | !! References : |
---|
| 57 | !! Madec et al. 1988, Ocean Modelling, issue 78, 1-6. |
---|
| 58 | !! Beare and Stevens 1997 Ann. Geophysicae 15, 1369-1377 |
---|
| 59 | !! |
---|
| 60 | !! History : |
---|
| 61 | !! ! 90-10 (G. Madec) Original code |
---|
| 62 | !! ! 91-11 (G. Madec) |
---|
| 63 | !! 7.1 ! 93-04 (G. Madec) time filter |
---|
| 64 | !! ! 96-05 (G. Madec) merge sor and pcg formulations |
---|
| 65 | !! ! 96-11 (A. Weaver) correction to preconditioning |
---|
| 66 | !! 9.0 ! 03-04 (C. Deltel, G. Madec) Red-Black SOR in free form |
---|
| 67 | !! 9.0 ! 05-09 (R. Benshila, G. Madec) MPI optimization |
---|
| 68 | !!---------------------------------------------------------------------- |
---|
| 69 | !! * Arguments |
---|
| 70 | INTEGER, INTENT( inout ) :: kindic ! solver indicator, < 0 if the conver- |
---|
| 71 | ! ! gence is not reached: the model is |
---|
| 72 | ! ! stopped in step |
---|
| 73 | ! ! set to zero before the call of solsor |
---|
| 74 | !! * Local declarations |
---|
| 75 | INTEGER :: ji, jj, jn ! dummy loop indices |
---|
| 76 | INTEGER :: ishift, icount |
---|
| 77 | REAL(wp) :: ztmp, zres, zres2 |
---|
| 78 | |
---|
| 79 | INTEGER :: ijmppodd, ijmppeven |
---|
| 80 | INTEGER :: ijpr2d |
---|
| 81 | !!---------------------------------------------------------------------- |
---|
| 82 | |
---|
| 83 | ijmppeven = MOD(nimpp+njmpp+jpr2di+jpr2dj,2) |
---|
| 84 | ijmppodd = MOD(nimpp+njmpp+jpr2di+jpr2dj+1,2) |
---|
| 85 | ijpr2d = MAX(jpr2di,jpr2dj) |
---|
| 86 | icount = 0 |
---|
| 87 | ! ! ============== |
---|
| 88 | DO jn = 1, nmax ! Iterative loop |
---|
| 89 | ! ! ============== |
---|
| 90 | |
---|
| 91 | ! applied the lateral boundary conditions |
---|
| 92 | IF( MOD(icount,ijpr2d+1) == 0 ) CALL lbc_lnk_e( gcx, c_solver_pt, 1. ) |
---|
| 93 | |
---|
| 94 | ! Residus |
---|
| 95 | ! ------- |
---|
| 96 | |
---|
| 97 | ! Guess black update |
---|
| 98 | DO jj = 2-jpr2dj, nlcj-1+jpr2dj |
---|
| 99 | ishift = MOD( jj-ijmppodd-jpr2dj, 2 ) |
---|
| 100 | DO ji = 2-jpr2di+ishift, nlci-1+jpr2di, 2 |
---|
| 101 | ztmp = gcb(ji ,jj ) & |
---|
| 102 | & - gcp(ji,jj,1) * gcx(ji ,jj-1) & |
---|
| 103 | & - gcp(ji,jj,2) * gcx(ji-1,jj ) & |
---|
| 104 | & - gcp(ji,jj,3) * gcx(ji+1,jj ) & |
---|
| 105 | & - gcp(ji,jj,4) * gcx(ji ,jj+1) |
---|
| 106 | ! Estimate of the residual |
---|
| 107 | zres = ztmp - gcx(ji,jj) |
---|
| 108 | gcr(ji,jj) = zres * gcdmat(ji,jj) * zres |
---|
| 109 | ! Guess update |
---|
| 110 | gcx(ji,jj) = sor * ztmp + (1-sor) * gcx(ji,jj) |
---|
| 111 | END DO |
---|
| 112 | END DO |
---|
| 113 | icount = icount + 1 |
---|
| 114 | |
---|
| 115 | ! applied the lateral boundary conditions |
---|
| 116 | IF( MOD(icount,ijpr2d+1) == 0 ) CALL lbc_lnk_e( gcx, c_solver_pt, 1. ) |
