bool runTestType(cl::Context context, cl::CommandQueue queue)
{
cl_uint size = (1 << 16) + 16384;
std::vector<T> input(size);
std::cout << "##Testing AMD radix sort for " << input.size() << " elements and type "
<< magnet::CL::detail::traits<T>::kernel_type();
for(size_t i = 0; i < input.size(); ++i)
input[i] = input.size() - i - 1;
// create input buffer using pinned memory
cl::Buffer bufferIn(context, CL_MEM_ALLOC_HOST_PTR |
CL_MEM_COPY_HOST_PTR | CL_MEM_READ_WRITE,
sizeof(T) * input.size(), &input[0]);
magnet::CL::radixSortAMD<T> radixSortFunctor;
radixSortFunctor.build(queue, context);
radixSortFunctor(bufferIn, bufferIn);
std::vector<T> output(size);
queue.enqueueReadBuffer(bufferIn, CL_TRUE, 0, input.size() *
sizeof(T), &output[0]);
bool failed = !testOutput(input, output);
std::cout << " key(only) " << (failed ? "FAILED" : "PASSED") << ", ";
//Now test with some data!
//Refresh the input array
queue.enqueueWriteBuffer(bufferIn, CL_TRUE, 0, input.size() *
sizeof(T), &input[0]);
//Write a data array
std::vector<cl_uint> data(size);
for(size_t i = 0; i < input.size(); ++i)
data[i] = i;
cl::Buffer dataIn(context, CL_MEM_ALLOC_HOST_PTR |
CL_MEM_COPY_HOST_PTR | CL_MEM_READ_WRITE,
sizeof(cl_uint) * data.size(), &data[0])
;
radixSortFunctor(bufferIn, dataIn, bufferIn, dataIn);
queue.enqueueReadBuffer(dataIn, CL_TRUE, 0, data.size() *
sizeof(cl_uint), &data[0]);
bool keyfail = !testOutput(input, output);
std::cout << " key " << (keyfail ? "FAILED" : "PASSED");
bool datafail = false;
for(size_t i = 0; i < input.size(); ++i)
if (data[i] != input.size() - 1 - i)
datafail = true;
std::cout << " data " << (datafail ? "FAILED" : "PASSED") << std::endl;
return failed || keyfail || datafail;
}
int main(void)
{
printf( "Testing output and speed of MersenneTwister.h\n" );
printf( "\nTest of random integer generation:\n" );
MTRand mtrand1( 4357U );
for( int i = 0; i < 1000; ++i )
{
printf( "%10lu ", mtrand1.randInt() );
if( i % 5 == 4 ) printf("\n");
}
printf( "\nTest of random real number [0,1] generation:\n" );
MTRand mtrand2( 4357U );
for( int i = 0; i < 1000; ++i )
{
printf( "%10.8f ", mtrand2.rand() );
if( i % 5 == 4 ) printf("\n");
}
printf( "\nTest of random real number [0,1) generation:\n" );
MTRand mtrand3( 4357U );
for( int i = 0; i < 1000; ++i )
{
printf( "%10.8f ", mtrand3.randExc() );
if( i % 5 == 4 ) printf("\n");
}
printf( "\nTest of time to generate 300 million random integers:\n" );
MTRand mtrand4( 4357U );
unsigned long junk;
clock_t startClock = clock();
for( long i = 0; i < 3e+8; ++i )
{
junk = mtrand4.randInt();
}
clock_t stopClock = clock();
printf( "Time elapsed = " );
printf( "%8.3f", double( stopClock - startClock ) / CLOCKS_PER_SEC );
printf( " s\n" );
printf( "\nTests of functionality:\n" );
// Array save/load test
bool saveArrayFailure = false;
MTRand mtrand5;
unsigned long pass1[5], pass2[5];
MTRand::uint32 saveArray[ MTRand::SAVE ];
mtrand5.save( saveArray );
