bool VariablePacker::CheckVariablesWithinPackingLimits(int maxVectors, const TVariableInfoList& in_variables)
{
ASSERT(maxVectors > 0);
maxRows_ = maxVectors;
topNonFullRow_ = 0;
bottomNonFullRow_ = maxRows_ - 1;
TVariableInfoList variables(in_variables);
// As per GLSL 1.017 Appendix A, Section 7 variables are packed in specific
// order by type, then by size of array, largest first.
std::sort(variables.begin(), variables.end(), TVariableInfoComparer());
rows_.clear();
rows_.resize(maxVectors, 0);
// Packs the 4 column variables.
size_t ii = 0;
for (; ii < variables.size(); ++ii) {
const TVariableInfo& variable = variables[ii];
if (GetNumComponentsPerRow(variable.type) != 4) {
break;
}
topNonFullRow_ += GetNumRows(variable.type) * variable.size;
}
if (topNonFullRow_ > maxRows_) {
return false;
}
// Packs the 3 column variables.
int num3ColumnRows = 0;
for (; ii < variables.size(); ++ii) {
const TVariableInfo& variable = variables[ii];
if (GetNumComponentsPerRow(variable.type) != 3) {
break;
}
num3ColumnRows += GetNumRows(variable.type) * variable.size;
}
if (topNonFullRow_ + num3ColumnRows > maxRows_) {
return false;
}
fillColumns(topNonFullRow_, num3ColumnRows, 0, 3);
// Packs the 2 column variables.
int top2ColumnRow = topNonFullRow_ + num3ColumnRows;
int twoColumnRowsAvailable = maxRows_ - top2ColumnRow;
int rowsAvailableInColumns01 = twoColumnRowsAvailable;
int rowsAvailableInColumns23 = twoColumnRowsAvailable;
for (; ii < variables.size(); ++ii) {
const TVariableInfo& variable = variables[ii];
if (GetNumComponentsPerRow(variable.type) != 2) {
break;
}
int numRows = GetNumRows(variable.type) * variable.size;
if (numRows <= rowsAvailableInColumns01) {
rowsAvailableInColumns01 -= numRows;
} else if (numRows <= rowsAvailableInColumns23) {
rowsAvailableInColumns23 -= numRows;
} else {
return false;
}
}
int numRowsUsedInColumns01 =
twoColumnRowsAvailable - rowsAvailableInColumns01;
int numRowsUsedInColumns23 =
twoColumnRowsAvailable - rowsAvailableInColumns23;
fillColumns(top2ColumnRow, numRowsUsedInColumns01, 0, 2);
fillColumns(maxRows_ - numRowsUsedInColumns23, numRowsUsedInColumns23,
2, 2);
// Packs the 1 column variables.
for (; ii < variables.size(); ++ii) {
const TVariableInfo& variable = variables[ii];
ASSERT(1 == GetNumComponentsPerRow(variable.type));
int numRows = GetNumRows(variable.type) * variable.size;
int smallestColumn = -1;
int smallestSize = maxRows_ + 1;
int topRow = -1;
for (int column = 0; column < kNumColumns; ++column) {
int row = 0;
int size = 0;
if (searchColumn(column, numRows, &row, &size)) {
if (size < smallestSize) {
smallestSize = size;
smallestColumn = column;
topRow = row;
}
}
}
if (smallestColumn < 0) {
return false;
}
fillColumns(topRow, numRows, smallestColumn, 1);
}
ASSERT(variables.size() == ii);
return true;
}
TEST_F(test_move, independents)
{
independent_variables independents(1, 2);
independents(1) = 1.5;
independents(2) = 3.5;
gradient_structure gs;
ASSERT_EQ(0, gradient_structure::GRAD_STACK1->total());
ASSERT_EQ(1750, gradient_structure::GRAD_LIST->total_addresses());
dvar_vector variables(independents);
dvariable a(variables(1));
dvariable b(variables(2));
auto sum = [&a, &b]()
{
return a + b;
};
dvariable result = sum();
ASSERT_EQ(4, gradient_structure::GRAD_STACK1->total());
ASSERT_EQ(1753, gradient_structure::GRAD_LIST->total_addresses());
ASSERT_DOUBLE_EQ(value(result), 5.0);
grad_stack_entry* ptr = gradient_structure::GRAD_STACK1->ptr;
ASSERT_TRUE(ptr->func == NULL);
--ptr;
ASSERT_TRUE(ptr->func == &default_evaluation1);
ASSERT_TRUE(ptr->dep_addr == gradient_structure::GRAD_LIST->get(1752));
ASSERT_TRUE(ptr->ind_addr1 == &(gradient_structure::RETURN_PTR->v->x));
ASSERT_TRUE(ptr->ind_addr2 == NULL);
ASSERT_DOUBLE_EQ(ptr->mult1, 0.0);
ASSERT_DOUBLE_EQ(ptr->mult2, 0.0);
--ptr;
ASSERT_TRUE(ptr->func == &default_evaluation4);
ASSERT_TRUE(ptr->dep_addr == &(gradient_structure::RETURN_PTR->v->x));
ASSERT_TRUE(ptr->ind_addr1 == gradient_structure::GRAD_LIST->get(1750));
ASSERT_TRUE(ptr->ind_addr2 == gradient_structure::GRAD_LIST->get(1751));
ASSERT_DOUBLE_EQ(ptr->mult1, 0.0);
ASSERT_DOUBLE_EQ(ptr->mult2, 0.0);
--ptr;
ASSERT_TRUE(ptr->func == &default_evaluation1);
ASSERT_TRUE(ptr->dep_addr == gradient_structure::GRAD_LIST->get(1751));
ASSERT_TRUE(ptr->ind_addr1 == &(variables(2).v->x));
ASSERT_TRUE(ptr->ind_addr2 == NULL);
ASSERT_DOUBLE_EQ(ptr->mult1, 0.0);
ASSERT_DOUBLE_EQ(ptr->mult2, 0.0);
--ptr;
ASSERT_TRUE(ptr->func == &default_evaluation1);
ASSERT_TRUE(ptr->dep_addr == gradient_structure::GRAD_LIST->get(1750));
ASSERT_TRUE(ptr->ind_addr1 == &(variables(1).v->x));
ASSERT_TRUE(ptr->ind_addr2 == NULL);
ASSERT_DOUBLE_EQ(ptr->mult1, 0.0);
ASSERT_DOUBLE_EQ(ptr->mult2, 0.0);
}
TEST_F(test_move, independents_gradmanual)
{
independent_variables independents(1, 2);
