| 1 |
C $Header: /u/gcmpack/MITgcm_contrib/dgoldberg/streamice/streamice_cg_solve.F,v 1.7 2013/04/06 17:43:41 dgoldberg Exp $ |
| 2 |
C $Name: $ |
| 3 |
|
| 4 |
#include "STREAMICE_OPTIONS.h" |
| 5 |
|
| 6 |
C---+----1----+----2----+----3----+----4----+----5----+----6----+----7-|--+----| |
| 7 |
|
| 8 |
CBOP |
| 9 |
SUBROUTINE STREAMICE_CG_SOLVE( |
| 10 |
U cg_Uin, ! x-velocities |
| 11 |
U cg_Vin, ! y-velocities |
| 12 |
I cg_Bu, ! force in x dir |
| 13 |
I cg_Bv, ! force in y dir |
| 14 |
I A_uu, ! section of matrix that multiplies u and projects on u |
| 15 |
I A_uv, ! section of matrix that multiplies v and projects on u |
| 16 |
I A_vu, ! section of matrix that multiplies u and projects on v |
| 17 |
I A_vv, ! section of matrix that multiplies v and projects on v |
| 18 |
I tolerance, |
| 19 |
O iters, |
| 20 |
I myThid ) |
| 21 |
C /============================================================\ |
| 22 |
C | SUBROUTINE | |
| 23 |
C | o | |
| 24 |
C |============================================================| |
| 25 |
C | | |
| 26 |
C \============================================================/ |
| 27 |
IMPLICIT NONE |
| 28 |
|
| 29 |
#include "SIZE.h" |
| 30 |
#include "EEPARAMS.h" |
| 31 |
#include "PARAMS.h" |
| 32 |
#include "STREAMICE.h" |
| 33 |
#include "STREAMICE_CG.h" |
| 34 |
|
| 35 |
|
| 36 |
|
| 37 |
#ifdef ALLOW_PETSC |
| 38 |
#include "finclude/petsc.h" |
| 39 |
#include "finclude/petscvec.h" |
| 40 |
#include "finclude/petscmat.h" |
| 41 |
#include "finclude/petscksp.h" |
| 42 |
#include "finclude/petscpc.h" |
| 43 |
#endif |
| 44 |
C === Global variables === |
| 45 |
|
| 46 |
|
| 47 |
C !INPUT/OUTPUT ARGUMENTS |
| 48 |
C cg_Uin, cg_Vin - input and output velocities |
| 49 |
C cg_Bu, cg_Bv - driving stress |
| 50 |
INTEGER myThid |
| 51 |
INTEGER iters |
| 52 |
_RL tolerance |
| 53 |
_RL cg_Uin (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
| 54 |
_RL cg_Vin (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
| 55 |
_RL cg_Bu (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
| 56 |
_RL cg_Bv (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
| 57 |
_RL |
| 58 |
& A_uu (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy,-1:1,-1:1), |
| 59 |
& A_vu (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy,-1:1,-1:1), |
| 60 |
& A_uv (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy,-1:1,-1:1), |
| 61 |
& A_vv (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy,-1:1,-1:1) |
| 62 |
|
| 63 |
C LOCAL VARIABLES |
| 64 |
INTEGER i, j, bi, bj, cg_halo, conv_flag |
| 65 |
INTEGER iter, is, js, ie, je, colx, coly, k |
| 66 |
_RL dot_p1, dot_p2, alpha_k, beta_k, resid, resid_0 |
| 67 |
_RL dot_p1_tile (nSx,nSy) |
| 68 |
_RL dot_p2_tile (nSx,nSy) |
| 69 |
CHARACTER*(MAX_LEN_MBUF) msgBuf |
| 70 |
|
| 71 |
|
| 72 |
#ifdef ALLOW_PETSC |
| 73 |
INTEGER indices(2*(snx*nsx*sny*nsy)) |
| 74 |
INTEGER n_dofs_cum_sum (0:nPx*nPy-1), idx(1) |
| 75 |
_RL rhs_values(2*(snx*nsx*sny*nsy)) |
| 76 |
_RL solution_values(2*(snx*nsx*sny*nsy)) |
| 77 |
! _RL mat_values (2*Nx*Ny,2*(snx*nsx*sny*nsy)) |
| 78 |
_RL mat_values (18,1), mat_val_return(1) |
| 79 |
INTEGER indices_col(18) |
| 80 |