for( int i = 0; i < 5; ++i )
pass1[i] = mtrand5.randInt();
mtrand5.load( saveArray );
for( int i = 0; i < 5; ++i )
{
pass2[i] = mtrand5.randInt();
if( pass2[i] != pass1[i] )
saveArrayFailure = true;
}
if( saveArrayFailure )
printf( "Error - Failed array save/load test\n" );
else
printf( "Passed array save/load test\n" );
// Stream save/load test
bool saveStreamFailure = false;
std::ofstream dataOut( "state.data" );
if( dataOut )
{
dataOut << mtrand5; // comment out if compiler does not support
dataOut.close();
}
for( int i = 0; i < 5; ++i )
pass1[i] = mtrand5.randInt();
std::ifstream dataIn( "state.data" );
if( dataIn )
{
dataIn >> mtrand5; // comment out if compiler does not support
dataIn.close();
}
for( int i = 0; i < 5; ++i )
{
pass2[i] = mtrand5.randInt();
if( pass2[i] != pass1[i] )
saveStreamFailure = true;
}
if( saveStreamFailure )
printf( "Error - Failed stream save/load test\n" );
else
printf( "Passed stream save/load test\n" );
// Integer range test
MTRand mtrand6;
bool integerRangeFailure = false;
bool gotMax = false;
for( int i = 0; i < 10000; ++i )
{
int r = mtrand6.randInt(17);
if( r < 0 || r > 17 )
integerRangeFailure = true;
if( r == 17 )
gotMax = true;
}
if( !gotMax )
integerRangeFailure = true;
if( integerRangeFailure )
printf( "Error - Failed integer range test\n" );
else
printf( "Passed integer range test\n" );
// Float range test
MTRand mtrand7;
bool floatRangeFailure = false;
for( int i = 0; i < 10000; ++i )
{
float r = mtrand7.rand(0.3183);
if( r < 0.0 || r > 0.3183 )
floatRangeFailure = true;
}
if( floatRangeFailure )
printf( "Error - Failed float range test\n" );
else
printf( "Passed float range test\n" );
// Auto-seed uniqueness test
MTRand mtrand8a, mtrand8b, mtrand8c;
double r8a = mtrand8a();
double r8b = mtrand8b();
double r8c = mtrand8c();
if( r8a == r8b || r8a == r8c || r8b == r8c )
printf( "Error - Failed auto-seed uniqueness test\n" );
else
printf( "Passed auto-seed uniqueness test\n" );
return 0;
}
int main(int argc, char* argv[])
{
std::cerr << "STARTING\n";
std::cerr << "Opening config file" << argv[1] << "\n";
std::ifstream inFile(argv[1],std::ios::in);
std::vector<double> dataIn(7,0.0F);
std::vector< std::vector<double> > vectorIn(1000,dataIn);
int column = 0;
int row = 0;
int family = 0;
bool comment = false;
std::string in;
std::cerr << "Parsing config file" << argv[1] << "\n";
while (inFile >> in)
{
if(!in.compare("#")) //comment code # found
{
if(!comment)
{
comment = true;
}
else //end of comment
{
comment = false;
}
}
if((!comment) && in.compare("#")) //real line found
{
if(column == 1)
{
//vectorIn[row][family] = atof(in.c_str());
//std::cerr << "Adding " << in << "[" << row << "]"
// << "[" << family*2 << "]\n";
}
if(column == 3)
{
vectorIn[row][family] = atof(in.c_str());
//std::cerr << "Adding " << in << "[" << row << "]"
// << "[" << (family*2)+1 << "]\n";
}
column++;
if(column == 5)
{
column = 0;
row++;
}
if(!in.compare("%"))
{
row = 0;
