1 |
C $Header$ |
C $Header$ |
2 |
|
C $Name$ |
3 |
|
|
4 |
|
#include "PACKAGES_CONFIG.h" |
5 |
#include "CPP_OPTIONS.h" |
#include "CPP_OPTIONS.h" |
6 |
|
|
7 |
#define VERBOSE |
#define VERBOSE |
8 |
|
|
9 |
|
CBOP |
10 |
|
C !ROUTINE: CG3D |
11 |
|
C !INTERFACE: |
12 |
SUBROUTINE CG3D( |
SUBROUTINE CG3D( |
13 |
|
I cg3d_b, |
14 |
|
U cg3d_x, |
15 |
|
O firstResidual, |
16 |
|
O lastResidual, |
17 |
|
U numIters, |
18 |
I myThid ) |
I myThid ) |
19 |
C /==========================================================\ |
C !DESCRIPTION: \bv |
20 |
C | SUBROUTINE CG3D | |
C *==========================================================* |
21 |
C | o Three-dimensional grid problem conjugate-gradient | |
C | SUBROUTINE CG3D |
22 |
C | inverter (with preconditioner). | |
C | o Three-dimensional grid problem conjugate-gradient |
23 |
C |==========================================================| |
C | inverter (with preconditioner). |
24 |
C | Con. grad is an iterative procedure for solving Ax = b. | |
C *==========================================================* |
25 |
C | It requires the A be symmetric. | |
C | Con. grad is an iterative procedure for solving Ax = b. |
26 |
C | This implementation assumes A is a five-diagonal | |
C | It requires the A be symmetric. |
27 |
C | matrix of the form that arises in the discrete | |
C | This implementation assumes A is a seven-diagonal |
28 |
C | representation of the del^2 operator in a | |
C | matrix of the form that arises in the discrete |
29 |
C | two-dimensional space. | |
C | representation of the del^2 operator in a |
30 |
C | Notes: | |
C | three-dimensional space. |
31 |
C | ====== | |
C | Notes: |
32 |
C | This implementation can support shared-memory | |
C | ====== |
33 |
C | multi-threaded execution. In order to do this COMMON | |
C | This implementation can support shared-memory |
34 |
C | blocks are used for many of the arrays - even ones that | |
C | multi-threaded execution. In order to do this COMMON |
35 |
C | are only used for intermedaite results. This design is | |
C | blocks are used for many of the arrays - even ones that |
36 |
C | OK if you want to all the threads to collaborate on | |
C | are only used for intermedaite results. This design is |
37 |
C | solving the same problem. On the other hand if you want | |
C | OK if you want to all the threads to collaborate on |
38 |
C | the threads to solve several different problems | |
C | solving the same problem. On the other hand if you want |
39 |
C | concurrently this implementation will not work. | |
C | the threads to solve several different problems |
40 |
C \==========================================================/ |
C | concurrently this implementation will not work. |
41 |
IMPLICIT NONE |
C *==========================================================* |
42 |
|
C \ev |
43 |
|
|
44 |
|
C !USES: |
45 |
|
IMPLICIT NONE |
46 |
C === Global data === |