INTEGER local_dofs, global_dofs, dof_index, dof_index_col |
| 81 |
INTEGER local_offset |
| 82 |
Mat matrix |
| 83 |
KSP ksp |
| 84 |
PC pc |
| 85 |
Vec rhs |
| 86 |
Vec solution |
| 87 |
PetscErrorCode ierr |
| 88 |
#ifdef ALLOW_USE_MPI |
| 89 |
integer mpiRC, mpiMyWid |
| 90 |
#endif |
| 91 |
#endif |
| 92 |
|
| 93 |
|
| 94 |
#ifdef ALLOW_STREAMICE |
| 95 |
|
| 96 |
|
| 97 |
|
| 98 |
CALL TIMER_START ('STREAMICE_CG_SOLVE',myThid) |
| 99 |
#ifndef STREAMICE_SERIAL_TRISOLVE |
| 100 |
|
| 101 |
#ifdef ALLOW_PETSC |
| 102 |
|
| 103 |
#ifdef ALLOW_USE_MPI |
| 104 |
|
| 105 |
|
| 106 |
CALL MPI_COMM_RANK( MPI_COMM_WORLD, mpiMyWId, mpiRC ) |
| 107 |
local_dofs = n_dofs_process (mpiMyWid) |
| 108 |
global_dofs = 0 |
| 109 |
|
| 110 |
n_dofs_cum_sum(0) = 0 |
| 111 |
DO i=0,nPx*nPy-1 |
| 112 |
global_dofs = global_dofs + n_dofs_process (i) |
| 113 |
if (i.ge.1) THEN |
| 114 |
n_dofs_cum_sum(i) = n_dofs_cum_sum(i-1)+ |
| 115 |
& n_dofs_process(i-1) |
| 116 |
endif |
| 117 |
ENDDO |
| 118 |
local_offset = n_dofs_cum_sum(mpimywid) |
| 119 |
|
| 120 |
#else |
| 121 |
|
| 122 |
local_dofs = n_dofs_process (0) |
| 123 |
global_dofs = local_dofs |
| 124 |
local_offset = 0 |
| 125 |
|
| 126 |
#endif |
| 127 |
|
| 128 |
! call petscInitialize(PETSC_NULL_CHARACTER,ierr) |
| 129 |
|
| 130 |
!---------------------- |
| 131 |
|
| 132 |
call VecCreate(PETSC_COMM_WORLD, rhs, ierr) |
| 133 |
call VecSetSizes(rhs, local_dofs, global_dofs, ierr) |
| 134 |
call VecSetType(rhs, VECMPI, ierr) |
| 135 |
|
| 136 |
call VecCreate(PETSC_COMM_WORLD, solution, ierr) |
| 137 |
call VecSetSizes(solution, local_dofs, global_dofs, ierr) |
| 138 |
call VecSetType(solution, VECMPI, ierr) |
| 139 |
|
| 140 |
do i=1,local_dofs |
| 141 |
indices(i) = i-1 + local_offset |
| 142 |
end do |
| 143 |
do i=1,2*nSx*nSy*sNx*sNy |
| 144 |
rhs_values (i) = 0. _d 0 |
| 145 |
solution_values (i) = 0. _d 0 |
| 146 |
enddo |
| 147 |
|
| 148 |
! gather rhs and initial guess values to populate petsc vectors |
| 149 |
|
| 150 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 151 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 152 |
DO j=1,sNy |
| 153 |
DO i=1,sNx |
| 154 |
|
| 155 |
dof_index = INT(streamice_petsc_dofs_u(i,j,bi,bj)) |
| 156 |
& - local_offset |
| 157 |
|
| 158 |
if (dof_index.ge.0) THEN |
| 159 |
|
| 160 |
rhs_values(dof_index+1) = cg_Bu(i,j,bi,bj) |
| 161 |
solution_values(dof_index+1) = cg_Uin(i,j,bi,bj) |
| 162 |
|
| 163 |
endif |
| 164 |
|
| 165 |
!--------------- |
| 166 |
|
| 167 |
dof_index = INT(streamice_petsc_dofs_v(i,j,bi,bj)) |
| 168 |
& - local_offset |
| 169 |
|
| 170 |
if (dof_index.ge.0) THEN |
| 171 |
|
| 172 |
rhs_values(dof_index+1) = cg_Bv(i,j,bi,bj) |
| 173 |
solution_values(dof_index+1) = cg_Vin(i,j,bi,bj) |
| 174 |
|
| 175 |
endif |
| 176 |
|
| 177 |
ENDDO |
| 178 |
ENDDO |
| 179 |
ENDDO |
| 180 |
ENDDO |
| 181 |
|
| 182 |
|
| 183 |
call VecSetValues(rhs, local_dofs, indices, rhs_values, |
| 184 |
& INSERT_VALUES, ierr) |
| 185 |
call VecAssemblyBegin(rhs, ierr) |
| 186 |
call VecAssemblyEnd(rhs, ierr) |
| 187 |
|
| 188 |
|
| 189 |
call VecSetValues(solution, local_dofs, indices, |
| 190 |
& solution_values, INSERT_VALUES, ierr) |