independents(1) = 1.5;
independents(2) = 3.5;
gradient_structure gs;
ASSERT_EQ(0, gradient_structure::GRAD_STACK1->total());
ASSERT_EQ(1750, gradient_structure::GRAD_LIST->total_addresses());
dvar_vector variables(independents);
auto sum = [&variables]()
{
dvariable a(variables(1));
dvariable b(variables(2));
return a + b;
};
dvariable result = sum();
ASSERT_EQ(4, gradient_structure::GRAD_STACK1->total());
ASSERT_EQ(1752, gradient_structure::GRAD_LIST->total_addresses());
ASSERT_DOUBLE_EQ(value(result), 5.0);
--gradient_structure::GRAD_STACK1->ptr;
double_and_int* ptr = (double_and_int*)gradient_structure::get_ARRAY_MEMBLOCK_BASE();
unsigned long int imax = gradient_structure::ARR_LIST1->get_max_last_offset() / sizeof(double_and_int);
for (unsigned int i = 0; i < imax; ++i)
{
ptr->x = 0.0;
++ptr;
}
gradient_structure::GRAD_LIST->initialize();
*gradient_structure::GRAD_STACK1->ptr->dep_addr = 1.0;
//Sum
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->dep_addr), 1.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr1), 0.0);
(*(gradient_structure::GRAD_STACK1->ptr->func))();
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->dep_addr), 0.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr1), 1.0);
ASSERT_TRUE(gradient_structure::GRAD_STACK1->ptr->ind_addr2 == NULL);
ASSERT_DOUBLE_EQ(gradient_structure::GRAD_STACK1->ptr->mult1, 0.0);
ASSERT_DOUBLE_EQ(gradient_structure::GRAD_STACK1->ptr->mult2, 0.0);
//Addition
--gradient_structure::GRAD_STACK1->ptr;
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->dep_addr), 1.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr1), 0.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr2), 0.0);
(*(gradient_structure::GRAD_STACK1->ptr->func))();
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->dep_addr), 0.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr1), 1.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr2), 1.0);
ASSERT_DOUBLE_EQ(gradient_structure::GRAD_STACK1->ptr->mult1, 0.0);
ASSERT_DOUBLE_EQ(gradient_structure::GRAD_STACK1->ptr->mult2, 0.0);
//Constructor
--gradient_structure::GRAD_STACK1->ptr;
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->dep_addr), 1.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr1), 0.0);
(*(gradient_structure::GRAD_STACK1->ptr->func))();
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->dep_addr), 0.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr1), 1.0);
ASSERT_TRUE(gradient_structure::GRAD_STACK1->ptr->ind_addr2 == NULL);
ASSERT_DOUBLE_EQ(gradient_structure::GRAD_STACK1->ptr->mult1, 0.0);
ASSERT_DOUBLE_EQ(gradient_structure::GRAD_STACK1->ptr->mult2, 0.0);
//Constructor
--gradient_structure::GRAD_STACK1->ptr;
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->dep_addr), 1.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr1), 0.0);
(*(gradient_structure::GRAD_STACK1->ptr->func))();
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->dep_addr), 0.0);
ASSERT_DOUBLE_EQ(*(gradient_structure::GRAD_STACK1->ptr->ind_addr1), 1.0);
ASSERT_TRUE(gradient_structure::GRAD_STACK1->ptr->ind_addr2 == NULL);
ASSERT_DOUBLE_EQ(gradient_structure::GRAD_STACK1->ptr->mult1, 0.0);
ASSERT_DOUBLE_EQ(gradient_structure::GRAD_STACK1->ptr->mult2, 0.0);
//ASSERT_DOUBLE_EQ(*gradient_structure::INDVAR_LIST->get_address(1), 0.0);
//ASSERT_DOUBLE_EQ(*gradient_structure::INDVAR_LIST->get_address(2), 0.0);
ASSERT_DOUBLE_EQ(value(variables(1)), 1.0);
ASSERT_DOUBLE_EQ(value(variables(2)), 1.0);
}
void write_realization( const circuit& circ, std::ostream& os, const write_realization_settings& settings )
{
unsigned oldsize = 0;
if ( !settings.header.empty() )
{
std::string header = settings.header;
boost::algorithm::replace_all( header, "\n", "\n# " );
os << "# " << header << std::endl;
}
if ( !settings.version.empty() )
{
os << ".version " << settings.version << std::endl;
}
os << ".numvars " << circ.lines() << std::endl;
std::vector<std::string> variables( circ.lines() );
for ( unsigned i = 0u; i < circ.lines(); ++i )
{
variables[i] = boost::str( boost::format( "x%d" ) % i );
}
std::vector<std::string> _inputs( circ.inputs().begin(), circ.inputs().end() );
oldsize = _inputs.size();
_inputs.resize( circ.lines() );
for ( unsigned i = oldsize; i < circ.lines(); ++i )
{
_inputs[i] = boost::str( boost::format( "i%d" ) % i );
}
std::vector<std::string> _outputs( circ.outputs().begin(), circ.outputs().end() );
oldsize = _outputs.size();
_outputs.resize( circ.lines() );
for ( unsigned i = oldsize; i < circ.lines(); ++i )
{
_outputs[i] = boost::str( boost::format( "o%d" ) % i );
}
os << ".variables " << boost::algorithm::join( variables, " " ) << std::endl;
namespace karma = boost::spirit::karma;
namespace ascii = boost::spirit::ascii;
os << ".inputs";
//std::ostream_iterator<char> outit( os );
