const char* CppSQLite3Table::fieldName(int nCol)
{
checkResults();
if (nCol < 0 || nCol > mnCols-1)
{
throw CppSQLite3Exception(CPPSQLITE_ERROR,"Invalid field index requested",DONT_DELETE_MSG);
}
return mpaszResults[nCol];
}
void show_sched_param(pthread_attr_t *a) {
int rc = 0;
struct sched_param param;
rc = mythread_attr_getschedparam(a, ¶m);
checkResults("pthread_attr_getschedparam()\n", rc);
printf("param.sched_priority = %d\n",
param.__sched_priority);
return;
}
void *theThread(void *parm)
{
int rc;
threadSpecific_data_t *gData;
printf("Thread %.8x %.8x: Entered\n", pthread_self());
gData = (threadSpecific_data_t *)parm;
rc = pthread_setspecific(threadSpecificKey, gData);
checkResults("pthread_setspecific()\n", rc);
foo();
return NULL;
}
void *theThread(void *parm)
{
int rc;
rc = pthread_mutex_lock(&mutex);
printf("Thread %u: Entered\n",(unsigned int ) pthread_self());
checkResults("pthread_mutex_lock()\n", rc);
/********** Critical Section *******************/
printf("Thread %u: Start critical section, holding lock\n",(unsigned int ) pthread_self());
/* Access to shared data goes here */
sharedData++;
sharedData2--;
printf("Thread %u: End critical section, release lock\n", (unsigned int )pthread_self());
/********** Critical Section *******************/
rc = pthread_mutex_unlock(&mutex);
checkResults("pthread_mutex_unlock()\n", rc);
return NULL;
}
const char* CppSQLite3Table::fieldValue(int nField)
{
checkResults();
if (nField < 0 || nField > mnCols-1)
{
IwError(("Invalid field index requested"));
}
int nIndex = (mnCurrentRow*mnCols) + mnCols + nField;
return mpaszResults[nIndex];
}
const char* CppSQLite3Table::fieldValue(int nField)
{
checkResults();
if (nField < 0 || nField > mnCols-1)
{
throw CppSQLite3Exception(CPPSQLITE_ERROR,"Invalid field index requested",DONT_DELETE_MSG);
}
int nIndex = (mnCurrentRow*mnCols) + mnCols + nField;
return mpaszResults[nIndex];
}
void CppSQLite3Table::setRow(int nRow)
{
checkResults();
if (nRow < 0 || nRow > mnRows-1)
{
throw CppSQLite3Exception(CPPSQLITE_ERROR,
"Invalid row index requested",
DONT_DELETE_MSG);
}
mnCurrentRow = nRow;
}
int main(int argc, char **argv)
{
int rc=0;
int i;
pthread_t threadid[NTHREADS];
printf("Enter Testcase - %s\n", argv[0]);
printf("Create %d threads\n", NTHREADS);
for(i=0; i<NTHREADS; ++i) {
rc = pthread_create(&threadid[i], NULL, threadfunc, NULL);
checkResults("pthread_create()\n", rc);
}
sleep(5); /* Sleep is not a very robust way to serialize threads */
rc = pthread_mutex_lock(&mutex);
checkResults("pthread_mutex_lock()\n", rc);
/* The condition has occured. Set the flag and wake up any waiting threads */
conditionMet = 1;
printf("Wake up all waiting threads...\n");
rc = pthread_cond_broadcast(&cond);
checkResults("pthread_cond_broadcast()\n", rc);
rc = pthread_mutex_unlock(&mutex);
checkResults("pthread_mutex_unlock()\n", rc);
printf("Wait for threads and cleanup\n");
for (i=0; i<NTHREADS; ++i) {
rc = pthread_join(threadid[i], NULL);
checkResults("pthread_join()\n", rc);
}
pthread_cond_destroy(&cond);
pthread_mutex_destroy(&mutex);
printf("Main completed\n");
return 0;
}
void *threadfunc(void *parm)
{
int rc;
while (1) {
/* Usually worker threads will loop on these operations */
rc = pthread_mutex_lock(&g_mutex);
checkResults("pthread_mutex_lock()\n", rc);
while (not workToDo) {
printf("Thread blocked\n");
rc = pthread_cond_wait(&g_cond, &g_mutex);
checkResults("pthread_cond_wait()\n", rc);
}