column = 0;
family++;
std::cerr << "New Family " << family << "\n";
}
}
}
// start timer
Timer tim;
tim.reset();
uint64 t0 = tim.get(); // to measure display time
// get test image
if(argc > 2)
itemNumber = atoi(argv[2]);
else
itemNumber = -666;
// create operating objects
configIn.openFile("NPclassify.conf");
polySet.openFile("polySet.conf");
NPclassify NP(configIn,polySet,true);
if(argc > 3)
{
NP.inputCommandLineSettings(atof(argv[3]),atof(argv[4]),atof(argv[5]),
atof(argv[6]),atoi(argv[7]),atoi(argv[8]),
atof(argv[9]),atof(argv[10]),atof(argv[11]));
}
//inport data
std::vector<double> feature(2,0);
std::vector<long> roots;
std::vector<long> parents;
std::vector<double> density;
NP.addSpace(vectorIn,(row-1));
NP.classifySpaceNew();
uint64 t1 = tim.get();
t0 = t1 - t0;
roots = NP.getStems();
parents = NP.getParents();
density = NP.getDensity();
std::vector<std::vector<long> > theReturn = NP.getChildren();
for(int i = 0; i < NP.getStemNumber(); i++)
{
if(NP.getClassSize(i) > NP.getMinClassSize())
{
LINFO("CLASS %d size %ld",i,NP.getClassSize(i));
}
}
//outport classification data
NP.metaClassify(itemNumber);
std::ofstream outfile("feature_trian.dat",std::ios::out);
std::ofstream matfile("feature_trian_mat.dat",std::ios::out);
for(int i = 0; i < NP.getStemNumber(); i++)
{
for(int j = 0; j < NP.getClassSize(i); j++)
{
long item = NP.getClass(i,j);
outfile << item << "\t" << i << "\t" << j << "\t"
<< NP.getFeature(item,0) << "\t"
<< NP.getFeature(item,1) << "\n";
matfile << item << "\t" << i << "\n";
}
}
}
int main(int argc, char ** argv) {
MPI_Init(&argc, &argv);
NcError error(NcError::silent_nonfatal);
try {
// Input filename
std::string strInputFile;
// Output filename
std::string strOutputFile;
// Separate topography file
std::string strTopographyFile;
// List of variables to extract
std::string strVariables;
// Extract geopotential height
bool fGeopotentialHeight;
// Pressure levels to extract
std::string strPressureLevels;
// Height levels to extract
std::string strHeightLevels;
// Extract variables at the surface
bool fExtractSurface;
// Extract total energy
bool fExtractTotalEnergy;
// Parse the command line
BeginCommandLine()
CommandLineString(strInputFile, "in", "");
CommandLineString(strOutputFile, "out", "");
CommandLineString(strVariables, "var", "");
CommandLineBool(fGeopotentialHeight, "output_z");
CommandLineBool(fExtractTotalEnergy, "output_energy");
CommandLineString(strPressureLevels, "p", "");
CommandLineString(strHeightLevels, "z", "");
CommandLineBool(fExtractSurface, "surf");
ParseCommandLine(argc, argv);
EndCommandLine(argv)
AnnounceBanner();
// Check command line arguments
if (strInputFile == "") {
_EXCEPTIONT("No input file specified");
}
if (strOutputFile == "") {
_EXCEPTIONT("No output file specified");
}
if (strVariables == "") {
_EXCEPTIONT("No variables specified");
}
// Parse variable string
std::vector< std::string > vecVariableStrings;
