void *work_thread (void *lp) {
int task_id = *((int *) lp);
int begin, end;
struct timeval start, finish;
int i;
/*get the divided task*/
begin = (nsize * task_id) / task_num + 1;
end = (nsize * (task_id + 1)) / task_num;
if(task_id==0) gettimeofday (&start, NULL);
fprintf (stderr, "thread %d: begin %d, end %d\n", task_id, begin, end);
barrier (task_num);
/* initialization */
if (task_id == 0)
initRHS(nsize, begin, end);
barrier (task_num);
initResult(nsize, begin, end);
barrier (task_num);
/* Gauss Compute */
computeGauss(nsize, task_id);
barrier (task_num);
/* Gauss computation done */
if(task_id==0) {
/* Since there are dependencies in computing equation stage
the upper part need the results of upper part), it should be done by thread 0 solely. */
solveGauss(nsize);
gettimeofday (&finish, NULL);
printf ("Elapsed time: %.2f seconds\n",
(((finish.tv_sec * 1000000.0) + finish.tv_usec) -
((start.tv_sec * 1000000.0) + start.tv_usec)) / 1000000.0);
}
}
void ABoxDlg::onStubCol( wxListEvent& event ) {
if(m_SortCol == event.GetColumn())
ms_Sort = !ms_Sort;
else
ms_Sort = true;
m_SortCol = event.GetColumn();
initResult();
}
int main(int argc, char *argv[])
{
int i;
struct timeval start, finish;
double error;
pthread_attr_t attr;
pthread_t *tid;
int *id;
if (argc < 2) {
fprintf(stderr, "usage: %s <matrixfile>\n", argv[0]);
exit(-1);
}
// for getting the threads
if(argc == 3) {
task_num = strtol(argv[2], NULL, 10);
}
nsize = initMatrix(argv[1]);
initRHS(nsize);
initResult(nsize);
// create threads
id = (int *) malloc (sizeof (int) * task_num);
tid = (pthread_t *) malloc (sizeof (pthread_t) * task_num);
if (!id || !tid)
errexit ("out of shared memory");
pthread_attr_init (&attr);
pthread_attr_setscope (&attr, PTHREAD_SCOPE_SYSTEM);
for (i = 1; i < task_num; i++) {
id[i] = i;
pthread_create (&tid[i], &attr, work_thread, &id[i]);
}
id[0]=0;
work_thread(&id[0]);
// wait for all threads to finish
for (i = 1; i < task_num; i++)
pthread_join (tid[i], NULL);
solveGauss(nsize);
error = 0.0;
for (i = 0; i < nsize; i++) {
double error__ = (X__[i]==0.0) ? 1.0 : fabs((X[i]-X__[i])/X__[i]);
if (error < error__) {
error = error__;
}
}
fprintf(stdout, "Error: %e\n", error);
}
int init()
{
int x1, x2;
matrix1 = readMatrix("in1.txt", &m, &x1);
if(matrix1 == NULL)
return MATRIX1_ERROR;
matrix2 = readMatrix("in2.txt", &x2, &n);
if(matrix2 == NULL)
return MATRIX2_ERROR;
if(x1 != x2)
return INCOMPATIBLE_ERROR;
x = x1;
resultMatrix = initResult();
return SUCCESS;
}
void ABoxDlg::onShowPath( wxCommandEvent& event ) {
wABox.setshowpath(m_Ini, m_ShowPath->IsChecked()?True:False);
initResult();
}
void ABoxDlg::event(iONode node) {
char* s = NodeOp.base.toString(node);
TraceOp.trc( "aboxdlg", TRCLEVEL_INFO, __LINE__, 9999, "event: %.256s", s );
StrOp.free(s);
if( wDataReq.getcmd(node) == wDataReq.abox_addlink ) {
const char* uid = wDataReq.getid(node);
iONode direntry = wDataReq.getdirentry(node);
if( direntry != NULL && wDirEntry.getfileentry(direntry) != NULL ) {
iONode fileentry = wDirEntry.getfileentry(direntry);
TraceOp.trc( "aboxdlg", TRCLEVEL_INFO, __LINE__, 9999, "uid=%s for [%s] added=[%s]", uid, wFileEntry.getfname(fileentry), m_AddedFilename );
/*
char* s = StrOp.fmt("%s:\nUID=%s\n%s=%s", wxGetApp().getCMsg("upload"), uid, wxGetApp().getCMsg("file"), wFileEntry.getfname(fileentry) );
