1 |
C $Header: /u/gcmpack/MITgcm/model/src/cg3d.F,v 1.23 2010/01/17 21:55:48 jmc Exp $ |
2 |
C $Name: $ |
3 |
|
4 |
#include "CPP_OPTIONS.h" |
5 |
#ifdef TARGET_NEC_SX |
6 |
C set a sensible default for the outer loop unrolling parameter that can |
7 |
C be overriden in the Makefile with the DEFINES macro or in CPP_OPTIONS.h |
8 |
#ifndef CG3D_OUTERLOOPITERS |
9 |
# define CG3D_OUTERLOOPITERS 10 |
10 |
#endif |
11 |
#endif /* TARGET_NEC_SX */ |
12 |
|
13 |
CBOP |
14 |
C !ROUTINE: CG3D |
15 |
C !INTERFACE: |
16 |
SUBROUTINE CG3D( |
17 |
I cg3d_b, |
18 |
U cg3d_x, |
19 |
O firstResidual, |
20 |
O lastResidual, |
21 |
U numIters, |
22 |
I myIter, myThid ) |
23 |
C !DESCRIPTION: \bv |
24 |
C *==========================================================* |
25 |
C | SUBROUTINE CG3D |
26 |
C | o Three-dimensional grid problem conjugate-gradient |
27 |
C | inverter (with preconditioner). |
28 |
C *==========================================================* |
29 |
C | Con. grad is an iterative procedure for solving Ax = b. |
30 |
C | It requires the A be symmetric. |
31 |
C | This implementation assumes A is a seven-diagonal |
32 |
C | matrix of the form that arises in the discrete |
33 |
C | representation of the del^2 operator in a |
34 |
C | three-dimensional space. |
35 |
C | Notes: |
36 |
C | ====== |
37 |
C | This implementation can support shared-memory |
38 |
C | multi-threaded execution. In order to do this COMMON |
39 |
C | blocks are used for many of the arrays - even ones that |
40 |
C | are only used for intermedaite results. This design is |
41 |
C | OK if you want to all the threads to collaborate on |
42 |
C | solving the same problem. On the other hand if you want |
43 |
C | the threads to solve several different problems |
44 |
C | concurrently this implementation will not work. |
45 |
C *==========================================================* |
46 |
C \ev |
47 |
|
48 |
C !USES: |
49 |
IMPLICIT NONE |
50 |
C === Global data === |
51 |
#include "SIZE.h" |
52 |
#include "EEPARAMS.h" |
53 |
#include "PARAMS.h" |
54 |
#include "GRID.h" |
55 |
#include "SURFACE.h" |
56 |
#include "CG3D.h" |
57 |
|
58 |
C !INPUT/OUTPUT PARAMETERS: |
59 |
C === Routine arguments === |
60 |
C cg3d_b :: The source term or "right hand side" |
61 |
C cg3d_x :: The solution |
62 |
C firstResidual :: the initial residual before any iterations |
63 |
C lastResidual :: the actual residual reached |
64 |
C numIters :: Entry: the maximum number of iterations allowed |
65 |
C :: Exit: the actual number of iterations used |
66 |
C myIter :: Current iteration number in simulation |
67 |
C myThid :: my Thread Id number |
