2 |
C $Name$ |
C $Name$ |
3 |
|
|
4 |
#include "CPP_OPTIONS.h" |
#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 CG2D_OUTERLOOPITERS |
9 |
|
# define CG2D_OUTERLOOPITERS 10 |
10 |
|
#endif |
11 |
|
#endif /* TARGET_NEC_SX */ |
12 |
|
|
13 |
CBOP |
CBOP |
14 |
C !ROUTINE: CG2D |
C !ROUTINE: CG2D |
15 |
C !INTERFACE: |
C !INTERFACE: |
16 |
SUBROUTINE CG2D( |
SUBROUTINE CG2D( |
17 |
I cg2d_b, |
U cg2d_b, cg2d_x, |
18 |
U cg2d_x, |
O firstResidual, minResidualSq, lastResidual, |
19 |
O firstResidual, |
U numIters, nIterMin, |
|
O lastResidual, |
|
|
U numIters, |
|
20 |
I myThid ) |
I myThid ) |
21 |
C !DESCRIPTION: \bv |
C !DESCRIPTION: \bv |
22 |
C *==========================================================* |
C *==========================================================* |
23 |
C | SUBROUTINE CG2D |
C | SUBROUTINE CG2D |
24 |
C | o Two-dimensional grid problem conjugate-gradient |
C | o Two-dimensional grid problem conjugate-gradient |
25 |
C | inverter (with preconditioner). |
C | inverter (with preconditioner). |
26 |
C *==========================================================* |
C *==========================================================* |
27 |
C | Con. grad is an iterative procedure for solving Ax = b. |
C | Con. grad is an iterative procedure for solving Ax = b. |
28 |
C | It requires the A be symmetric. |
C | It requires the A be symmetric. |
29 |
C | This implementation assumes A is a five-diagonal |
C | This implementation assumes A is a five-diagonal |
30 |
C | matrix of the form that arises in the discrete |
C | matrix of the form that arises in the discrete |
31 |
C | representation of the del^2 operator in a |
C | representation of the del^2 operator in a |
32 |
C | two-dimensional space. |
C | two-dimensional space. |
33 |
C | Notes: |
C | Notes: |
34 |
C | ====== |
C | ====== |
35 |
C | This implementation can support shared-memory |
C | This implementation can support shared-memory |
36 |
C | multi-threaded execution. In order to do this COMMON |
C | multi-threaded execution. In order to do this COMMON |
37 |
C | blocks are used for many of the arrays - even ones that |
C | blocks are used for many of the arrays - even ones that |
38 |
C | are only used for intermedaite results. This design is |
C | are only used for intermedaite results. This design is |
39 |
C | OK if you want to all the threads to collaborate on |
C | OK if you want to all the threads to collaborate on |
40 |
C | solving the same problem. On the other hand if you want |
C | solving the same problem. On the other hand if you want |
41 |
C | the threads to solve several different problems |
C | the threads to solve several different problems |
42 |
C | concurrently this implementation will not work. |
C | concurrently this implementation will not work. |
43 |
C *==========================================================* |
C *==========================================================* |
44 |
C \ev |
C \ev |
45 |
|
|
49 |
#include "SIZE.h" |
#include "SIZE.h" |
50 |
#include "EEPARAMS.h" |
#include "EEPARAMS.h" |
51 |
#include "PARAMS.h" |
#include "PARAMS.h" |
|
#include "GRID.h" |
|
52 |