| 191 |
call VecAssemblyBegin(solution, ierr) |
| 192 |
call VecAssemblyEnd(solution, ierr) |
| 193 |
|
| 194 |
|
| 195 |
call MatCreateMPIAIJ (PETSC_COMM_WORLD, |
| 196 |
& local_dofs, local_dofs, |
| 197 |
& global_dofs, global_dofs, |
| 198 |
& 18, PETSC_NULL_INTEGER, |
| 199 |
& 18, PETSC_NULL_INTEGER, |
| 200 |
& matrix, ierr) |
| 201 |
|
| 202 |
|
| 203 |
! populate petsc matrix |
| 204 |
|
| 205 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 206 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 207 |
DO j=1,sNy |
| 208 |
DO i=1,sNx |
| 209 |
|
| 210 |
dof_index = INT(streamice_petsc_dofs_u(i,j,bi,bj)) |
| 211 |
! & - local_offset |
| 212 |
|
| 213 |
IF (dof_index .ge. 0) THEN |
| 214 |
|
| 215 |
DO k=1,18 |
| 216 |
indices_col(k) = 0 |
| 217 |
mat_values(k,1) = 0. _d 0 |
| 218 |
ENDDO |
| 219 |
k=0 |
| 220 |
|
| 221 |
DO coly=-1,1 |
| 222 |
DO colx=-1,1 |
| 223 |
|
| 224 |
dof_index_col = streamice_petsc_dofs_u(i+colx,j+coly,bi,bj) |
| 225 |
|
| 226 |
if (dof_index_col.ge.0) THEN |
| 227 |
! pscal = A_uu(i,j,bi,bj,colx,coly) |
| 228 |
! CALL MatSetValue (matrix,dof_index, dof_index_col, |
| 229 |
! & pscal,INSERT_VALUES,ierr) |
| 230 |
k=k+1 |
| 231 |
mat_values (k,1) = A_uu(i,j,bi,bj,colx,coly) |
| 232 |
indices_col (k) = dof_index_col |
| 233 |
endif |
| 234 |
|
| 235 |
dof_index_col = streamice_petsc_dofs_v(i+colx,j+coly,bi,bj) |
| 236 |
|
| 237 |
if (dof_index_col.ge.0) THEN |
| 238 |
! CALL MatSetValue (matrix,dof_index, dof_index_col, |
| 239 |
! & A_uv(i,j,bi,bj,colx,coly),INSERT_VALUES,ierr) |
| 240 |
k=k+1 |
| 241 |
mat_values (k,1) = A_uv(i,j,bi,bj,colx,coly) |
| 242 |
indices_col (k) = dof_index_col |
| 243 |
endif |
| 244 |
|
| 245 |
ENDDO |
| 246 |
ENDDO |
| 247 |
|
| 248 |
call matSetValues (matrix, 1, dof_index, k, indices_col, |
| 249 |
& mat_values,INSERT_VALUES,ierr) |
| 250 |
|
| 251 |
|
| 252 |
ENDIF |
| 253 |
|
| 254 |
! ---------------------------------------------- |
| 255 |
|
| 256 |
dof_index = INT(streamice_petsc_dofs_v(i,j,bi,bj)) |
| 257 |
! & - local_offset |
| 258 |
|
| 259 |
IF (dof_index .ge. 0) THEN |
| 260 |
|
| 261 |
DO k=1,18 |
| 262 |
indices_col(k) = 0 |
| 263 |
mat_values(k,1) = 0. _d 0 |
| 264 |
ENDDO |
| 265 |
k=0 |
| 266 |
|
| 267 |
DO coly=-1,1 |
| 268 |
DO colx=-1,1 |
| 269 |
|
| 270 |
dof_index_col = streamice_petsc_dofs_u(i+colx,j+coly,bi,bj) |
| 271 |
|
| 272 |
if (dof_index_col.ge.0) THEN |
| 273 |
! CALL MatSetValue (matrix,dof_index, dof_index_col, |
| 274 |
! & A_vu(i,j,bi,bj,colx,coly),INSERT_VALUES,ierr) |
| 275 |
k=k+1 |
| 276 |
mat_values (k,1) = A_vu(i,j,bi,bj,colx,coly) |
| 277 |
indices_col (k) = dof_index_col |
| 278 |
endif |
| 279 |
|
| 280 |
dof_index_col = streamice_petsc_dofs_v(i+colx,j+coly,bi,bj) |
| 281 |
|
| 282 |
if (dof_index_col.ge.0) THEN |
| 283 |
! CALL MatSetValue (matrix,dof_index, dof_index_col, |
| 284 |
! & A_vv(i,j,bi,bj,colx,coly),INSERT_VALUES,ierr) |
| 285 |
k=k+1 |
| 286 |
mat_values (k,1) = A_vv(i,j,bi,bj,colx,coly) |
| 287 |
indices_col (k) = dof_index_col |
| 288 |
endif |
| 289 |
|
| 290 |
ENDDO |
| 291 |
ENDDO |
| 292 |
|
| 293 |
call matSetValues (matrix, 1, dof_index, k, indices_col, |
| 294 |