//karma::generate_delimited( outit, *( karma::no_delimit['"' << karma::string] << '"' ), ascii::space, _inputs );
for ( const auto& _input : _inputs )
{
std::string quote = ( _input.find( " " ) != std::string::npos ) ? "\"" : "";
os << boost::format( " %s%s%s" ) % quote % _input % quote;
}
os << std::endl;
os << ".outputs";
//karma::generate_delimited( outit, *( karma::no_delimit['"' << karma::string] << '"' ), ascii::space, _outputs );
for ( const auto& _output : _outputs )
{
std::string quote = ( _output.find( " " ) != std::string::npos ) ? "\"" : "";
os << boost::format( " %s%s%s" ) % quote % _output % quote;
}
os << std::endl;
std::string _constants( circ.lines(), '-' );
std::transform( circ.constants().begin(), circ.constants().end(), _constants.begin(), constant_to_char() );
std::string _garbage( circ.lines(), '-' );
std::transform( circ.garbage().begin(), circ.garbage().end(), _garbage.begin(), garbage_to_char() );
os << ".constants " << _constants << std::endl
<< ".garbage " << _garbage << std::endl;
for ( const auto& bus : circ.inputbuses().buses() )
{
std::vector<std::string> lines;
std::transform( bus.second.begin(), bus.second.end(), std::back_inserter( lines ), line_to_variable() );
os << ".inputbus " << bus.first << " " << boost::algorithm::join( lines, " " ) << std::endl;
}
for ( const auto& bus : circ.outputbuses().buses() )
{
std::vector<std::string> lines;
std::transform( bus.second.begin(), bus.second.end(), std::back_inserter( lines ), line_to_variable() );
os << ".outputbus " << bus.first << " " << boost::algorithm::join( lines, " " ) << std::endl;
}
for ( const auto& bus : circ.statesignals().buses() )
{
std::vector<std::string> lines;
std::transform( bus.second.begin(), bus.second.end(), std::back_inserter( lines ), line_to_variable() );
os << ".state " << bus.first << " " << boost::algorithm::join( lines, " " ) << std::endl;
}
for ( const auto& module : circ.modules() )
{
os << ".module " << module.first << std::endl;
write_realization_settings module_settings;
module_settings.version.clear();
module_settings.header.clear();
write_realization( *module.second, os, module_settings );
}
os << ".begin" << std::endl;
std::string cmd;
for ( const auto& g : circ )
{
cmd = settings.type_label( g );
std::vector<std::string> lines;
// Peres is special
boost::transform( g.controls(), std::back_inserter( lines ), line_to_variable() );
boost::transform( g.targets(), std::back_inserter( lines ), line_to_variable() );
os << cmd << " " << boost::algorithm::join( lines, " " );
boost::optional<const std::map<std::string, std::string>&> annotations = circ.annotations( g );
if ( annotations )
{
std::string sannotations;
for ( const auto& p : *annotations )
{
sannotations += boost::str( boost::format( " %s=\"%s\"" ) % p.first % p.second );
}
os << " #@" << sannotations;
}
os << std::endl;
}
os << ".end" << std::endl;
}
void plotMakerConstBin(string htag,string newDatasetName, unsigned int nNewFiles)
{
if (newDatasetName=="")
{
cerr << "no new dataset name? what the hell do you want me to do? Returning..." << endl;
return;
}
// variables to plot are in defined in plotVars.h
std::vector<string> variables(plotVars, plotVars + sizeof plotVars / sizeof plotVars[0]);
TChain * newFileChain=new TChain("CollectionTree");
if(nNewFiles == 0)
{
cerr << "No files in new dataset? This shouldn't happen. Something has gone wrong. Check inputs" << endl;
return;
}
else if(nNewFiles == 1)
{
// if its only 1 file, then its not a folder
newFileChain->Add(newDatasetName.c_str());
}
else // it's folder!
{
for(int i = 1; i <= nNewFiles;i++)
{
size_t found = newDatasetName.find_last_of("/");
string baseFileName=newDatasetName.substr(found+1);
size_t f = baseFileName.find(".root");
string fileNum="";
if(i < 10) fileNum="00"+std::to_string(i);
else if (i < 100) fileNum="0"+std::to_string(i);
else fileNum=std::to_string(i);
baseFileName.replace(f, baseFileName.length(), "."+fileNum+".root");
string path=newDatasetName+"/"+baseFileName;
newFileChain->Add(path.c_str());
}
}
float progress=0.0;
size_t found = newDatasetName.find_last_of("/");
string baseFileName=newDatasetName.substr(found+1);
string outfilename="samples/"+htag+"/"+baseFileName;
TFile f(outfilename.c_str(),"recreate");
for(unsigned int i =0; i< variables.size();i++)
{
//cout << variables[i] << endl;
//if (variables[i] != "HGamEventInfoAuxDyn.m_yy_truthVertex" ) continue;
int barWidth = 70;
std::cout << "[";
int pos = barWidth * progress;
for (int i = 0; i < barWidth; ++i) {
if (i < pos) std::cout << "=";
else if (i == pos) std::cout << ">";
else std::cout << " ";
}
std::cout << "] " << int(progress * 100.0) << " %\r";
std::cout.flush();
//size_t found = newDatasetName.find_last_of("/");
//string baseFileName=newDatasetName.substr(found+1);
//cout << "1" << endl;
TCanvas * c1 =new TCanvas("c1");
