1 |
C $Header: /u/gcmpack/MITgcm_contrib/dgoldberg/streamice/streamice_cg_solve.F,v 1.6 2012/07/26 16:22:58 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 |
#ifdef ALLOW_PETSC |
35 |
#include "finclude/petsc.h" |
36 |
!#include "finclude/petscvec.h" |
37 |
!#include "finclude/petscmat.h" |
38 |
!#include "finclude/petscksp.h" |
39 |
!#include "finclude/petscpc.h" |
40 |
#endif |
41 |
C === Global variables === |
42 |
|
43 |
|
44 |
C !INPUT/OUTPUT ARGUMENTS |
45 |
C cg_Uin, cg_Vin - input and output velocities |
46 |
C cg_Bu, cg_Bv - driving stress |
47 |
INTEGER myThid |
48 |
INTEGER iters |
49 |
_RL tolerance |
50 |
_RL cg_Uin (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
51 |
_RL cg_Vin (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
52 |
_RL cg_Bu (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
53 |
_RL cg_Bv (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
54 |
_RL |
55 |
& A_uu (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy,-1:1,-1:1), |
56 |
& A_vu (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy,-1:1,-1:1), |
57 |
& A_uv (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy,-1:1,-1:1), |
58 |
& A_vv (1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy,-1:1,-1:1) |
59 |
|
60 |
C LOCAL VARIABLES |
61 |
INTEGER i, j, bi, bj, cg_halo, conv_flag |
62 |
INTEGER iter, is, js, ie, je, colx, coly, k |
63 |
_RL dot_p1, dot_p2, alpha_k, beta_k, resid, resid_0 |
64 |
_RL dot_p1_tile (nSx,nSy) |
65 |
_RL dot_p2_tile (nSx,nSy) |
66 |
CHARACTER*(MAX_LEN_MBUF) msgBuf |
67 |
|
68 |
|
69 |
#ifdef ALLOW_PETSC |
70 |
INTEGER indices(2*(snx*nsx*sny*nsy)) |
71 |
INTEGER n_dofs_cum_sum (0:nPx*nPy-1), idx(1) |
72 |
_RL rhs_values(2*(snx*nsx*sny*nsy)) |
73 |
_RL solution_values(2*(snx*nsx*sny*nsy)) |
74 |
! _RL mat_values (2*Nx*Ny,2*(snx*nsx*sny*nsy)) |
75 |
_RL mat_values (18,1), mat_val_return(1) |
76 |
INTEGER indices_col(18) |
77 |
INTEGER local_dofs, global_dofs, dof_index, dof_index_col |
78 |
INTEGER local_offset |
79 |
Mat matrix |
80 |
KSP ksp |
81 |
PC pc |
82 |
Vec rhs |
83 |
Vec solution |
84 |
PetscErrorCode ierr |
85 |
#ifdef ALLOW_USE_MPI |
86 |
integer mpiRC, mpiMyWid |
87 |
#endif |
88 |
#endif |
89 |
|
90 |
|
91 |
#ifdef ALLOW_STREAMICE |
92 |
|
93 |
CALL TIMER_START ('STREAMICE_CG_SOLVE',myThid) |
94 |
|
95 |
#ifdef ALLOW_PETSC |
96 |
|
97 |
#ifdef ALLOW_USE_MPI |
98 |
|
99 |
|
100 |
CALL MPI_COMM_RANK( MPI_COMM_WORLD, mpiMyWId, mpiRC ) |
101 |
local_dofs = n_dofs_process (mpiMyWid) |
102 |
global_dofs = 0 |
103 |
|
104 |
n_dofs_cum_sum(0) = 0 |
105 |
DO i=0,nPx*nPy-1 |
106 |
global_dofs = global_dofs + n_dofs_process (i) |
107 |
if (i.ge.1) THEN |
108 |
n_dofs_cum_sum(i) = n_dofs_cum_sum(i-1)+ |
109 |
& n_dofs_process(i-1) |
110 |
endif |
111 |