printf("Thread awake, finish work!\n");
/* Under protection of the lock, complete or remove the work */
/* from whatever worker queue we have. Here its simply a flag */
workToDo = 0;
rc = pthread_mutex_unlock(&g_mutex);
checkResults("pthread_mutex_lock()\n", rc);
}
return NULL;
}
const char* CppSQLite3Table::fieldValue(const char* szField)
{
checkResults();
if (szField)
{
for (int nField = 0; nField < mnCols; nField++)
{
if (strcmp(szField, mpaszResults[nField]) == 0)
{
int nIndex = (mnCurrentRow*mnCols) + mnCols + nField;
return mpaszResults[nIndex];
}
}
}
throw CppSQLite3Exception(CPPSQLITE_ERROR,"Invalid field name requested",DONT_DELETE_MSG);
}
const char* CppSQLite3Table::fieldValue(const char* szField)
{
checkResults();
if (szField)
{
for (int nField = 0; nField < mnCols; nField++)
{
if (strcmp(szField, mpaszResults[nField]) == 0)
{
int nIndex = (mnCurrentRow*mnCols) + mnCols + nField;
return mpaszResults[nIndex];
}
}
}
IwError(("Invalid field name requested"));
return 0;
}
bool CppSQLite3Table::fieldIsNull(int nField)
{
checkResults();
return (fieldValue(nField) == 0);
}
int CppSQLite3Table::numRows()
{
checkResults();
return mnRows;
}
int CppSQLite3Table::numFields()
{
checkResults();
return mnCols;
}
int main (int argc,char * argv[])
{
pthread_t *threads;
pthread_attr_t pta;
int iteration,THREADS,ret_count;
char * CLASS;
double time_start, time_end;
struct timeval tv;
struct timezone tz;
int counter;
printf("\n\t\t Objective : To compute the minimum of a list of numbers using mutex. \n ");
if( argc != 3 ) // check command line arguments
{
printf("\t\t Very Few Arguments\n ");
printf("\t\t Syntax : exec <NumElements> <Threads>\n");
exit(-1);
} // user input
else {
NumElements =atoi( argv[1]);
THREADS = atoi(argv[2]);
}
NumThreads = THREADS;
printf("\n\t\t Array Size : %ld",NumElements);
printf("\n\t\t Threads : %d",NumThreads);
printf("\n");
if (NumThreads < 1 )
{
printf("\n Number of threads must be greater than zero. Aborting ...\n");
return 0;
}
if ((NumThreads != 1) && (NumThreads != 2) && (NumThreads != 4) && (NumThreads != 8))
{
printf("\n Number of Threads must be 1 or 2 or 4 or 8. Aborting ...\n");
return 0;
}
if ( ( NumElements % NumThreads ) != 0 )
{
printf("\n Number of threads not a factor of Integer List size. Aborting.\n");
return 0 ;
}
partial_list_size = NumElements / NumThreads; // define partial size for each thread
list = (long int *)malloc(sizeof(long int) * NumElements);
for(counter = 0 ; counter < NumElements ; counter++)
{
srand48((unsigned int)NumElements); // random number generation
list[counter] = (rand()%1000)+1.0;
}
threads = (pthread_t *)malloc(sizeof(pthread_t)*NumThreads);
minimum_value = list[0];
ret_count=pthread_mutex_init(&minimum_value_lock, NULL);
checkResults("pthread_mutex_init()\n", ret_count);
ret_count=pthread_attr_init(&pta);
checkResults("pthread_attr_init()\n", ret_count);
pthread_attr_setscope(&pta,PTHREAD_SCOPE_SYSTEM);
gettimeofday(&tv, &tz);
time_start = (double)tv.tv_sec + (double)tv.tv_usec / 1000000.0;
for(counter = 0 ; counter < NumThreads ; counter++)
{
ret_count=pthread_create(&threads[counter],&pta,(void *(*) (void *)) find_min_rwlock,(void *) (counter+1)); // call find_min_rwlock function
checkResults("pthread_create()\n", ret_count);
}
gettimeofday(&tv, &tz);
time_end = (double)tv.tv_sec + (double)tv.tv_usec / 1000000.0;
for(counter = 0 ; counter < NumThreads ; counter++)
{
ret_count=pthread_join(threads[counter],NULL);