C === Global data === |
47 |
#include "SIZE.h" |
#include "SIZE.h" |
48 |
#include "EEPARAMS.h" |
#include "EEPARAMS.h" |
50 |
#include "GRID.h" |
#include "GRID.h" |
51 |
#include "CG3D.h" |
#include "CG3D.h" |
52 |
|
|
53 |
|
C !INPUT/OUTPUT PARAMETERS: |
54 |
C === Routine arguments === |
C === Routine arguments === |
55 |
C myThid - Thread on which I am working. |
C myThid - Thread on which I am working. |
56 |
|
C cg2d_b - The source term or "right hand side" |
57 |
|
C cg2d_x - The solution |
58 |
|
C firstResidual - the initial residual before any iterations |
59 |
|
C lastResidual - the actual residual reached |
60 |
|
C numIters - Entry: the maximum number of iterations allowed |
61 |
|
C Exit: the actual number of iterations used |
62 |
|
_RL cg3d_b(1-OLx:sNx+OLx,1-OLy:sNy+OLy,Nr,nSx,nSy) |
63 |
|
_RL cg3d_x(1-OLx:sNx+OLx,1-OLy:sNy+OLy,Nr,nSx,nSy) |
64 |
|
_RL firstResidual |
65 |
|
_RL lastResidual |
66 |
|
INTEGER numIters |
67 |
INTEGER myThid |
INTEGER myThid |
68 |
|
|
69 |
|
|
70 |
|
#ifdef ALLOW_NONHYDROSTATIC |
71 |
|
|
72 |
|
C !LOCAL VARIABLES: |
73 |
C === Local variables ==== |
C === Local variables ==== |
74 |
C actualIts - Number of iterations taken |
C actualIts - Number of iterations taken |
75 |
C actualResidual - residual |
C actualResidual - residual |
76 |
C bi - Block index in X and Y. |
C bi - Block index in X and Y. |
77 |
C bj |
C bj |
78 |
C etaN - Used in computing search directions |
C eta_qrN - Used in computing search directions |
79 |
C etaNM1 suffix N and NM1 denote current and |
C eta_qrNM1 suffix N and NM1 denote current and |
80 |
C cgBeta previous iterations respectively. |
C cgBeta previous iterations respectively. |
81 |
C alpha |
C alpha |
82 |
C sumRHS - Sum of right-hand-side. Sometimes this is a |
C sumRHS - Sum of right-hand-side. Sometimes this is a |
90 |
INTEGER bi, bj |
INTEGER bi, bj |
91 |
INTEGER I, J, K, it3d |
INTEGER I, J, K, it3d |
92 |
INTEGER KM1, KP1 |
INTEGER KM1, KP1 |
93 |
_RL err |
_RL err, errTile |
94 |
_RL etaN |
_RL eta_qrN, eta_qrNtile |
95 |
_RL etaNM1 |
_RL eta_qrNM1 |
96 |
_RL cgBeta |
_RL cgBeta |
97 |
_RL alpha |
_RL alpha , alphaTile |
98 |
_RL sumRHS |
_RL sumRHS, sumRHStile |
99 |
_RL rhsMax |
_RL rhsMax |
100 |
_RL rhsNorm |
_RL rhsNorm |
101 |
|
|
106 |
INTEGER exchWidthX |
INTEGER exchWidthX |
107 |
INTEGER exchWidthY |
INTEGER exchWidthY |
108 |
INTEGER myNz |
INTEGER myNz |
109 |
_RL topLevFac |
_RL topLevTerm |
110 |
|
CEOP |
111 |
|
|
112 |
CcnhDebugStarts |
ceh3 needs an IF ( useNONHYDROSTATIC ) THEN |
|
CHARACTER*(MAX_LEN_FNAM) suff |
|
|
CcnhDebugEnds |
|
113 |
|
|
|
#ifdef ALLOW_NONHYDROSTATIC |
|
114 |
|
|
115 |
C-- Initialise inverter |
C-- Initialise inverter |
116 |
etaNM1 = 1. D0 |
eta_qrNM1 = 1. D0 |
117 |
|
|
118 |
C-- Normalise RHS |
C-- Normalise RHS |
119 |
rhsMax = 0. _d 0 |
rhsMax = 0. _d 0 |
149 |
_EXCH_XYZ_R8( cg3d_b, myThid ) |
_EXCH_XYZ_R8( cg3d_b, myThid ) |
150 |