& mat_values,INSERT_VALUES,ierr) |
| 295 |
ENDIF |
| 296 |
|
| 297 |
ENDDO |
| 298 |
ENDDO |
| 299 |
ENDDO |
| 300 |
ENDDO |
| 301 |
|
| 302 |
call MatAssemblyBegin(matrix,MAT_FINAL_ASSEMBLY,ierr) |
| 303 |
call MatAssemblyEnd(matrix,MAT_FINAL_ASSEMBLY,ierr) |
| 304 |
|
| 305 |
|
| 306 |
call KSPCreate(PETSC_COMM_WORLD, ksp, ierr) |
| 307 |
call KSPSetOperators(ksp, matrix, matrix, |
| 308 |
& DIFFERENT_NONZERO_PATTERN, ierr) |
| 309 |
|
| 310 |
SELECT CASE (PETSC_SOLVER_TYPE) |
| 311 |
CASE ('CG') |
| 312 |
PRINT *, "PETSC SOLVER: SELECTED CG" |
| 313 |
call KSPSetType(ksp, KSPCG, ierr) |
| 314 |
CASE ('GMRES') |
| 315 |
PRINT *, "PETSC SOLVER: SELECTED GMRES" |
| 316 |
call KSPSetType(ksp, KSPGMRES, ierr) |
| 317 |
CASE ('BICG') |
| 318 |
PRINT *, "PETSC SOLVER: SELECTED BICG" |
| 319 |
call KSPSetType(ksp, KSPBICG, ierr) |
| 320 |
CASE DEFAULT |
| 321 |
PRINT *, "PETSC SOLVER: SELECTED DEFAULT" |
| 322 |
call KSPSetType(ksp, KSPCG, ierr) |
| 323 |
END SELECT |
| 324 |
|
| 325 |
call KSPGetPC(ksp, pc, ierr) |
| 326 |
call KSPSetTolerances(ksp,tolerance, |
| 327 |
& PETSC_DEFAULT_DOUBLE_PRECISION, |
| 328 |
& PETSC_DEFAULT_DOUBLE_PRECISION, |
| 329 |
& streamice_max_cg_iter,ierr) |
| 330 |
|
| 331 |
SELECT CASE (PETSC_PRECOND_TYPE) |
| 332 |
CASE ('BLOCKJACOBI') |
| 333 |
PRINT *, "PETSC PRECOND: SELECTED BJACOBI" |
| 334 |
call PCSetType(pc, PCBJACOBI, ierr) |
| 335 |
CASE ('JACOBI') |
| 336 |
PRINT *, "PETSC PRECOND: SELECTED JACOBI" |
| 337 |
call PCSetType(pc, PCJACOBI, ierr) |
| 338 |
CASE ('ILU') |
| 339 |
PRINT *, "PETSC PRECOND: SELECTED ILU" |
| 340 |
call PCSetType(pc, PCILU, ierr) |
| 341 |
CASE DEFAULT |
| 342 |
PRINT *, "PETSC PRECOND: SELECTED DEFAULT" |
| 343 |
call PCSetType(pc, PCBJACOBI, ierr) |
| 344 |
END SELECT |
| 345 |
|
| 346 |
|
| 347 |
call KSPSolve(ksp, rhs, solution, ierr) |
| 348 |
call KSPGetIterationNumber(ksp,iters,ierr) |
| 349 |
|
| 350 |
call VecGetValues(solution,local_dofs,indices, |
| 351 |
& solution_values,ierr) |
| 352 |
|
| 353 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 354 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 355 |
DO j=1,sNy |
| 356 |
DO i=1,sNx |
| 357 |
|
| 358 |
dof_index = INT(streamice_petsc_dofs_u(i,j,bi,bj)) |
| 359 |
& - local_offset |
| 360 |
if (dof_index.ge.0) THEN |
| 361 |
cg_Uin(i,j,bi,bj) = solution_values(dof_index+1) |
| 362 |
endif |
| 363 |
|
| 364 |
dof_index = INT(streamice_petsc_dofs_v(i,j,bi,bj)) |
| 365 |
& - local_offset |
| 366 |
if (dof_index.ge.0) THEN |
| 367 |
cg_Vin(i,j,bi,bj) = solution_values(dof_index+1) |
| 368 |
endif |
| 369 |
|
| 370 |
ENDDO |
| 371 |
ENDDO |
| 372 |
ENDDO |
| 373 |
ENDDO |
| 374 |
|
| 375 |
|
| 376 |
|
| 377 |
#else /* ALLOW_PETSC */ |
| 378 |
|
| 379 |
|
| 380 |
iters = streamice_max_cg_iter |
| 381 |
conv_flag = 0 |
| 382 |
|
| 383 |
|
| 384 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 385 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 386 |
DO j=1,sNy |
| 387 |
DO i=1,sNx |
| 388 |
Zu_SI (i,j,bi,bj) = 0. _d 0 |
| 389 |
Zv_SI (i,j,bi,bj) = 0. _d 0 |
| 390 |
Ru_SI (i,j,bi,bj) = 0. _d 0 |
| 391 |
Rv_SI (i,j,bi,bj) = 0. _d 0 |
| 392 |
Au_SI (i,j,bi,bj) = 0. _d 0 |
| 393 |
