/[escript]/branches/arrayview_from_1695_trunk/finley/src/Assemble_PDE_System2_C.c
ViewVC logotype

Annotation of /branches/arrayview_from_1695_trunk/finley/src/Assemble_PDE_System2_C.c

Parent Directory Parent Directory | Revision Log Revision Log


Revision 853 - (hide annotations)
Wed Sep 20 05:56:36 2006 UTC (12 years, 11 months ago) by gross
Original Path: trunk/finley/src/Assemble_PDE_System2_C.c
File MIME type: text/plain
File size: 8085 byte(s)
some performance problems wit openmp fixed.
1 gross 798 /*
2     ************************************************************
3     * Copyright 2006 by ACcESS MNRF *
4     * *
5     * http://www.access.edu.au *
6     * Primary Business: Queensland, Australia *
7     * Licensed under the Open Software License version 3.0 *
8     * http://www.opensource.org/licenses/osl-3.0.php *
9     * *
10     ************************************************************
11     */
12    
13     /**************************************************************/
14    
15     /* assembles the system of numEq PDEs into the stiffness matrix S right hand side F */
16    
17     /* D_{k,m} u_m and Y_k */
18    
19     /* u has p.numComp components in a 3D domain. The shape functions for test and solution must be identical */
20     /* and 2* row_NS == row_NN (contact elements) */
21    
22     /* Shape of the coefficients: */
23    
24     /* D = p.numEqu x p.numComp */
25     /* Y = p.numEqu */
26    
27    
28     /**************************************************************/
29    
30     /* Author: gross@access.edu.au */
31     /* Version: $Id:$ */
32    
33     /**************************************************************/
34    
35    
36     #include "Assemble.h"
37     #include "Util.h"
38 gross 853 #ifdef _OPENMP
39     #include <omp.h>
40     #endif
41 gross 798
42 gross 853
43 gross 798 /**************************************************************/
44    
45     void Finley_Assemble_PDE_System2_C(Assemble_Parameters p, Finley_ElementFile* elements,
46     Paso_SystemMatrix* Mat, escriptDataC* F, escriptDataC* D, escriptDataC* Y) {
47    
48     index_t color;
49     dim_t e;
50 gross 853 double *EM_S, *EM_F, *Vol, *D_p, *Y_p;
51     index_t *row_index;
52     register dim_t q, s,r,k,m;
53     register double rtmp, rtmp_D;
54     bool_t add_EM_F, add_EM_S;
55    
56 gross 798 bool_t extendedD=isExpanded(D);
57     bool_t extendedY=isExpanded(Y);
58     double *F_p=getSampleData(F,0);
59     double *S=p.row_jac->ReferenceElement->S;
60    
61    
62 gross 853 #pragma omp parallel private(color,EM_S, EM_F, Vol, D_p, Y_p,row_index,q, s,r,k,m,rtmp, rtmp_D,add_EM_F, add_EM_S)
63 gross 798 {
64 gross 853 EM_S=THREAD_MEMALLOC(p.row_NN*p.col_NN*p.numEqu*p.numComp,double);
65     EM_F=THREAD_MEMALLOC(p.row_NN*p.numEqu,double);
66     row_index=THREAD_MEMALLOC(p.row_NN,index_t);
67    
68     if (!Finley_checkPtr(EM_S) && !Finley_checkPtr(EM_F) && !Finley_checkPtr(row_index) ) {
69    
70     #ifndef PASO_MPI
71     for (color=elements->minColor;color<=elements->maxColor;color++) {
72     /* open loop over all elements: */
73     #pragma omp for private(e) schedule(static)
74     for(e=0;e<elements->numElements;e++){
75     if (elements->Color[e]==color) {
76     #else
77     {
78     for(e=0;e<elements->numElements;e++) {
79     {
80     #endif
81     Vol=&(p.row_jac->volume[INDEX2(0,e,p.numQuad)]);
82     add_EM_F=FALSE;
83     add_EM_S=FALSE;
84     /************************************************************* */