#include "CG2D.h" |
#include "CG2D.h" |
|
#include "SURFACE.h" |
|
53 |
|
|
54 |
C !INPUT/OUTPUT PARAMETERS: |
C !INPUT/OUTPUT PARAMETERS: |
55 |
C === Routine arguments === |
C === Routine arguments === |
56 |
C myThid - Thread on which I am working. |
C cg2d_b :: The source term or "right hand side" (output: normalised RHS) |
57 |
C cg2d_b - The source term or "right hand side" |
C cg2d_x :: The solution (input: first guess) |
58 |
C cg2d_x - The solution |
C firstResidual :: the initial residual before any iterations |
59 |
C firstResidual - the initial residual before any iterations |
C minResidualSq :: the lowest residual reached (squared) |
60 |
C lastResidual - the actual residual reached |
C lastResidual :: the actual residual reached |
61 |
C numIters - Entry: the maximum number of iterations allowed |
C numIters :: Inp: the maximum number of iterations allowed |
62 |
C Exit: the actual number of iterations used |
C Out: the actual number of iterations used |
63 |
|
C nIterMin :: Inp: decide to store (if >=0) or not (if <0) lowest res. sol. |
64 |
|
C Out: iteration number corresponding to lowest residual |
65 |
|
C myThid :: Thread on which I am working. |
66 |
_RL cg2d_b(1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
_RL cg2d_b(1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
67 |
_RL cg2d_x(1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
_RL cg2d_x(1-OLx:sNx+OLx,1-OLy:sNy+OLy,nSx,nSy) |
68 |
_RL firstResidual |
_RL firstResidual |
69 |
|
_RL minResidualSq |
70 |
_RL lastResidual |
_RL lastResidual |
71 |
INTEGER numIters |
INTEGER numIters |
72 |
|
INTEGER nIterMin |
73 |
INTEGER myThid |
INTEGER myThid |
74 |
|
|
75 |
C !LOCAL VARIABLES: |
C !LOCAL VARIABLES: |
76 |
C === Local variables ==== |
C === Local variables ==== |
77 |
C actualIts - Number of iterations taken |
C bi, bj :: tile index in X and Y. |
78 |
C actualResidual - residual |
C i, j, it2d :: Loop counters ( it2d counts CG iterations ) |
79 |
C bi - Block index in X and Y. |
C actualIts :: actual CG iteration number |
80 |
C bj |
C err_sq :: Measure of the square of the residual of Ax - b. |
81 |
C eta_qrN - Used in computing search directions |
C eta_qrN :: Used in computing search directions; suffix N and NM1 |
82 |
C eta_qrNM1 suffix N and NM1 denote current and |
C eta_qrNM1 denote current and previous iterations respectively. |
83 |
C cgBeta previous iterations respectively. |
C cgBeta :: coeff used to update conjugate direction vector "s". |
84 |
C alpha |
C alpha :: coeff used to update solution & residual |
85 |
C sumRHS - Sum of right-hand-side. Sometimes this is a |
C sumRHS :: Sum of right-hand-side. Sometimes this is a useful |
86 |
C useful debuggin/trouble shooting diagnostic. |
C debugging/trouble shooting diagnostic. For neumann problems |
87 |
C For neumann problems sumRHS needs to be ~0. |
C sumRHS needs to be ~0 or it converge at a non-zero residual. |
88 |
C or they converge at a non-zero residual. |
C cg2d_min :: used to store solution corresponding to lowest residual. |
89 |
C err - Measure of residual of Ax - b, usually the norm. |
C msgBuf :: Informational/error message buffer |