checkResults("pthread_join()\n", ret_count);
}
ret_count=pthread_attr_destroy(&pta);
checkResults("pthread_attr_destroy()\n", ret_count);
printf("\n\t\t..........................................................................\n");
printf("\n\t\t Minimum Value found in the Integer list : %d",minimum_value);
printf("\n\t\t Time Taken in Seconds (T) : %lf Seconds",( time_end - time_start));
printf("\n\t\t..........................................................................\n");
free(list);
free(threads);
return 0;
}
int main(int argc, char *argv[]) {
int i;
#ifdef EPETRA_MPI
// Initialize MPI
MPI_Init(&argc,&argv);
Epetra_MpiComm comm(MPI_COMM_WORLD);
#else
Epetra_SerialComm comm;
#endif
// Uncomment to debug in parallel int tmp; if (comm.MyPID()==0) cin >> tmp; comm.Barrier();
bool verbose = false;
bool veryVerbose = false;
// Check if we should print results to standard out
if (argc>1) if (argv[1][0]=='-' && argv[1][1]=='v') verbose = true;
if (!verbose) comm.SetTracebackMode(0); // This should shut down any error traceback reporting
if (verbose) cout << comm << endl << flush;
bool verbose1 = verbose;
if (verbose) verbose = (comm.MyPID()==0);
int nx = 4;
int ny = comm.NumProc()*nx; // Scale y grid with number of processors
// Create funky stencil to make sure the matrix is non-symmetric (transpose non-trivial):
// (i-1,j-1) (i-1,j )
// (i ,j-1) (i ,j ) (i ,j+1)
// (i+1,j-1) (i+1,j )
int npoints = 2;
int xoff[] = {-1, 0, 1, -1, 0, 1, 0};
int yoff[] = {-1, -1, -1, 0, 0, 0, 1};
Epetra_Map * map;
Epetra_CrsMatrix * A;
Epetra_Vector * x, * b, * xexact;
Trilinos_Util_GenerateCrsProblem(nx, ny, npoints, xoff, yoff, comm, map, A, x, b, xexact);
if (verbose)
cout << "npoints = " << npoints << " nx = " << nx << " ny = " << ny << endl ;
if (verbose && nx<6 ) {
cout << *A << endl;
cout << "B = " << endl << *b << endl;
}
// Construct linear problem object
Epetra_LinearProblem origProblem(A, x, b);
Epetra_LinearProblem *redistProblem;
Epetra_Time timer(comm);
// Construct redistor object, use all processors and replicate full problem on each
double start = timer.ElapsedTime();
Epetra_LinearProblemRedistor redistor(&origProblem, comm.NumProc(), true);
if (verbose) cout << "\nTime to construct redistor = "
<< timer.ElapsedTime() - start << endl;
bool ConstructTranspose = true;
bool MakeDataContiguous = true;
start = timer.ElapsedTime();
redistor.CreateRedistProblem(ConstructTranspose, MakeDataContiguous, redistProblem);
if (verbose) cout << "\nTime to create redistributed problem = "
<< timer.ElapsedTime() - start << endl;
// Now test output of redistor by performing matvecs
int ierr = 0;
ierr += checkResults( ConstructTranspose, &redistor, &origProblem,
redistProblem, verbose);
// Now change values in original rhs and test update facility of redistor
// Multiply b by 2
double Value = 2.0;
b->Scale(Value); // b = 2*b
redistor.UpdateRedistRHS(b);
if (verbose) cout << "\nTime to update redistributed RHS = "
<< timer.ElapsedTime() - start << endl;
ierr += checkResults( ConstructTranspose, &redistor,
&origProblem, redistProblem, verbose);
// Now change values in original matrix and test update facility of redistor
#define CREATE_CONST_MATRIX
#ifdef CREATE_CONST_MATRIX
// The easiest way that I could find to change the matrix without EPETRA_CHK_ERRs
A->PutScalar(13.0);
#else
// This has no effect on matrices, such as when nx = 4, that have no
// diagonal entries. However, it does cause many EPETRA_CHK_ERR prints.