68 |
_RL cg3d_b(1-OLx:sNx+OLx,1-OLy:sNy+OLy,Nr,nSx,nSy) |
69 |
_RL cg3d_x(1-OLx:sNx+OLx,1-OLy:sNy+OLy,Nr,nSx,nSy) |
70 |
_RL firstResidual |
71 |
_RL lastResidual |
72 |
INTEGER numIters |
73 |
INTEGER myIter |
74 |
INTEGER myThid |
75 |
|
76 |
#ifdef ALLOW_NONHYDROSTATIC |
77 |
|
78 |
C !LOCAL VARIABLES: |
79 |
C === Local variables ==== |
80 |
C actualIts :: Number of iterations taken |
81 |
C actualResidual :: residual |
82 |
C bi,bj :: tile indices |
83 |
C eta_qrN :: Used in computing search directions |
84 |
C eta_qrNM1 suffix N and NM1 denote current and |
85 |
C cgBeta previous iterations respectively. |
86 |
C alpha |
87 |
C sumRHS :: Sum of right-hand-side. Sometimes this is a |
88 |
C useful debuggin/trouble shooting diagnostic. |
89 |
C For neumann problems sumRHS needs to be ~0. |
90 |
C or they converge at a non-zero residual. |
91 |
C err :: Measure of residual of Ax - b, usually the norm. |
92 |
C i, j, k :: Loop counters |
93 |
C it3d :: Loop counter for CG iterations |
94 |
C msgBuf :: Informational/error message buffer |
95 |
INTEGER actualIts |
96 |
_RL actualResidual |
97 |
INTEGER bi, bj |
98 |
INTEGER i, j, k, it3d |
99 |
INTEGER km1, kp1 |
100 |
_RL maskM1, maskP1 |
101 |
_RL err, errTile(nSx,nSy) |
102 |
_RL eta_qrN,eta_qrNtile(nSx,nSy) |
103 |
_RL eta_qrNM1 |
104 |
_RL cgBeta |
105 |
_RL alpha , alphaTile(nSx,nSy) |
106 |
_RL sumRHS, sumRHStile(nSx,nSy) |
107 |
_RL rhsMax |
108 |
_RL rhsNorm |
109 |
CHARACTER*(MAX_LEN_MBUF) msgBuf |
110 |
LOGICAL printResidual |
111 |
_RL surfFac |
112 |
#ifdef NONLIN_FRSURF |
113 |
INTEGER ks |
114 |
_RL surfTerm(sNx,sNy) |
115 |
#endif /* NONLIN_FRSURF */ |
116 |
CEOP |
117 |
|
118 |
IF ( select_rStar .NE. 0 ) THEN |
119 |
surfFac = freeSurfFac |
120 |
ELSE |
121 |
surfFac = 0. |
122 |
ENDIF |
123 |
#ifdef NONLIN_FRSURF |
124 |
DO j=1,sNy |
125 |
DO i=1,sNx |
126 |
surfTerm(i,j) = 0. |
127 |
ENDDO |
128 |
ENDDO |
129 |
#endif /* NONLIN_FRSURF */ |
130 |
|
131 |
C-- Initialise inverter |
132 |
eta_qrNM1 = 1. _d 0 |
133 |
|
134 |
C-- Normalise RHS |
135 |
rhsMax = 0. _d 0 |
136 |
DO bj=myByLo(myThid),myByHi(myThid) |
137 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
138 |
DO k=1,Nr |
139 |
DO j=1,sNy |
140 |
DO i=1,sNx |
141 |
cg3d_b(i,j,k,bi,bj) = cg3d_b(i,j,k,bi,bj)*cg3dNorm |
142 |
& * maskC(i,j,k,bi,bj) |
143 |
rhsMax = MAX(ABS(cg3d_b(i,j,k,bi,bj)),rhsMax) |
144 |
ENDDO |
145 |
ENDDO |
146 |
ENDDO |
147 |
ENDDO |
148 |
ENDDO |
149 |
_GLOBAL_MAX_RL( rhsMax, myThid ) |
150 |
rhsNorm = 1. _d 0 |
151 |
IF ( rhsMax .NE. 0. ) rhsNorm = 1. _d 0 / rhsMax |
152 |
DO bj=myByLo(myThid),myByHi(myThid) |