---|
| 117 | |
---|
| 118 | ! Guess red update |
---|
| 119 | DO jj = 2-jpr2dj, nlcj-1+jpr2dj |
---|
| 120 | ishift = MOD( jj-ijmppeven-jpr2dj, 2 ) |
---|
| 121 | DO ji = 2-jpr2di+ishift, nlci-1+jpr2di, 2 |
---|
| 122 | ztmp = gcb(ji ,jj ) & |
---|
| 123 | & - gcp(ji,jj,1) * gcx(ji ,jj-1) & |
---|
| 124 | & - gcp(ji,jj,2) * gcx(ji-1,jj ) & |
---|
| 125 | & - gcp(ji,jj,3) * gcx(ji+1,jj ) & |
---|
| 126 | & - gcp(ji,jj,4) * gcx(ji ,jj+1) |
---|
| 127 | ! Estimate of the residual |
---|
| 128 | zres = ztmp - gcx(ji,jj) |
---|
| 129 | gcr(ji,jj) = zres * gcdmat(ji,jj) * zres |
---|
| 130 | ! Guess update |
---|
| 131 | gcx(ji,jj) = sor * ztmp + (1-sor) * gcx(ji,jj) |
---|
| 132 | END DO |
---|
| 133 | END DO |
---|
| 134 | icount = icount + 1 |
---|
| 135 | |
---|
| 136 | ! test of convergence |
---|
| 137 | IF ( jn > nmin .AND. MOD( jn-nmin, nmod ) == 0 ) then |
---|
| 138 | |
---|
| 139 | SELECT CASE ( nsol_arp ) |
---|
| 140 | CASE ( 0 ) ! absolute precision (maximum value of the residual) |
---|
| 141 | zres2 = MAXVAL( gcr(2:nlci-1,2:nlcj-1) ) |
---|
| 142 | IF( lk_mpp ) CALL mpp_max( zres2 ) ! max over the global domain |
---|
| 143 | ! test of convergence |
---|
| 144 | IF( zres2 < resmax .OR. jn == nmax ) THEN |
---|
| 145 | res = SQRT( zres2 ) |
---|
| 146 | niter = jn |
---|
| 147 | ncut = 999 |
---|
| 148 | ENDIF |
---|
| 149 | CASE ( 1 ) ! relative precision |
---|
| 150 | rnorme = SUM( gcr(2:nlci-1,2:nlcj-1) ) |
---|
| 151 | IF( lk_mpp ) CALL mpp_sum( rnorme ) ! sum over the global domain |
---|
| 152 | ! test of convergence |
---|
| 153 | IF( rnorme < epsr .OR. jn == nmax ) THEN |
---|
| 154 | res = SQRT( rnorme ) |
---|
| 155 | niter = jn |
---|
| 156 | ncut = 999 |
---|
| 157 | ENDIF |
---|
| 158 | END SELECT |
---|
| 159 | |
---|
| 160 | !**** |
---|
| 161 | ! IF(lwp)WRITE(numsol,9300) jn, res, sqrt( epsr ) / eps |
---|
| 162 | 9300 FORMAT(' niter :',i4,' res :',e20.10,' b :',e20.10) |
---|
| 163 | !**** |
---|
| 164 | |
---|
| 165 | ENDIF |
---|
| 166 | ! indicator of non-convergence or explosion |
---|
| 167 | IF( jn == nmax .OR. SQRT(epsr)/eps > 1.e+20 ) kindic = -2 |
---|
| 168 | IF( ncut == 999 ) GOTO 999 |
---|
| 169 | |
---|
| 170 | ! ! ===================== |
---|
| 171 | END DO ! END of iterative loop |
---|
| 172 | ! ! ===================== |
---|
| 173 | |
---|
| 174 | 999 CONTINUE |
---|
| 175 | |
---|
| 176 | |
---|
| 177 | ! Output in gcx |
---|
| 178 | ! ------------- |
---|
| 179 | |
---|
| 180 | CALL lbc_lnk_e( gcx, c_solver_pt, 1. ) ! boundary conditions |
---|
| 181 | |
---|
| 182 | |
---|
| 183 | END SUBROUTINE sol_sor_e |
---|
| 184 | |
---|
| 185 | !!===================================================================== |
---|
| 186 | END MODULE solsor_e |
---|