_EXCH_XYZ_R8( cg3d_x, myThid ) |
_EXCH_XYZ_R8( cg3d_x, myThid ) |
151 |
|
|
152 |
#ifdef NONO |
C-- Initial residual calculation (with free-Surface term) |
|
CcnhDebugStarts |
|
|
C-- Initial residual calculation |
|
|
err = 0. _d 0 |
|
|
sumRHS = 0. _d 0 |
|
|
DO bj=myByLo(myThid),myByHi(myThid) |
|
|
DO bi=myBxLo(myThid),myBxHi(myThid) |
|
|
DO J=1,sNy |
|
|
DO I=1,sNx |
|
|
alpha = 0. _d 0 |
|
|
DO K=1,Nr |
|
|
KM1 = K-1 |
|
|
IF ( KM1 .EQ. 0 ) KM1 = 1 |
|
|
KP1 = K+1 |
|
|
IF ( KP1 .EQ. Nr+1 ) KP1 = 1 |
|
|
cg3d_r(I,J,K,bi,bj) = cg3d_b(I,J,K,bi,bj) -( 0. |
|
|
& +aW3d(I ,J ,K ,bi,bj)*cg3d_x(I-1,J ,K ,bi,bj) |
|
|
& +aW3d(I+1,J ,K ,bi,bj)*cg3d_x(I+1,J ,K ,bi,bj) |
|
|
& +aS3d(I ,J ,K ,bi,bj)*cg3d_x(I ,J-1,K ,bi,bj) |
|
|
& +aS3d(I ,J+1,K ,bi,bj)*cg3d_x(I ,J+1,K ,bi,bj) |
|
|
& +aV3d(I ,J ,K ,bi,bj)*cg3d_x(I ,J ,KM1,bi,bj) |
|
|
& +aV3d(I ,J ,KP1,bi,bj)*cg3d_x(I ,J ,KP1,bi,bj) |
|
|
& -aW3d(I ,J ,K ,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
|
|
& -aW3d(I+1,J ,K ,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
|
|
& -aS3d(I ,J ,K ,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
|
|
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
|
|
& -aV3d(I ,J ,K ,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
|
|
& -aV3d(I ,J ,KP1,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
|
|
& ) |
|
|
alpha = alpha |
|
|
& +cg3d_r(I,J,K,bi,bj) |
|
|
sumRHS = sumRHS |
|
|
& +cg3d_b(I,J,K,bi,bj) |
|
|
ENDDO |
|
|
err = err + alpha*alpha |
|
|
ENDDO |
|
|
ENDDO |
|
|
ENDDO |
|
|
ENDDO |
|
|
WRITE(6,*) 'DEBUG mythid, err = ', mythid, SQRT(err) |
|
|
_GLOBAL_SUM_R8( err , myThid ) |
|
|
_GLOBAL_SUM_R8( sumRHS , myThid ) |
|
|
_BEGIN_MASTER( myThid ) |
|
|
write(0,*) 'DEBUG cg3d: Sum(rhs) = ',sumRHS |
|
|
_END_MASTER( ) |
|
|
actualIts = 0 |
|
|
actualResidual = SQRT(err) |
|
|
C _BARRIER |
|
|
_BEGIN_MASTER( myThid ) |
|
|
WRITE(0,'(A,I6,1PE30.14)') 'DEBUG CG3D iters, err = ', |
|
|
& actualIts, actualResidual |
|
|
_END_MASTER( ) |
|
|
CcnhDebugEnds |
|
|
#endif |
|
|
|
|
|
C-- Initial residual calculation |
|
153 |
err = 0. _d 0 |
err = 0. _d 0 |
154 |
sumRHS = 0. _d 0 |
sumRHS = 0. _d 0 |
155 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
156 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
157 |
|
errTile = 0. _d 0 |
158 |
|
sumRHStile = 0. _d 0 |
159 |
DO K=1,Nr |
DO K=1,Nr |
160 |
KM1 = K-1 |
KM1 = K-1 |
161 |
IF ( K .EQ. 1 ) KM1 = 1 |
IF ( K .EQ. 1 ) KM1 = 1 |
162 |
KP1 = K+1 |
KP1 = K+1 |
163 |
IF ( K .EQ. Nr ) KP1 = 1 |
IF ( K .EQ. Nr ) KP1 = 1 |
164 |
topLevFac = 0. |
topLevTerm = 0. |
165 |
IF ( K .EQ. 1) topLevFac = 1. |
IF ( K .EQ. 1) topLevTerm = freeSurfFac*cg3dNorm* |
166 |
|
& (horiVertRatio/gravity)/deltaTMom/deltaTMom |
167 |
DO J=1,sNy |
DO J=1,sNy |
168 |
DO I=1,sNx |
DO I=1,sNx |
169 |
cg3d_s(I,J,K,bi,bj) = 0. |
cg3d_s(I,J,K,bi,bj) = 0. |
180 |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
181 |