// Add 2 to the diagonal of each row
for (i=0; i< A->NumMyRows(); i++) {
// for (i=0; i < 1; i++)
cout << " i = " << i ;
A->SumIntoMyValues(i, 1, &Value, &i);
}
#endif
start = timer.ElapsedTime();
redistor.UpdateRedistProblemValues(&origProblem);
if (verbose) cout << "\nTime to update redistributed problem = "
<< timer.ElapsedTime() - start << endl;
ierr += checkResults(ConstructTranspose, &redistor, &origProblem, redistProblem, verbose);
delete A;
delete b;
delete x;
delete xexact;
delete map;
#ifdef EPETRA_MPI
MPI_Finalize();
#endif
return ierr;
}
// EXCEPTIONS: CProcessFailure
void CPCPATRDllProcess::processANAFile(CProcessStatus& status)
{
CModelFilesSet* pSourceMFS = status.getInputMFS();
ASSERTX(pSourceMFS);
CModelFilesSet* pTargetMFS = status.getOutputMFS(); // may be null
try
{
IPatrParser *pPatr;
CLSID clsid;
HRESULT hr;
#ifndef hab245
CString sBase;
sBase.Format(_T("%s%d"), (LPCTSTR)getTempFileNameBase(status), status.m_iProcNumber);
#else // hab245
CString sBase = getTempFileNameBase(status);
#endif // hab245
m_sOutPath = status.makeTempPath(sBase, _T(".ana") );
m_sLOGPath = status.makeTempPath(sBase, _T(".log") );
m_sInPath = status.sANAPath;
m_sOutPath.deleteFile();
m_sLOGPath.deleteFile();
hr = CoInitialize(NULL);
#ifndef hab245
checkResults(hr, _T("coinitialize"), (IPatrParser *)NULL);
hr = CLSIDFromProgID(L"SIL.CARLA.PatrParser.1", &clsid);
checkResults(hr, _T("get CLSID"), (IPatrParser *)NULL);
hr = CoCreateInstance(clsid, NULL, CLSCTX_ALL,
IID_IPatrParser, (void **) &pPatr);
checkResults(hr, _T("do CoCreateInstance"), (IPatrParser *)NULL);
CString s = m_sLOGPath.getQuotedPath();
CString sLogFile = s.Mid(1, s.GetLength()-2);
hr = pPatr->OpenLog(sLogFile.AllocSysString());
checkResults(hr, _T("set log file"), (IPatrParser *)NULL);
hr = pPatr->Clear();
checkResults(hr, _T("clear"), pPatr);
hr = pPatr->put_CommentChar(status.getInputMFS()->getCommentChar());
checkResults(hr, _T("set comment character"), pPatr);
hr = pPatr->put_AmplePropertyIsFeature(m_bAmplePropertyIsFeature);
checkResults(hr, _T("set AmplePropertyIsFeature"), pPatr);
hr = pPatr->put_PromoteDefaultAtoms(m_bPromoteDefaultAtomicValues);
checkResults(hr, _T("set PromoteDefaultAtoms"), pPatr);
hr = pPatr->put_Failures(m_bShowFailures);
checkResults(hr, _T("set Failures"), pPatr);
hr = pPatr->put_DisplayFeatures(m_bDisplayFeatures);
checkResults(hr, _T("set DisplayFeatures"), pPatr);
hr = pPatr->put_TopFeatureOnly(!m_bAllFeatures);
checkResults(hr, _T("set TopFeatureOnly"), pPatr);
hr = pPatr->put_Gloss(m_bDisplayGloss);
checkResults(hr, _T("set Gloss"), pPatr);
hr = pPatr->put_TrimEmptyFeatures(m_bTrimEmptyFeatures);
checkResults(hr, _T("set TrimEmptyFeatures"), pPatr);
hr = pPatr->put_Unification(m_bPerformUnification);
checkResults(hr, _T("set Unification"), pPatr);
hr = pPatr->put_WriteAmpleParses(m_bWriteAmpleParses);
checkResults(hr, _T("set WriteAmpleParses"), pPatr);
hr = pPatr->put_MaxAmbiguity(m_uiMaxAmbiguities);
checkResults(hr, _T("set MaxAmbiguity"), pPatr);
#ifndef rde273
hr = pPatr->put_CodePage(GetCSAcp());
#endif // rde273
hr = pPatr->put_SentenceFinalPunctuation(
m_sSentenceFinalPunctuation.AllocSysString());
checkResults(hr, _T("set SentenceFinalPunctuation"), pPatr);
hr = pPatr->put_TimeLimit(m_uiTimeLimit);
checkResults(hr, _T("set TimeLimit"), pPatr);
hr = pPatr->put_TreeDisplay(m_iTreeDisplayFormat);
checkResults(hr, _T("set TreeDisplay"), pPatr);
hr = pPatr->put_FlatFeatureDisplay(m_bFlatFeatureDisplay);
checkResults(hr, _T("set FlatFeatureDisplay"), pPatr);
#ifndef hab262
if (m_iRootGlossSetting > 0)