int action = wxMessageDialog( this, wxString(s,wxConvUTF8), _T("Rocrail"), wxOK ).ShowModal();
StrOp.free(s);
*/
if( StrOp.equals( wFileEntry.getfname(fileentry), m_AddedFilename ) ) {
TraceOp.trc( "aboxdlg", TRCLEVEL_INFO, __LINE__, 9999, "start upload of [%s]...", wFileEntry.getfname(fileentry) );
StrOp.copy( m_AddedUID, uid );
iONode cmd = NodeOp.inst( wDataReq.name(), NULL, ELEMENT_NODE );
wDataReq.setcmd( cmd, wDataReq.abox_filedata );
wDataReq.setid( cmd, uid );
wDataReq.setcategory(cmd, wFileEntry.getcategory(fileentry));
wDataReq.setfilename(cmd, FileOp.ripPath(m_AddedFilename));
wDataReq.setdatapart(cmd, 0);
if( readDataBlock(wFileEntry.getfname(fileentry), cmd, 0) ) {
m_AddedFilename[0] = '\0';
m_AddedUID[0] = '\0';
EnableDlg(true);
}
wxGetApp().sendToRocrail( cmd );
cmd->base.del(cmd);
}
}
}
else if( wDataReq.getcmd(node) == wDataReq.abox_getdata ) {
TraceOp.trc( "aboxdlg", TRCLEVEL_INFO, __LINE__, 9999, "getdata..." );
/*
* <datareq cmd="11" id="20150816093859345" category="Zomaar" filename="paspoort-2014.jpeg" datapart="0" controlcode="" server="infw5601F5A0" data="FFD8FFE0
*/
if( StrOp.equals(m_DownloadUID, wDataReq.getid(node)) ) {
if(wDataReq.getrc(node) != 0) {
m_DownloadFilename[0] = '\0';
m_DownloadUID[0] = '\0';
m_DownloadPart = -1;
m_labDownloadState->SetLabel(wxT(""));
EnableDlg(true);
int action = wxMessageDialog( this, wxT("Error getting file data!"), _T("Rocrail"), wxOK | wxICON_EXCLAMATION ).ShowModal();
return;
}
wxString tempdir = wxFileName::GetTempDir();
char* filepath = StrOp.fmt("%s%c%s", (const char*)tempdir.mb_str(wxConvUTF8), SystemOp.getFileSeparator(), wDataReq.getfilename(node) );
iOFile f = FileOp.inst(filepath, OPEN_APPEND);
if( f != NULL ) {
const char* byteStr = wDataReq.getdata(node);
byte* filedata = StrOp.strToByte( byteStr );
int len = StrOp.len(byteStr)/2;
FileOp.write(f, (char*)filedata, len);
FileOp.base.del(f);
freeMem(filedata);
}
if( !wDataReq.isack(node) ) {
// get next part
iONode cmd = NodeOp.inst( wDataReq.name(), NULL, ELEMENT_NODE );
wDataReq.setcmd( cmd, wDataReq.abox_getdata );
wDataReq.setid( cmd, wDataReq.getid(node) );
wDataReq.setcategory( cmd, wDataReq.getcategory(node) );
wDataReq.setfilename(cmd, wDataReq.getfilename(node) );
m_DownloadPart++;
m_labDownloadState->SetLabel(wxString::Format(wxT("%d"), m_DownloadPart));
wDataReq.setdatapart(cmd, m_DownloadPart);
wxGetApp().sendToRocrail( cmd );
cmd->base.del(cmd);
}
else {
m_DownloadFilename[0] = '\0';
m_DownloadUID[0] = '\0';
m_DownloadPart = -1;
m_labDownloadState->SetLabel(wxT(""));
EnableDlg(true);
executeStub(filepath);
}
StrOp.free(filepath);
}
}
else if( wDataReq.getcmd(node) == wDataReq.abox_filedata ) {
TraceOp.trc( "aboxdlg", TRCLEVEL_INFO, __LINE__, 9999, "filedata..." );
// <datareq cmd="10" id="20150816081059871" category="Zomaar" datapart="0" dataparts="33" totalsize="794973" data=""
// controlcode="" server="infw2368A4D0" ack="true"/>
if( wDataReq.isack(node) ) {
if( StrOp.equals(m_AddedUID, wDataReq.getid(node) ) ) {
iONode cmd = NodeOp.inst( wDataReq.name(), NULL, ELEMENT_NODE );
wDataReq.setcmd( cmd, wDataReq.abox_filedata );
wDataReq.setid( cmd, wDataReq.getid(node) );
wDataReq.setcategory(cmd, wDataReq.getcategory(node));
wDataReq.setfilename(cmd, wDataReq.getfilename(node));