Av_SI (i,j,bi,bj) = 0. _d 0 |
| 394 |
Du_SI (i,j,bi,bj) = 0. _d 0 |
| 395 |
Dv_SI (i,j,bi,bj) = 0. _d 0 |
| 396 |
ENDDO |
| 397 |
ENDDO |
| 398 |
ENDDO |
| 399 |
ENDDO |
| 400 |
|
| 401 |
C FIND INITIAL RESIDUAL, and initialize r |
| 402 |
|
| 403 |
! #ifdef STREAMICE_CONSTRUCT_MATRIX |
| 404 |
|
| 405 |
|
| 406 |
|
| 407 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 408 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 409 |
DO j=1,sNy |
| 410 |
DO i=1,sNx |
| 411 |
DO colx=-1,1 |
| 412 |
DO coly=-1,1 |
| 413 |
Au_SI(i,j,bi,bj) = Au_SI(i,j,bi,bj) + |
| 414 |
& A_uu(i,j,bi,bj,colx,coly)* |
| 415 |
& cg_Uin(i+colx,j+coly,bi,bj)+ |
| 416 |
& A_uv(i,j,bi,bj,colx,coly)* |
| 417 |
& cg_Vin(i+colx,j+coly,bi,bj) |
| 418 |
|
| 419 |
|
| 420 |
Av_SI(i,j,bi,bj) = Av_SI(i,j,bi,bj) + |
| 421 |
& A_vu(i,j,bi,bj,colx,coly)* |
| 422 |
& cg_Uin(i+colx,j+coly,bi,bj)+ |
| 423 |
& A_vv(i,j,bi,bj,colx,coly)* |
| 424 |
& cg_Vin(i+colx,j+coly,bi,bj) |
| 425 |
ENDDO |
| 426 |
ENDDO |
| 427 |
ENDDO |
| 428 |
ENDDO |
| 429 |
ENDDO |
| 430 |
ENDDO |
| 431 |
|
| 432 |
|
| 433 |
_EXCH_XY_RL( Au_SI, myThid ) |
| 434 |
_EXCH_XY_RL( Av_SI, myThid ) |
| 435 |
|
| 436 |
|
| 437 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 438 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 439 |
DO j=1-OLy,sNy+OLy |
| 440 |
DO i=1-OLx,sNx+OLx |
| 441 |
Ru_SI(i,j,bi,bj)=cg_Bu(i,j,bi,bj)- |
| 442 |
& Au_SI(i,j,bi,bj) |
| 443 |
Rv_SI(i,j,bi,bj)=cg_Bv(i,j,bi,bj)- |
| 444 |
& Av_SI(i,j,bi,bj) |
| 445 |
ENDDO |
| 446 |
ENDDO |
| 447 |
dot_p1_tile(bi,bj) = 0. _d 0 |
| 448 |
dot_p2_tile(bi,bj) = 0. _d 0 |
| 449 |
ENDDO |
| 450 |
ENDDO |
| 451 |
|
| 452 |
|
| 453 |
|
| 454 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 455 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 456 |
DO j=1,sNy |
| 457 |
DO i=1,sNx |
| 458 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) |
| 459 |
& dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Ru_SI(i,j,bi,bj)**2 |
| 460 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) |
| 461 |
& dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Rv_SI(i,j,bi,bj)**2 |
| 462 |
ENDDO |
| 463 |
ENDDO |
| 464 |
ENDDO |
| 465 |
ENDDO |
| 466 |
|
| 467 |
CALL GLOBAL_SUM_TILE_RL( dot_p1_tile, dot_p1, myThid ) |
| 468 |
resid_0 = sqrt(dot_p1) |
| 469 |
|
| 470 |
WRITE(msgBuf,'(A,E14.7)') 'CONJ GRAD INIT RESID, ', |
| 471 |
& resid_0 |
| 472 |
CALL PRINT_MESSAGE( msgBuf, standardMessageUnit, |
| 473 |
& SQUEEZE_RIGHT , 1) |
| 474 |
|
| 475 |
C CCCCCCCCCCCCCCCCCCCC |
| 476 |
|
| 477 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 478 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 479 |
DO j=1-OLy,sNy+OLy |
| 480 |
DO i=1-OLx,sNx+OLx |
| 481 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) |
| 482 |
& Zu_SI(i,j,bi,bj)=Ru_SI(i,j,bi,bj) / DIAGu_SI(i,j,bi,bj) |
| 483 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) |
| 484 |
& Zv_SI(i,j,bi,bj)=Rv_SI(i,j,bi,bj) / DIAGv_SI(i,j,bi,bj) |
| 485 |
ENDDO |
| 486 |
ENDDO |
| 487 |
ENDDO |
| 488 |
ENDDO |
| 489 |
|
| 490 |
cg_halo = min(OLx-1,OLy-1) |
| 491 |
conv_flag = 0 |
| 492 |
|
| 493 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 494 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 495 |