153 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
154 |
DO k=1,Nr |
155 |
DO j=1,sNy |
156 |
DO i=1,sNx |
157 |
cg3d_b(i,j,k,bi,bj) = cg3d_b(i,j,k,bi,bj)*rhsNorm |
158 |
cg3d_x(i,j,k,bi,bj) = cg3d_x(i,j,k,bi,bj)*rhsNorm |
159 |
ENDDO |
160 |
ENDDO |
161 |
ENDDO |
162 |
ENDDO |
163 |
ENDDO |
164 |
|
165 |
C-- Update overlaps |
166 |
c _EXCH_XYZ_RL( cg3d_b, myThid ) |
167 |
_EXCH_XYZ_RL( cg3d_x, myThid ) |
168 |
|
169 |
C-- Initial residual calculation (with free-Surface term) |
170 |
err = 0. _d 0 |
171 |
sumRHS = 0. _d 0 |
172 |
DO bj=myByLo(myThid),myByHi(myThid) |
173 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
174 |
errTile(bi,bj) = 0. _d 0 |
175 |
sumRHStile(bi,bj) = 0. _d 0 |
176 |
#ifdef NONLIN_FRSURF |
177 |
IF ( select_rStar .NE. 0 ) THEN |
178 |
DO j=1,sNy |
179 |
DO i=1,sNx |
180 |
surfTerm(i,j) = 0. |
181 |
ENDDO |
182 |
ENDDO |
183 |
DO k=1,Nr |
184 |
DO j=1,sNy |
185 |
DO i=1,sNx |
186 |
surfTerm(i,j) = surfTerm(i,j) |
187 |
& +cg3d_x(i,j,k,bi,bj)*drF(k)*h0FacC(i,j,k,bi,bj) |
188 |
ENDDO |
189 |
ENDDO |
190 |
ENDDO |
191 |
DO j=1,sNy |
192 |
DO i=1,sNx |
193 |
ks = kSurfC(i,j,bi,bj) |
194 |
surfTerm(i,j) = surfTerm(i,j)*cg3dNorm |
195 |
& *recip_Rcol(i,j,bi,bj)*recip_Rcol(i,j,bi,bj) |
196 |
& *rA(i,j,bi,bj)*deepFac2F(ks) |
197 |
& *recip_Bo(i,j,bi,bj)/deltaTMom/deltaTfreesurf |
198 |
ENDDO |
199 |
ENDDO |
200 |
ENDIF |
201 |
#endif /* NONLIN_FRSURF */ |
202 |
DO k=1,Nr |
203 |
km1 = MAX(k-1, 1 ) |
204 |
kp1 = MIN(k+1, Nr) |
205 |
maskM1 = 1. _d 0 |
206 |
maskP1 = 1. _d 0 |
207 |
IF ( k .EQ. 1 ) maskM1 = 0. _d 0 |
208 |
IF ( k .EQ. Nr) maskP1 = 0. _d 0 |
209 |
#ifdef TARGET_NEC_SX |
210 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
211 |
#endif /* TARGET_NEC_SX */ |
212 |
DO j=1,sNy |
213 |
DO i=1,sNx |
214 |
cg3d_r(i,j,k,bi,bj) = cg3d_b(i,j,k,bi,bj) |
215 |
& -( 0. |
216 |
& +aW3d( i, j, k, bi,bj)*cg3d_x(i-1,j, k, bi,bj) |
217 |
& +aW3d(i+1,j, k, bi,bj)*cg3d_x(i+1,j, k, bi,bj) |
218 |
& +aS3d( i, j, k, bi,bj)*cg3d_x( i,j-1,k, bi,bj) |
219 |
& +aS3d( i,j+1,k, bi,bj)*cg3d_x( i,j+1,k, bi,bj) |
220 |
& +aV3d( i, j, k, bi,bj)*cg3d_x( i, j,km1,bi,bj)*maskM1 |
221 |
& +aV3d( i, j,kp1,bi,bj)*cg3d_x( i, j,kp1,bi,bj)*maskP1 |
222 |
& +aC3d( i, j, k, bi,bj)*cg3d_x( i, j, k, bi,bj) |
223 |
#ifdef NONLIN_FRSURF |
224 |
& -surfFac*surfTerm(i,j)*drF(k)*h0FacC(i,j,k,bi,bj) |
225 |
#endif /* NONLIN_FRSURF */ |
226 |
& ) |
227 |
errTile(bi,bj) = errTile(bi,bj) |
228 |
& +cg3d_r(i,j,k,bi,bj)*cg3d_r(i,j,k,bi,bj) |
229 |
sumRHStile(bi,bj) = sumRHStile(bi,bj)+cg3d_b(i,j,k,bi,bj) |
230 |
ENDDO |
231 |
ENDDO |
232 |
DO J=1-1,sNy+1 |
233 |
DO I=1-1,sNx+1 |
234 |
cg3d_s(i,j,k,bi,bj) = 0. |
235 |