m_iRootGlossSetting++; // need to adjust value to get correct setting
hr = pPatr->put_RootGlossFeature(m_iRootGlossSetting);
if (m_iRootGlossSetting > 0)
m_iRootGlossSetting--; // adjust the value back
checkResults(hr, _T("set RootGlossFeature"), pPatr);
#endif // hab262
hr = pPatr->LoadGrammarFile(m_sGrammarFileName.AllocSysString());
checkResults(hr, _T("load grammar file.\nSee Log File."), pPatr);
s = m_sOutPath.getQuotedPath();;
CString sOutFile = s.Mid(1, s.GetLength()-2);
hr = pPatr->DisambiguateAnaFile(m_sInPath.getShortPath().AllocSysString(),
sOutFile.AllocSysString());
checkResults(hr, _T("disambiguate ANA file"), pPatr);
hr = pPatr->CloseLog();
checkResults(hr, _T("close log file"), pPatr);
#else // hab245
checkResults(hr, _T("coinitialize"));
hr = CLSIDFromProgID(L"SIL.CARLA.PatrParser.1", &clsid);
checkResults(hr, "get CLSID");
hr = CoCreateInstance(clsid, NULL, CLSCTX_ALL,
IID_IPatrParser, (void **) &pPatr);
checkResults(hr, "do CoCreateInstance");
hr = pPatr->Clear();
checkResults(hr, "clear");
hr = pPatr->put_CommentChar(status.getInputMFS()->getCommentChar());
checkResults(hr, "set comment character");
hr = pPatr->put_AmplePropertyIsFeature(m_bAmplePropertyIsFeature);
checkResults(hr, "set AmplePropertyIsFeature");
hr = pPatr->put_PromoteDefaultAtoms(m_bPromoteDefaultAtomicValues);
checkResults(hr, "set PromoteDefaultAtoms");
hr = pPatr->put_Failures(m_bShowFailures);
checkResults(hr, "set Failures");
hr = pPatr->put_DisplayFeatures(m_bDisplayFeatures);
checkResults(hr, "set DisplayFeatures");
hr = pPatr->put_TopFeatureOnly(!m_bAllFeatures);
checkResults(hr, "set TopFeatureOnly");
hr = pPatr->put_Gloss(m_bDisplayGloss);
checkResults(hr, "set Gloss");
hr = pPatr->put_TrimEmptyFeatures(m_bTrimEmptyFeatures);
checkResults(hr, "set TrimEmptyFeatures");
hr = pPatr->put_Unification(m_bPerformUnification);
checkResults(hr, "set Unification");
hr = pPatr->put_WriteAmpleParses(m_bWriteAmpleParses);
checkResults(hr, "set WriteAmpleParses");
hr = pPatr->put_MaxAmbiguity(m_uiMaxAmbiguities);
checkResults(hr, "set MaxAmbiguity");
hr = pPatr->put_SentenceFinalPunctuation(
m_sSentenceFinalPunctuation.AllocSysString());
checkResults(hr, "set SentenceFinalPunctuation");
hr = pPatr->put_TimeLimit(m_uiTimeLimit);
checkResults(hr, "set TimeLimit");
hr = pPatr->put_TreeDisplay(m_iTreeDisplayFormat);
checkResults(hr, "set TreeDisplay");
hr = pPatr->put_FlatFeatureDisplay(m_bFlatFeatureDisplay);
checkResults(hr, "set FlatFeatureDisplay");
hr = pPatr->LoadGrammarFile(m_sGrammarFileName.AllocSysString());
checkResults(hr, "load grammar file");
CString s = m_sLOGPath.getQuotedPath();
CString sLogFile = s.Mid(1, s.GetLength()-2);
hr = pPatr->OpenLog(sLogFile.AllocSysString());
checkResults(hr, "set log file");
s = m_sOutPath.getQuotedPath();;
CString sOutFile = s.Mid(1, s.GetLength()-2);
hr = pPatr->DisambiguateAnaFile(m_sInPath.getShortPath().AllocSysString(),
sOutFile.AllocSysString());
checkResults(hr, "disambiguate ANA file");
hr = pPatr->CloseLog();
checkResults(hr, "close log file");
#endif // hab245
pPatr->Release();
pPatr = NULL;
CoUninitialize();
waitForCompletion(&status, &m_sLOGPath, &m_sOutPath);
// register ana output so the user can view it
CString sShortDesc, sLongDesc, sTabLabel;
CCarlaLanguage *pOutANALang=NULL;
sShortDesc = "PC-PATR Disambiguated ANA";
sLongDesc = "ANA of the source-language file, after filtering through PC-PATR Syntactic Grammar rules.";
sTabLabel = "PATR-Disamb ANA";
pOutANALang = status.getInputLang();
CResultStreamFile* anaStream =
new CResultStreamFile(
new CResultStreamDescriptor(this,