ENDDO |
112 |
local_offset = n_dofs_cum_sum(mpimywid) |
113 |
|
114 |
#else |
115 |
|
116 |
local_dofs = n_dofs_process (0) |
117 |
global_dofs = local_dofs |
118 |
local_offset = 0 |
119 |
|
120 |
#endif |
121 |
|
122 |
! call petscInitialize(PETSC_NULL_CHARACTER,ierr) |
123 |
|
124 |
!---------------------- |
125 |
|
126 |
call VecCreate(PETSC_COMM_WORLD, rhs, ierr) |
127 |
call VecSetSizes(rhs, local_dofs, global_dofs, ierr) |
128 |
call VecSetType(rhs, VECMPI, ierr) |
129 |
|
130 |
call VecCreate(PETSC_COMM_WORLD, solution, ierr) |
131 |
call VecSetSizes(solution, local_dofs, global_dofs, ierr) |
132 |
call VecSetType(solution, VECMPI, ierr) |
133 |
|
134 |
do i=1,local_dofs |
135 |
indices(i) = i-1 + local_offset |
136 |
end do |
137 |
do i=1,2*nSx*nSy*sNx*sNy |
138 |
rhs_values (i) = 0. _d 0 |
139 |
solution_values (i) = 0. _d 0 |
140 |
enddo |
141 |
|
142 |
! gather rhs and initial guess values to populate petsc vectors |
143 |
|
144 |
DO bj = myByLo(myThid), myByHi(myThid) |
145 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
146 |
DO j=1,sNy |
147 |
DO i=1,sNx |
148 |
|
149 |
dof_index = INT(streamice_petsc_dofs_u(i,j,bi,bj)) |
150 |
& - local_offset |
151 |
|
152 |
if (dof_index.ge.0) THEN |
153 |
|
154 |
rhs_values(dof_index+1) = cg_Bu(i,j,bi,bj) |
155 |
solution_values(dof_index+1) = cg_Uin(i,j,bi,bj) |
156 |
|
157 |
endif |
158 |
|
159 |
!--------------- |
160 |
|
161 |
dof_index = INT(streamice_petsc_dofs_v(i,j,bi,bj)) |
162 |
& - local_offset |
163 |
|
164 |
if (dof_index.ge.0) THEN |
165 |
|
166 |
rhs_values(dof_index+1) = cg_Bv(i,j,bi,bj) |
167 |
solution_values(dof_index+1) = cg_Vin(i,j,bi,bj) |
168 |
|
169 |
endif |
170 |
|
171 |
ENDDO |
172 |
ENDDO |
173 |
ENDDO |
174 |
ENDDO |
175 |
|
176 |
|
177 |
call VecSetValues(rhs, local_dofs, indices, rhs_values, |
178 |
& INSERT_VALUES, ierr) |
179 |
call VecAssemblyBegin(rhs, ierr) |
180 |
call VecAssemblyEnd(rhs, ierr) |
181 |
|
182 |
|
183 |
call VecSetValues(solution, local_dofs, indices, |
184 |
& solution_values, INSERT_VALUES, ierr) |
185 |
call VecAssemblyBegin(solution, ierr) |
186 |
call VecAssemblyEnd(solution, ierr) |
187 |
|
188 |
|
189 |
call MatCreateAIJ (PETSC_COMM_WORLD, |
190 |
& local_dofs, local_dofs, |
191 |
& global_dofs, global_dofs, |
192 |
& 18, PETSC_NULL_INTEGER, |
193 |
& 18, PETSC_NULL_INTEGER, |
194 |
& matrix, ierr) |
195 |
|
196 |
! populate petsc matrix |
197 |
|
198 |
DO bj = myByLo(myThid), myByHi(myThid) |
199 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
200 |
DO j=1,sNy |
201 |
DO i=1,sNx |
202 |
|
203 |
dof_index = INT(streamice_petsc_dofs_u(i,j,bi,bj)) |
204 |
! & - local_offset |
205 |
|
206 |
IF (dof_index .ge. 0) THEN |
207 |
|
208 |
DO k=1,18 |
209 |
indices_col(k) = 0 |
210 |