ENDDO |
236 |
ENDDO |
237 |
ENDDO |
238 |
ENDDO |
239 |
ENDDO |
240 |
CALL EXCH_S3D_RL( cg3d_r, Nr, myThid ) |
241 |
c CALL EXCH_S3D_RL( cg3d_s, Nr, myThid ) |
242 |
CALL GLOBAL_SUM_TILE_RL( sumRHStile, sumRHS, myThid ) |
243 |
CALL GLOBAL_SUM_TILE_RL( errTile, err, myThid ) |
244 |
IF ( debugLevel.GE.debLevC .AND. diagFreq.GT.0. ) THEN |
245 |
CALL WRITE_FLD_S3D_RL( |
246 |
I 'cg3d_r_I', 'I10', 1, Nr, cg3d_r, myIter, myThid ) |
247 |
ENDIF |
248 |
|
249 |
printResidual = .FALSE. |
250 |
IF ( debugLevel .GE. debLevZero ) THEN |
251 |
_BEGIN_MASTER( myThid ) |
252 |
printResidual = printResidualFreq.GE.1 |
253 |
WRITE(standardmessageunit,'(A,1P2E22.14)') |
254 |
& ' cg3d: Sum(rhs),rhsMax = ',sumRHS,rhsMax |
255 |
_END_MASTER( myThid ) |
256 |
ENDIF |
257 |
|
258 |
actualIts = 0 |
259 |
actualResidual = SQRT(err) |
260 |
firstResidual=actualResidual |
261 |
|
262 |
C >>>>>>>>>>>>>>> BEGIN SOLVER <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< |
263 |
DO 10 it3d=1, numIters |
264 |
|
265 |
IF ( actualResidual .LT. cg3dTargetResidual ) GOTO 11 |
266 |
C-- Solve preconditioning equation and update |
267 |
C-- conjugate direction vector "s". |
268 |
C Note. On the next two loops over all tiles the inner loop ranges |
269 |
C in sNx and sNy are expanded by 1 to avoid a communication |
270 |
C step. However this entails a bit of gynamastics because we only |
271 |
C want eta_qrN for the interior points. |
272 |
eta_qrN = 0. _d 0 |
273 |
DO bj=myByLo(myThid),myByHi(myThid) |
274 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
275 |
eta_qrNtile(bi,bj) = 0. _d 0 |
276 |
DO k=1,1 |
277 |
#ifdef TARGET_NEC_SX |
278 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
279 |
#endif /* TARGET_NEC_SX */ |
280 |
DO j=1-1,sNy+1 |
281 |
DO i=1-1,sNx+1 |
282 |
cg3d_q(i,j,k,bi,bj) = zMC(i,j,k,bi,bj) |
283 |
& *cg3d_r(i,j,k,bi,bj) |
284 |
ENDDO |
285 |
ENDDO |
286 |
ENDDO |
287 |
DO k=2,Nr |
288 |
#ifdef TARGET_NEC_SX |
289 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
290 |
#endif /* TARGET_NEC_SX */ |
291 |
DO j=1-1,sNy+1 |
292 |
DO i=1-1,sNx+1 |
293 |
cg3d_q(i,j,k,bi,bj) = zMC(i,j,k,bi,bj) |
294 |
& *( cg3d_r(i,j,k,bi,bj) |
295 |
& -zML(i,j,k,bi,bj)*cg3d_q(i,j,k-1,bi,bj) |
296 |
& ) |
297 |
ENDDO |
298 |
ENDDO |
299 |
ENDDO |
300 |
DO k=Nr,Nr |
301 |
#ifdef TARGET_NEC_SX |
302 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
303 |
#endif /* TARGET_NEC_SX */ |
304 |
DO j=1,sNy |
305 |
DO i=1,sNx |
306 |
eta_qrNtile(bi,bj) = eta_qrNtile(bi,bj) |
307 |
& +cg3d_q(i,j,k,bi,bj)*cg3d_r(i,j,k,bi,bj) |
308 |
ENDDO |
309 |
ENDDO |
310 |
ENDDO |
311 |
DO k=Nr-1,1,-1 |
312 |
#ifdef TARGET_NEC_SX |