ParseVariableList(strVariables, vecVariableStrings);
// Check variables
if (vecVariableStrings.size() == 0) {
_EXCEPTIONT("No variables specified");
}
// Parse pressure level string
std::vector<double> vecPressureLevels;
ParseLevelArray(strPressureLevels, vecPressureLevels);
int nPressureLevels = (int)(vecPressureLevels.size());
for (int k = 0; k < nPressureLevels; k++) {
if (vecPressureLevels[k] <= 0.0) {
_EXCEPTIONT("Non-positive pressure values not allowed");
}
}
// Parse height level string
std::vector<double> vecHeightLevels;
ParseLevelArray(strHeightLevels, vecHeightLevels);
int nHeightLevels = (int)(vecHeightLevels.size());
// Check pressure levels
if ((nPressureLevels == 0) &&
(nHeightLevels == 0) &&
(!fExtractSurface)
) {
_EXCEPTIONT("No pressure / height levels to process");
}
// Open input file
AnnounceStartBlock("Loading input file");
NcFile ncdf_in(strInputFile.c_str(), NcFile::ReadOnly);
if (!ncdf_in.is_valid()) {
_EXCEPTION1("Unable to open file \"%s\" for reading",
strInputFile.c_str());
}
// Load time array
Announce("Time");
NcVar * varTime = ncdf_in.get_var("time");
if (varTime == NULL) {
_EXCEPTION1("File \"%s\" does not contain variable \"time\"",
strInputFile.c_str());
}
int nTime = varTime->get_dim(0)->size();
DataArray1D<double> dTime(nTime);
varTime->set_cur((long)0);
varTime->get(&(dTime[0]), nTime);
// Load latitude array
Announce("Latitude");
NcVar * varLat = ncdf_in.get_var("lat");
if (varLat == NULL) {
_EXCEPTION1("File \"%s\" does not contain variable \"lat\"",
strInputFile.c_str());
}
int nLat = varLat->get_dim(0)->size();
DataArray1D<double> dLat(nLat);
varLat->set_cur((long)0);
varLat->get(&(dLat[0]), nLat);
// Load longitude array
Announce("Longitude");
NcVar * varLon = ncdf_in.get_var("lon");
if (varLon == NULL) {
_EXCEPTION1("File \"%s\" does not contain variable \"lon\"",
strInputFile.c_str());
}
int nLon = varLon->get_dim(0)->size();
DataArray1D<double> dLon(nLon);
varLon->set_cur((long)0);
varLon->get(&(dLon[0]), nLon);
// Load level array
Announce("Level");
NcVar * varLev = ncdf_in.get_var("lev");
if (varLev == NULL) {
_EXCEPTION1("File \"%s\" does not contain variable \"lev\"",
strInputFile.c_str());
}
int nLev = varLev->get_dim(0)->size();
DataArray1D<double> dLev(nLev);
varLev->set_cur((long)0);
varLev->get(&(dLev[0]), nLev);
// Load level interface array
Announce("Interface");
NcVar * varILev = ncdf_in.get_var("ilev");
int nILev = 0;
DataArray1D<double> dILev;
if (varILev == NULL) {
Announce("Warning: Variable \"ilev\" not found");
} else {
nILev = varILev->get_dim(0)->size();
if (nILev != nLev + 1) {
_EXCEPTIONT("Variable \"ilev\" must have size lev+1");
}
dILev.Allocate(nILev);
varILev->set_cur((long)0);
varILev->get(&(dILev[0]), nILev);
}
// Load topography
Announce("Topography");
NcVar * varZs = ncdf_in.get_var("Zs");
if (varZs == NULL) {
_EXCEPTION1("File \"%s\" does not contain variable \"Zs\"",
strInputFile.c_str());
}
DataArray2D<double> dZs(nLat, nLon);