sShortDesc, //short description
sLongDesc, // long description
sTabLabel // tab label
),
m_sOutPath.getFullPath(),
pOutANALang);
status.registerResultStream(anaStream);
registerLog(status, m_sLOGPath, status.getInputLang());
status.sANAPath = m_sOutPath;
}
catch(CProcessFailure failure)
{
registerLog(status, m_sLOGPath);
throw(failure);
}
catch(CString sError)
{
registerLog(status, m_sLOGPath);
throw(CProcessFailure(this, sError));
}
}
void Database_PostgreSQL::endSave()
{
checkResults(PQexec(m_conn, "COMMIT;"));
}
bool CppSQLite3Table::fieldIsNull(const char* szField)
{
checkResults();
return (fieldValue(szField) == 0);
}
void benchmark(int samples = 1)
{
std::cout << std::string(70, '-') << std::endl
<< getTypeName<T>()
<< " conversion (size = " << sizeof(T)*8 << " bits) - ";
std::vector<std::string> strings = generateRandStrings<T>(numNumbers);
std::cout << std::string(70, '-') << std::endl;
////
std::cout << " stringstream: ";
if (checkResults(strings, stringstream_convert<T>))
time_run<T>(strings, stringstream_convert<T>, samples);
else
std::cout << "FAILURE" << std::endl;
////
std::cout << " sscanf: ";
if (checkResults(strings, sscanf_convert<T>))
time_run<T>(strings, sscanf_convert<T>, samples);
else
std::cout << "FAILURE" << std::endl;
////
std::cout << " stoi: ";
if (sizeof(T) > sizeof(int))
std::cout << "SKIPPED" << std::endl;
else if (checkResults(strings, stoi_convert<T>))
time_run<T>(strings, stoi_convert<T>, samples);
else
std::cout << "FAILURE" << std::endl;
////
std::cout << " strtol: ";
if (sizeof(T) > sizeof(long))
std::cout << "SKIPPED" << std::endl;
else if (checkResults(strings, strtol_convert<T>))
time_run<T>(strings, strtol_convert<T>, samples);
else
std::cout << "FAILURE" << std::endl;
std::cout << " stol: ";
if (sizeof(T) > sizeof(long))
std::cout << "SKIPPED" << std::endl;
else if (checkResults(strings, stol_convert<T>))
time_run<T>(strings, stol_convert<T>, samples);
else
std::cout << "FAILURE" << std::endl;
std::cout << " strtoll: ";
if (sizeof(T) > sizeof(long long))
std::cout << "SKIPPED" << std::endl;
else if (checkResults(strings, strtoll_convert<T>))
time_run<T>(strings, strtoll_convert<T>, samples);
else
std::cout << "FAILURE" << std::endl;
std::cout << " stoll: ";
if (sizeof(T) > sizeof(long long))
std::cout << "SKIPPED" << std::endl;
else if (checkResults(strings, stoll_convert<T>))
time_run<T>(strings, stoll_convert<T>, samples);
else
std::cout << "FAILURE" << std::endl;
std::cout << " handcoded: ";
if (checkResults(strings, handcoded_convert<T>))
time_run<T>(strings, handcoded_convert<T>, samples);
else
std::cout << "FAILURE" << std::endl;
std::cout << " hybrid: ";
if (checkResults(strings, hybrid_convert<T>))
time_run<T>(strings, hybrid_convert<T>, samples);
else
std::cout << "FAILURE" << std::endl;
}
int main(int argc, char *argv[])
{
int i;
#ifdef EPETRA_MPI
// Initialize MPI
MPI_Init(&argc,&argv);
Epetra_MpiComm comm(MPI_COMM_WORLD);
#else
Epetra_SerialComm comm;
#endif
// Uncomment to debug in parallel int tmp; if (comm.MyPID()==0) cin >> tmp; comm.Barrier();
bool verbose = false;
// Check if we should print results to standard out
if (argc>1) if (argv[1][0]=='-' && argv[1][1]=='v') verbose = true;
if (!verbose) comm.SetTracebackMode(0); // This should shut down any error traceback reporting
if (verbose) cout << comm << endl << flush;
if (verbose) verbose = (comm.MyPID()==0);
if (verbose)
cout << EpetraExt::EpetraExt_Version() << endl << endl;
int nx = 128;
int ny = comm.NumProc()*nx; // Scale y grid with number of processors