DO j=1-OLy,sNy+OLy |
| 496 |
DO i=1-OLx,sNx+OLx |
| 497 |
Du_SI(i,j,bi,bj)=Zu_SI(i,j,bi,bj) |
| 498 |
Dv_SI(i,j,bi,bj)=Zv_SI(i,j,bi,bj) |
| 499 |
ENDDO |
| 500 |
ENDDO |
| 501 |
ENDDO |
| 502 |
ENDDO |
| 503 |
|
| 504 |
resid = resid_0 |
| 505 |
iters = 0 |
| 506 |
|
| 507 |
c !!!!!!!!!!!!!!!!!! |
| 508 |
c !! !! |
| 509 |
c !! MAIN CG LOOP !! |
| 510 |
c !! !! |
| 511 |
c !!!!!!!!!!!!!!!!!! |
| 512 |
|
| 513 |
|
| 514 |
|
| 515 |
|
| 516 |
c ! initially, b-grid data is valid up to 3 halo nodes out -- right? (check for MITgcm!!) |
| 517 |
|
| 518 |
WRITE(msgBuf,'(A)') 'BEGINNING MAIN CG LOOP' |
| 519 |
CALL PRINT_MESSAGE( msgBuf, standardMessageUnit, |
| 520 |
& SQUEEZE_RIGHT , 1) |
| 521 |
|
| 522 |
! IF(STREAMICE_construct_matrix) CALL STREAMICE_CG_MAKE_A(myThid) |
| 523 |
|
| 524 |
|
| 525 |
do iter = 1, streamice_max_cg_iter |
| 526 |
if (resid .gt. tolerance*resid_0) then |
| 527 |
|
| 528 |
c to avoid using "exit" |
| 529 |
iters = iters + 1 |
| 530 |
|
| 531 |
is = 1 - cg_halo |
| 532 |
ie = sNx + cg_halo |
| 533 |
js = 1 - cg_halo |
| 534 |
je = sNy + cg_halo |
| 535 |
|
| 536 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 537 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 538 |
DO j=1-OLy,sNy+OLy |
| 539 |
DO i=1-OLx,sNx+OLx |
| 540 |
Au_SI(i,j,bi,bj) = 0. _d 0 |
| 541 |
Av_SI(i,j,bi,bj) = 0. _d 0 |
| 542 |
ENDDO |
| 543 |
ENDDO |
| 544 |
ENDDO |
| 545 |
ENDDO |
| 546 |
|
| 547 |
! IF (STREAMICE_construct_matrix) THEN |
| 548 |
|
| 549 |
! #ifdef STREAMICE_CONSTRUCT_MATRIX |
| 550 |
|
| 551 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 552 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 553 |
DO j=js,je |
| 554 |
DO i=is,ie |
| 555 |
DO colx=-1,1 |
| 556 |
DO coly=-1,1 |
| 557 |
Au_SI(i,j,bi,bj) = Au_SI(i,j,bi,bj) + |
| 558 |
& A_uu(i,j,bi,bj,colx,coly)* |
| 559 |
& Du_SI(i+colx,j+coly,bi,bj)+ |
| 560 |
& A_uv(i,j,bi,bj,colx,coly)* |
| 561 |
& Dv_SI(i+colx,j+coly,bi,bj) |
| 562 |
Av_SI(i,j,bi,bj) = Av_SI(i,j,bi,bj) + |
| 563 |
& A_vu(i,j,bi,bj,colx,coly)* |
| 564 |
& Du_SI(i+colx,j+coly,bi,bj)+ |
| 565 |
& A_vv(i,j,bi,bj,colx,coly)* |
| 566 |
& Dv_SI(i+colx,j+coly,bi,bj) |
| 567 |
ENDDO |
| 568 |
ENDDO |
| 569 |
ENDDO |
| 570 |
ENDDO |
| 571 |
ENDDO |
| 572 |
ENDDO |
| 573 |
|
| 574 |
! else |
| 575 |
! #else |
| 576 |
! |
| 577 |
! CALL STREAMICE_CG_ACTION( myThid, |
| 578 |
! O Au_SI, |
| 579 |
! O Av_SI, |
| 580 |
! I Du_SI, |
| 581 |
! I Dv_SI, |
| 582 |
! I is,ie,js,je) |
| 583 |
! |
| 584 |
! ! ENDIF |
| 585 |
! |
| 586 |
! #endif |
| 587 |
|
| 588 |
|
| 589 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 590 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 591 |
dot_p1_tile(bi,bj) = 0. _d 0 |
| 592 |
dot_p2_tile(bi,bj) = 0. _d 0 |
| 593 |
ENDDO |
| 594 |
ENDDO |
| 595 |
|
| 596 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 597 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 598 |
DO j=1,sNy |
| 599 |
DO i=1,sNx |
| 600 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) THEN |
| 601 |
dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Zu_SI(i,j,bi,bj)* |
| 602 |
& Ru_SI(i,j,bi,bj) |
