1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
|
/* CalculiX - A 3-dimensional finite element program */
/* Copyright (C) 1998-2015 Guido Dhondt */
/* This program is free software; you can redistribute it and/or */
/* modify it under the terms of the GNU General Public License as */
/* published by the Free Software Foundation(version 2); */
/* */
/* This program is distributed in the hope that it will be useful, */
/* but WITHOUT ANY WARRANTY; without even the implied warranty of */
/* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the */
/* GNU General Public License for more details. */
/* You should have received a copy of the GNU General Public License */
/* along with this program; if not, write to the Free Software */
/* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
#include <unistd.h>
#include <stdio.h>
#include <math.h>
#include <stdlib.h>
#include <pthread.h>
#include "CalculiX.h"
static char *lakonf1;
static ITG num_cpus,*nef1,*ipnei1,*neifa1,*neiel1,*jq1,*irow1,*nzs1,
*ielfa1,*ifabou1,*nbody1,*neq1,*nactdohinv1;
static double *au1=NULL,*ad1=NULL,*b1=NULL,*vfa1,*xxn1,*area1,*vel1,
*umel1,*xlet1,*xle1,*gradtfa1,*xxi1,*body1,*volume1,*dtimef1,*velo1,
*veloo1,*cvfa1,*hcfa1,*cvel1,*gradvel1,*xload1,*gammat1,*xrlfa1,
*xxj1,*a11,*a21,*a31,*flux1;
void mafilltmain(ITG *nef,ITG *ipnei,ITG *neifa,
ITG *neiel,double *vfa,double *xxn,double *area,
double *au,double *ad,ITG *jq,ITG *irow,ITG *nzs,
double *b,double *vel,double *umel,double *xlet,
double *xle,double *gradtfa,double *xxi,double *body,
double *volume,ITG *ielfa,char *lakonf,
ITG *ifabou,ITG *nbody,ITG *neq,double *dtimef,double *velo,
double *veloo,double *cvfa,double *hcfa,double *cvel,
double *gradvel,double *xload,double *gammat,double *xrlfa,
double *xxj,ITG *nactdohinv,double *a1,double *a2,double *a3,
double *flux){
ITG i,j;
/* variables for multithreading procedure */
ITG sys_cpus,*ithread=NULL;
char *env,*envloc,*envsys;
num_cpus = 0;
sys_cpus=0;
/* explicit user declaration prevails */
envsys=getenv("NUMBER_OF_CPUS");
if(envsys){
sys_cpus=atoi(envsys);
if(sys_cpus<0) sys_cpus=0;
}
/* automatic detection of available number of processors */
if(sys_cpus==0){
sys_cpus = getSystemCPUs();
if(sys_cpus<1) sys_cpus=1;
}
/* local declaration prevails, if strictly positive */
envloc = getenv("CCX_NPROC_CFD");
if(envloc){
num_cpus=atoi(envloc);
if(num_cpus<0){
num_cpus=0;
}else if(num_cpus>sys_cpus){
num_cpus=sys_cpus;
}
}
/* else global declaration, if any, applies */
env = getenv("OMP_NUM_THREADS");
if(num_cpus==0){
if (env)
num_cpus = atoi(env);
if (num_cpus < 1) {
num_cpus=1;
}else if(num_cpus>sys_cpus){
num_cpus=sys_cpus;
}
}
// next line is to be inserted in a similar way for all other parallel parts
if(*nef<num_cpus) num_cpus=*nef;
pthread_t tid[num_cpus];
/* allocating fields for lhs and rhs matrix */
NNEW(ad1,double,num_cpus**neq);
NNEW(au1,double,(long long)num_cpus*2**nzs);
NNEW(b1,double,num_cpus**neq);
/* calculating the stiffness and/or mass matrix
(symmetric part) */
nef1=nef;ipnei1=ipnei;neifa1=neifa;neiel1=neiel;vfa1=vfa;xxn1=xxn;
area1=area;
jq1=jq;irow1=irow;nzs1=nzs;vel1=vel;umel1=umel;xlet1=xlet;xle1=xle;
gradtfa1=gradtfa;xxi1=xxi;body1=body;volume1=volume;
ielfa1=ielfa;lakonf1=lakonf;ifabou1=ifabou;
nbody1=nbody;neq1=neq;dtimef1=dtimef;velo1=velo;veloo1=veloo;
cvfa1=cvfa;hcfa1=hcfa;cvel1=cvel;gradvel1=gradvel;xload1=xload;
gammat1=gammat;xrlfa1=xrlfa;xxj1=xxj;nactdohinv1=nactdohinv;a11=a1;
a21=a2;a31=a3;flux1=flux;
/* create threads and wait */
NNEW(ithread,ITG,num_cpus);
for(i=0; i<num_cpus; i++) {
ithread[i]=i;
pthread_create(&tid[i], NULL, (void *)mafilltmt, (void *)&ithread[i]);
}
for(i=0; i<num_cpus; i++) pthread_join(tid[i], NULL);
SFREE(ithread);
/* copying and accumulating the stiffnes and/or mass matrix */
#pragma omp parallel \
default(none) \
shared(neq,ad,ad1,num_cpus,nzs,au,au1,b,b1) \
private(i,j)
{
#pragma omp for
for(i=0;i<*neq;i++){
ad[i]=ad1[i];
for(j=1;j<num_cpus;j++){
ad[i]+=ad1[i+j**neq];
}
}
#pragma omp for
for(i=0;i<2**nzs;i++){
au[i]=au1[i];
for(j=1;j<num_cpus;j++){
au[i]+=au1[i+(long long)j*2**nzs];
}
}
#pragma omp for
for(i=0;i<*neq;i++){
b[i]=b1[i];
for(j=1;j<num_cpus;j++){
b[i]+=b1[i+j**neq];
}
}
}
SFREE(ad1);
SFREE(au1);
SFREE(b1);
return;
}
/* subroutine for multithreading of mafillt */
void *mafilltmt(ITG *i){
ITG indexad,indexb,nefa,nefb,nefdelta;
long long indexau;
indexad=*i**neq1;
indexau=(long long)*i*2**nzs1;
indexb=*i**neq1;
// ceil -> floor
nefdelta=(ITG)floor(*nef1/(double)num_cpus);
nefa=*i*nefdelta+1;
nefb=(*i+1)*nefdelta;
// next line! -> all parallel sections
if((*i==num_cpus-1)&&(nefb<*nef1)) nefb=*nef1;
FORTRAN(mafillt,(nef1,ipnei1,neifa1,neiel1,vfa1,xxn1,area1,
&au1[indexau],&ad1[indexad],jq1,irow1,nzs1,
&b1[indexb],vel1,umel1,xlet1,xle1,gradtfa1,xxi1,
body1,volume1,ielfa1,lakonf1,ifabou1,
nbody1,neq1,dtimef1,velo1,veloo1,cvfa1,hcfa1,cvel1,
gradvel1,xload1,gammat1,xrlfa1,xxj1,nactdohinv1,
a11,a21,a31,flux1,&nefa,&nefb));
return NULL;
}
|