wDataReq.setdatapart(cmd, wDataReq.getdatapart(node)+1);
m_labUploadState->SetLabel(wxString::Format(wxT("%d"), wDataReq.getdatapart(cmd)));
if( readDataBlock(m_AddedFilename, cmd, wDataReq.getdatapart(cmd)) ) {
m_AddedFilename[0] = '\0';
m_AddedUID[0] = '\0';
m_labUploadState->SetLabel(wxT(""));
EnableDlg(true);
}
wxGetApp().sendToRocrail( cmd );
cmd->base.del(cmd);
}
}
}
else if( wDataReq.getcmd(node) == wDataReq.abox_getcategories ) {
m_ReadOnly = wDataReq.isreadonly(node)?true:false;
if( m_Enable )
m_Add->Enable(!m_ReadOnly);
m_Category->Clear();
wxString findtext = m_FindText->GetValue();
m_FindText->Clear();
m_FindText->SetValue(findtext);
iOStrTok tok = StrTokOp.inst( wDataReq.getcategory( node ), ',' );
while( StrTokOp.hasMoreTokens(tok) ) {
const char* category = StrTokOp.nextToken( tok );
m_Category->Append( wxString(category,wxConvUTF8) );
m_FindText->Append( wxString(category,wxConvUTF8) );
}
}
else if( wDataReq.getcmd(node) == wDataReq.abox_find ) {
m_SelectedStub = wxNOT_FOUND;
m_Open->Enable(false);
m_Modify->Enable(false);
m_Delete->Enable(false);
m_Stubs->DeleteAllItems();
clearStubList();
char* s = NodeOp.base.toString(node);
TraceOp.trc( "aboxdlg", TRCLEVEL_INFO, __LINE__, 9999, "found [%s]", s );
StrOp.free(s);
iONode stub = NodeOp.findNode( node, "stub");
while( stub != NULL ) {
iONode clone = (iONode)NodeOp.base.clone(stub);
ListOp.add(m_StubList, (obj)clone);
stub = NodeOp.findNextNode( node, stub);
}
initResult();
if(wDataReq.istoomanyhits(node)) {
// Show warning.
int action = wxMessageDialog( this, wxGetApp().getMsg("toomanyhits"), _T("Rocrail"), wxOK | wxICON_EXCLAMATION ).ShowModal();
}
}
}
int main(int argc, char **argv){
setbuf(stdout,0);
int num = sysconf(_SC_NPROCESSORS_CONF);
printf("system has %d processor(s)\n", num);
cpu_set_t mask,get;
CPU_ZERO(&mask);
CPU_SET(0, &mask);
int i,j ;
int ite = 0;
ite_converge = 0;
hrtimer_converge = 0.0;
int retcode,allone;
void *retval;
double hrtimer_tmp,hrtime_start,hrtime_end,l2normtotal;
if ( argc != 4) {printf("[usage] SOR <matrix> <thread> <w>\n"); return 0;}
if (pthread_setaffinity_np(pthread_self(), sizeof(mask), &mask) < 0) {
fprintf(stderr, "set thread affinity failed\n");
}
thread_num = atoi(argv[2]);
w = atof(argv[3]);
nsize = initMatrix(argv[1]);
initRHS(nsize);
initResult(nsize);
int workload = (int)ceil((double)nsize/(double)thread_num) ;
pthread_t thread[thread_num];
int *seq = (int *)malloc(thread_num*sizeof(int));
for(i=0;i<thread_num;i++)
seq[i] = i;
shuffle(seq,thread_num);
for(j=1;j<thread_num;j++)
{
retcode = pthread_create(&thread[j], NULL, task, new int(j));
if (retcode != 0)
fprintf (stderr, "create thread failed %d\n", retcode);
}
sense_reverse_barrier(0); // start at the same time
hrtime_start = gethrtime_x86();
printf("Finish reading file. Start computing...\n");
while(converge == 0)
{
ticket_acquire(&flags[0].tlock);
if(flags[0].test == 1)
{
solve(0,workload-1,0,1);
flags[0].flag = 0;
flags[0].test = 0;
sense_reverse_barrier(0);
}
else
{
hrtimer_tmp = gethrtime_x86();
ite ++;
solve(0,workload-1,0,0);
hrtimer_compute[0] += gethrtime_x86() - hrtimer_tmp;
}
if(token == 0 && flags[0].flag == 1&&converge==0)
{
//observe all flags
allone = 1;
for(j=0;j<thread_num;j++)
{
allone = allone & flags[j].flag;
}
if(allone == 1)
{
ite_converge ++;