if(newFileChain->Draw(variables[i].c_str(),"") == -1)
{
cout << "Draw command failed!" << endl;
progress += 1.0/variables.size();
continue;
}
//cout << "2" << endl;
TH1F *htemp = (TH1F*)gPad->GetPrimitive("htemp");
if (htemp==0)
{
cout << "histo is empty for var " << variables[i] << ", must have no entries" << endl;
continue;
}
//cout << "2" << endl;
//cout << htemp << endl;
htemp->SetLineColor( kRed);
float MeanVal=htemp->GetMean();
float stdev =htemp->GetStdDev();
delete c1;
float xmin = MeanVal-2*stdev;
long xminRound=floor(xmin/1000)*1000;
if(xmin<-10000) // case for variables like Energy where there is no reason to have a negative xmin
{
xminRound = 0;
}
float xmax = MeanVal+2*stdev;
long xmaxRound=floor(xmax/1000)*1000;
if(xmin==xmax) // case for where there is no std dev for variables like m_ee in ggH125_small samples, so xmin==xmax and root gets confused
{
xmin = xmin - 1000;
xmax = xmax + 1000;
}
string lowCut=variables[i]+" > "+std::to_string(xmin);
string highCut=variables[i]+" < "+std::to_string(xmax);
string fullCut=lowCut+" && "+highCut+" && HGamEventInfoAuxDyn.isPassedPreselection";
long nBins=(xmaxRound-xminRound)/1000;
TCanvas * c2 =new TCanvas("c2");
string plotVar=variables[i]+">>htemp("+to_string(nBins)+","+to_string(xminRound)+","+to_string(xmaxRound)+")";
newFileChain->Draw(plotVar.c_str(),fullCut.c_str(),"");
htemp = (TH1F*)gPad->GetPrimitive("htemp");
//cout << "3" << endl;
int maxITER = 5;
int iter=0;
//while( (htemp == 0 || htemp->Integral() < 5 ) && iter <= maxITER )
//{
// fullCut = increaseCut(variables[i],MeanVal,stdev,xmin,xmax);
// newFileChain->Draw(variables[i].c_str(),fullCut.c_str());
// htemp = (TH1F*)gPad->GetPrimitive("htemp");
// iter++;
//
// if(iter==maxITER) cout << "max iterations reached, might be no plot for " << variables[i] << endl;
//}
htemp->SetNameTitle(variables[i].c_str(),variables[i].c_str());
htemp->Write();
delete c2;
progress += 1.0/variables.size();
}
int barWidth=70;
std::cout << "[";
int pos = barWidth * 1;
for (int i = 0; i < barWidth; ++i) {
if (i < pos) std::cout << "=";
else if (i == pos) std::cout << ">";
else std::cout << " ";
}
std::cout << "] " << int(1 * 100.0) << " %\r";
std::cout.flush();
cout << endl;
f.Close();
}
void perform_write(
hpx::lcos::local::channel<void>& sync
, octree_server& e
, DBfile* file
, std::vector<std::string> const& directory_names
, std::string const& variable_name
, boost::uint64_t variable_index
)
{ // {{{
boost::uint64_t const bw = science().ghost_zone_length;
boost::uint64_t const gnx = config().grid_node_length;
boost::uint64_t level = e.get_level();
int nnodes[] = {
int(gnx - 2 * bw + 1)
, int(gnx - 2 * bw + 1)
, int(gnx - 2 * bw + 1)
};
int nzones[] = {
int(gnx - 2 * bw)
, int(gnx - 2 * bw)
, int(gnx - 2 * bw)
};
//char* coordinate_names[] = { (char*) "X", (char*) "Y", (char*) "Z" };
boost::scoped_array<double*> coordinates(new double* [3]);
coordinates[0] = new double [nnodes[0]];
coordinates[1] = new double [nnodes[1]];
coordinates[2] = new double [nnodes[2]];
boost::scoped_array<double> variables
(new double [nzones[0] * nzones[1] * nzones[2]]);
for (boost::uint64_t i = bw; i < (gnx - bw + 1); ++i)
{
coordinates[0][i - bw] = e.x_face(i);
coordinates[1][i - bw] = e.y_face(i);
coordinates[2][i - bw] = e.z_face(i);
}
for (boost::uint64_t i = bw; i < (gnx - bw); ++i)
for (boost::uint64_t j = bw; j < (gnx - bw); ++j)
for (boost::uint64_t k = bw; k < (gnx - bw); ++k)
{
boost::uint64_t index = (i - bw)
+ (j - bw) * nzones[0]
+ (k - bw) * nzones[0] * nzones[1];
variables[index] = e(i, j, k)[variable_index];
}
array<boost::uint64_t, 3> location = e.get_location();
std::string mesh_name
= boost::str( boost::format("mesh_L%i_%i_%i_%i")
% level
% location[0]
% location[1]
% location[2]);
std::string value_name
= boost::str( boost::format("%s_L%i_%i_%i_%i")
% variable_name
% level
% location[0]
% location[1]
% location[2]);
int error = DBSetDir(file, directory_names[level].c_str());
OCTOPUS_ASSERT(error == 0);
{
DBoptlist* optlist = DBMakeOptlist(1);
// REVIEW: Verify this.
int type = DB_ROWMAJOR;
DBAddOption(optlist, DBOPT_MAJORORDER, &type);
error = DBPutQuadmesh(file
, mesh_name.c_str()
//, coordinate_names
, NULL // SILO docs say this is ignored.
, coordinates.get()
, nnodes
, 3, DB_DOUBLE, DB_COLLINEAR, optlist);
OCTOPUS_ASSERT(error == 0);
}
{
DBoptlist* optlist = DBMakeOptlist(3);
// REVIEW: Verify this.
int type = DB_ROWMAJOR;
DBAddOption(optlist, DBOPT_MAJORORDER, &type);
error = DBPutQuadvar1(file
, value_name.c_str()
, mesh_name.c_str()
, variables.get()
, nzones
, 3, NULL, 0, DB_DOUBLE, DB_ZONECENT, optlist);
OCTOPUS_ASSERT(error == 0);
}
delete[] coordinates[0];
delete[] coordinates[1];
delete[] coordinates[2];
sync.post();
} // }}}
/*!
Returns all the variables declared in the global context.