| 603 |
dot_p2_tile(bi,bj)=dot_p2_tile(bi,bj)+Du_SI(i,j,bi,bj)* |
| 604 |
& Au_SI(i,j,bi,bj) |
| 605 |
ENDIF |
| 606 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) THEN |
| 607 |
dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Zv_SI(i,j,bi,bj)* |
| 608 |
& Rv_SI(i,j,bi,bj) |
| 609 |
dot_p2_tile(bi,bj)=dot_p2_tile(bi,bj)+Dv_SI(i,j,bi,bj)* |
| 610 |
& Av_SI(i,j,bi,bj) |
| 611 |
ENDIF |
| 612 |
ENDDO |
| 613 |
ENDDO |
| 614 |
ENDDO |
| 615 |
ENDDO |
| 616 |
|
| 617 |
CALL GLOBAL_SUM_TILE_RL( dot_p1_tile, dot_p1, myThid ) |
| 618 |
CALL GLOBAL_SUM_TILE_RL( dot_p2_tile, dot_p2, myThid ) |
| 619 |
alpha_k = dot_p1/dot_p2 |
| 620 |
|
| 621 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 622 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 623 |
DO j=1-OLy,sNy+OLy |
| 624 |
DO i=1-OLx,sNx+OLx |
| 625 |
|
| 626 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) THEN |
| 627 |
cg_Uin(i,j,bi,bj)=cg_Uin(i,j,bi,bj)+ |
| 628 |
& alpha_k*Du_SI(i,j,bi,bj) |
| 629 |
Ru_old_SI(i,j,bi,bj) = Ru_SI(i,j,bi,bj) |
| 630 |
Zu_old_SI(i,j,bi,bj) = Zu_SI(i,j,bi,bj) |
| 631 |
Ru_SI(i,j,bi,bj) = Ru_SI(i,j,bi,bj)- |
| 632 |
& alpha_k*Au_SI(i,j,bi,bj) |
| 633 |
Zu_SI(i,j,bi,bj) = Ru_SI(i,j,bi,bj) / |
| 634 |
& DIAGu_SI(i,j,bi,bj) |
| 635 |
ENDIF |
| 636 |
|
| 637 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) THEN |
| 638 |
cg_Vin(i,j,bi,bj)=cg_Vin(i,j,bi,bj)+ |
| 639 |
& alpha_k*Dv_SI(i,j,bi,bj) |
| 640 |
Rv_old_SI(i,j,bi,bj) = Rv_SI(i,j,bi,bj) |
| 641 |
Zv_old_SI(i,j,bi,bj) = Zv_SI(i,j,bi,bj) |
| 642 |
Rv_SI(i,j,bi,bj) = Rv_SI(i,j,bi,bj)- |
| 643 |
& alpha_k*Av_SI(i,j,bi,bj) |
| 644 |
Zv_SI(i,j,bi,bj) = Rv_SI(i,j,bi,bj) / |
| 645 |
& DIAGv_SI(i,j,bi,bj) |
| 646 |
|
| 647 |
ENDIF |
| 648 |
ENDDO |
| 649 |
ENDDO |
| 650 |
ENDDO |
| 651 |
ENDDO |
| 652 |
|
| 653 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 654 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 655 |
dot_p1_tile(bi,bj) = 0. _d 0 |
| 656 |
dot_p2_tile(bi,bj) = 0. _d 0 |
| 657 |
ENDDO |
| 658 |
ENDDO |
| 659 |
|
| 660 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 661 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 662 |
DO j=1,sNy |
| 663 |
DO i=1,sNx |
| 664 |
|
| 665 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) THEN |
| 666 |
dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Zu_SI(i,j,bi,bj)* |
| 667 |
& Ru_SI(i,j,bi,bj) |
| 668 |
dot_p2_tile(bi,bj)=dot_p2_tile(bi,bj)+Zu_old_SI(i,j,bi,bj)* |
| 669 |
& Ru_old_SI(i,j,bi,bj) |
| 670 |
ENDIF |
| 671 |
|
| 672 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) THEN |
| 673 |
dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Zv_SI(i,j,bi,bj)* |
| 674 |
& Rv_SI(i,j,bi,bj) |
| 675 |
dot_p2_tile(bi,bj)=dot_p2_tile(bi,bj)+Zv_old_SI(i,j,bi,bj)* |
| 676 |
& Rv_old_SI(i,j,bi,bj) |
| 677 |
ENDIF |
| 678 |
|
| 679 |
ENDDO |
| 680 |
ENDDO |
| 681 |
ENDDO |
| 682 |
ENDDO |
| 683 |
|
| 684 |
CALL GLOBAL_SUM_TILE_RL( dot_p1_tile, dot_p1, myThid ) |
| 685 |
CALL GLOBAL_SUM_TILE_RL( dot_p2_tile, dot_p2, myThid ) |
| 686 |
|
| 687 |
beta_k = dot_p1/dot_p2 |
| 688 |
|
| 689 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 690 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 691 |