varZs->set_cur((long)0, (long)0);
varZs->get(&(dZs[0][0]), nLat, nLon);
AnnounceEndBlock("Done");
// Open output file
AnnounceStartBlock("Constructing output file");
NcFile ncdf_out(strOutputFile.c_str(), NcFile::Replace);
if (!ncdf_out.is_valid()) {
_EXCEPTION1("Unable to open file \"%s\" for writing",
strOutputFile.c_str());
}
CopyNcFileAttributes(&ncdf_in, &ncdf_out);
// Output time array
Announce("Time");
NcDim * dimOutTime = ncdf_out.add_dim("time");
NcVar * varOutTime = ncdf_out.add_var("time", ncDouble, dimOutTime);
varOutTime->set_cur((long)0);
varOutTime->put(&(dTime[0]), nTime);
CopyNcVarAttributes(varTime, varOutTime);
// Output pressure array
NcDim * dimOutP = NULL;
NcVar * varOutP = NULL;
if (nPressureLevels > 0) {
Announce("Pressure");
dimOutP = ncdf_out.add_dim("p", nPressureLevels);
varOutP = ncdf_out.add_var("p", ncDouble, dimOutP);
varOutP->set_cur((long)0);
varOutP->put(&(vecPressureLevels[0]), nPressureLevels);
}
// Output height array
NcDim * dimOutZ = NULL;
NcVar * varOutZ = NULL;
if (nHeightLevels > 0) {
Announce("Height");
dimOutZ = ncdf_out.add_dim("z", nHeightLevels);
varOutZ = ncdf_out.add_var("z", ncDouble, dimOutZ);
varOutZ->set_cur((long)0);
varOutZ->put(&(vecHeightLevels[0]), nHeightLevels);
}
// Output latitude and longitude array
Announce("Latitude");
NcDim * dimOutLat = ncdf_out.add_dim("lat", nLat);
NcVar * varOutLat = ncdf_out.add_var("lat", ncDouble, dimOutLat);
varOutLat->set_cur((long)0);
varOutLat->put(&(dLat[0]), nLat);
CopyNcVarAttributes(varLat, varOutLat);
Announce("Longitude");
NcDim * dimOutLon = ncdf_out.add_dim("lon", nLon);
NcVar * varOutLon = ncdf_out.add_var("lon", ncDouble, dimOutLon);
varOutLon->set_cur((long)0);
varOutLon->put(&(dLon[0]), nLon);
CopyNcVarAttributes(varLon, varOutLon);
// Output topography
Announce("Topography");
NcVar * varOutZs = ncdf_out.add_var(
"Zs", ncDouble, dimOutLat, dimOutLon);
varOutZs->set_cur((long)0, (long)0);
varOutZs->put(&(dZs[0][0]), nLat, nLon);
AnnounceEndBlock("Done");
// Done
AnnounceEndBlock("Done");
// Load all variables
Announce("Loading variables");
std::vector<NcVar *> vecNcVar;
for (int v = 0; v < vecVariableStrings.size(); v++) {
vecNcVar.push_back(ncdf_in.get_var(vecVariableStrings[v].c_str()));
if (vecNcVar[v] == NULL) {
_EXCEPTION1("Unable to load variable \"%s\" from file",
vecVariableStrings[v].c_str());
}
}
// Physical constants
Announce("Initializing thermodynamic variables");
NcAtt * attEarthRadius = ncdf_in.get_att("earth_radius");
double dEarthRadius = attEarthRadius->as_double(0);
NcAtt * attRd = ncdf_in.get_att("Rd");
double dRd = attRd->as_double(0);
NcAtt * attCp = ncdf_in.get_att("Cp");
double dCp = attCp->as_double(0);
double dGamma = dCp / (dCp - dRd);
NcAtt * attP0 = ncdf_in.get_att("P0");
double dP0 = attP0->as_double(0);
double dPressureScaling = dP0 * std::pow(dRd / dP0, dGamma);
NcAtt * attZtop = ncdf_in.get_att("Ztop");
double dZtop = attZtop->as_double(0);