\sa functions(), classes()
*/
QStringList QSInterpreter::variables(bool includeStatic, bool includeCustom,
bool includeMemberVariables) const
{
return variables(QString::null, includeStatic, includeCustom,
includeMemberVariables);
}
void EBPoissonOp::
restrictResidual(LevelData<EBCellFAB>& a_resCoar,
LevelData<EBCellFAB>& a_phiThisLevel,
const LevelData<EBCellFAB>& a_rhsThisLevel)
{
CH_TIME("EBPoissonOp::restrictResidual");
CH_assert(a_resCoar.nComp() == 1);
CH_assert(a_phiThisLevel.nComp() == 1);
CH_assert(a_rhsThisLevel.nComp() == 1);
LevelData<EBCellFAB> resThisLevel;
bool homogeneous = true;
EBCellFactory ebcellfactTL(m_eblg.getEBISL());
IntVect ghostVec = a_rhsThisLevel.ghostVect();
resThisLevel.define(m_eblg.getDBL(), 1, ghostVec, ebcellfactTL);
// Get the residual on the fine grid
residual(resThisLevel,a_phiThisLevel,a_rhsThisLevel,homogeneous);
// now use our nifty averaging operator
Interval variables(0, 0);
CH_assert(m_hasMGObjects);
m_ebAverageMG.average(a_resCoar, resThisLevel, variables);
#ifdef DO_EB_RHS_CORRECTION
// Apply error-correction modification to restricted RHS
// Right now this only works with Dirichlet BC's, and only makes sense for the EB
int correctionType = 0; // Change this to activate RHS correction
if (correctionType != 0)
{
for (DataIterator dit = a_resCoar.disjointBoxLayout().dataIterator(); dit.ok(); ++dit)
{
EBCellFAB& resFAB = a_resCoar[dit()]; // Extract the FAB for this box
const EBISBox& ebis = resFAB.getEBISBox(); // Get the set of all IntVect indices
// Iterate over the parts of the RHS corresponding to the irregular cells,
// correcting the RHS in each cell
VoFIterator ebvofit(ebis.getIrregIVS(ebis.getRegion()), ebis.getEBGraph());
for (ebvofit.reset(); ebvofit.ok(); ++ebvofit)
{
for (int icomp = 0; icomp < resFAB.nComp(); ++icomp) // For each component of the residual on this VoF
{
if (correctionType == 1)
{
// Setting the residual to zero on the irregular cells gives convergence,
// though the rate isn't as good as the full correction
resFAB(ebvofit(), icomp) = 0.0;
}
else if (correctionType == 2)
{
// Kludge valid only for pseudo-1D test case. May not work for you.
Real kappa = ebis.volFrac(ebvofit());
kappa = (kappa > 0.5 ? kappa : 0.5); // Floor on kappa to prevent dividing by a tiny number
Real rhoCoeff = (kappa + 0.5)/(2.0*kappa);
resFAB(ebvofit(), icomp) *= (1.0 - rhoCoeff);
}
// else silently do nothing
}
}
}
}
#endif
}
Statement* Expand::operator()(Each* e)
{
vector<string> variables(e->variables());
Expression* expr = e->list()->perform(eval->with(env, backtrace));
List* list = 0;
Map* map = 0;
if (expr->concrete_type() == Expression::MAP) {
map = static_cast<Map*>(expr);
}
else if (expr->concrete_type() != Expression::LIST) {
list = new (ctx.mem) List(expr->path(), expr->position(), 1, List::COMMA);
*list << expr;
}
else {
list = static_cast<List*>(expr);
}
Env new_env;
for (size_t i = 0, L = variables.size(); i < L; ++i) new_env[variables[i]] = 0;
new_env.link(env);
env = &new_env;
Block* body = e->block();
if (map) {
for (auto key : map->keys()) {
Expression* k = key->perform(eval->with(env, backtrace));
Expression* v = map->at(key)->perform(eval->with(env, backtrace));
if (variables.size() == 1) {
List* variable = new (ctx.mem) List(map->path(), map->position(), 2, List::SPACE);
*variable << k;
*variable << v;
(*env)[variables[0]] = variable;
} else {
(*env)[variables[0]] = k;
(*env)[variables[1]] = v;
}
append_block(body);
}
}
else {
for (size_t i = 0, L = list->length(); i < L; ++i) {
List* variable = 0;
if ((*list)[i]->concrete_type() != Expression::LIST || variables.size() == 1) {
variable = new (ctx.mem) List((*list)[i]->path(), (*list)[i]->position(), 1, List::COMMA);
*variable << (*list)[i];
}
else {
variable = static_cast<List*>((*list)[i]);
}
for (size_t j = 0, K = variables.size(); j < K; ++j) {
if (j < variable->length()) {
(*env)[variables[j]] = (*variable)[j]->perform(eval->with(env, backtrace));
}
else {
(*env)[variables[j]] = new (ctx.mem) Null(expr->path(), expr->position());
}
}
append_block(body);
}
}
env = new_env.parent();
return 0;
}
TString TMVAPredict(TString method_name, EnumPredictMode predictMode = EnumPredictMode::FINAL)
{
std::cout << "------------ predict with : " << method_name << " ------ " << std::endl;
std::vector<std::string> inputNames = {"training","test","check_correlation","check_agreement"};
std::map<std::string,std::vector<std::string>> varsForInput;
std::vector<std::string> variableOrder = {"id", "signal", "mass", "min_ANNmuon", "prediction"};
varsForInput["training"].emplace_back ("prediction");
if (predictMode != EnumPredictMode::INTERMEDIATE)
{
varsForInput["training"].emplace_back ("id");
varsForInput["training"].emplace_back ("signal");
varsForInput["training"].emplace_back ("mass");
varsForInput["training"].emplace_back ("min_ANNmuon");
varsForInput["test"].emplace_back ("prediction");
varsForInput["test"].emplace_back ("id");
varsForInput["check_agreement"].emplace_back ("signal");
varsForInput["check_agreement"].emplace_back ("weight");
varsForInput["check_agreement"].emplace_back ("prediction");
varsForInput["check_correlation"].emplace_back ("mass");
varsForInput["check_correlation"].emplace_back ("prediction");
}
std::map<std::string,std::vector<std::string>> createForInput;