313 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
314 |
#endif /* TARGET_NEC_SX */ |
315 |
DO j=1-1,sNy+1 |
316 |
DO i=1-1,sNx+1 |
317 |
cg3d_q(i,j,k,bi,bj) = cg3d_q(i,j,k,bi,bj) |
318 |
& -zMU(i,j,k,bi,bj)*cg3d_q(i,j,k+1,bi,bj) |
319 |
ENDDO |
320 |
ENDDO |
321 |
#ifdef TARGET_NEC_SX |
322 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
323 |
#endif /* TARGET_NEC_SX */ |
324 |
DO j=1,sNy |
325 |
DO i=1,sNx |
326 |
eta_qrNtile(bi,bj) = eta_qrNtile(bi,bj) |
327 |
& +cg3d_q(i,j,k,bi,bj)*cg3d_r(i,j,k,bi,bj) |
328 |
ENDDO |
329 |
ENDDO |
330 |
ENDDO |
331 |
ENDDO |
332 |
ENDDO |
333 |
|
334 |
CALL GLOBAL_SUM_TILE_RL( eta_qrNtile,eta_qrN,myThid ) |
335 |
CcnhDebugStarts |
336 |
C WRITE(*,*) ' CG3D: Iteration ',it3d-1,' eta_qrN = ',eta_qrN |
337 |
CcnhDebugEnds |
338 |
cgBeta = eta_qrN/eta_qrNM1 |
339 |
CcnhDebugStarts |
340 |
C WRITE(*,*) ' CG3D: Iteration ',it3d-1,' beta = ',cgBeta |
341 |
CcnhDebugEnds |
342 |
eta_qrNM1 = eta_qrN |
343 |
|
344 |
DO bj=myByLo(myThid),myByHi(myThid) |
345 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
346 |
DO k=1,Nr |
347 |
DO j=1-1,sNy+1 |
348 |
DO i=1-1,sNx+1 |
349 |
cg3d_s(i,j,k,bi,bj) = cg3d_q(i,j,k,bi,bj) |
350 |
& + cgBeta*cg3d_s(i,j,k,bi,bj) |
351 |
ENDDO |
352 |
ENDDO |
353 |
ENDDO |
354 |
ENDDO |
355 |
ENDDO |
356 |
|
357 |
C== Evaluate laplace operator on conjugate gradient vector |
358 |
C== q = A.s |
359 |
alpha = 0. _d 0 |
360 |
DO bj=myByLo(myThid),myByHi(myThid) |
361 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
362 |
alphaTile(bi,bj) = 0. _d 0 |
363 |
#ifdef NONLIN_FRSURF |
364 |
IF ( select_rStar .NE. 0 ) THEN |
365 |
DO j=1,sNy |
366 |
DO i=1,sNx |
367 |
surfTerm(i,j) = 0. |
368 |
ENDDO |
369 |
ENDDO |
370 |
DO k=1,Nr |
371 |
DO j=1,sNy |
372 |
DO i=1,sNx |
373 |
surfTerm(i,j) = surfTerm(i,j) |
374 |
& +cg3d_s(i,j,k,bi,bj)*drF(k)*h0FacC(i,j,k,bi,bj) |
375 |
ENDDO |
376 |
ENDDO |
377 |
ENDDO |
378 |
DO j=1,sNy |
379 |
DO i=1,sNx |
380 |
ks = kSurfC(i,j,bi,bj) |
381 |
surfTerm(i,j) = surfTerm(i,j)*cg3dNorm |
382 |
& *recip_Rcol(i,j,bi,bj)*recip_Rcol(i,j,bi,bj) |
383 |
& *rA(i,j,bi,bj)*deepFac2F(ks) |
384 |
& *recip_Bo(i,j,bi,bj)/deltaTMom/deltaTfreesurf |
385 |
ENDDO |
386 |
ENDDO |
387 |
ENDIF |
388 |
#endif /* NONLIN_FRSURF */ |
389 |
IF ( Nr .GT. 1 ) THEN |
390 |
k=1 |
391 |
#ifdef TARGET_NEC_SX |
392 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
393 |
#endif /* TARGET_NEC_SX */ |
394 |
DO j=1,sNy |
395 |
DO i=1,sNx |
396 |
cg3d_q(i,j,k,bi,bj) = |
397 |
& aW3d( i, j, k, bi,bj)*cg3d_s(i-1,j, k, bi,bj) |
398 |
& +aW3d(i+1,j, k, bi,bj)*cg3d_s(i+1,j, k, bi,bj) |
399 |
& +aS3d( i, j, k, bi,bj)*cg3d_s( i,j-1,k, bi,bj) |