mat_values(k,1) = 0. _d 0 |
211 |
ENDDO |
212 |
k=0 |
213 |
|
214 |
DO coly=-1,1 |
215 |
DO colx=-1,1 |
216 |
|
217 |
dof_index_col = streamice_petsc_dofs_u(i+colx,j+coly,bi,bj) |
218 |
|
219 |
if (dof_index_col.ge.0) THEN |
220 |
! pscal = A_uu(i,j,bi,bj,colx,coly) |
221 |
! CALL MatSetValue (matrix,dof_index, dof_index_col, |
222 |
! & pscal,INSERT_VALUES,ierr) |
223 |
k=k+1 |
224 |
mat_values (k,1) = A_uu(i,j,bi,bj,colx,coly) |
225 |
indices_col (k) = dof_index_col |
226 |
endif |
227 |
|
228 |
dof_index_col = streamice_petsc_dofs_v(i+colx,j+coly,bi,bj) |
229 |
|
230 |
if (dof_index_col.ge.0) THEN |
231 |
! CALL MatSetValue (matrix,dof_index, dof_index_col, |
232 |
! & A_uv(i,j,bi,bj,colx,coly),INSERT_VALUES,ierr) |
233 |
k=k+1 |
234 |
mat_values (k,1) = A_uv(i,j,bi,bj,colx,coly) |
235 |
indices_col (k) = dof_index_col |
236 |
endif |
237 |
|
238 |
ENDDO |
239 |
ENDDO |
240 |
|
241 |
call matSetValues (matrix, 1, dof_index, k, indices_col, |
242 |
& mat_values,INSERT_VALUES,ierr) |
243 |
|
244 |
|
245 |
ENDIF |
246 |
|
247 |
! ---------------------------------------------- |
248 |
|
249 |
dof_index = INT(streamice_petsc_dofs_v(i,j,bi,bj)) |
250 |
! & - local_offset |
251 |
|
252 |
IF (dof_index .ge. 0) THEN |
253 |
|
254 |
DO k=1,18 |
255 |
indices_col(k) = 0 |
256 |
mat_values(k,1) = 0. _d 0 |
257 |
ENDDO |
258 |
k=0 |
259 |
|
260 |
DO coly=-1,1 |
261 |
DO colx=-1,1 |
262 |
|
263 |
dof_index_col = streamice_petsc_dofs_u(i+colx,j+coly,bi,bj) |
264 |
|
265 |
if (dof_index_col.ge.0) THEN |
266 |
! CALL MatSetValue (matrix,dof_index, dof_index_col, |
267 |
! & A_vu(i,j,bi,bj,colx,coly),INSERT_VALUES,ierr) |
268 |
k=k+1 |
269 |
mat_values (k,1) = A_vu(i,j,bi,bj,colx,coly) |
270 |
indices_col (k) = dof_index_col |
271 |
endif |
272 |
|
273 |
dof_index_col = streamice_petsc_dofs_v(i+colx,j+coly,bi,bj) |
274 |
|
275 |
if (dof_index_col.ge.0) THEN |
276 |
! CALL MatSetValue (matrix,dof_index, dof_index_col, |
277 |
! & A_vv(i,j,bi,bj,colx,coly),INSERT_VALUES,ierr) |
278 |
k=k+1 |
279 |
mat_values (k,1) = A_vv(i,j,bi,bj,colx,coly) |
280 |
indices_col (k) = dof_index_col |
281 |
endif |
282 |
|
283 |
ENDDO |
284 |
ENDDO |
285 |
|
286 |
call matSetValues (matrix, 1, dof_index, k, indices_col, |
287 |
& mat_values,INSERT_VALUES,ierr) |
288 |
ENDIF |
289 |
|
290 |
ENDDO |
291 |
ENDDO |
292 |
ENDDO |
293 |
ENDDO |
294 |
|
295 |
call MatAssemblyBegin(matrix,MAT_FINAL_ASSEMBLY,ierr) |
296 |
call MatAssemblyEnd(matrix,MAT_FINAL_ASSEMBLY,ierr) |
297 |
|
298 |
|
299 |
call KSPCreate(PETSC_COMM_WORLD, ksp, ierr) |
300 |
call KSPSetOperators(ksp, matrix, matrix, |
301 |
& DIFFERENT_NONZERO_PATTERN, ierr) |
302 |
|
303 |
SELECT CASE (PETSC_SOLVER_TYPE) |
304 |
CASE ('CG') |
305 |
PRINT *, "PETSC SOLVER: SELECTED CG" |
306 |
