void ClusterBenchmark::testInsertionPerformance()
{
Cluster storage;
Cell* cell = 0;
qDebug() << "measuring loading-like insertion...";
Time::tval start = 0;
Time::tval ticks = 0;
int col = 1;
int row = 1;
int cols = 100;
int rows = 10000;
long counter = 0;
start = Time::stamp();
for (int r = row; r <= rows; ++r) {
for (int c = col; c <= cols; c += 1) {
storage.insert(cell, c, r);
counter += 1;
}
}
ticks = Time::elapsed(start);
qDebug() << qPrintable(Time::printAverage(ticks, counter));
qDebug() << "measuring random singular insertion...";
storage.clear();
counter = 0;
while (counter < Time::iterations) {
col = 1 + rand() % 1000;
row = 1 + rand() % 1000;
cols = col + 1;
rows = row + 1;
start = Time::stamp();
for (int r = row; r <= rows; ++r) {
for (int c = col; c <= cols && counter < Time::iterations; c += 1) {
storage.insert(cell, c, r);
counter += 1;
}
}
ticks += Time::elapsed(start);
}
qDebug() << qPrintable(Time::printAverage(ticks, counter));
}
/*
* Identify all clusters in the system.
*/
void ClusterIdentifier::identifyClusters()
{
// Initialize all data structures:
// Setup a grid of empty cells
cellList_.setup(system().boundary(), cutoff_);
// Clear clusters array and all links
clusters_.clear();
for (int i = 0; i < links_.capacity(); ++i) {
links_[i].clear();
}
// Build the cellList, associate Molecule with ClusterLink.
// Iterate over molecules of species speciesId_
System::MoleculeIterator molIter;
Molecule::AtomIterator atomIter;
system().begin(speciesId_, molIter);
for ( ; molIter.notEnd(); ++molIter) {
// Associate this Molecule with a ClusterLink
links_[molIter->id()].setMolecule(*molIter.get());
// Add atoms of type = atomTypeId_ to the CellList
for (molIter->begin(atomIter); atomIter.notEnd(); ++atomIter) {
if (atomIter->typeId() == atomTypeId_) {
system().boundary().shift(atomIter->position());
cellList_.addAtom(*atomIter);
}
}
}
// Identify all clusters
Cluster* clusterPtr;
ClusterLink* linkPtr;
int clusterId = 0;
system().begin(speciesId_, molIter);
for ( ; molIter.notEnd(); ++molIter) {
// Find the link with same index as this molecule
linkPtr = &(links_[molIter->id()]);
assert (&(linkPtr->molecule()) == molIter.get());
// If this link is not in a cluster, begin a new cluster
if (linkPtr->clusterId() == -1) {
// Add a new empty cluster to clusters_ array
clusters_.resize(clusterId+1);
clusterPtr = &clusters_[clusterId];
clusterPtr->clear();
clusterPtr->setId(clusterId);
// Identify molecules in this cluster
clusterPtr->addLink(*linkPtr);
workStack_.push(*linkPtr);
while (workStack_.size() > 0) {
processNextMolecule(*clusterPtr);
}
clusterId++;
}
}
// Validity check - throws exception on failure.
