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
|
#include <tjutils/tjthread.h>
#include <tjutils/tjlog.h>
template<typename In, typename Out, typename Local>
void ThreadedLoop<In,Out,Local>::WorkThread::run() {
Log<ThreadComponent> odinlog("WorkThread","run");
while(1) {
process.wait();
process.reset(); // Reset for next event
ODINLOG(odinlog,normalDebug) << "catched process signal, cont=" << tloop->cont << STD_endl;
if(!tloop->cont) break;
ODINLOG(odinlog,normalDebug) << "processing thread " << begin << "/" << end << STD_endl;
status=tloop->kernel(*tloop->in_cache, *out_cache, local, begin, end);
ODINLOG(odinlog,normalDebug) << "signaling finished=" << status << STD_endl;
finished.signal();
if(!status) break;
}
}
//////////////////////////////////////////////////////////////////////////
template<typename In, typename Out, typename Local>
bool ThreadedLoop<In,Out,Local>::init(unsigned int numof_threads, unsigned int loopsize) {
Log<ThreadComponent> odinlog("ThreadedLoop","init");
mainbegin=0;
mainend=loopsize;
#ifndef NO_THREADS
destroy(); // stop old threads
ODINLOG(odinlog,normalDebug) << "numof_threads=" << numof_threads << STD_endl;
if(numof_threads>1) {
threads.resize(numof_threads-1); // the main thread is also used
unsigned int onesize=loopsize/numof_threads;
unsigned int rest=loopsize%numof_threads;
unsigned int count=0;
for(unsigned int i=0; i<(numof_threads-1); i++) {
threads[i]=new WorkThread(this);
threads[i]->begin=count;
count+=onesize;
if(i<rest) count++;
threads[i]->end=count;
threads[i]->start();
}
mainbegin=count;
count+=onesize;
if((numof_threads-1)<rest) count++;
mainend=count;
}
#endif
return true;
}
template<typename In, typename Out, typename Local>
void ThreadedLoop<In,Out,Local>::destroy() {
Log<ThreadComponent> odinlog("ThreadedLoop","destroy");
#ifndef NO_THREADS
cont=false; // Stop threads
for(unsigned int i=0; i<threads.size(); i++) {
threads[i]->process.signal();
threads[i]->wait();
delete threads[i];
}
threads.resize(0);
#endif
}
template<typename In, typename Out, typename Local>
bool ThreadedLoop<In,Out,Local>::execute(const In& in, STD_vector<Out>& outvec) {
Log<ThreadComponent> odinlog("ThreadedLoop","execute");
#ifdef NO_THREADS
outvec.resize(1);
return kernel(in, outvec[0], mainlocal, mainbegin, mainend);
#else
unsigned int nthreads=threads.size();
outvec.resize(nthreads+1);
if(nthreads) {
in_cache=∈
cont=true;
for(unsigned int i=0; i<nthreads; i++) {
threads[i]->out_cache=&(outvec[i]);
threads[i]->status=true;
threads[i]->process.signal();
}
}
bool result=kernel(in, outvec[nthreads], mainlocal, mainbegin, mainend); // Use mainthread also for kernel
if(nthreads) {
for(unsigned int i=0; i<nthreads; i++) {
threads[i]->finished.wait(); // Wait for result
threads[i]->finished.reset();
if(!threads[i]->status) result=false;
}
ODINLOG(odinlog,normalDebug) << "finished.wait() done" << STD_endl;
}
return result;
#endif
}
|