call KSPSetType(ksp, KSPCG, ierr) |
307 |
CASE ('GMRES') |
308 |
PRINT *, "PETSC SOLVER: SELECTED GMRES" |
309 |
call KSPSetType(ksp, KSPGMRES, ierr) |
310 |
CASE ('BICG') |
311 |
PRINT *, "PETSC SOLVER: SELECTED BICG" |
312 |
call KSPSetType(ksp, KSPBICG, ierr) |
313 |
CASE DEFAULT |
314 |
PRINT *, "PETSC SOLVER: SELECTED DEFAULT" |
315 |
call KSPSetType(ksp, KSPCG, ierr) |
316 |
END SELECT |
317 |
|
318 |
call KSPGetPC(ksp, pc, ierr) |
319 |
call KSPSetTolerances(ksp,tolerance, |
320 |
& PETSC_DEFAULT_DOUBLE_PRECISION, |
321 |
& PETSC_DEFAULT_DOUBLE_PRECISION, |
322 |
& streamice_max_cg_iter,ierr) |
323 |
|
324 |
SELECT CASE (PETSC_PRECOND_TYPE) |
325 |
CASE ('BLOCKJACOBI') |
326 |
PRINT *, "PETSC PRECOND: SELECTED BJACOBI" |
327 |
call PCSetType(pc, PCBJACOBI, ierr) |
328 |
CASE ('JACOBI') |
329 |
PRINT *, "PETSC PRECOND: SELECTED JACOBI" |
330 |
call PCSetType(pc, PCJACOBI, ierr) |
331 |
CASE ('ILU') |
332 |
PRINT *, "PETSC PRECOND: SELECTED ILU" |
333 |
call PCSetType(pc, PCILU, ierr) |
334 |
CASE DEFAULT |
335 |
PRINT *, "PETSC PRECOND: SELECTED DEFAULT" |
336 |
call PCSetType(pc, PCBJACOBI, ierr) |
337 |
END SELECT |
338 |
|
339 |
|
340 |
call KSPSolve(ksp, rhs, solution, ierr) |
341 |
call KSPGetIterationNumber(ksp,iters,ierr) |
342 |
|
343 |
call VecGetValues(solution,local_dofs,indices, |
344 |
& solution_values,ierr) |
345 |
|
346 |
DO bj = myByLo(myThid), myByHi(myThid) |
347 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
348 |
DO j=1,sNy |
349 |
DO i=1,sNx |
350 |
|
351 |
dof_index = INT(streamice_petsc_dofs_u(i,j,bi,bj)) |
352 |
& - local_offset |
353 |
if (dof_index.ge.0) THEN |
354 |
cg_Uin(i,j,bi,bj) = solution_values(dof_index+1) |
355 |
endif |
356 |
|
357 |
dof_index = INT(streamice_petsc_dofs_v(i,j,bi,bj)) |
358 |
& - local_offset |
359 |
if (dof_index.ge.0) THEN |
360 |
cg_Vin(i,j,bi,bj) = solution_values(dof_index+1) |
361 |
endif |
362 |
|
363 |
ENDDO |
364 |
ENDDO |
365 |
ENDDO |
366 |
ENDDO |
367 |
|
368 |
|
369 |
|
370 |
#else |
371 |
|
372 |
|
373 |
iters = streamice_max_cg_iter |
374 |
conv_flag = 0 |
375 |
|
376 |
|
377 |
DO bj = myByLo(myThid), myByHi(myThid) |
378 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
379 |
DO j=1,sNy |
380 |
DO i=1,sNx |
381 |
Zu_SI (i,j,bi,bj) = 0. _d 0 |
382 |
Zv_SI (i,j,bi,bj) = 0. _d 0 |
383 |
Ru_SI (i,j,bi,bj) = 0. _d 0 |
384 |
Rv_SI (i,j,bi,bj) = 0. _d 0 |
385 |
Au_SI (i,j,bi,bj) = 0. _d 0 |
386 |
Av_SI (i,j,bi,bj) = 0. _d 0 |
387 |
Du_SI (i,j,bi,bj) = 0. _d 0 |
388 |
Dv_SI (i,j,bi,bj) = 0. _d 0 |
389 |
ENDDO |
390 |
ENDDO |
391 |
ENDDO |
392 |
ENDDO |
393 |
|
394 |
C FIND INITIAL RESIDUAL, and initialize r |
395 |
|
396 |
! #ifdef STREAMICE_CONSTRUCT_MATRIX |
397 |
|
398 |
|
399 |
|
400 |
DO bj = myByLo(myThid), myByHi(myThid) |