& -aV3d(I ,J ,K ,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
& -aV3d(I ,J ,K ,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
182 |
& -aV3d(I ,J ,KP1,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
& -aV3d(I ,J ,KP1,bi,bj)*cg3d_x(I ,J ,K ,bi,bj) |
183 |
& -freeSurfFac*_rA(i,j,bi,bj)* horiVertRatio* |
& -topLevTerm*_rA(I,J,bi,bj)*cg3d_x(I,J,K,bi,bj) |
|
& cg3d_x(I ,J ,K,bi,bj)/deltaTMom/deltaTMom*cg3dNorm |
|
|
& *topLevFac |
|
184 |
& ) |
& ) |
185 |
err = err |
errTile = errTile |
186 |
& +cg3d_r(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
& +cg3d_r(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
187 |
sumRHS = sumRHS |
sumRHStile = sumRHStile |
188 |
& +cg3d_b(I,J,K,bi,bj) |
& +cg3d_b(I,J,K,bi,bj) |
189 |
ENDDO |
ENDDO |
190 |
ENDDO |
ENDDO |
191 |
ENDDO |
ENDDO |
192 |
|
err = err + errTile |
193 |
|
sumRHS = sumRHS + sumRHStile |
194 |
ENDDO |
ENDDO |
195 |
ENDDO |
ENDDO |
196 |
C _EXCH_XYZ_R8( cg3d_r, myThid ) |
C _EXCH_XYZ_R8( cg3d_r, myThid ) |
220 |
_GLOBAL_SUM_R8( sumRHS, myThid ) |
_GLOBAL_SUM_R8( sumRHS, myThid ) |
221 |
_GLOBAL_SUM_R8( err , myThid ) |
_GLOBAL_SUM_R8( err , myThid ) |
222 |
|
|
223 |
_BEGIN_MASTER( myThid ) |
IF ( debugLevel .GE. debLevZero ) THEN |
224 |
write(0,'(A,1PE30.14)') ' cg3d: Sum(rhs) = ',sumRHS |
_BEGIN_MASTER( myThid ) |
225 |
_END_MASTER( ) |
write(*,'(A,1P2E22.14)') |
226 |
|
& ' cg3d: Sum(rhs),rhsMax = ',sumRHS,rhsMax |
227 |
|
_END_MASTER( myThid ) |
228 |
|
ENDIF |
229 |
|
|
230 |
actualIts = 0 |
actualIts = 0 |
231 |
actualResidual = SQRT(err) |
actualResidual = SQRT(err) |
232 |
C _BARRIER |
C _BARRIER |
233 |
_BEGIN_MASTER( myThid ) |
c _BEGIN_MASTER( myThid ) |
234 |
WRITE(0,'(A,I6,1PE30.14)') ' CG3D iters, err = ', |
c WRITE(*,'(A,I6,1PE30.14)') ' CG3D iters, err = ', |
235 |
& actualIts, actualResidual |
c & actualIts, actualResidual |
236 |
_END_MASTER( ) |
c _END_MASTER( myThid ) |
237 |
|
firstResidual=actualResidual |
238 |
|
|
239 |
C >>>>>>>>>>>>>>> BEGIN SOLVER <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< |
C >>>>>>>>>>>>>>> BEGIN SOLVER <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< |
240 |
DO 10 it3d=1, cg3dMaxIters |
DO 10 it3d=1, cg3dMaxIters |
241 |
|
|
242 |
CcnhDebugStarts |
CcnhDebugStarts |
243 |
#ifdef VERBOSE |
#ifdef VERBOSE |
244 |
IF ( mod(it3d-1,10).EQ.0) |
c IF ( mod(it3d-1,10).EQ.0) |
245 |
& WRITE(0,*) ' CG3D: Iteration ',it3d-1, |
c & WRITE(*,*) ' CG3D: Iteration ',it3d-1, |
246 |
& ' residual = ',actualResidual |
c & ' residual = ',actualResidual |
247 |
#endif |
#endif |
248 |
CcnhDebugEnds |
CcnhDebugEnds |
249 |
IF ( actualResidual .LT. cg3dTargetResidual ) GOTO 11 |
IF ( actualResidual .LT. cg3dTargetResidual ) GOTO 11 |
252 |
C Note. On the next to loops over all tiles the inner loop ranges |
C Note. On the next to loops over all tiles the inner loop ranges |
253 |
C in sNx and sNy are expanded by 1 to avoid a communication |