90 |
C I, J, N - Loop counters ( N counts CG iterations ) |
INTEGER bi, bj |
91 |
|
INTEGER i, j, it2d |
92 |
INTEGER actualIts |
INTEGER actualIts |
93 |
_RL actualResidual |
_RL cg2dTolerance_sq |
94 |
INTEGER bi, bj |
_RL err_sq, errTile(nSx,nSy) |
95 |
INTEGER I, J, it2d |
_RL eta_qrN, eta_qrNtile(nSx,nSy) |
|
_RL err |
|
|
_RL eta_qrN |
|
96 |
_RL eta_qrNM1 |
_RL eta_qrNM1 |
97 |
_RL cgBeta |
_RL cgBeta |
98 |
_RL alpha |
_RL alpha, alphaTile(nSx,nSy) |
99 |
_RL sumRHS |
_RL sumRHS, sumRHStile(nSx,nSy) |
100 |
_RL rhsMax |
_RL rhsMax |
101 |
_RL rhsNorm |
_RL rhsNorm |
102 |
|
_RL cg2d_min(1:sNx,1:sNy,nSx,nSy) |
103 |
INTEGER OLw |
#ifdef CG2D_SINGLECPU_SUM |
104 |
INTEGER OLe |
_RL localBuf(1:sNx,1:sNy,nSx,nSy) |
105 |
INTEGER OLn |
#endif |
106 |
INTEGER OLs |
CHARACTER*(MAX_LEN_MBUF) msgBuf |
107 |
INTEGER exchWidthX |
LOGICAL printResidual |
|
INTEGER exchWidthY |
|
|
INTEGER myNz |
|
108 |
CEOP |
CEOP |
109 |
|
|
110 |
|
C-- Initialise auxiliary constant, some output variable and inverter |
111 |
CcnhDebugStarts |
cg2dTolerance_sq = cg2dTolerance*cg2dTolerance |
112 |
C CHARACTER*(MAX_LEN_FNAM) suff |
minResidualSq = -1. _d 0 |
113 |
CcnhDebugEnds |
eta_qrNM1 = 1. _d 0 |
|
|
|
|
|
|
|
C-- Initialise inverter |
|
|
eta_qrNM1 = 1. _d 0 |
|
|
|
|
|
CcnhDebugStarts |
|
|
C _EXCH_XY_R8( cg2d_b, myThid ) |
|
|
C CALL PLOT_FIELD_XYRL( cg2d_b, 'CG2D.0 CG2D_B' , 1, myThid ) |
|
|
C suff = 'unnormalised' |
|
|
C CALL WRITE_FLD_XY_RL ( 'cg2d_b.',suff, cg2d_b, 1, myThid) |
|
|
C STOP |
|
|
CcnhDebugEnds |
|
114 |
|
|
115 |
C-- Normalise RHS |
C-- Normalise RHS |
116 |
rhsMax = 0. _d 0 |
rhsMax = 0. _d 0 |
117 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
118 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
119 |
DO J=1,sNy |
DO j=1,sNy |
120 |
DO I=1,sNx |
DO i=1,sNx |
121 |
cg2d_b(I,J,bi,bj) = cg2d_b(I,J,bi,bj)*cg2dNorm |
cg2d_b(i,j,bi,bj) = cg2d_b(i,j,bi,bj)*cg2dNorm |
122 |
rhsMax = MAX(ABS(cg2d_b(I,J,bi,bj)),rhsMax) |
rhsMax = MAX(ABS(cg2d_b(i,j,bi,bj)),rhsMax) |
123 |
ENDDO |
ENDDO |
124 |
ENDDO |
ENDDO |
125 |
ENDDO |
ENDDO |
127 |
|
|
128 |
IF (cg2dNormaliseRHS) THEN |
IF (cg2dNormaliseRHS) THEN |
129 |
C- Normalise RHS : |
C- Normalise RHS : |
130 |
#ifdef LETS_MAKE_JAM |
_GLOBAL_MAX_RL( rhsMax, myThid ) |
|
C _GLOBAL_MAX_R8( rhsMax, myThid ) |
|
|
rhsMax=1. |
|
|
#else |
|
|
_GLOBAL_MAX_R8( rhsMax, myThid ) |
|
|
Catm rhsMax=1. |
|
|
#endif |
|
131 |
rhsNorm = 1. _d 0 |
rhsNorm = 1. _d 0 |
132 |
IF ( rhsMax .NE. 0. ) rhsNorm = 1. _d 0 / rhsMax |
IF ( rhsMax .NE. 0. ) rhsNorm = 1. _d 0 / rhsMax |
133 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
134 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
135 |
DO J=1,sNy |
DO j=1,sNy |
136 |
DO I=1,sNx |
DO i=1,sNx |
137 |
cg2d_b(I,J,bi,bj) = cg2d_b(I,J,bi,bj)*rhsNorm |
cg2d_b(i,j,bi,bj) = cg2d_b(i,j,bi,bj)*rhsNorm |
138 |
cg2d_x(I,J,bi,bj) = cg2d_x(I,J,bi,bj)*rhsNorm |
cg2d_x(i,j,bi,bj) = cg2d_x(i,j,bi,bj)*rhsNorm |
139 |
ENDDO |
ENDDO |
140 |
ENDDO |
ENDDO |
141 |
ENDDO |
ENDDO |
144 |
ENDIF |
ENDIF |
145 |
|
|
146 |
C-- Update overlaps |
C-- Update overlaps |
147 |
_EXCH_XY_R8( cg2d_b, myThid ) |