401 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
402 |
DO j=1,sNy |
403 |
DO i=1,sNx |
404 |
DO colx=-1,1 |
405 |
DO coly=-1,1 |
406 |
Au_SI(i,j,bi,bj) = Au_SI(i,j,bi,bj) + |
407 |
& A_uu(i,j,bi,bj,colx,coly)* |
408 |
& cg_Uin(i+colx,j+coly,bi,bj)+ |
409 |
& A_uv(i,j,bi,bj,colx,coly)* |
410 |
& cg_Vin(i+colx,j+coly,bi,bj) |
411 |
|
412 |
|
413 |
Av_SI(i,j,bi,bj) = Av_SI(i,j,bi,bj) + |
414 |
& A_vu(i,j,bi,bj,colx,coly)* |
415 |
& cg_Uin(i+colx,j+coly,bi,bj)+ |
416 |
& A_vv(i,j,bi,bj,colx,coly)* |
417 |
& cg_Vin(i+colx,j+coly,bi,bj) |
418 |
ENDDO |
419 |
ENDDO |
420 |
ENDDO |
421 |
ENDDO |
422 |
ENDDO |
423 |
ENDDO |
424 |
|
425 |
|
426 |
_EXCH_XY_RL( Au_SI, myThid ) |
427 |
_EXCH_XY_RL( Av_SI, myThid ) |
428 |
|
429 |
|
430 |
DO bj = myByLo(myThid), myByHi(myThid) |
431 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
432 |
DO j=1-OLy,sNy+OLy |
433 |
DO i=1-OLx,sNx+OLx |
434 |
Ru_SI(i,j,bi,bj)=cg_Bu(i,j,bi,bj)- |
435 |
& Au_SI(i,j,bi,bj) |
436 |
Rv_SI(i,j,bi,bj)=cg_Bv(i,j,bi,bj)- |
437 |
& Av_SI(i,j,bi,bj) |
438 |
ENDDO |
439 |
ENDDO |
440 |
dot_p1_tile(bi,bj) = 0. _d 0 |
441 |
dot_p2_tile(bi,bj) = 0. _d 0 |
442 |
ENDDO |
443 |
ENDDO |
444 |
|
445 |
|
446 |
|
447 |
DO bj = myByLo(myThid), myByHi(myThid) |
448 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
449 |
DO j=1,sNy |
450 |
DO i=1,sNx |
451 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) |
452 |
& dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Ru_SI(i,j,bi,bj)**2 |
453 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) |
454 |
& dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Rv_SI(i,j,bi,bj)**2 |
455 |
ENDDO |
456 |
ENDDO |
457 |
ENDDO |
458 |
ENDDO |
459 |
|
460 |
CALL GLOBAL_SUM_TILE_RL( dot_p1_tile, dot_p1, myThid ) |
461 |
resid_0 = sqrt(dot_p1) |
462 |
|
463 |
WRITE(msgBuf,'(A,E14.7)') 'CONJ GRAD INIT RESID, ', |
464 |
& resid_0 |
465 |
CALL PRINT_MESSAGE( msgBuf, standardMessageUnit, |
466 |
& SQUEEZE_RIGHT , 1) |
467 |
|
468 |
C CCCCCCCCCCCCCCCCCCCC |
469 |
|
470 |
DO bj = myByLo(myThid), myByHi(myThid) |
471 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
472 |
DO j=1-OLy,sNy+OLy |
473 |
DO i=1-OLx,sNx+OLx |
474 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) |
475 |
& Zu_SI(i,j,bi,bj)=Ru_SI(i,j,bi,bj) / DIAGu_SI(i,j,bi,bj) |
476 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) |
477 |
& Zv_SI(i,j,bi,bj)=Rv_SI(i,j,bi,bj) / DIAGv_SI(i,j,bi,bj) |
478 |
ENDDO |
479 |
ENDDO |
480 |
ENDDO |
481 |
ENDDO |
482 |
|
483 |
cg_halo = min(OLx-1,OLy-1) |
484 |
conv_flag = 0 |
485 |
|
486 |
DO bj = myByLo(myThid), myByHi(myThid) |
487 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
488 |
DO j=1-OLy,sNy+OLy |
489 |
DO i=1-OLx,sNx+OLx |
490 |
Du_SI(i,j,bi,bj)=Zu_SI(i,j,bi,bj) |