85     /* process D */
86     /**************************************************************/
87     D_p=getSampleData(D,e);
88     if (NULL!=D_p) {
89     add_EM_S=TRUE;
90     if (extendedD) {
91     for (s=0;s<p.row_NS;s++) {
92     for (r=0;r<p.col_NS;r++) {
93     for (k=0;k<p.numEqu;k++) {
94     for (m=0;m<p.numComp;m++) {
95     rtmp=0;
96     for (q=0;q<p.numQuad;q++) {
97     rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*D_p[INDEX3(k,m,q,p.numEqu,p.numComp)]*S[INDEX2(r,q,p.row_NS)];
98     }
99     EM_S[INDEX4(k,m,s ,r ,p.numEqu,p.numComp,p.row_NN)]= rtmp;
100     EM_S[INDEX4(k,m,s ,r+p.col_NS,p.numEqu,p.numComp,p.row_NN)]=-rtmp;
101     EM_S[INDEX4(k,m,s+p.row_NS,r ,p.numEqu,p.numComp,p.row_NN)]=-rtmp;
102     EM_S[INDEX4(k,m,s+p.row_NS,r+p.col_NS,p.numEqu,p.numComp,p.row_NN)]= rtmp;
103     }
104     }
105     }
106     }
107     } else {
108     for (s=0;s<p.row_NS;s++) {
109     for (r=0;r<p.col_NS;r++) {
110     rtmp=0;
111     for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*S[INDEX2(r,q,p.row_NS)];
112     for (k=0;k<p.numEqu;k++) {
113     for (m=0;m<p.numComp;m++) {
114     rtmp_D=rtmp*D_p[INDEX2(k,m,p.numEqu)];
115     EM_S[INDEX4(k,m,s ,r ,p.numEqu,p.numComp,p.row_NN)]= rtmp_D;
116     EM_S[INDEX4(k,m,s ,r+p.col_NS,p.numEqu,p.numComp,p.row_NN)]=-rtmp_D;
117     EM_S[INDEX4(k,m,s+p.row_NS,r ,p.numEqu,p.numComp,p.row_NN)]=-rtmp_D;
118     EM_S[INDEX4(k,m,s+p.row_NS,r+p.col_NS,p.numEqu,p.numComp,p.row_NN)]= rtmp_D;
119     }
120     }
121     }
122     }
123     }
124     }
125     /**************************************************************/
126     /* process Y: */
127     /**************************************************************/
128     Y_p=getSampleData(Y,e);
129     if (NULL!=Y_p) {
130     add_EM_F=TRUE;
131     if (extendedY) {
132     for (s=0;s<p.row_NS;s++) {
133     for (k=0;k<p.numEqu;k++) {
134 gross 798 rtmp=0;
135 gross 853 for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)]*Y_p[INDEX2(k,q,p.numEqu)];
136     EM_F[INDEX2(k,s ,p.numEqu)]=-rtmp;
137     EM_F[INDEX2(k,s+p.row_NS,p.numEqu)]= rtmp;
138     }
139 gross 798 }
140 gross 853 } else {
141     for (s=0;s<p.row_NS;s++) {
142 gross 798 rtmp=0;
143 gross 853 for (q=0;q<p.numQuad;q++) rtmp+=Vol[q]*S[INDEX2(s,q,p.row_NS)];
144 gross 798 for (k=0;k<p.numEqu;k++) {
145 gross 853 rtmp_D=rtmp*Y_p[k];
146     EM_F[INDEX2(k,s ,p.numEqu)]=-rtmp_D;
147     EM_F[INDEX2(k,s+p.row_NS,p.numEqu)]= rtmp_D;
148 gross 798 }
149     }
150     }
151 gross 853 }
152     /***********************************************************************************************/
153     /* add the element matrices onto the matrix and right hand side */
154     /***********************************************************************************************/
155     for (q=0;q<p.row_NN;q++) row_index[q]=p.row_DOF[elements->Nodes[INDEX2(p.row_node[q],e,p.row_NN)]];
156     if (add_EM_F) Finley_Util_AddScatter(p.row_NN,row_index,p.numEqu,EM_F,F_p, p.row_DOF_UpperBound);
157     if (add_EM_S) Finley_Assemble_addToSystemMatrix(Mat,p.row_NN,row_index,p.numEqu,p.col_NN,row_index,p.numComp,EM_S);
158    
159     } /* end color check */
160     } /* end element loop */
161     } /* end color loop */
162    
163     THREAD_MEMFREE(EM_S);
164     THREAD_MEMFREE(EM_F);
165     THREAD_MEMFREE(row_index);
166 gross 798
167 gross 853 } /* end of pointer check */
168 gross 798 } /* end parallel region */
169     }
170     /*
171     * $Log$
172     */

  ViewVC Help
Powered by ViewVC 1.1.26