// Create funky stencil to make sure the matrix is non-symmetric (transpose non-trivial):
// (i-1,j-1) (i-1,j )
// (i ,j-1) (i ,j ) (i ,j+1)
// (i+1,j-1) (i+1,j )
int npoints = 7;
int xoff[] = {-1, 0, 1, -1, 0, 1, 0};
int yoff[] = {-1, -1, -1, 0, 0, 0, 1};
Epetra_Map * map;
Epetra_CrsMatrix * A;
Epetra_Vector * x, * b, * xexact;
Trilinos_Util_GenerateCrsProblem(nx, ny, npoints, xoff, yoff, comm, map, A, x, b, xexact);
if (nx<8)
{
cout << *A << endl;
cout << "X exact = " << endl << *xexact << endl;
cout << "B = " << endl << *b << endl;
}
// Construct transposer
Epetra_Time timer(comm);
double start = timer.ElapsedTime();
//bool IgnoreNonLocalCols = false;
bool MakeDataContiguous = true;
EpetraExt::RowMatrix_Transpose transposer( MakeDataContiguous );
if (verbose) cout << "\nTime to construct transposer = " << timer.ElapsedTime() - start << endl;
Epetra_CrsMatrix & transA = dynamic_cast<Epetra_CrsMatrix&>(transposer(*A));
start = timer.ElapsedTime();
if (verbose) cout << "\nTime to create transpose matrix = " << timer.ElapsedTime() - start << endl;
// Now test output of transposer by performing matvecs
int ierr = 0;
ierr += checkResults(A, &transA, xexact, verbose);
// Now change values in original matrix and test update facility of transposer
// Add 2 to the diagonal of each row
double Value = 2.0;
for (i=0; i< A->NumMyRows(); i++)
A->SumIntoMyValues(i, 1, &Value, &i);
start = timer.ElapsedTime();
transposer.fwd();
if (verbose) cout << "\nTime to update transpose matrix = " << timer.ElapsedTime() - start << endl;
ierr += checkResults(A, &transA, xexact, verbose);
delete A;
delete b;
delete x;
delete xexact;
delete map;
if (verbose) cout << endl << "Checking transposer for VbrMatrix objects" << endl<< endl;
int nsizes = 4;
int sizes[] = {4, 6, 5, 3};
Epetra_VbrMatrix * Avbr;
Epetra_BlockMap * bmap;
Trilinos_Util_GenerateVbrProblem(nx, ny, npoints, xoff, yoff, nsizes, sizes,
comm, bmap, Avbr, x, b, xexact);
if (nx<8)
{
cout << *Avbr << endl;
cout << "X exact = " << endl << *xexact << endl;
cout << "B = " << endl << *b << endl;
}
start = timer.ElapsedTime();
EpetraExt::RowMatrix_Transpose transposer1( MakeDataContiguous );
Epetra_CrsMatrix & transA1 = dynamic_cast<Epetra_CrsMatrix&>(transposer1(*Avbr));
if (verbose) cout << "\nTime to create transpose matrix = " << timer.ElapsedTime() - start << endl;
// Now test output of transposer by performing matvecs
;
ierr += checkResults(Avbr, &transA1, xexact, verbose);
// Now change values in original matrix and test update facility of transposer
// Scale matrix on the left by rowsums
Epetra_Vector invRowSums(Avbr->RowMap());
Avbr->InvRowSums(invRowSums);
Avbr->LeftScale(invRowSums);
start = timer.ElapsedTime();
transposer1.fwd();
if (verbose) cout << "\nTime to update transpose matrix = " << timer.ElapsedTime() - start << endl;
ierr += checkResults(Avbr, &transA1, xexact, verbose);
delete Avbr;
delete b;
delete x;
delete xexact;
delete bmap;
#ifdef EPETRA_MPI
MPI_Finalize();
#endif
return ierr;
}
int main(int argc, char* argv[]) {
printf("\n NODE_BIND:%d, NUMA:%d, CPU_BIND:%d, FIRST_TOUCH:%d\n",NODE_BIND, NUMA, CPU_BIND, FIRST_TOUCH);
int repetitions, // number of repetition
maxThreads, // max number of threads
it,
N; // array size;
int bitCount = 1;
int * key; // array of keys
long * dataIn; // input data
long * dataSTL; // input stl data
long * dataRadix; // input radix data
repetitions = 1;
#pragma omp parallel
maxThreads = omp_get_num_threads();
if(argc ==1 ){
printf("prog input_file number_of_elements bit_count number_of_repetitions\n");
printf("NO INPUT FILE");
return 0;
}