createForInput["training"].emplace_back ("root");
if (predictMode != EnumPredictMode::INTERMEDIATE)
{
createForInput["training"].emplace_back ("csv");
createForInput["test"].emplace_back ("csv");
createForInput["check_agreement"].emplace_back ("csv");
createForInput["check_correlation"].emplace_back ("csv");
}
// -------- prepare the Reader ------
TMVA::Tools::Instance();
std::cout << "==> Start TMVAPredict" << std::endl;
TMVA::Reader *reader = new TMVA::Reader( "!Color:!Silent" );
std::vector<Float_t> variables (variableNames.size ());
auto itVar = begin (variables);
for (auto varName : variableNames)
{
Float_t* pVar = &(*itVar);
auto localVarName = varName;
localVarName.substr(0,localVarName.find(":="));
reader->AddVariable(varName.c_str(), pVar);
(*itVar) = 0.0;
++itVar;
}
// spectators not known for the reader (in test.csv)
for (auto varName : spectatorNames)
{
Float_t spectator (0.0);
reader->AddSpectator (varName.c_str(), &spectator);
++itVar;
}
TString dir = "weights/";
TString prefix = "TMVAClassification";
TString weightfile = dir + prefix + TString("_") + method_name + TString(".weights.xml");
std::cout << "weightfile name : " << weightfile.Data () << std::endl;
reader->BookMVA( method_name, weightfile );
// --------- for each of the input files
for (auto inputName : inputNames)
{
// --- define variables
Int_t id;
Float_t prediction;
Float_t weight;
Float_t min_ANNmuon;
Float_t mass;
Float_t signal;
// --- open input file
TFile *input(0);
std::stringstream infilename;
infilename << pathToData.Data () << inputName << ".root";
std::cout << "infilename = " << infilename.str ().c_str () << std::endl;
input = TFile::Open (infilename.str ().c_str ());
TTree* tree = (TTree*)input->Get("data");
// --- prepare branches on input file
// id field if needed
if (contains (varsForInput, inputName, "id"))
tree->SetBranchAddress("id", &id);
// signal field if needed
if (contains (varsForInput, inputName, "signal"))
tree->SetBranchAddress("signal", &signal);
// min_ANNmuon field if needed
if (contains (varsForInput, inputName, "min_ANNmuon"))
tree->SetBranchAddress("min_ANNmuon", &min_ANNmuon);
// mass field if needed
if (contains (varsForInput, inputName, "mass"))
tree->SetBranchAddress("mass", &mass);
// weight field if needed
if (contains (varsForInput, inputName, "weight"))
tree->SetBranchAddress("weight", &weight);
// variables for prediction
itVar = begin (variables);
for (auto inputName : variableNames)
{
Float_t* pVar = &(*itVar);
tree->SetBranchAddress(inputName.c_str(), pVar);
++itVar;
}
// ---- make ROOT file
TString rootFileName;
TFile* outRootFile (NULL);
TTree* outTree (NULL);
if (contains (createForInput, inputName, "root"))
{
rootFileName = TString (inputName.c_str ()) + TString ("_prediction__") + method_name + TString (".root");
outRootFile = new TFile (rootFileName.Data (), "RECREATE");
outTree = new TTree("data","data");
if (contains (varsForInput, inputName, "id"))
outTree->Branch ("id", &id, "F");
if (contains (varsForInput, inputName, "signal"))
outTree->Branch ("signal", &signal, "F");
if (contains (varsForInput, inputName, "min_ANNmuon"))
outTree->Branch ("min_ANNmuon", &min_ANNmuon, "F");
if (contains (varsForInput, inputName, "mass"))
outTree->Branch ("mass", &mass, "F");
if (contains (varsForInput, inputName, "weight"))
outTree->Branch ("weight", &weight, "F");
if (contains (varsForInput, inputName, "prediction"))
outTree->Branch ("prediction", &prediction, "F");
}
// ---- prepare csv file
std::ofstream outfile;
if (contains (createForInput, inputName, "csv"))
{
std::stringstream outfilename;
outfilename << inputName << "_prediction__" << method_name.Data () << ".csv";
std::cout << outfilename.str () << std::endl;
/* return; */
outfile.open (outfilename.str ());
bool isFirst = true;
for (auto varName : variableOrder)
{
if (contains (varsForInput, inputName, varName))
{
if (!isFirst)
outfile << ",";
isFirst = false;
outfile << varName;
}
}
outfile << "\n";
}
bool doCSV = contains (createForInput, inputName, "csv");
bool doROOT = contains (createForInput, inputName, "root");
for (Long64_t ievt=0; ievt < tree->GetEntries(); ievt++)
{
tree->GetEntry(ievt);
// predict
prediction = reader->EvaluateMVA (method_name);
prediction = std::max<double> (0.0, std::min<double> (1.0, prediction));
//prediction = (prediction + 1.0)/2.0;
if (doCSV)
{
for (auto varName : variableOrder)
{
if (varName == "id" && contains (varsForInput, inputName, "id"))
outfile << id << ",";
if (varName == "signal" && contains (varsForInput, inputName, "signal"))
outfile << signal << ",";
if (varName == "min_ANNmuon" && contains (varsForInput, inputName, "min_ANNmuon"))
outfile << min_ANNmuon << ",";
if (varName == "mass" && contains (varsForInput, inputName, "mass"))
outfile << mass << ",";
if (varName == "weight" && contains (varsForInput, inputName, "weight"))
outfile << weight << ",";
if (varName == "prediction" && contains (varsForInput, inputName, "prediction"))
outfile << prediction;
}
outfile << "\n";
}
if (doROOT)
{
outTree->Fill ();
}
}
outfile.close();
input->Close();
if (doROOT)
{
outRootFile->Write ();
}
if (predictMode == EnumPredictMode::INTERMEDIATE)
{
delete reader;
std::cout << "DONE predict INTERMEDIATE" << std::endl;
return rootFileName;
}
}