shuffle(seq,thread_num);
for(i=0;i<thread_num;i++)
if(seq[i] != 0)
{
ticket_acquire(&flags[seq[i]].tlock);
flags[seq[i]].test = 1;
ticket_release(&flags[seq[i]].tlock);
}
hrtimer_tmp = gethrtime_x86();
solve(0,workload-1,0,1);
hrtimer_converge += gethrtime_x86() - hrtimer_tmp;
sense_reverse_barrier(0);
l2normtotal = 0.0;
for(j=0;j<thread_num;j++)
l2normtotal += flags[j].l2norm;
if(l2normtotal < EPS)
{
converge = 1;
}
else
{
token = (token + 1) % thread_num;
}
}
}
ticket_release(&flags[0].tlock);
}
printf ("#%d finished\n",0);
iteration[0] = ite;
for(j=1;j<thread_num;j++)
{
retcode = pthread_join(thread[j], &retval);
iteration[j] = *(int *)retval;
if (retcode != 0)
fprintf (stderr, "join failed %d\n", retcode);
}
hrtime_end = gethrtime_x86();
CPU_ZERO(&get);
if (pthread_getaffinity_np(pthread_self(), sizeof(get), &get) < 0)
fprintf(stderr, "get thread affinity failed\n");
for (j = 0; j < 24; j++)
if (CPU_ISSET(j, &get))
printf("thread %d is running in processor %d sched_getcpu %d\n", -1, j,sched_getcpu());
#ifdef VERIFY
FILE *res;
res = fopen ("ASOR result", "w");
for (i = 0; i < nsize; i++) {
fprintf (res, "[%d] = %lf\n", i+1, X[i]);
}
#endif
double total = hrtime_end-hrtime_start;
printf("Total time:%.16lfs\n",total);
printf("Computation time:%.16lfs\n",max(thread_num,hrtimer_compute));
printf("Convergence time:%.16lfs\n",hrtimer_converge);
printf("Syn time:%.16lfs\n",total-max(thread_num,hrtimer_compute)-hrtimer_converge);
for(i=0;i<thread_num;i++)
printf("Iteration[%d]: %d\t\tComputation time:%.16lfs\n",i,iteration[i],hrtimer_compute[i]);
printf("#Converge iteration:%d\n",ite_converge);
printf("sizeof(cflag) = %lu\n",sizeof(cflag));
return 0;
}
int main(int argc, char *argv[])
{
int i;
struct timeval start, finish;
double error;
pthread_attr_t attr;
pthread_t *tid;
int *id;
//Ensure that the program takes two parameters
// first param is the input matrix and the second is the number of processors for the parallel run
if (argc != 3) {
fprintf(stderr, "usage: %s <matrixfile> <#ofProcesses>\n", argv[0]);
exit(-1);
}
task_num = atoi(argv[2]);
nsize = initMatrix(argv[1]);
initRHS(nsize);
initResult(nsize);
gettimeofday(&start, 0);
// create threads
id = (int *) malloc (sizeof (int) * task_num);
tid = (pthread_t *) malloc (sizeof (pthread_t) * task_num);
if (!id || !tid)
errexit ("out of shared memory");
pthread_attr_init (&attr);
pthread_attr_setscope (&attr, PTHREAD_SCOPE_SYSTEM);
for (i = 1; i < task_num; i++) {
id[i] = i;
pthread_create (&tid[i], &attr, work_thread, &id[i]);
}
id[0]=0;
work_thread(&id[0]);
// wait for all threads to finish
for (i = 1; i < task_num; i++)
pthread_join (tid[i], NULL);
gettimeofday(&finish, 0);
solveGauss(nsize);
fprintf(stdout, "Time: %f seconds\n", (finish.tv_sec - start.tv_sec) + (finish.tv_usec - start.tv_usec)*0.000001);
error = 0.0;
for (i = 0; i < nsize; i++) {
double error__ = (X__[i]==0.0) ? 1.0 : fabs((X[i]-X__[i])/X__[i]);
if (error < error__) {
error = error__;
}
}
fprintf(stdout, "Error: %e\n", error);
//File output
// float timeOutput = ((finish.tv_sec - start.tv_sec) + (finish.tv_usec - start.tv_usec)*0.000001);
// FILE *pthreadTimingFile = fopen("pthread1Timing.txt", "a");
// fprintf(pthreadTimingFile, "%s %d %f %G\n", argv[1], task_num, timeOutput, error);
// fclose(pthreadTimingFile);
return 0;
}