CALL EXCH_XY_RL( cg2d_x, myThid ) |
|
_EXCH_XY_R8( cg2d_x, myThid ) |
|
|
CcnhDebugStarts |
|
|
C CALL PLOT_FIELD_XYRL( cg2d_b, 'CG2D.1 CG2D_B' , 1, myThid ) |
|
|
C suff = 'normalised' |
|
|
C CALL WRITE_FLD_XY_RL ( 'cg2d_b.',suff, cg2d_b, 1, myThid) |
|
|
CcnhDebugEnds |
|
148 |
|
|
149 |
C-- Initial residual calculation |
C-- Initial residual calculation |
|
err = 0. _d 0 |
|
|
sumRHS = 0. _d 0 |
|
150 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
151 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
152 |
DO J=1,sNy |
IF ( nIterMin.GE.0 ) THEN |
153 |
DO I=1,sNx |
DO j=1,sNy |
154 |
cg2d_s(I,J,bi,bj) = 0. |
DO i=1,sNx |
155 |
cg2d_r(I,J,bi,bj) = cg2d_b(I,J,bi,bj) - |
cg2d_min(i,j,bi,bj) = cg2d_x(i,j,bi,bj) |
156 |
& (aW2d(I ,J ,bi,bj)*cg2d_x(I-1,J ,bi,bj) |
ENDDO |
157 |
& +aW2d(I+1,J ,bi,bj)*cg2d_x(I+1,J ,bi,bj) |
ENDDO |
158 |
& +aS2d(I ,J ,bi,bj)*cg2d_x(I ,J-1,bi,bj) |
ENDIF |
159 |
& +aS2d(I ,J+1,bi,bj)*cg2d_x(I ,J+1,bi,bj) |
DO j=0,sNy+1 |
160 |
& -aW2d(I ,J ,bi,bj)*cg2d_x(I ,J ,bi,bj) |
DO i=0,sNx+1 |
161 |
& -aW2d(I+1,J ,bi,bj)*cg2d_x(I ,J ,bi,bj) |
cg2d_s(i,j,bi,bj) = 0. |
162 |
& -aS2d(I ,J ,bi,bj)*cg2d_x(I ,J ,bi,bj) |
ENDDO |
163 |
& -aS2d(I ,J+1,bi,bj)*cg2d_x(I ,J ,bi,bj) |
ENDDO |
164 |
& -freeSurfFac*_rA(i,j,bi,bj)*recip_Bo(i,j,bi,bj)* |
sumRHStile(bi,bj) = 0. _d 0 |
165 |
& cg2d_x(I ,J ,bi,bj)/deltaTMom/deltaTfreesurf*cg2dNorm |
errTile(bi,bj) = 0. _d 0 |
166 |
|
#ifdef TARGET_NEC_SX |
167 |
|
!CDIR OUTERUNROLL=CG2D_OUTERLOOPITERS |
168 |
|
#endif /* TARGET_NEC_SX */ |
169 |
|
DO j=1,sNy |
170 |
|
DO i=1,sNx |
171 |
|
cg2d_r(i,j,bi,bj) = cg2d_b(i,j,bi,bj) - |
172 |
|
& (aW2d(i ,j ,bi,bj)*cg2d_x(i-1,j ,bi,bj) |
173 |
|
& +aW2d(i+1,j ,bi,bj)*cg2d_x(i+1,j ,bi,bj) |
174 |
|
& +aS2d(i ,j ,bi,bj)*cg2d_x(i ,j-1,bi,bj) |
175 |
|
& +aS2d(i ,j+1,bi,bj)*cg2d_x(i ,j+1,bi,bj) |
176 |
|
& +aC2d(i ,j ,bi,bj)*cg2d_x(i ,j ,bi,bj) |
177 |
& ) |
& ) |
178 |
err = err + |
#ifdef CG2D_SINGLECPU_SUM |
179 |
& cg2d_r(I,J,bi,bj)*cg2d_r(I,J,bi,bj) |
localBuf(i,j,bi,bj) = cg2d_r(i,j,bi,bj)*cg2d_r(i,j,bi,bj) |
180 |
sumRHS = sumRHS + |
#else |
181 |
& cg2d_b(I,J,bi,bj) |
errTile(bi,bj) = errTile(bi,bj) |
182 |
|
& + cg2d_r(i,j,bi,bj)*cg2d_r(i,j,bi,bj) |
183 |
|
sumRHStile(bi,bj) = sumRHStile(bi,bj) + cg2d_b(i,j,bi,bj) |
184 |
|
#endif |
185 |
ENDDO |
ENDDO |
186 |
ENDDO |
ENDDO |
187 |
ENDDO |
ENDDO |
188 |
ENDDO |
ENDDO |
189 |
C _EXCH_XY_R8( cg2d_r, myThid ) |
CALL EXCH_S3D_RL( cg2d_r, 1, myThid ) |
190 |
#ifdef LETS_MAKE_JAM |
#ifdef CG2D_SINGLECPU_SUM |
191 |
CALL EXCH_XY_O1_R8_JAM( cg2d_r ) |
CALL GLOBAL_SUM_SINGLECPU_RL(localBuf, err_sq, 0, 0, myThid) |
192 |
#else |
CALL GLOBAL_SUM_SINGLECPU_RL(cg2d_b, sumRHS, OLx, OLy, myThid) |
|
CALL EXCH_XY_RL( cg2d_r, myThid ) |
|
|
#endif |
|
|
C _EXCH_XY_R8( cg2d_s, myThid ) |
|
|
#ifdef LETS_MAKE_JAM |
|
|
CALL EXCH_XY_O1_R8_JAM( cg2d_s ) |
|
193 |
#else |
#else |
194 |
CALL EXCH_XY_RL( cg2d_s, myThid ) |
CALL GLOBAL_SUM_TILE_RL( errTile, err_sq, myThid ) |
195 |
|
CALL GLOBAL_SUM_TILE_RL( sumRHStile, sumRHS, myThid ) |
196 |
#endif |
#endif |
197 |
_GLOBAL_SUM_R8( sumRHS, myThid ) |
actualIts = 0 |
198 |