C in sNx and sNy are expanded by 1 to avoid a communication |
254 |
C step. However this entails a bit of gynamastics because we only |
C step. However this entails a bit of gynamastics because we only |
255 |
C want etaN for the interior points. |
C want eta_qrN for the interior points. |
256 |
etaN = 0. _d 0 |
eta_qrN = 0. _d 0 |
257 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
258 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
259 |
|
eta_qrNtile = 0. _d 0 |
260 |
DO K=1,1 |
DO K=1,1 |
261 |
DO J=1-1,sNy+1 |
DO J=1-1,sNy+1 |
262 |
DO I=1-1,sNx+1 |
DO I=1-1,sNx+1 |
286 |
caja ENDIF |
caja ENDIF |
287 |
DO J=1,sNy |
DO J=1,sNy |
288 |
DO I=1,sNx |
DO I=1,sNx |
289 |
etaN = etaN |
eta_qrNtile = eta_qrNtile |
290 |
& +cg3d_q(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
& +cg3d_q(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
291 |
ENDDO |
ENDDO |
292 |
ENDDO |
ENDDO |
301 |
ENDDO |
ENDDO |
302 |
DO J=1,sNy |
DO J=1,sNy |
303 |
DO I=1,sNx |
DO I=1,sNx |
304 |
etaN = etaN |
eta_qrNtile = eta_qrNtile |
305 |
& +cg3d_q(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
& +cg3d_q(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
306 |
ENDDO |
ENDDO |
307 |
ENDDO |
ENDDO |
308 |
ENDDO |
ENDDO |
309 |
|
eta_qrN = eta_qrN + eta_qrNtile |
310 |
ENDDO |
ENDDO |
311 |
ENDDO |
ENDDO |
312 |
caja |
caja |
313 |
caja etaN=0. |
caja eta_qrN=0. |
314 |
caja DO bj=myByLo(myThid),myByHi(myThid) |
caja DO bj=myByLo(myThid),myByHi(myThid) |
315 |
caja DO bi=myBxLo(myThid),myBxHi(myThid) |
caja DO bi=myBxLo(myThid),myBxHi(myThid) |
316 |
caja DO K=1,Nr |
caja DO K=1,Nr |
317 |
caja DO J=1,sNy |
caja DO J=1,sNy |
318 |
caja DO I=1,sNx |
caja DO I=1,sNx |
319 |
caja etaN = etaN |
caja eta_qrN = eta_qrN |
320 |
caja & +cg3d_q(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
caja & +cg3d_q(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
321 |
caja ENDDO |
caja ENDDO |
322 |
caja ENDDO |
caja ENDDO |
325 |
caja ENDDO |
caja ENDDO |
326 |
caja |
caja |
327 |
|
|
328 |
_GLOBAL_SUM_R8(etaN, myThid) |
_GLOBAL_SUM_R8(eta_qrN, myThid) |
329 |
CcnhDebugStarts |
CcnhDebugStarts |
330 |
C WRITE(0,*) ' CG3D: Iteration ',it3d-1,' etaN = ',etaN |
C WRITE(*,*) ' CG3D: Iteration ',it3d-1,' eta_qrN = ',eta_qrN |
331 |
CcnhDebugEnds |
CcnhDebugEnds |
332 |
cgBeta = etaN/etaNM1 |
cgBeta = eta_qrN/eta_qrNM1 |
333 |
CcnhDebugStarts |
CcnhDebugStarts |
334 |
C WRITE(0,*) ' CG3D: Iteration ',it3d-1,' beta = ',cgBeta |
C WRITE(*,*) ' CG3D: Iteration ',it3d-1,' beta = ',cgBeta |
335 |
CcnhDebugEnds |
CcnhDebugEnds |
336 |
etaNM1 = etaN |
eta_qrNM1 = eta_qrN |
337 |
|
|
338 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
339 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
351 |
C== Evaluate laplace operator on conjugate gradient vector |
C== Evaluate laplace operator on conjugate gradient vector |
352 |
C== q = A.s |
C== q = A.s |
353 |
alpha = 0. _d 0 |
alpha = 0. _d 0 |
354 |
|