// Input data
DataArray3D<double> dataIn(nLev, nLat, nLon);
DataArray3D<double> dataInt(nILev, nLat, nLon);
// Output data
DataArray2D<double> dataOut(nLat, nLon);
// Pressure in column
DataArray1D<double> dataColumnP(nLev);
// Height in column
DataArray1D<double> dataColumnZ(nLev);
DataArray1D<double> dataColumnIZ(nILev);
// Column weights
DataArray1D<double> dW(nLev);
DataArray1D<double> dIW(nILev);
// Loop through all times, pressure levels and variables
AnnounceStartBlock("Interpolating");
// Add energy variable
NcVar * varEnergy;
if (fExtractTotalEnergy) {
varEnergy = ncdf_out.add_var("TE", ncDouble, dimOutTime);
}
// Create output pressure variables
std::vector<NcVar *> vecOutNcVarP;
if (nPressureLevels > 0) {
for (int v = 0; v < vecVariableStrings.size(); v++) {
vecOutNcVarP.push_back(
ncdf_out.add_var(
vecVariableStrings[v].c_str(), ncDouble,
dimOutTime, dimOutP, dimOutLat, dimOutLon));
// Copy attributes
CopyNcVarAttributes(vecNcVar[v], vecOutNcVarP[v]);
}
}
// Create output height variables
std::vector<NcVar *> vecOutNcVarZ;
if (nHeightLevels > 0) {
for (int v = 0; v < vecVariableStrings.size(); v++) {
std::string strVarName = vecVariableStrings[v];
if (nPressureLevels > 0) {
strVarName += "z";
}
vecOutNcVarZ.push_back(
ncdf_out.add_var(
strVarName.c_str(), ncDouble,
dimOutTime, dimOutZ, dimOutLat, dimOutLon));
// Copy attributes
CopyNcVarAttributes(vecNcVar[v], vecOutNcVarZ[v]);
}
}
// Create output surface variable
std::vector<NcVar *> vecOutNcVarS;
if (fExtractSurface) {
for (int v = 0; v < vecVariableStrings.size(); v++) {
std::string strVarName = vecVariableStrings[v];
strVarName += "S";
vecOutNcVarS.push_back(
ncdf_out.add_var(
strVarName.c_str(), ncDouble,
dimOutTime, dimOutLat, dimOutLon));
// Copy attributes
CopyNcVarAttributes(vecNcVar[v], vecOutNcVarS[v]);
}
}
// Loop over all times
for (int t = 0; t < nTime; t++) {
char szAnnounce[256];
sprintf(szAnnounce, "Time %i", t);
AnnounceStartBlock(szAnnounce);
// Rho
DataArray3D<double> dataRho(nLev, nLat, nLon);
NcVar * varRho = ncdf_in.get_var("Rho");
if (varRho == NULL) {
_EXCEPTIONT("Unable to load variable \"Rho\" from file");
}
varRho->set_cur(t, 0, 0, 0);
varRho->get(&(dataRho[0][0][0]), 1, nLev, nLat, nLon);
// Pressure
DataArray3D<double> dataP(nLev, nLat, nLon);
if (nPressureLevels != 0) {
NcVar * varP = ncdf_in.get_var("P");
if (varP == NULL) {
_EXCEPTIONT("Unable to load variable \"P\" from file");
}
varP->set_cur(t, 0, 0, 0);
varP->get(&(dataP[0][0][0]), 1, nLev, nLat, nLon);
}
/*
// Populate pressure array
if (nPressureLevels > 0) {
// Calculate pointwise pressure
for (int k = 0; k < nLev; k++) {
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
dataP[k][i][j] = dPressureScaling
* exp(log(dataRho[k][i][j] * dataP[k][i][j]) * dGamma);
}
}
}
}
*/
// Height everywhere
DataArray3D<double> dataZ(nLev, nLat, nLon);
DataArray3D<double> dataIZ;
if (nILev != 0) {
dataIZ.Allocate(nILev, nLat, nLon);
}
// Populate height array