491 |
Dv_SI(i,j,bi,bj)=Zv_SI(i,j,bi,bj) |
492 |
ENDDO |
493 |
ENDDO |
494 |
ENDDO |
495 |
ENDDO |
496 |
|
497 |
resid = resid_0 |
498 |
iters = 0 |
499 |
|
500 |
c !!!!!!!!!!!!!!!!!! |
501 |
c !! !! |
502 |
c !! MAIN CG LOOP !! |
503 |
c !! !! |
504 |
c !!!!!!!!!!!!!!!!!! |
505 |
|
506 |
|
507 |
|
508 |
|
509 |
c ! initially, b-grid data is valid up to 3 halo nodes out -- right? (check for MITgcm!!) |
510 |
|
511 |
WRITE(msgBuf,'(A)') 'BEGINNING MAIN CG LOOP' |
512 |
CALL PRINT_MESSAGE( msgBuf, standardMessageUnit, |
513 |
& SQUEEZE_RIGHT , 1) |
514 |
|
515 |
! IF(STREAMICE_construct_matrix) CALL STREAMICE_CG_MAKE_A(myThid) |
516 |
|
517 |
|
518 |
do iter = 1, streamice_max_cg_iter |
519 |
if (resid .gt. tolerance*resid_0) then |
520 |
|
521 |
c to avoid using "exit" |
522 |
iters = iters + 1 |
523 |
|
524 |
is = 1 - cg_halo |
525 |
ie = sNx + cg_halo |
526 |
js = 1 - cg_halo |
527 |
je = sNy + cg_halo |
528 |
|
529 |
DO bj = myByLo(myThid), myByHi(myThid) |
530 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
531 |
DO j=1-OLy,sNy+OLy |
532 |
DO i=1-OLx,sNx+OLx |
533 |
Au_SI(i,j,bi,bj) = 0. _d 0 |
534 |
Av_SI(i,j,bi,bj) = 0. _d 0 |
535 |
ENDDO |
536 |
ENDDO |
537 |
ENDDO |
538 |
ENDDO |
539 |
|
540 |
! IF (STREAMICE_construct_matrix) THEN |
541 |
|
542 |
! #ifdef STREAMICE_CONSTRUCT_MATRIX |
543 |
|
544 |
DO bj = myByLo(myThid), myByHi(myThid) |
545 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
546 |
DO j=js,je |
547 |
DO i=is,ie |
548 |
DO colx=-1,1 |
549 |
DO coly=-1,1 |
550 |
Au_SI(i,j,bi,bj) = Au_SI(i,j,bi,bj) + |
551 |
& A_uu(i,j,bi,bj,colx,coly)* |
552 |
& Du_SI(i+colx,j+coly,bi,bj)+ |
553 |
& A_uv(i,j,bi,bj,colx,coly)* |
554 |
& Dv_SI(i+colx,j+coly,bi,bj) |
555 |
Av_SI(i,j,bi,bj) = Av_SI(i,j,bi,bj) + |
556 |
& A_vu(i,j,bi,bj,colx,coly)* |
557 |
& Du_SI(i+colx,j+coly,bi,bj)+ |
558 |
& A_vv(i,j,bi,bj,colx,coly)* |
559 |
& Dv_SI(i+colx,j+coly,bi,bj) |
560 |
ENDDO |
561 |
ENDDO |
562 |
ENDDO |
563 |
ENDDO |
564 |
ENDDO |
565 |
ENDDO |
566 |
|
567 |
! else |
568 |
! #else |
569 |
! |
570 |
! CALL STREAMICE_CG_ACTION( myThid, |
571 |
! O Au_SI, |
572 |
! O Av_SI, |
573 |
! I Du_SI, |
574 |
! I Dv_SI, |
575 |
! I is,ie,js,je) |
576 |
! |
577 |
! ! ENDIF |
578 |
! |
579 |
! #endif |
580 |
|
581 |
|
582 |
DO bj = myByLo(myThid), myByHi(myThid) |
583 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
584 |
dot_p1_tile(bi,bj) = 0. _d 0 |
585 |
dot_p2_tile(bi,bj) = 0. _d 0 |
586 |
ENDDO |
587 |
ENDDO |
588 |
|
589 |
DO bj = myByLo(myThid), myByHi(myThid) |
590 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
591 |
DO j=1,sNy |
592 |
DO i=1,sNx |
593 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) THEN |
594 |
dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Zu_SI(i,j,bi,bj)* |