400 |
& +aS3d( i,j+1,k, bi,bj)*cg3d_s( i,j+1,k, bi,bj) |
401 |
& +aV3d( i, j,k+1,bi,bj)*cg3d_s( i, j,k+1,bi,bj) |
402 |
& +aC3d( i, j, k, bi,bj)*cg3d_s( i, j, k, bi,bj) |
403 |
#ifdef NONLIN_FRSURF |
404 |
& -surfFac*surfTerm(i,j)*drF(k)*h0FacC(i,j,k,bi,bj) |
405 |
#endif /* NONLIN_FRSURF */ |
406 |
alphaTile(bi,bj) = alphaTile(bi,bj) |
407 |
& +cg3d_s(i,j,k,bi,bj)*cg3d_q(i,j,k,bi,bj) |
408 |
ENDDO |
409 |
ENDDO |
410 |
ELSE |
411 |
k=1 |
412 |
#ifdef TARGET_NEC_SX |
413 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
414 |
#endif /* TARGET_NEC_SX */ |
415 |
DO j=1,sNy |
416 |
DO i=1,sNx |
417 |
cg3d_q(i,j,k,bi,bj) = |
418 |
& aW3d( i, j, k, bi,bj)*cg3d_s(i-1,j, k, bi,bj) |
419 |
& +aW3d(i+1,j, k, bi,bj)*cg3d_s(i+1,j, k, bi,bj) |
420 |
& +aS3d( i, j, k, bi,bj)*cg3d_s( i,j-1,k, bi,bj) |
421 |
& +aS3d( i,j+1,k, bi,bj)*cg3d_s( i,j+1,k, bi,bj) |
422 |
& +aC3d( i, j, k, bi,bj)*cg3d_s( i, j, k, bi,bj) |
423 |
#ifdef NONLIN_FRSURF |
424 |
& -surfFac*surfTerm(i,j)*drF(k)*h0FacC(i,j,k,bi,bj) |
425 |
#endif /* NONLIN_FRSURF */ |
426 |
alphaTile(bi,bj) = alphaTile(bi,bj) |
427 |
& +cg3d_s(i,j,k,bi,bj)*cg3d_q(i,j,k,bi,bj) |
428 |
ENDDO |
429 |
ENDDO |
430 |
ENDIF |
431 |
DO k=2,Nr-1 |
432 |
#ifdef TARGET_NEC_SX |
433 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
434 |
#endif /* TARGET_NEC_SX */ |
435 |
DO j=1,sNy |
436 |
DO i=1,sNx |
437 |
cg3d_q(i,j,k,bi,bj) = |
438 |
& aW3d( i, j, k, bi,bj)*cg3d_s(i-1,j, k, bi,bj) |
439 |
& +aW3d(i+1,j, k, bi,bj)*cg3d_s(i+1,j, k, bi,bj) |
440 |
& +aS3d( i, j, k, bi,bj)*cg3d_s( i,j-1,k, bi,bj) |
441 |
& +aS3d( i,j+1,k, bi,bj)*cg3d_s( i,j+1,k, bi,bj) |
442 |
& +aV3d( i, j, k, bi,bj)*cg3d_s( i, j,k-1,bi,bj) |
443 |
& +aV3d( i, j,k+1,bi,bj)*cg3d_s( i, j,k+1,bi,bj) |
444 |
& +aC3d( i, j, k, bi,bj)*cg3d_s( i, j, k, bi,bj) |
445 |
#ifdef NONLIN_FRSURF |
446 |
& -surfFac*surfTerm(i,j)*drF(k)*h0FacC(i,j,k,bi,bj) |
447 |
#endif /* NONLIN_FRSURF */ |
448 |
alphaTile(bi,bj) = alphaTile(bi,bj) |
449 |
& +cg3d_s(i,j,k,bi,bj)*cg3d_q(i,j,k,bi,bj) |
450 |
ENDDO |
451 |
ENDDO |
452 |
ENDDO |
453 |
IF ( Nr .GT. 1 ) THEN |
454 |
k=Nr |
455 |
#ifdef TARGET_NEC_SX |
456 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
457 |
#endif /* TARGET_NEC_SX */ |
458 |
DO j=1,sNy |
459 |
DO i=1,sNx |
460 |
cg3d_q(i,j,k,bi,bj) = |
461 |
& aW3d( i, j, k, bi,bj)*cg3d_s(i-1,j, k, bi,bj) |
462 |
& +aW3d(i+1,j, k, bi,bj)*cg3d_s(i+1,j, k, bi,bj) |
463 |
& +aS3d( i, j, k, bi,bj)*cg3d_s( i,j-1,k, bi,bj) |
464 |
& +aS3d( i,j+1,k, bi,bj)*cg3d_s( i,j+1,k, bi,bj) |
465 |
& +aV3d( i, j, k, bi,bj)*cg3d_s( i, j,k-1,bi,bj) |
466 |
& +aC3d( i, j, k, bi,bj)*cg3d_s( i, j, k, bi,bj) |
467 |