if ((nHeightLevels > 0) || (fGeopotentialHeight)) {
for (int k = 0; k < nLev; k++) {
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
dataZ[k][i][j] = dZs[i][j] + dLev[k] * (dZtop - dZs[i][j]);
}
}
}
for (int k = 0; k < nILev; k++) {
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
dataIZ[k][i][j] = dZs[i][j] + dILev[k] * (dZtop - dZs[i][j]);
}
}
}
}
// Loop through all pressure levels and variables
for (int v = 0; v < vecNcVar.size(); v++) {
bool fOnInterfaces = false;
// Load in the data array
vecNcVar[v]->set_cur(t, 0, 0, 0);
if (vecNcVar[v]->get_dim(1)->size() == nLev) {
vecNcVar[v]->get(&(dataIn[0][0][0]), 1, nLev, nLat, nLon);
Announce("%s (n)", vecVariableStrings[v].c_str());
} else if (vecNcVar[v]->get_dim(1)->size() == nILev) {
vecNcVar[v]->get(&(dataInt[0][0][0]), 1, nILev, nLat, nLon);
fOnInterfaces = true;
Announce("%s (i)", vecVariableStrings[v].c_str());
} else {
_EXCEPTION1("Variable \"%s\" has invalid level dimension",
vecVariableStrings[v].c_str());
}
// At the physical surface
if (fExtractSurface) {
if (fOnInterfaces) {
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
dataOut[i][j] = dataInt[0][i][j];
}
}
} else {
int kBegin = 0;
int kEnd = 3;
PolynomialInterp::LagrangianPolynomialCoeffs(
3, dLev, dW, 0.0);
// Loop thorugh all latlon indices
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
// Interpolate in the vertical
dataOut[i][j] = 0.0;
for (int k = kBegin; k < kEnd; k++) {
dataOut[i][j] += dW[k] * dataIn[k][i][j];
}
}
}
}
// Write variable
vecOutNcVarS[v]->set_cur(t, 0, 0);
vecOutNcVarS[v]->put(&(dataOut[0][0]), 1, nLat, nLon);
}
// Loop through all pressure levels
for (int p = 0; p < nPressureLevels; p++) {
// Loop thorugh all latlon indices
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
// Store column pressure
for (int k = 0; k < nLev; k++) {
dataColumnP[k] = dataP[k][i][j];
}
// Find weights
int kBegin = 0;
int kEnd = 0;
// On a pressure surface
InterpolationWeightsLinear(
vecPressureLevels[p],
dataColumnP,
kBegin,
kEnd,
dW);
// Interpolate in the vertical
dataOut[i][j] = 0.0;
for (int k = kBegin; k < kEnd; k++) {
dataOut[i][j] += dW[k] * dataIn[k][i][j];
}
}
}
// Write variable
vecOutNcVarP[v]->set_cur(t, p, 0, 0);
vecOutNcVarP[v]->put(&(dataOut[0][0]), 1, 1, nLat, nLon);
}
// Loop through all height levels
for (int z = 0; z < nHeightLevels; z++) {
// Loop thorugh all latlon indices
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
// Find weights
int kBegin = 0;
int kEnd = 0;
// Interpolate from levels to z surfaces
if (!fOnInterfaces) {
for (int k = 0; k < nLev; k++) {
dataColumnZ[k] = dataZ[k][i][j];
}
InterpolationWeightsLinear(
vecHeightLevels[z],
dataColumnZ,
kBegin,
kEnd,
dW);
dataOut[i][j] = 0.0;
for (int k = kBegin; k < kEnd; k++) {
dataOut[i][j] += dW[k] * dataIn[k][i][j];
}
// Interpolate from interfaces to z surfaces
} else {
for (int k = 0; k < nILev; k++) {
dataColumnIZ[k] = dataIZ[k][i][j];
}