DO j=1-OLy,sNy+OLy |
| 692 |
DO i=1-OLx,sNx+OLx |
| 693 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) |
| 694 |
& Du_SI(i,j,bi,bj)=beta_k*Du_SI(i,j,bi,bj)+ |
| 695 |
& Zu_SI(i,j,bi,bj) |
| 696 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) |
| 697 |
& Dv_SI(i,j,bi,bj)=beta_k*Dv_SI(i,j,bi,bj)+ |
| 698 |
& Zv_SI(i,j,bi,bj) |
| 699 |
ENDDO |
| 700 |
ENDDO |
| 701 |
ENDDO |
| 702 |
ENDDO |
| 703 |
|
| 704 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 705 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 706 |
dot_p1_tile(bi,bj) = 0. _d 0 |
| 707 |
ENDDO |
| 708 |
ENDDO |
| 709 |
|
| 710 |
DO bj = myByLo(myThid), myByHi(myThid) |
| 711 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
| 712 |
DO j=1,sNy |
| 713 |
DO i=1,sNx |
| 714 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) |
| 715 |
& dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Ru_SI(i,j,bi,bj)**2 |
| 716 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) |
| 717 |
& dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Rv_SI(i,j,bi,bj)**2 |
| 718 |
ENDDO |
| 719 |
ENDDO |
| 720 |
ENDDO |
| 721 |
ENDDO |
| 722 |
|
| 723 |
CALL GLOBAL_SUM_TILE_RL( dot_p1_tile, dot_p1, myThid ) |
| 724 |
resid = sqrt(dot_p1) |
| 725 |
|
| 726 |
! IF (iter .eq. 1) then |
| 727 |
! print *, alpha_k, beta_k, resid |
| 728 |
! ENDIF |
| 729 |
|
| 730 |
cg_halo = cg_halo - 1 |
| 731 |
|
| 732 |
if (cg_halo .eq. 0) then |
| 733 |
cg_halo = min(OLx-1,OLy-1) |
| 734 |
_EXCH_XY_RL( Du_SI, myThid ) |
| 735 |
_EXCH_XY_RL( Dv_SI, myThid ) |
| 736 |
_EXCH_XY_RL( Ru_SI, myThid ) |
| 737 |
_EXCH_XY_RL( Rv_SI, myThid ) |
| 738 |
_EXCH_XY_RL( cg_Uin, myThid ) |
| 739 |
_EXCH_XY_RL( cg_Vin, myThid ) |
| 740 |
endif |
| 741 |
|
| 742 |
|
| 743 |
endif |
| 744 |
enddo ! end of CG loop |
| 745 |
|
| 746 |
c to avoid using "exit" |
| 747 |
c if iters has reached max_iters there is no convergence |
| 748 |
|
| 749 |
IF (iters .lt. streamice_max_cg_iter) THEN |
| 750 |
conv_flag = 1 |
| 751 |
ENDIF |
| 752 |
|
| 753 |
! DO bj = myByLo(myThid), myByHi(myThid) |
| 754 |
! DO bi = myBxLo(myThid), myBxHi(myThid) |
| 755 |
! DO j=1-OLy,sNy+OLy |
| 756 |
! DO i=1-OLy,sNx+OLy |
| 757 |
! IF (STREAMICE_umask(i,j,bi,bj).eq.3.0) |
| 758 |
! & cg_Uin(i,j,bi,bj)=u_bdry_values_SI(i,j,bi,bj) |
| 759 |
! IF (STREAMICE_vmask(i,j,bi,bj).eq.3.0) |
| 760 |
! & cg_Vin(i,j,bi,bj)=v_bdry_values_SI(i,j,bi,bj) |
| 761 |
! ENDDO |
| 762 |
! ENDDO |
| 763 |
! ENDDO |
| 764 |
! ENDDO |
| 765 |
! |
| 766 |
! _EXCH_XY_RL( cg_Uin, myThid ) |
| 767 |
! _EXCH_XY_RL( cg_Vin, myThid ) |
| 768 |
|
| 769 |
#endif /* ifndef ALLOW_PETSC */ |
| 770 |
|
| 771 |
#else /* STREAMICE_SERIAL_TRISOLVE */ |
| 772 |
|
| 773 |
CALL STREAMICE_TRIDIAG_SOLVE( |
| 774 |
U cg_Uin, ! x-velocities |
| 775 |
U cg_Vin, |
| 776 |
U cg_Bu, ! force in x dir |
| 777 |
I A_uu, ! section of matrix that multiplies u and projects on u |
| 778 |
I STREAMICE_umask, |
| 779 |
I myThid ) |
| 780 |
|
| 781 |
#endif |
| 782 |
|
| 783 |
CALL TIMER_STOP ('STREAMICE_CG_SOLVE',myThid) |
| 784 |
|
| 785 |
|
| 786 |
#endif |
| 787 |
RETURN |
| 788 |
END |
Parent Directory
| 
Revision Log
| 
Revision Graph