_GLOBAL_SUM_R8( err , myThid ) |
firstResidual = SQRT(err_sq) |
199 |
err = SQRT(err) |
IF ( nIterMin.GE.0 ) THEN |
200 |
actualIts = 0 |
nIterMin = 0 |
201 |
actualResidual = err |
minResidualSq = err_sq |
202 |
|
ENDIF |
203 |
IF ( debugLevel .GE. debLevA ) THEN |
|
204 |
|
printResidual = .FALSE. |
205 |
|
IF ( debugLevel .GE. debLevZero ) THEN |
206 |
_BEGIN_MASTER( myThid ) |
_BEGIN_MASTER( myThid ) |
207 |
write(*,'(A,1P2E22.14)')' cg2d: Sum(rhs),rhsMax = ', |
printResidual = printResidualFreq.GE.1 |
208 |
& sumRHS,rhsMax |
WRITE(standardmessageunit,'(A,1P2E22.14)') |
209 |
_END_MASTER( ) |
& ' cg2d: Sum(rhs),rhsMax = ', sumRHS,rhsMax |
210 |
ENDIF |
_END_MASTER( myThid ) |
211 |
C _BARRIER |
ENDIF |
|
c _BEGIN_MASTER( myThid ) |
|
|
c WRITE(*,'(A,I6,1PE30.14)') ' CG2D iters, err = ', |
|
|
c & actualIts, actualResidual |
|
|
c _END_MASTER( ) |
|
|
firstResidual=actualResidual |
|
212 |
|
|
213 |
IF ( err .LT. cg2dTolerance ) GOTO 11 |
IF ( err_sq .LT. cg2dTolerance_sq ) GOTO 11 |
214 |
|
|
215 |
C >>>>>>>>>>>>>>> BEGIN SOLVER <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< |
C >>>>>>>>>>>>>>> BEGIN SOLVER <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<< |
216 |
DO 10 it2d=1, numIters |
DO 10 it2d=1, numIters |
217 |
|
|
|
CcnhDebugStarts |
|
|
C WRITE(*,*) ' CG2D: Iteration ',it2d-1,' residual = ', |
|
|
C & actualResidual |
|
|
CcnhDebugEnds |
|
218 |
C-- Solve preconditioning equation and update |
C-- Solve preconditioning equation and update |
219 |
C-- conjugate direction vector "s". |
C-- conjugate direction vector "s". |
|
eta_qrN = 0. _d 0 |
|
220 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
221 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
222 |
DO J=1,sNy |
eta_qrNtile(bi,bj) = 0. _d 0 |
223 |
DO I=1,sNx |
#ifdef TARGET_NEC_SX |
224 |
cg2d_q(I,J,bi,bj) = |
!CDIR OUTERUNROLL=CG2D_OUTERLOOPITERS |
225 |
& pC(I ,J ,bi,bj)*cg2d_r(I ,J ,bi,bj) |
#endif /* TARGET_NEC_SX */ |
226 |
& +pW(I ,J ,bi,bj)*cg2d_r(I-1,J ,bi,bj) |
DO j=1,sNy |
227 |
& +pW(I+1,J ,bi,bj)*cg2d_r(I+1,J ,bi,bj) |
DO i=1,sNx |
228 |
& +pS(I ,J ,bi,bj)*cg2d_r(I ,J-1,bi,bj) |
cg2d_q(i,j,bi,bj) = |
229 |
& +pS(I ,J+1,bi,bj)*cg2d_r(I ,J+1,bi,bj) |
& pC(i ,j ,bi,bj)*cg2d_r(i ,j ,bi,bj) |
230 |
|
& +pW(i ,j ,bi,bj)*cg2d_r(i-1,j ,bi,bj) |
231 |
|
& +pW(i+1,j ,bi,bj)*cg2d_r(i+1,j ,bi,bj) |
232 |
|
& +pS(i ,j ,bi,bj)*cg2d_r(i ,j-1,bi,bj) |
233 |
|
& +pS(i ,j+1,bi,bj)*cg2d_r(i ,j+1,bi,bj) |
234 |
CcnhDebugStarts |
CcnhDebugStarts |
235 |
C cg2d_q(I,J,bi,bj) = cg2d_r(I ,J ,bi,bj) |
c cg2d_q(i,j,bi,bj) = cg2d_r(j ,j ,bi,bj) |
236 |
CcnhDebugEnds |
CcnhDebugEnds |
237 |
eta_qrN = eta_qrN |
#ifdef CG2D_SINGLECPU_SUM |
238 |
& +cg2d_q(I,J,bi,bj)*cg2d_r(I,J,bi,bj) |
localBuf(i,j,bi,bj) = |
239 |
|
& cg2d_q(i,j,bi,bj)*cg2d_r(i,j,bi,bj) |
240 |
|
#else |
241 |
|
eta_qrNtile(bi,bj) = eta_qrNtile(bi,bj) |
242 |
|
& +cg2d_q(i,j,bi,bj)*cg2d_r(i,j,bi,bj) |
243 |
|
#endif |
244 |
ENDDO |
ENDDO |
245 |
ENDDO |
ENDDO |
246 |
ENDDO |
ENDDO |
247 |
ENDDO |
ENDDO |
248 |
|
|
249 |
_GLOBAL_SUM_R8(eta_qrN, myThid) |
#ifdef CG2D_SINGLECPU_SUM |
250 |
CcnhDebugStarts |
CALL GLOBAL_SUM_SINGLECPU_RL( localBuf,eta_qrN,0,0,myThid ) |