#ifdef NONLIN_FRSURF |
468 |
& -surfFac*surfTerm(i,j)*drF(k)*h0FacC(i,j,k,bi,bj) |
469 |
#endif /* NONLIN_FRSURF */ |
470 |
alphaTile(bi,bj) = alphaTile(bi,bj) |
471 |
& +cg3d_s(i,j,k,bi,bj)*cg3d_q(i,j,k,bi,bj) |
472 |
ENDDO |
473 |
ENDDO |
474 |
ENDIF |
475 |
ENDDO |
476 |
ENDDO |
477 |
CALL GLOBAL_SUM_TILE_RL( alphaTile, alpha, myThid ) |
478 |
CcnhDebugStarts |
479 |
C WRITE(*,*) ' CG3D: Iteration ',it3d-1,' SUM(s*q)= ',alpha |
480 |
CcnhDebugEnds |
481 |
alpha = eta_qrN/alpha |
482 |
CcnhDebugStarts |
483 |
C WRITE(*,*) ' CG3D: Iteration ',it3d-1,' alpha= ',alpha |
484 |
CcnhDebugEnds |
485 |
|
486 |
C== Update solution and residual vectors |
487 |
C Now compute "interior" points. |
488 |
err = 0. _d 0 |
489 |
DO bj=myByLo(myThid),myByHi(myThid) |
490 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
491 |
errTile(bi,bj) = 0. _d 0 |
492 |
DO k=1,Nr |
493 |
#ifdef TARGET_NEC_SX |
494 |
!CDIR OUTERUNROLL=CG3D_OUTERLOOPITERS |
495 |
#endif /* TARGET_NEC_SX */ |
496 |
DO j=1,sNy |
497 |
DO i=1,sNx |
498 |
cg3d_x(i,j,k,bi,bj)=cg3d_x(i,j,k,bi,bj) |
499 |
& +alpha*cg3d_s(i,j,k,bi,bj) |
500 |
cg3d_r(i,j,k,bi,bj)=cg3d_r(i,j,k,bi,bj) |
501 |
& -alpha*cg3d_q(i,j,k,bi,bj) |
502 |
errTile(bi,bj) = errTile(bi,bj) |
503 |
& +cg3d_r(i,j,k,bi,bj)*cg3d_r(i,j,k,bi,bj) |
504 |
ENDDO |
505 |
ENDDO |
506 |
ENDDO |
507 |
ENDDO |
508 |
ENDDO |
509 |
|
510 |
CALL GLOBAL_SUM_TILE_RL( errTile, err, myThid ) |
511 |
err = SQRT(err) |
512 |
actualIts = it3d |
513 |
actualResidual = err |
514 |
IF ( printResidual ) THEN |
515 |
IF ( MOD( it3d-1, printResidualFreq ).EQ.0 ) THEN |
516 |
WRITE(msgBuf,'(A,I6,A,1PE21.14)') |
517 |
& ' cg3d: iter=', actualIts, ' ; resid.= ', actualResidual |
518 |
CALL PRINT_MESSAGE( msgBuf, standardMessageUnit, |
519 |
& SQUEEZE_RIGHT, myThid ) |
520 |
ENDIF |
521 |
ENDIF |
522 |
IF ( actualResidual .LT. cg3dTargetResidual ) GOTO 11 |
523 |
CALL EXCH_S3D_RL( cg3d_r, Nr, myThid ) |
524 |
|
525 |
10 CONTINUE |
526 |
11 CONTINUE |
527 |
|
528 |
IF ( debugLevel.GE.debLevC .AND. diagFreq.GT.0. ) THEN |
529 |
CALL WRITE_FLD_S3D_RL( |
530 |
I 'cg3d_r_F', 'I10', 1, Nr, cg3d_r, myIter, myThid ) |
531 |
ENDIF |
532 |
|
533 |
C-- Un-normalise the answer |
534 |
DO bj=myByLo(myThid),myByHi(myThid) |
535 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
536 |
DO k=1,Nr |
537 |
DO j=1,sNy |
538 |
DO i=1,sNx |
539 |
cg3d_x(i,j,k,bi,bj) = cg3d_x(i,j,k,bi,bj)/rhsNorm |
540 |
ENDDO |
541 |
ENDDO |
542 |
ENDDO |
543 |
ENDDO |
544 |
ENDDO |
545 |
|
546 |
lastResidual = actualResidual |
547 |
numIters = actualIts |
548 |
|
549 |
#endif /* ALLOW_NONHYDROSTATIC */ |
550 |
|
551 |
RETURN |
552 |
END |