if(argc == 2){
printf("prog input_file number_of_elements bit_count number_of_repetitions\n");
printf("NO ELEMENT COUNT\n");
return 0;
}
if(argc >2 ){
N = (int) strtol(argv[2], NULL, 10);
}
if(argc >3){
int tmp;
tmp = (int) strtol(argv[3], NULL, 10);
if ((tmp > 0) && (tmp<=16 )) // limit bit count
bitCount = tmp;
}
if(argc >4){
int tmp;
tmp = (int) strtol(argv[4], NULL, 10);
if ((tmp > 0) && (tmp<=10000 )) // limit repetitions
repetitions = tmp;
}
int *input;
size_t N2;
printf( "Reading data from file.\n" );
if( readIntArrayFile( argv[1], &input, &N2 ) )
return 1;
printf( "Data reading done.\n" );
if( (N2<(size_t)N) || (N<=0) )
N = N2;
printf( "\nPARALLEL STL SORT for N=%d, max threads = %d, test repetitions: %d\n", N, maxThreads, repetitions);
dataIn = new long[N];
dataSTL = new long[N];
#ifdef _WIN32
dataRadix = new long[N];
key = new int[N];
#endif
#ifdef linux
key = new int[N];
#if NUMA==0
dataRadix = new long[N];
#elif NUMA==1
dataRadix = (long*) numa_alloc_interleaved(N * sizeof(long));
#elif NUMA==2
dataRadix = (long*)numa_alloc_onnode(sizeof(long)*N,1);
#endif
#endif
VTimer stlTimes(maxThreads);
VTimer radixTimes(maxThreads);
#if TIME_COUNT==1
VTimer partTimes(TIMERS_COUNT);
#endif
#if FLUSH_CACHE==1
#ifdef linux
CacheFlusher cf;
#endif
#endif
for(long i=0;i<N;i++)
dataIn[i]=input[i];
delete[] input;
// loop from 1 to maxThreads
for (int t = 1; t <= maxThreads; t++) {
int i;
#if TIME_COUNT==1
partTimes.reset();
#endif
#if CALC_REF==1
// parallel STL
for (it = 0; it < repetitions; it++) {
setThreadsNo(t, maxThreads);
#pragma omp parallel for private(i)
for (i = 0; i < N; i++)
dataSTL[i] = dataIn[i];
#if FLUSH_CACHE==1
#ifdef linux
cf.flush();
#endif
#endif
stlTimes.timerStart(t-1);
#ifdef linux
__gnu_parallel::sort(dataSTL, dataSTL + N);
#endif
#ifdef _WIN32
std::sort(dataSTL, dataSTL + N);
#endif
stlTimes.timerEnd(t-1);
}
#if FLUSH_CACHE==1
#ifdef linux
cf.flush();
#endif
#endif
#endif
// radix sort V1
for (it = 0; it < repetitions; it++) {
setThreadsNo(t, maxThreads);
#pragma omp parallel for private(i) default(shared)
for (i = 0; i < N; i++){
dataRadix[i] = dataIn[i];
key[i]=i;
}
#if FLUSH_CACHE==1
#ifdef linux
cf.flush();
#endif
#endif
omp_set_num_threads(t);
radixTimes.timerStart(t-1);
#if TIME_COUNT==1
prsort::pradsort<long,int>(dataRadix,key, N, bitCount,&partTimes);
#else
prsort::pradsort<long,int>(dataRadix,key, N,bitCount,NULL);
#endif
radixTimes.timerEnd(t-1);
}
#if CALC_REF==1
printf("|STL SORT(th=%2d) : %1.3fs |\t", t,
stlTimes.getTime(t-1));
#endif
#if TIME_COUNT==1
for (int i = 0; i < TIMERS_COUNT; i++) {
#if CREATE_OUTPUT==1
printf("%d %d %d %d %d %d %d %f\n", NUMA, NODE_BIND, CPU_BIND, FIRST_TOUCH,bitCount , t, i ,partTimes.getTime(i));
#else
printf("part%d :%f ", i, partTimes.getTime(i));
#endif
}
#endif
#if CREATE_OUTPUT ==1
printf("%d %d %d %d %d %d calosc %1.3f", NUMA,NODE_BIND,CPU_BIND,FIRST_TOUCH,bitCount, t ,radixTimes.getTime(t-1));
#else
printf("|RADIX SORT (th=%2d) : %1.3fs |\t", t,
radixTimes.getTime(t-1));
#endif
// Attention: checking result only from the last function usage
#if CALC_REF==1
checkResults(dataSTL, dataRadix, N);
#else
printf("\n");
#endif
#if CHECK_KEY==1
if(checkKey(dataIn,dataRadix,key,N))printf("Keys are good\n");
#endif
}
#ifdef linux
delete[] key;
#if NUMA>0
numa_free(dataRadix, sizeof(long) * N);
#else
delete[] dataRadix;
#endif
#endif
#ifdef _WIN32
delete[] dataRadix;
#endif
delete[] dataIn;
delete[] dataSTL;
#if TIME_COUNT==1
#endif
return 0;
}
void Database_PostgreSQL::beginSave()
{
verifyDatabase();
checkResults(PQexec(m_conn, "BEGIN;"));
}