251 |
C WRITE(*,*) ' CG2D: Iteration ',it2d-1,' eta_qrN = ',eta_qrN |
#else |
252 |
CcnhDebugEnds |
CALL GLOBAL_SUM_TILE_RL( eta_qrNtile,eta_qrN,myThid ) |
253 |
|
#endif |
254 |
cgBeta = eta_qrN/eta_qrNM1 |
cgBeta = eta_qrN/eta_qrNM1 |
255 |
CcnhDebugStarts |
CcnhDebugStarts |
256 |
C WRITE(*,*) ' CG2D: Iteration ',it2d-1,' beta = ',cgBeta |
c WRITE(*,*) ' CG2D: Iteration ', it2d-1, |
257 |
|
c & ' eta_qrN=', eta_qrN, ' beta=', cgBeta |
258 |
CcnhDebugEnds |
CcnhDebugEnds |
259 |
eta_qrNM1 = eta_qrN |
eta_qrNM1 = eta_qrN |
260 |
|
|
261 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
262 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
263 |
DO J=1,sNy |
DO j=1,sNy |
264 |
DO I=1,sNx |
DO i=1,sNx |
265 |
cg2d_s(I,J,bi,bj) = cg2d_q(I,J,bi,bj) |
cg2d_s(i,j,bi,bj) = cg2d_q(i,j,bi,bj) |
266 |
& + cgBeta*cg2d_s(I,J,bi,bj) |
& + cgBeta*cg2d_s(i,j,bi,bj) |
267 |
ENDDO |
ENDDO |
268 |
ENDDO |
ENDDO |
269 |
ENDDO |
ENDDO |
270 |
ENDDO |
ENDDO |
271 |
|
|
272 |
C-- Do exchanges that require messages i.e. between |
C-- Do exchanges that require messages i.e. between processes. |
273 |
C-- processes. |
CALL EXCH_S3D_RL( cg2d_s, 1, myThid ) |
|
C _EXCH_XY_R8( cg2d_s, myThid ) |
|
|
#ifdef LETS_MAKE_JAM |
|
|
CALL EXCH_XY_O1_R8_JAM( cg2d_s ) |
|
|
#else |
|
|
CALL EXCH_XY_RL( cg2d_s, myThid ) |
|
|
#endif |
|
274 |
|
|
275 |
C== Evaluate laplace operator on conjugate gradient vector |
C== Evaluate laplace operator on conjugate gradient vector |
276 |
C== q = A.s |
C== q = A.s |
|
alpha = 0. _d 0 |
|
277 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
278 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
279 |
DO J=1,sNy |
alphaTile(bi,bj) = 0. _d 0 |
280 |
DO I=1,sNx |
#ifdef TARGET_NEC_SX |
281 |
cg2d_q(I,J,bi,bj) = |
!CDIR OUTERUNROLL=CG2D_OUTERLOOPITERS |
282 |
& aW2d(I ,J ,bi,bj)*cg2d_s(I-1,J ,bi,bj) |
#endif /* TARGET_NEC_SX */ |
283 |
& +aW2d(I+1,J ,bi,bj)*cg2d_s(I+1,J ,bi,bj) |
DO j=1,sNy |
284 |
& +aS2d(I ,J ,bi,bj)*cg2d_s(I ,J-1,bi,bj) |
DO i=1,sNx |
285 |
& +aS2d(I ,J+1,bi,bj)*cg2d_s(I ,J+1,bi,bj) |
cg2d_q(i,j,bi,bj) = |
286 |
& -aW2d(I ,J ,bi,bj)*cg2d_s(I ,J ,bi,bj) |
& aW2d(i ,j ,bi,bj)*cg2d_s(i-1,j ,bi,bj) |
287 |
& -aW2d(I+1,J ,bi,bj)*cg2d_s(I ,J ,bi,bj) |
& +aW2d(i+1,j ,bi,bj)*cg2d_s(i+1,j ,bi,bj) |
288 |
& -aS2d(I ,J ,bi,bj)*cg2d_s(I ,J ,bi,bj) |
& +aS2d(i ,j ,bi,bj)*cg2d_s(i ,j-1,bi,bj) |
289 |
& -aS2d(I ,J+1,bi,bj)*cg2d_s(I ,J ,bi,bj) |
& +aS2d(i ,j+1,bi,bj)*cg2d_s(i ,j+1,bi,bj) |
290 |
& -freeSurfFac*_rA(i,j,bi,bj)*recip_Bo(i,j,bi,bj)* |
& +aC2d(i ,j ,bi,bj)*cg2d_s(i ,j ,bi,bj) |
291 |
& cg2d_s(I ,J ,bi,bj)/deltaTMom/deltaTfreesurf*cg2dNorm |
#ifdef CG2D_SINGLECPU_SUM |
292 |
alpha = alpha+cg2d_s(I,J,bi,bj)*cg2d_q(I,J,bi,bj) |
localBuf(i,j,bi,bj) = cg2d_s(i,j,bi,bj)*cg2d_q(i,j,bi,bj) |
293 |
|
#else |
294 |
|
alphaTile(bi,bj) = alphaTile(bi,bj) |
295 |
|
& + cg2d_s(i,j,bi,bj)*cg2d_q(i,j,bi,bj) |
296 |
|
#endif |
297 |
ENDDO |
ENDDO |
298 |
ENDDO |
ENDDO |
299 |
ENDDO |
ENDDO |
300 |
ENDDO |
ENDDO |
301 |
_GLOBAL_SUM_R8(alpha,myThid) |
#ifdef CG2D_SINGLECPU_SUM |
302 |
|
CALL GLOBAL_SUM_SINGLECPU_RL(localBuf, alpha, 0, 0, myThid) |
303 |
|