delete reader;
if (predictMode == EnumPredictMode::FINAL)
{
std::cout << "DONE predict FINAL" << std::endl;
TString cmd (".! python tests.py ");
cmd += method_name;
gROOT->ProcessLine (cmd);
}
return method_name;
}
void createCDF ()
{
std::cout << "==> create CDF" << std::endl;
std::vector<std::string> inputNames = {"training"};
for (auto inputName : inputNames)
{
std::stringstream outfilename;
outfilename << inputName << "_cdf__" << inputName << ".root";
std::cout << outfilename.str () << std::endl;
/* return; */
std::stringstream infilename;
infilename << pathToData.Data () << inputName << ".root";
TFile *input(0);
std::cout << "infilename = " << infilename.str ().c_str () << std::endl;
input = TFile::Open (infilename.str ().c_str ());
TTree* tree = (TTree*)input->Get("data");
// variables for prediction
std::cout << "prepare variables" << std::endl;
auto localVariableNames = variableNames+additionalVariableNames;
std::vector<Float_t> variables (localVariableNames.size ());
auto itVar = begin (variables);
for (auto inputName : localVariableNames)
{
Float_t* pVar = &(*itVar);
tree->SetBranchAddress(inputName.c_str(), pVar);
++itVar;
}
Int_t id;
// id field
tree->SetBranchAddress("id", &id);
Long64_t ievtEnd = tree->GetEntries ();
ievtEnd = 100;
std::cout << "process entries #" << ievtEnd << std::endl;
std::vector<double> sumSmaller (ievtEnd, 0.0);
struct Vars
{
typedef std::vector<Float_t>::const_iterator iterator;
Vars (iterator itBegin, iterator itEnd, Float_t _weight, Int_t _id, Long64_t _order)
: variables (itBegin, itEnd)
, weight (_weight)
, id (_id)
, order (_order)
{
}
std::vector<Float_t> variables;
Int_t id;
Float_t weight;
Float_t cdf;
Long64_t order;
bool operator< (const Vars& other) const
{
for (auto itOther = begin (other.variables), it = begin (variables),
itOtherEnd = end (other.variables), itEnd = end (variables);
it != itEnd && itOther != itOtherEnd; ++itOther, ++it)
{
//std::cout << "(" << *it << "," << *itOther << ")" << std::flush;
if (*it >= *itOther)
{
// std::cout << "X" << std::flush;
return false;
}
else
std::cout << "D" << std::flush;
}
std::cout << "U" << std::flush;
return true;
}
};
Float_t weightSum (0.0);
std::vector<Vars> vars;
for (Long64_t ievt=0; ievt < ievtEnd; ievt++)
{
tree->GetEntry (ievt);
std::cout << "." << std::flush;
Float_t weight = 1.0;
vars.emplace_back (begin (variables), end (variables), weight, id, ievt);
weightSum += weight;
}
std::cout << "provide values" << std::endl;
for (auto it = begin (vars), itEnd = end (vars); it != itEnd; ++it)
{
std::cout << "-" << std::flush;
for (auto itCmp = begin (vars), itCmpEnd = end (vars); itCmp != itCmpEnd; ++itCmp)
{
if (*it < *itCmp)
{
std::cout << "!" << std::flush;
break;
}
else
{
std::cout << "+" << std::flush;
(*it).cdf += (*itCmp).weight;
}
}
}
std::cout << "normalize" << std::endl;
for_each (begin (vars), end (vars), [weightSum](Vars& v) {
v.cdf /= weightSum;
});
// sort by order
std::sort (begin (vars), end (vars), [](const Vars& lhs, const Vars& rhs){
return lhs.order < rhs.order;
});
input->Close();
std::cout << "store data" << std::endl;
TFile* outFile = new TFile (outfilename.str ().c_str (), "RECREATE");
TTree* outTree = new TTree("cdf_raw","cdf_raw");
Float_t cdf (0.0);
outTree->Branch ("id", &id, "F");
outTree->Branch ("cdf", &cdf, "F");
for (auto v : vars)
{
id = v.id;
cdf = v.cdf;
outTree->Fill ();
}
outFile->Write ();
outFile->Close ();
}
}
TString useAutoencoder (TString method_name)
{
TMVA::Tools::Instance();
std::cout << "==> Start useAutoencoder" << std::endl;
TMVA::Reader *reader = new TMVA::Reader( "!Color:!Silent" );
Float_t signal = 0.0;
Float_t outSignal = 0.0;
Float_t inSignal = 0.0;
std::vector<std::string> localVariableNames (variableNames+additionalVariableNames);
std::vector<Float_t> variables (localVariableNames.size ());
auto itVar = begin (variables);
for (auto varName : localVariableNames)
{
Float_t* pVar = &(*itVar);
reader->AddVariable(varName.c_str(), pVar);
(*itVar) = 0.0;
++itVar;
}
int idxSignal = std::distance (localVariableNames.begin (),
std::find (localVariableNames.begin (), localVariableNames.end (),std::string ("signal")));
TString dir = "weights/";
TString prefix = "TMVAAutoencoder";
TString weightfile = dir + prefix + TString("_") + method_name + TString(".weights.xml");
TString outPrefix = "transformed";
TString outfilename = pathToData + outPrefix + TString("_") + method_name + TString(".root");
reader->BookMVA( method_name, weightfile );
TFile* outFile = new TFile (outfilename.Data (), "RECREATE");
std::vector<std::string> inputNames = {"training"};
std::map<std::string,std::vector<std::string>> varsForInput;
varsForInput["training"].emplace_back ("id");
varsForInput["training"].emplace_back ("signal");
for (auto inputName : inputNames)
{
std::stringstream outfilename;
outfilename << inputName << "_transformed__" << method_name.Data () << ".root";
std::cout << outfilename.str () << std::endl;
/* return; */
std::stringstream infilename;
infilename << pathToData.Data () << inputName << ".root";
TTree* outTree = new TTree("transformed","transformed");
std::vector<Float_t> outVariables (localVariableNames.size ());
itVar = begin (variables);
auto itOutVar = begin (outVariables);
for (auto varName : localVariableNames)
{