topLevTerm = freeSurfFac*cg3dNorm* |
355 |
|
& (horiVertRatio/gravity)/deltaTMom/deltaTMom |
356 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
357 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
358 |
|
alphaTile = 0. _d 0 |
359 |
IF ( Nr .GT. 1 ) THEN |
IF ( Nr .GT. 1 ) THEN |
360 |
DO K=1,1 |
DO K=1,1 |
361 |
DO J=1,sNy |
DO J=1,sNy |
371 |
& -aS3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aS3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
372 |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
373 |
& -aV3d(I ,J ,K+1,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aV3d(I ,J ,K+1,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
374 |
& -freeSurfFac*_rA(i,j,bi,bj)* horiVertRatio* |
& -topLevTerm*_rA(I,J,bi,bj)*cg3d_s(I,J,K,bi,bj) |
375 |
& cg3d_s(I ,J ,K,bi,bj)/deltaTMom/deltaTMom*cg3dNorm |
alphaTile = alphaTile |
376 |
alpha = alpha+cg3d_s(I,J,K,bi,bj)*cg3d_q(I,J,K,bi,bj) |
& +cg3d_s(I,J,K,bi,bj)*cg3d_q(I,J,K,bi,bj) |
377 |
ENDDO |
ENDDO |
378 |
ENDDO |
ENDDO |
379 |
ENDDO |
ENDDO |
390 |
& -aW3d(I+1,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aW3d(I+1,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
391 |
& -aS3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aS3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
392 |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
393 |
& -freeSurfFac*_rA(i,j,bi,bj)* horiVertRatio* |
& -topLevTerm*_rA(I,J,bi,bj)*cg3d_s(I,J,K,bi,bj) |
394 |
& cg3d_s(I ,J ,K,bi,bj)/deltaTMom/deltaTMom*cg3dNorm |
alphaTile = alphaTile |
395 |
alpha = alpha+cg3d_s(I,J,K,bi,bj)*cg3d_q(I,J,K,bi,bj) |
& +cg3d_s(I,J,K,bi,bj)*cg3d_q(I,J,K,bi,bj) |
396 |
ENDDO |
ENDDO |
397 |
ENDDO |
ENDDO |
398 |
ENDDO |
ENDDO |
413 |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
414 |
& -aV3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aV3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
415 |
& -aV3d(I ,J ,K+1,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aV3d(I ,J ,K+1,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
416 |
alpha = alpha+cg3d_s(I,J,K,bi,bj)*cg3d_q(I,J,K,bi,bj) |
alphaTile = alphaTile |
417 |
|
& +cg3d_s(I,J,K,bi,bj)*cg3d_q(I,J,K,bi,bj) |
418 |
ENDDO |
ENDDO |
419 |
ENDDO |
ENDDO |
420 |
ENDDO |
ENDDO |
433 |
& -aS3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aS3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
434 |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aS3d(I ,J+1,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
435 |
& -aV3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
& -aV3d(I ,J ,K ,bi,bj)*cg3d_s(I ,J ,K ,bi,bj) |
436 |
alpha = alpha+cg3d_s(I,J,K,bi,bj)*cg3d_q(I,J,K,bi,bj) |
alphaTile = alphaTile |
437 |
|
& +cg3d_s(I,J,K,bi,bj)*cg3d_q(I,J,K,bi,bj) |
438 |
ENDDO |
ENDDO |
439 |
ENDDO |
ENDDO |
440 |
ENDDO |
ENDDO |
441 |
ENDIF |
ENDIF |
442 |
|
alpha = alpha + alphaTile |
443 |
ENDDO |
ENDDO |
444 |
ENDDO |
ENDDO |
445 |