#else |
304 |
|
CALL GLOBAL_SUM_TILE_RL( alphaTile, alpha, myThid ) |
305 |
|
#endif |
306 |
CcnhDebugStarts |
CcnhDebugStarts |
307 |
C WRITE(*,*) ' CG2D: Iteration ',it2d-1,' SUM(s*q)= ',alpha |
c WRITE(*,*) ' CG2D: Iteration ', it2d-1, |
308 |
|
c & ' SUM(s*q)=', alpha, ' alpha=', eta_qrN/alpha |
309 |
CcnhDebugEnds |
CcnhDebugEnds |
310 |
alpha = eta_qrN/alpha |
alpha = eta_qrN/alpha |
311 |
CcnhDebugStarts |
|
312 |
C WRITE(*,*) ' CG2D: Iteration ',it2d-1,' alpha= ',alpha |
C== Update simultaneously solution and residual vectors (and Iter number) |
|
CcnhDebugEnds |
|
|
|
|
|
C== Update solution and residual vectors |
|
313 |
C Now compute "interior" points. |
C Now compute "interior" points. |
|
err = 0. _d 0 |
|
314 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
315 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
316 |
DO J=1,sNy |
errTile(bi,bj) = 0. _d 0 |
317 |
DO I=1,sNx |
DO j=1,sNy |
318 |
cg2d_x(I,J,bi,bj)=cg2d_x(I,J,bi,bj)+alpha*cg2d_s(I,J,bi,bj) |
DO i=1,sNx |
319 |
cg2d_r(I,J,bi,bj)=cg2d_r(I,J,bi,bj)-alpha*cg2d_q(I,J,bi,bj) |
cg2d_x(i,j,bi,bj)=cg2d_x(i,j,bi,bj)+alpha*cg2d_s(i,j,bi,bj) |
320 |
err = err+cg2d_r(I,J,bi,bj)*cg2d_r(I,J,bi,bj) |
cg2d_r(i,j,bi,bj)=cg2d_r(i,j,bi,bj)-alpha*cg2d_q(i,j,bi,bj) |
321 |
|
#ifdef CG2D_SINGLECPU_SUM |
322 |
|
localBuf(i,j,bi,bj) = cg2d_r(i,j,bi,bj)*cg2d_r(i,j,bi,bj) |
323 |
|
#else |
324 |
|
errTile(bi,bj) = errTile(bi,bj) |
325 |
|
& + cg2d_r(i,j,bi,bj)*cg2d_r(i,j,bi,bj) |
326 |
|
#endif |
327 |
ENDDO |
ENDDO |
328 |
ENDDO |
ENDDO |
329 |
ENDDO |
ENDDO |
330 |
ENDDO |
ENDDO |
331 |
|
actualIts = it2d |
332 |
|
|
333 |
_GLOBAL_SUM_R8( err , myThid ) |
#ifdef CG2D_SINGLECPU_SUM |
334 |
err = SQRT(err) |
CALL GLOBAL_SUM_SINGLECPU_RL(localBuf, err_sq, 0, 0, myThid) |
|
actualIts = it2d |
|
|
actualResidual = err |
|
|
IF ( err .LT. cg2dTolerance ) GOTO 11 |
|
|
C _EXCH_XY_R8(cg2d_r, myThid ) |
|
|
#ifdef LETS_MAKE_JAM |
|
|
CALL EXCH_XY_O1_R8_JAM( cg2d_r ) |
|
335 |
#else |
#else |
336 |
CALL EXCH_XY_RL( cg2d_r, myThid ) |
CALL GLOBAL_SUM_TILE_RL( errTile, err_sq, myThid ) |
337 |
#endif |
#endif |
338 |
|
IF ( printResidual ) THEN |
339 |
|
IF ( MOD( it2d-1, printResidualFreq ).EQ.0 ) THEN |
340 |
|
WRITE(msgBuf,'(A,I6,A,1PE21.14)') |
341 |
|
& ' cg2d: iter=', it2d, ' ; resid.= ', SQRT(err_sq) |
342 |
|
CALL PRINT_MESSAGE( msgBuf, standardMessageUnit, |
343 |
|
& SQUEEZE_RIGHT, myThid ) |
344 |
|
ENDIF |
345 |
|
ENDIF |
346 |
|
IF ( err_sq .LT. cg2dTolerance_sq ) GOTO 11 |
347 |
|
IF ( err_sq .LT. minResidualSq ) THEN |
348 |
|
C- Store lowest residual solution |
349 |
|
minResidualSq = err_sq |
350 |
|
nIterMin = it2d |
351 |
|
DO bj=myByLo(myThid),myByHi(myThid) |
352 |
|
DO bi=myBxLo(myThid),myBxHi(myThid) |
353 |
|
DO j=1,sNy |
354 |
|
DO i=1,sNx |
355 |
|
cg2d_min(i,j,bi,bj) = cg2d_x(i,j,bi,bj) |
356 |
|
ENDDO |
357 |
|
ENDDO |
358 |
|
ENDDO |
359 |
|
ENDDO |
360 |
|
ENDIF |
361 |
|
|
362 |
|
CALL EXCH_S3D_RL( cg2d_r, 1, myThid ) |
363 |
|
|
364 |
10 CONTINUE |
10 CONTINUE |
365 |
11 CONTINUE |
11 CONTINUE |
366 |
|
|
367 |
|
IF ( nIterMin.GE.0 .AND. err_sq .GT. minResidualSq ) THEN |
368 |
|