isValid();
}
void run(std::istream& in)
{
vector<string> tokens;
string line;
map<string,auto_vector<KnownGene> >::iterator rchrom;
while(getline(in,line,'\n'))
{
if(AbstractApplication::stopping()) break;
if(line.empty() || line[0]=='#') continue;
WHERE(line);
tokenizer.split(line,tokens);
if(chromCol>=(int)tokens.size())
{
cerr << "Column CHROM out of range in "<< line << endl;
continue;
}
if((rchrom=chrom2genes.find(tokens[chromCol]))==chrom2genes.end())
{
continue;
}
if(posCol>=(int)tokens.size())
{
cerr << "Column POS out of range in "<< line << endl;
continue;
}
if(sampleCol!=-1 && sampleCol>=(int)tokens.size())
{
cerr << "Column SAMPLE out of range in "<< line << endl;
continue;
}
if(colorCol!=-1 && colorCol>=(int)tokens.size())
{
cerr << "Column COLOR out of range in "<< line << endl;
continue;
}
char* p2;
int pos=(int)strtol(tokens[posCol].c_str(),&p2,10);
if(pos <0 || *p2!=0)
{
cerr << "Bad POS "<< tokens[posCol] << " in "<<line << endl;
continue;
}
WHERE(line);
if(cluster.chromEnd <=pos)
{
print();
cluster.clear();
auto_vector<KnownGene>& v= rchrom->second;
size_t i=0;
while(i< v.size())
{
KnownGene* g=v.at(i);
WHERE((g==NULL));
if(cluster.genes.empty())
{
if(g->txEnd <=pos || g->txStart> pos )
{
i++;
continue;
}
cluster.chromStart= g->txStart;
cluster.chromEnd= g->txEnd;
cluster.genes.push_back(v.release(i));
}
else
{
if(!(g->txStart>cluster.chromEnd || g->txEnd<= cluster.chromStart))
{
cluster.chromStart=min(cluster.chromStart,g->txStart);
cluster.chromEnd=max(cluster.chromEnd,g->txEnd);
cluster.genes.push_back(v.release(i));
}
else
{
++i;
}
}
}
}
if(!(cluster.chromStart<=pos && pos<cluster.chromEnd))
{
continue;
}
cluster.positions.insert(pos);
if(sampleCol!=-1)
{
SampleData data;
data.pos=pos;
if(colorCol!=-1)
{
Color color(tokens[colorCol].c_str());
data.rgb=color.asInt();
}
else
{
Color color(0.0f);
data.rgb=color.asInt();
}
cluster.sample2positions[ tokens[sampleCol] ].insert(data);
}
}
print();
}
void print()
{
if(cluster.positions.empty())
{
cluster.clear();
return;
}
++count_pages_printed;
cout << "\n%%Page: " << count_pages_printed << " "<<count_pages_printed << "\n";
WHERE(cluster.genes.size());
double fHeight=20;
float midy=fHeight/2.0f;
double cdsHeight=fHeight*0.4;
double exonHeight=fHeight*0.9;
cout << "2 " << (pageHeight()-10 )<<" moveto (" << cluster.genes.back()->chrom << ":"
<< cluster.chromStart << "-"
<< cluster.chromEnd << ") show\n"
;
cout << "1 0 0 setrgbcolor\n";
cout << "0.3 setlinewidth\n";
for(set<int32_t>::iterator r=cluster.positions.begin();
r!=cluster.positions.end();
++r)
{
cout << "newpath "<< toPixel(*r) << " 0 moveto 0 "<< pageHeight() << " rlineto stroke\n";
cout << toPixel(*r) << " " << (pageHeight()-5) <<" moveto -90 rotate (" << (*r) << ") show 90 rotate\n";
}
for(size_t i=0;i< cluster.genes.size();++i)
{
KnownGene* g=cluster.genes.at(i);
cout << "gsave\n";
cout <<"0 "<< pageHeight()-margin_top-(fHeight*i)<<" translate\n";
double x1=toPixel(g->txStart);
double x2=toPixel(g->txEnd);
cout << "0 0 0 setrgbcolor\n";
NEWPATH;
MOVETO(x1,midy);
LINETO(x2,midy);
STROKE;
//draw ticks
cout << "0.2 setlinewidth\n";
cout << "newpath\n";
cout << x1 <<" " << midy << " moveto\n";
cout << x1 << " " << x2 << (g->isForward()?" forticksF":" forticksR")<< endl;
cout << "closepath stroke\n";
cout << "0.5 setlinewidth\n";
//draw txStart/txEnd
cout << "0.1 0.1 0.5 setrgbcolor\n"
"newpath\n"
<< toPixel(g->cdsStart) << " "