_GLOBAL_SUM_R8(alpha,myThid) |
_GLOBAL_SUM_R8(alpha,myThid) |
446 |
CcnhDebugStarts |
CcnhDebugStarts |
447 |
C WRITE(0,*) ' CG3D: Iteration ',it3d-1,' SUM(s*q)= ',alpha |
C WRITE(*,*) ' CG3D: Iteration ',it3d-1,' SUM(s*q)= ',alpha |
448 |
CcnhDebugEnds |
CcnhDebugEnds |
449 |
alpha = etaN/alpha |
alpha = eta_qrN/alpha |
450 |
CcnhDebugStarts |
CcnhDebugStarts |
451 |
C WRITE(0,*) ' CG3D: Iteration ',it3d-1,' alpha= ',alpha |
C WRITE(*,*) ' CG3D: Iteration ',it3d-1,' alpha= ',alpha |
452 |
CcnhDebugEnds |
CcnhDebugEnds |
453 |
|
|
454 |
C== Update solution and residual vectors |
C== Update solution and residual vectors |
456 |
err = 0. _d 0 |
err = 0. _d 0 |
457 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
458 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
459 |
|
errTile = 0. _d 0 |
460 |
DO K=1,Nr |
DO K=1,Nr |
461 |
DO J=1,sNy |
DO J=1,sNy |
462 |
DO I=1,sNx |
DO I=1,sNx |
464 |
& +alpha*cg3d_s(I,J,K,bi,bj) |
& +alpha*cg3d_s(I,J,K,bi,bj) |
465 |
cg3d_r(I,J,K,bi,bj)=cg3d_r(I,J,K,bi,bj) |
cg3d_r(I,J,K,bi,bj)=cg3d_r(I,J,K,bi,bj) |
466 |
& -alpha*cg3d_q(I,J,K,bi,bj) |
& -alpha*cg3d_q(I,J,K,bi,bj) |
467 |
err = err+cg3d_r(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
errTile = errTile |
468 |
|
& +cg3d_r(I,J,K,bi,bj)*cg3d_r(I,J,K,bi,bj) |
469 |
ENDDO |
ENDDO |
470 |
ENDDO |
ENDDO |
471 |
ENDDO |
ENDDO |
472 |
|
err = err + errTile |
473 |
ENDDO |
ENDDO |
474 |
ENDDO |
ENDDO |
475 |
|
|
507 |
ENDDO |
ENDDO |
508 |
ENDDO |
ENDDO |
509 |
|
|
510 |
_EXCH_XYZ_R8(cg3d_x, myThid ) |
Cadj _EXCH_XYZ_R8(cg3d_x, myThid ) |
511 |
_BEGIN_MASTER( myThid ) |
c _BEGIN_MASTER( myThid ) |
512 |
WRITE(0,'(A,I6,1PE30.14)') ' CG3D iters, err = ', |
c WRITE(*,'(A,I6,1PE30.14)') ' CG3D iters, err = ', |
513 |
& actualIts, actualResidual |
c & actualIts, actualResidual |
514 |
_END_MASTER( ) |
c _END_MASTER( myThid ) |
515 |
|
lastResidual=actualResidual |
516 |
CcnhDebugStarts |
numIters=actualIts |
|
C CALL PLOT_FIELD_XYRL( cg2d_x, 'CALC_MOM_RHS CG2D_X' , 1, myThid ) |
|
|
C err = 0. _d 0 |
|
|
C DO bj=myByLo(myThid),myByHi(myThid) |
|
|
C DO bi=myBxLo(myThid),myBxHi(myThid) |
|
|
C DO J=1,sNy |
|
|
C DO I=1,sNx |
|
|
C cg2d_r(I,J,bi,bj) = cg2d_b(I,J,bi,bj) - |
|
|
C & (aW2d(I ,J ,bi,bj)*cg2d_x(I-1,J ,bi,bj) |
|
|
C & +aW2d(I+1,J ,bi,bj)*cg2d_x(I+1,J ,bi,bj) |
|
|
C & +aS2d(I ,J ,bi,bj)*cg2d_x(I ,J-1,bi,bj) |
|
|
C & +aS2d(I ,J+1,bi,bj)*cg2d_x(I ,J+1,bi,bj) |
|
|
C & -aW2d(I ,J ,bi,bj)*cg2d_x(I ,J ,bi,bj) |
|
|
C & -aW2d(I+1,J ,bi,bj)*cg2d_x(I ,J ,bi,bj) |
|
|
C & -aS2d(I ,J ,bi,bj)*cg2d_x(I ,J ,bi,bj) |
|
|
C & -aS2d(I ,J+1,bi,bj)*cg2d_x(I ,J ,bi,bj)) |
|
|
C err = err + |
|
|
C & cg2d_r(I,J,bi,bj)*cg2d_r(I,J,bi,bj) |
|
|
C ENDDO |
|
|
C ENDDO |
|
|
C ENDDO |
|
|
C ENDDO |
|
|
C _GLOBAL_SUM_R8( err , myThid ) |
|
|
C write(0,*) 'cg2d: Ax - b = ',SQRT(err) |
|
|
CcnhDebugEnds |
|
517 |
|
|
518 |
#endif /* ALLOW_NONHYDROSTATIC */ |
#endif /* ALLOW_NONHYDROSTATIC */ |
519 |
|
|