595 |
& Ru_SI(i,j,bi,bj) |
596 |
dot_p2_tile(bi,bj)=dot_p2_tile(bi,bj)+Du_SI(i,j,bi,bj)* |
597 |
& Au_SI(i,j,bi,bj) |
598 |
ENDIF |
599 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) THEN |
600 |
dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Zv_SI(i,j,bi,bj)* |
601 |
& Rv_SI(i,j,bi,bj) |
602 |
dot_p2_tile(bi,bj)=dot_p2_tile(bi,bj)+Dv_SI(i,j,bi,bj)* |
603 |
& Av_SI(i,j,bi,bj) |
604 |
ENDIF |
605 |
ENDDO |
606 |
ENDDO |
607 |
ENDDO |
608 |
ENDDO |
609 |
|
610 |
CALL GLOBAL_SUM_TILE_RL( dot_p1_tile, dot_p1, myThid ) |
611 |
CALL GLOBAL_SUM_TILE_RL( dot_p2_tile, dot_p2, myThid ) |
612 |
alpha_k = dot_p1/dot_p2 |
613 |
|
614 |
DO bj = myByLo(myThid), myByHi(myThid) |
615 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
616 |
DO j=1-OLy,sNy+OLy |
617 |
DO i=1-OLx,sNx+OLx |
618 |
|
619 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) THEN |
620 |
cg_Uin(i,j,bi,bj)=cg_Uin(i,j,bi,bj)+ |
621 |
& alpha_k*Du_SI(i,j,bi,bj) |
622 |
Ru_old_SI(i,j,bi,bj) = Ru_SI(i,j,bi,bj) |
623 |
Zu_old_SI(i,j,bi,bj) = Zu_SI(i,j,bi,bj) |
624 |
Ru_SI(i,j,bi,bj) = Ru_SI(i,j,bi,bj)- |
625 |
& alpha_k*Au_SI(i,j,bi,bj) |
626 |
Zu_SI(i,j,bi,bj) = Ru_SI(i,j,bi,bj) / |
627 |
& DIAGu_SI(i,j,bi,bj) |
628 |
ENDIF |
629 |
|
630 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) THEN |
631 |
cg_Vin(i,j,bi,bj)=cg_Vin(i,j,bi,bj)+ |
632 |
& alpha_k*Dv_SI(i,j,bi,bj) |
633 |
Rv_old_SI(i,j,bi,bj) = Rv_SI(i,j,bi,bj) |
634 |
Zv_old_SI(i,j,bi,bj) = Zv_SI(i,j,bi,bj) |
635 |
Rv_SI(i,j,bi,bj) = Rv_SI(i,j,bi,bj)- |
636 |
& alpha_k*Av_SI(i,j,bi,bj) |
637 |
Zv_SI(i,j,bi,bj) = Rv_SI(i,j,bi,bj) / |
638 |
& DIAGv_SI(i,j,bi,bj) |
639 |
|
640 |
ENDIF |
641 |
ENDDO |
642 |
ENDDO |
643 |
ENDDO |
644 |
ENDDO |
645 |
|
646 |
DO bj = myByLo(myThid), myByHi(myThid) |
647 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
648 |
dot_p1_tile(bi,bj) = 0. _d 0 |
649 |
dot_p2_tile(bi,bj) = 0. _d 0 |
650 |
ENDDO |
651 |
ENDDO |
652 |
|
653 |
DO bj = myByLo(myThid), myByHi(myThid) |
654 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
655 |
DO j=1,sNy |
656 |
DO i=1,sNx |
657 |
|
658 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) THEN |
659 |
dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Zu_SI(i,j,bi,bj)* |
660 |
& Ru_SI(i,j,bi,bj) |
661 |
dot_p2_tile(bi,bj)=dot_p2_tile(bi,bj)+Zu_old_SI(i,j,bi,bj)* |
662 |
& Ru_old_SI(i,j,bi,bj) |
663 |
ENDIF |
664 |
|
665 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) THEN |
666 |
dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Zv_SI(i,j,bi,bj)* |
667 |
& Rv_SI(i,j,bi,bj) |
668 |
dot_p2_tile(bi,bj)=dot_p2_tile(bi,bj)+Zv_old_SI(i,j,bi,bj)* |
669 |
& Rv_old_SI(i,j,bi,bj) |
670 |
ENDIF |
671 |
|
672 |
ENDDO |
673 |
ENDDO |
674 |
ENDDO |
675 |
ENDDO |
676 |
|
677 |