<< (midy-cdsHeight/2.0) <<" "
<< (toPixel(g->cdsEnd)-toPixel(g->cdsStart)) << " "
<< cdsHeight << " box closepath fill\n"
;
//draw each exon
for(int32_t j=0;j< g->countExons();++j)
{
cout << toPixel(g->exon(j)->start) << " "
<< (midy-exonHeight/2.0)<< " "
<< (toPixel(g->exon(j)->end)-toPixel(g->exon(j)->start)) << " "
<< exonHeight << " gradient\n"
;
}
//draw name
cout << "0 0 0 setrgbcolor\n";
cout << "10 " << midy << " moveto (" << g->name << ") show\n";
cout << "grestore\n";
}
if(sampleCol!=-1)
{
double y= pageHeight()-margin_top-(fHeight*(cluster.genes.size()+1));
for(map<string,set<SampleData>,SmartComparator >::iterator r=cluster.sample2positions.begin();
r!=cluster.sample2positions.end();
++r)
{
cout << "0.2 setlinewidth\n";
cout << "0 0 0 setrgbcolor\n";
cout << "10 " << (y-midy+5) << " moveto (" << r->first << ") show\n";
cout << "newpath "<< margin_left << " "<< y << " moveto\n"
<< width()<< " " << 0 << " rlineto stroke\n";
for(set<SampleData>::iterator r2= r->second.begin();
r2!=r->second.end();
++r2)
{
if(colorCol==-1)
{
cout << "0 setgray\n";
}
else
{
Color c((*r2).rgb);
cout << (int)c.r << " " << (int)c.g << " "<< (int)c.b << " setrgbcolor\n";
}
cout << "0.8 setlinewidth\n";
cout << "newpath "<<toPixel((*r2).pos)<<" "<< y << " circle closepath stroke\n";
}
y-=fHeight;
}
}
cout << "showpage\n";
cluster.clear();
}
void ClusterBenchmark::testLookupPerformance()
{
// row x column
const int scenarios[] = {
#if 1
1000, 1000 // large
#else
5, 5, // very small
30, 20, // fit to screen
100, 100, // medium
1000, 1000, // large
10000, 100, // typical data: more rows
10000, 2000, // and 20 times larger
100, 10000, // not really typical: more columns
8000, 8000 // hopelessly large
#endif
};
Cluster storage;
Cell* cell = 0;
for (uint sc = 0; sc < sizeof(scenarios) / sizeof(scenarios[0]) / 2; sc++) {
int maxrow = scenarios[sc*2];
int maxcol = scenarios[sc*2+1];
storage.clear();
#if 0
for (int r = 0; r < maxrow; ++r) {
for (int c = 0; c < maxcol; ++c) {
storage.insert(cell, c, r);
}
}
#else
storage.insert(cell, 1, 1);
storage.insert(cell, maxcol / 2, maxrow / 2);
storage.insert(cell, maxcol / 3, maxrow / 3);
storage.insert(cell, maxcol, maxrow);
#endif
// qDebug() << endl << qPrintable( storage.dump() );
QString prefix = QString("%1 x %2").arg(maxrow).arg(maxcol);
qDebug() << "start measuring..." << prefix;
Time::tval start = 0;
Time::tval ticks = 0;
Cell* v;
int col = 0;
int row = 0;
int cols = 0;
int rows = 0;
long counter = 0;
while (counter < Time::iterations) {
col = 1 + rand() % maxcol;
row = 1 + rand() % maxrow;
cols = col + 1 * (rand() % 10);
rows = row + rand() % 10;
start = Time::stamp();
for (int r = row; r <= rows && counter < Time::iterations; ++r) {
for (int c = col; c <= cols && counter < Time::iterations; c += 1) {
v = storage.lookup(c, r);
counter += 1;
}
}
ticks += Time::elapsed(start);
}
qDebug() << qPrintable(Time::printAverage(ticks, counter, prefix));
}
}
bool Soft_Cluster_Handler::EnforcedTransition(Cluster_List * clin) {
#ifdef AHAmomcheck
Vec4D checkbef(SumMomentum(clin));
#endif
size_t size(0);
std::vector<double> masses;
std::vector<Vec4D> momenta;
Cluster * cluster;
for (Cluster_Iterator cit=clin->begin();cit!=clin->end();cit++) {
cluster = (*cit);
switch (cluster->size()) {
case 1:
masses.push_back((*cluster)[0].HadMass());
momenta.push_back(cluster->Momentum());
++size;
break;
case 2:
case 0:
masses.push_back(sqrt(Max(cluster->GetTrip()->m_mom.Abs2(),0.)));
masses.push_back(sqrt(Max(cluster->GetAnti()->m_mom.Abs2(),0.)));
momenta.push_back(cluster->GetTrip()->m_mom);
momenta.push_back(cluster->GetAnti()->m_mom);
size+=2;
default:
break;
}
}
if (!hadpars->AdjustMomenta(size,&momenta.front(),&masses.front())) {
if (size>1 /*&& msg->LevelIsDebugging()*/) {
msg_Tracking()<<"Error in "<<METHOD<<" ("<<size<<" clusters) : \n"
<<" Could not adjust momenta for : \n";
int i(0);
for (Cluster_Iterator cit=clin->begin();cit!=clin->end();cit++) {
msg_Tracking()<<"Mass/Mom = "<<masses[i]<<"/"<<momenta[i];
if ((*cit)->size()==1) msg_Tracking()<<" ("<<((**cit)[0])<<" )";
msg_Tracking()<<" for \n"<<(**cit)<<"\n";
i++;
}
msg_Tracking()<<" Will possibly lead to retrying the event.\n";
}
return false;
}
int pos(0);
#ifdef AHAmomcheck
Vec4D checkaft(SumMomentum(clin));
#endif
for (Cluster_Iterator cit=clin->begin();cit!=clin->end();cit++) {
cluster = (*cit);
switch (cluster->size()) {
case 1:
cluster->SetFlav((*cluster)[0]);
cluster->SetMomentum(momenta[pos]);
break;
case 2:
#ifdef AHAmomcheck
checkaft += momenta[pos];
#endif
cluster->GetTrip()->m_mom=momenta[pos++];
cluster->GetAnti()->m_mom=momenta[pos];
cluster->Update();
if (cluster->Mass()<(*cluster)[0].HadMass()+(*cluster)[1].HadMass()) {
cluster->clear();
}
break;
case 0:
cluster->GetTrip()->m_mom=momenta[pos++];
cluster->GetAnti()->m_mom=momenta[pos];
cluster->Update();
break;
default:
break;
}
pos++;
}
#ifdef AHAmomcheck
double Q2(dabs((checkbef-checkaft).Abs2()));
if (Q2>1.e-12 || IsNan(Q2)) {
msg_tracking()<<METHOD<<" yields a momentum violation for "<<size<<" : \n"
<<" "<<checkbef<<" - "<<checkaft<<" --> "
<<(checkbef-checkaft).Abs2()<<"("<<size<<").\n"
<<(*clin)<<"\n";
}
else msg_Tracking()<<METHOD<<" satisfied four-momentum conservation.\n";
#endif
m_transitions += 1;
return true;
}
void ZimCreator::createClusters(ArticleSource& src, const std::string& tmpfname)
{
std::ofstream out(tmpfname.c_str());
Cluster cluster;
cluster.setCompression(compression);
DirentsType::size_type count = 0, progress = 0;
for (DirentsType::iterator di = dirents.begin(); out && di != dirents.end(); ++di, ++count)
{
while (progress < count * 100 / dirents.size() + 1)
{
INFO(progress << "% ready");
progress += 10;
}
if (di->isRedirect())
continue;
Blob blob = src.getData(di->getAid());
if (blob.size() > 0)
isEmpty = false;
if (di->isCompress())
{
di->setCluster(clusterOffsets.size(), cluster.count());
cluster.addBlob(blob);
if (cluster.size() >= minChunkSize * 1024)
{
log_info("compress cluster with " << cluster.count() << " articles, " << cluster.size() << " bytes; current title \"" << di->getTitle() << '\"');
clusterOffsets.push_back(out.tellp());
out << cluster;
log_debug("cluster compressed");
cluster.clear();
cluster.setCompression(compression);
}
}
else
{
if (cluster.count() > 0)
{
clusterOffsets.push_back(out.tellp());
cluster.setCompression(compression);
out << cluster;
cluster.clear();
cluster.setCompression(compression);
}
di->setCluster(clusterOffsets.size(), cluster.count());
clusterOffsets.push_back(out.tellp());
Cluster c;
c.addBlob(blob);
c.setCompression(zimcompNone);
out << c;
}
}
if (cluster.count() > 0)
{
clusterOffsets.push_back(out.tellp());
cluster.setCompression(compression);
out << cluster;
}
if (!out)
throw std::runtime_error("failed to write temporary cluster file");
clustersSize = out.tellp();
}