Float_t* pOutVar = &(*itOutVar);
outTree->Branch (varName.c_str (), pOutVar, "F");
(*itOutVar) = 0.0;
++itOutVar;
Float_t* pVar = &(*itVar);
std::stringstream svar;
svar << varName << "_in";
outTree->Branch (svar.str ().c_str (), pVar, "F");
(*itVar) = 0.0;
++itVar;
}
Float_t signal_original = 0.0;
outTree->Branch ("signal_original", &signal_original, "F");
TFile *input(0);
std::cout << "infilename = " << infilename.str ().c_str () << std::endl;
input = TFile::Open (infilename.str ().c_str ());
TTree* tree = (TTree*)input->Get("data");
Int_t ids;
// id field if needed
if (std::find (varsForInput[inputName].begin (), varsForInput[inputName].end (), "id") != varsForInput[inputName].end ())
tree->SetBranchAddress("id", &ids);
// variables for prediction
itVar = begin (variables);
for (auto inputName : localVariableNames)
{
Float_t* pVar = &(*itVar);
tree->SetBranchAddress (inputName.c_str(), pVar);
++itVar;
}
for (Long64_t ievt=0; ievt < tree->GetEntries(); ievt++)
{
tree->GetEntry(ievt);
// predict
signal_original = variables.at (idxSignal);
for (int forcedSignal = 0; forcedSignal <= 1; ++forcedSignal)
{
variables.at (idxSignal) = forcedSignal;
std::vector<Float_t> regressionValues = reader->EvaluateRegression (method_name);
size_t idx = 0;
for (auto it = std::begin (regressionValues), itEnd = std::end (regressionValues); it != itEnd; ++it)
{
outVariables.at (idx) = *it;
++idx;
}
outTree->Fill ();
}
}
outFile->Write ();
input->Close();
}
delete reader;
return outfilename;
}
QStringList pEnvironmentVariablesManager::variables( bool keepDisabled ) const
{
return pEnvironmentVariablesModel::variablesToStringList( variables(), keepDisabled );
}
ProblemInstance::ProblemInstance (RPHandle hndl, SharedPtr<ReaderPluginFunctionTable> functions)
: handle(hndl), ftable(functions)
{
int i;
m_variables = ftable->variables(handle);
m_constraints = ftable->constraints(handle);
m_objectives = ftable->objectives(handle);
m_variableTypes = new char[variables()+2*constraints()+2*objectives()];
m_functionTypes = &m_variableTypes[variables()];
m_constraintTypes = &m_functionTypes[constraints()+objectives()];
m_objectiveTypes = &m_constraintTypes[constraints()];
for (i = 0; i < variables(); ++i)
{
m_variableTypes[i] = ftable->variableType(handle, i);
}
for (i = 0; i < constraints(); ++i)
{
m_functionTypes[i] = ftable->functionType(handle, 'c', i);
}
for (i = 0; i < objectives(); ++i)
{
m_functionTypes[constraints()+i] = ftable->functionType(handle, 'o', i);
}
for (i = 0; i < constraints(); ++i)
{
m_constraintTypes[i] = ftable->constraintType(handle, i);
}
for (i = 0; i < objectives(); ++i)
{
m_objectiveTypes[i] = ftable->objectiveType(handle, i);
}
int* tmp = new int[std::max(objectives()+constraints()+2, variables())];
m_presenceIndexes = new unsigned[2*variables()+constraints()+objectives()];
for (i = 0; i < variables(); ++i)
{
ftable->variablePresence(handle, i, tmp, &tmp[constraints()]);
m_presences.reserve(m_presences.size()+2+tmp[0]+tmp[constraints()]);
varConstrPresenceIndex(i) = m_presences.size();
m_presences.insert(m_presences.end(), tmp, &tmp[tmp[0]+1]);
varObjPresenceIndex(i) = m_presences.size();
m_presences.insert(m_presences.end(), &tmp[constraints()], &tmp[constraints()+tmp[constraints()]+1]);
}
m_variableCounts = new int[4*(constraints()+objectives())];
std::fill(m_variableCounts, &m_variableCounts[4*(constraints()+objectives())], 0);
for (i = 0; i < constraints(); ++i)
{
int vars = ftable->constraintVariables(handle, i, tmp);
m_presences.reserve(m_presences.size()+1+vars);
constraintPresenceIndex(i) = m_presences.size();
m_presences.push_back(vars);
m_presences.insert(m_presences.end(), tmp, &tmp[vars]);
unsigned offset = 4*i;
m_variableCounts[offset] = vars;
for (int j = 0; j < vars; ++j)
{
switch (variableType(tmp[j])) {
case 'r': ++m_variableCounts[offset+1]; break;
case 'b': ++m_variableCounts[offset+2]; break;
case 'i': ++m_variableCounts[offset+3]; break;
}
}
}
for (i = 0; i < objectives(); ++i)
{
int vars = ftable->objectiveVariables(handle, i, tmp);
m_presences.reserve(m_presences.size()+1+vars);
objectivePresenceIndex(i) = m_presences.size();
m_presences.push_back(vars);
m_presences.insert(m_presences.end(), tmp, &tmp[vars]);
unsigned offset = 4*(constraints()+i);
m_variableCounts[offset] = vars;
for (int j = 0; j < vars; ++j)
{
switch (variableType(tmp[j])) {
case 'r': ++m_variableCounts[offset+1]; break;
case 'b': ++m_variableCounts[offset+2]; break;
case 'i': ++m_variableCounts[offset+3]; break;
}
}
}
delete [] tmp;
m_varBounds = new Real[2*(variables()+constraints())];
m_constrBounds = &m_varBounds[2*variables()];
for (i = 0; i < variables(); ++i)
{
ftable->variableBounds(handle, i, &m_varBounds[2*i], &m_varBounds[2*i+1]);
}
for (i = 0; i < constraints(); ++i)
{
ftable->constraintBounds(handle, i, &m_constrBounds[2*i], &m_constrBounds[2*i+1]);
switch (constraintType(i)) {
case 'l':
m_constrBounds[2*i] = -std::numeric_limits<Real>::infinity();
break;
case 'g':
m_constrBounds[2*i+1] = std::numeric_limits<Real>::infinity();
break;
case 'e':
m_constrBounds[2*i] = m_constrBounds[2*i+1];
break;
case 'u':
m_constrBounds[2*i] = -std::numeric_limits<Real>::infinity();
m_constrBounds[2*i+1] = std::numeric_limits<Real>::infinity();
break;
default:
break;
}
}
latestErrorType = 0;
}