CALL GLOBAL_SUM_TILE_RL( dot_p1_tile, dot_p1, myThid ) |
678 |
CALL GLOBAL_SUM_TILE_RL( dot_p2_tile, dot_p2, myThid ) |
679 |
|
680 |
beta_k = dot_p1/dot_p2 |
681 |
|
682 |
DO bj = myByLo(myThid), myByHi(myThid) |
683 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
684 |
DO j=1-OLy,sNy+OLy |
685 |
DO i=1-OLx,sNx+OLx |
686 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) |
687 |
& Du_SI(i,j,bi,bj)=beta_k*Du_SI(i,j,bi,bj)+ |
688 |
& Zu_SI(i,j,bi,bj) |
689 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) |
690 |
& Dv_SI(i,j,bi,bj)=beta_k*Dv_SI(i,j,bi,bj)+ |
691 |
& Zv_SI(i,j,bi,bj) |
692 |
ENDDO |
693 |
ENDDO |
694 |
ENDDO |
695 |
ENDDO |
696 |
|
697 |
DO bj = myByLo(myThid), myByHi(myThid) |
698 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
699 |
dot_p1_tile(bi,bj) = 0. _d 0 |
700 |
ENDDO |
701 |
ENDDO |
702 |
|
703 |
DO bj = myByLo(myThid), myByHi(myThid) |
704 |
DO bi = myBxLo(myThid), myBxHi(myThid) |
705 |
DO j=1,sNy |
706 |
DO i=1,sNx |
707 |
IF (STREAMICE_umask(i,j,bi,bj).eq.1.0) |
708 |
& dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Ru_SI(i,j,bi,bj)**2 |
709 |
IF (STREAMICE_vmask(i,j,bi,bj).eq.1.0) |
710 |
& dot_p1_tile(bi,bj)=dot_p1_tile(bi,bj)+Rv_SI(i,j,bi,bj)**2 |
711 |
ENDDO |
712 |
ENDDO |
713 |
ENDDO |
714 |
ENDDO |
715 |
|
716 |
CALL GLOBAL_SUM_TILE_RL( dot_p1_tile, dot_p1, myThid ) |
717 |
resid = sqrt(dot_p1) |
718 |
|
719 |
! IF (iter .eq. 1) then |
720 |
! print *, alpha_k, beta_k, resid |
721 |
! ENDIF |
722 |
|
723 |
cg_halo = cg_halo - 1 |
724 |
|
725 |
if (cg_halo .eq. 0) then |
726 |
cg_halo = min(OLx-1,OLy-1) |
727 |
_EXCH_XY_RL( Du_SI, myThid ) |
728 |
_EXCH_XY_RL( Dv_SI, myThid ) |
729 |
_EXCH_XY_RL( Ru_SI, myThid ) |
730 |
_EXCH_XY_RL( Rv_SI, myThid ) |
731 |
_EXCH_XY_RL( cg_Uin, myThid ) |
732 |
_EXCH_XY_RL( cg_Vin, myThid ) |
733 |
endif |
734 |
|
735 |
|
736 |
endif |
737 |
enddo ! end of CG loop |
738 |
|
739 |
c to avoid using "exit" |
740 |
c if iters has reached max_iters there is no convergence |
741 |
|
742 |
IF (iters .lt. streamice_max_cg_iter) THEN |
743 |
conv_flag = 1 |
744 |
ENDIF |
745 |
|
746 |
! DO bj = myByLo(myThid), myByHi(myThid) |
747 |
! DO bi = myBxLo(myThid), myBxHi(myThid) |
748 |
! DO j=1-OLy,sNy+OLy |
749 |
! DO i=1-OLy,sNx+OLy |
750 |
! IF (STREAMICE_umask(i,j,bi,bj).eq.3.0) |
751 |
! & cg_Uin(i,j,bi,bj)=u_bdry_values_SI(i,j,bi,bj) |
752 |
! IF (STREAMICE_vmask(i,j,bi,bj).eq.3.0) |
753 |
! & cg_Vin(i,j,bi,bj)=v_bdry_values_SI(i,j,bi,bj) |
754 |
! ENDDO |
755 |
! ENDDO |
756 |
! ENDDO |
757 |
! ENDDO |
758 |
! |
759 |
! _EXCH_XY_RL( cg_Uin, myThid ) |
760 |
! _EXCH_XY_RL( cg_Vin, myThid ) |
761 |
|
762 |
#endif |
763 |
|
764 |
CALL TIMER_STOP ('STREAMICE_CG_SOLVE',myThid) |
765 |
|
766 |
|
767 |
#endif |
768 |
RETURN |
769 |
END |