C- use the lowest residual solution (instead of current one = last residual) |
369 |
|
DO bj=myByLo(myThid),myByHi(myThid) |
370 |
|
DO bi=myBxLo(myThid),myBxHi(myThid) |
371 |
|
DO j=1,sNy |
372 |
|
DO i=1,sNx |
373 |
|
cg2d_x(i,j,bi,bj) = cg2d_min(i,j,bi,bj) |
374 |
|
ENDDO |
375 |
|
ENDDO |
376 |
|
ENDDO |
377 |
|
ENDDO |
378 |
|
ENDIF |
379 |
|
|
380 |
IF (cg2dNormaliseRHS) THEN |
IF (cg2dNormaliseRHS) THEN |
381 |
C-- Un-normalise the answer |
C-- Un-normalise the answer |
382 |
DO bj=myByLo(myThid),myByHi(myThid) |
DO bj=myByLo(myThid),myByHi(myThid) |
383 |
DO bi=myBxLo(myThid),myBxHi(myThid) |
DO bi=myBxLo(myThid),myBxHi(myThid) |
384 |
DO J=1,sNy |
DO j=1,sNy |
385 |
DO I=1,sNx |
DO i=1,sNx |
386 |
cg2d_x(I ,J ,bi,bj) = cg2d_x(I ,J ,bi,bj)/rhsNorm |
cg2d_x(i,j,bi,bj) = cg2d_x(i,j,bi,bj)/rhsNorm |
387 |
ENDDO |
ENDDO |
388 |
ENDDO |
ENDDO |
389 |
ENDDO |
ENDDO |
390 |
ENDDO |
ENDDO |
391 |
ENDIF |
ENDIF |
392 |
|
|
|
C The following exchange was moved up to solve_for_pressure |
|
|
C for compatibility with TAMC. |
|
|
C _EXCH_XY_R8(cg2d_x, myThid ) |
|
|
c _BEGIN_MASTER( myThid ) |
|
|
c WRITE(*,'(A,I6,1PE30.14)') ' CG2D iters, err = ', |
|
|
c & actualIts, actualResidual |
|
|
c _END_MASTER( ) |
|
|
|
|
393 |
C-- Return parameters to caller |
C-- Return parameters to caller |
394 |
lastResidual=actualResidual |
lastResidual = SQRT(err_sq) |
395 |
numIters=actualIts |
numIters = actualIts |
396 |
|
|
397 |
CcnhDebugStarts |
CcnhDebugStarts |
398 |
C CALL PLOT_FIELD_XYRL( cg2d_x, 'CALC_MOM_RHS CG2D_X' , 1, myThid ) |
c _EXCH_XY_RL(cg2d_x, myThid ) |
399 |
C err = 0. _d 0 |
c CALL PLOT_FIELD_XYRL( cg2d_x, 'CALC_MOM_RHS CG2D_X' , 1, myThid ) |
400 |
C DO bj=myByLo(myThid),myByHi(myThid) |
c err_sq = 0. _d 0 |
401 |
C DO bi=myBxLo(myThid),myBxHi(myThid) |
c DO bj=myByLo(myThid),myByHi(myThid) |
402 |
C DO J=1,sNy |
c DO bi=myBxLo(myThid),myBxHi(myThid) |
403 |
C DO I=1,sNx |
c DO j=1,sNy |
404 |
C cg2d_r(I,J,bi,bj) = cg2d_b(I,J,bi,bj) - |
c DO i=1,sNx |
405 |
C & (aW2d(I ,J ,bi,bj)*cg2d_x(I-1,J ,bi,bj) |
c cg2d_r(i,j,bi,bj) = cg2d_b(i,j,bi,bj) - |
406 |
C & +aW2d(I+1,J ,bi,bj)*cg2d_x(I+1,J ,bi,bj) |
c & (aW2d(i ,j ,bi,bj)*cg2d_x(i-1,j ,bi,bj) |
407 |
C & +aS2d(I ,J ,bi,bj)*cg2d_x(I ,J-1,bi,bj) |
c & +aW2d(i+1,j ,bi,bj)*cg2d_x(i+1,j ,bi,bj) |
408 |
C & +aS2d(I ,J+1,bi,bj)*cg2d_x(I ,J+1,bi,bj) |
c & +aS2d(i ,j ,bi,bj)*cg2d_x(i ,j-1,bi,bj) |
409 |
C & -aW2d(I ,J ,bi,bj)*cg2d_x(I ,J ,bi,bj) |
c & +aS2d(i ,j+1,bi,bj)*cg2d_x(i ,j+1,bi,bj) |
410 |
C & -aW2d(I+1,J ,bi,bj)*cg2d_x(I ,J ,bi,bj) |
c & +aC2d(i ,j ,bi,bj)*cg2d_x(i ,j ,bi,bj) |
411 |
C & -aS2d(I ,J ,bi,bj)*cg2d_x(I ,J ,bi,bj) |
c & ) |
412 |
C & -aS2d(I ,J+1,bi,bj)*cg2d_x(I ,J ,bi,bj)) |
c err_sq = err_sq + cg2d_r(i,j,bi,bj)*cg2d_r(i,j,bi,bj) |
413 |
C err = err + |
c ENDDO |
414 |
C & cg2d_r(I,J,bi,bj)*cg2d_r(I,J,bi,bj) |
c ENDDO |
415 |
C ENDDO |
c ENDDO |
416 |
C ENDDO |
c ENDDO |
417 |
C ENDDO |
c _GLOBAL_SUM_RL( err_sq, myThid ) |
418 |
C ENDDO |
c write(*,*) 'cg2d: Ax - b = ',SQRT(err_sq) |
|
C _GLOBAL_SUM_R8( err , myThid ) |
|
|
C write(*,*) 'cg2d: Ax - b = ',SQRT(err) |
|
419 |
CcnhDebugEnds |
CcnhDebugEnds |
420 |
|
|
421 |
RETURN |
RETURN |