InterpolationWeightsLinear(
vecHeightLevels[z],
dataColumnIZ,
kBegin,
kEnd,
dIW);
dataOut[i][j] = 0.0;
for (int k = kBegin; k < kEnd; k++) {
dataOut[i][j] += dIW[k] * dataInt[k][i][j];
}
}
}
}
// Write variable
vecOutNcVarZ[v]->set_cur(t, z, 0, 0);
vecOutNcVarZ[v]->put(&(dataOut[0][0]), 1, 1, nLat, nLon);
}
}
// Output geopotential height
if (fGeopotentialHeight) {
Announce("Geopotential height");
// Output variables
NcVar * varOutZ;
NcVar * varOutZs;
if (nPressureLevels > 0) {
varOutZ = ncdf_out.add_var(
"PHIZ", ncDouble, dimOutTime, dimOutP, dimOutLat, dimOutLon);
}
if (fExtractSurface) {
varOutZs = ncdf_out.add_var(
"PHIZS", ncDouble, dimOutTime, dimOutLat, dimOutLon);
}
// Interpolate onto pressure levels
for (int p = 0; p < nPressureLevels; p++) {
// Loop thorugh all latlon indices
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
int kBegin = 0;
int kEnd = 0;
for (int k = 0; k < nLev; k++) {
dataColumnP[k] = dataP[k][i][j];
}
InterpolationWeightsLinear(
vecPressureLevels[p],
dataColumnP,
kBegin,
kEnd,
dW);
// Interpolate in the vertical
dataOut[i][j] = 0.0;
for (int k = kBegin; k < kEnd; k++) {
dataOut[i][j] += dW[k] * dataZ[k][i][j];
}
}
}
// Write variable
varOutZ->set_cur(t, p, 0, 0);
varOutZ->put(&(dataOut[0][0]), 1, 1, nLat, nLon);
}
// Interpolate onto the physical surface
if (fExtractSurface) {
int kBegin = 0;
int kEnd = 3;
PolynomialInterp::LagrangianPolynomialCoeffs(
3, dLev, dW, 0.0);
// Loop thorugh all latlon indices
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
// Interpolate in the vertical
dataOut[i][j] = 0.0;
for (int k = kBegin; k < kEnd; k++) {
dataOut[i][j] += dW[k] * dataZ[k][i][j];
}
}
}
// Write variable
varOutZs->set_cur(t, 0, 0);
varOutZs->put(&(dataOut[0][0]), 1, nLat, nLon);
}
}
// Extract total energy
if (fExtractTotalEnergy) {
Announce("Total Energy");
// Zonal velocity
DataArray3D<double> dataU(nLev, nLat, nLon);
NcVar * varU = ncdf_in.get_var("U");
varU->set_cur(t, 0, 0, 0);
varU->get(&(dataU[0][0][0]), 1, nLev, nLat, nLon);
// Meridional velocity
DataArray3D<double> dataV(nLev, nLat, nLon);
NcVar * varV = ncdf_in.get_var("V");
varV->set_cur(t, 0, 0, 0);
varV->get(&(dataV[0][0][0]), 1, nLev, nLat, nLon);
// Vertical velocity
DataArray3D<double> dataW(nLev, nLat, nLon);
NcVar * varW = ncdf_in.get_var("W");
varW->set_cur(t, 0, 0, 0);
varW->get(&(dataW[0][0][0]), 1, nLev, nLat, nLon);
// Calculate total energy
double dTotalEnergy = 0.0;
double dElementRefArea =
dEarthRadius * dEarthRadius
* M_PI / static_cast<double>(nLat)
* 2.0 * M_PI / static_cast<double>(nLon);
for (int k = 0; k < nLev; k++) {
for (int i = 0; i < nLat; i++) {
for (int j = 0; j < nLon; j++) {
double dKineticEnergy =
0.5 * dataRho[k][i][j] *
( dataU[k][i][j] * dataU[k][i][j]
+ dataV[k][i][j] * dataV[k][i][j]
+ dataW[k][i][j] * dataW[k][i][j]);
double dInternalEnergy =
dataP[k][i][j] / (dGamma - 1.0);
dTotalEnergy +=
(dKineticEnergy + dInternalEnergy)
* std::cos(M_PI * dLat[i] / 180.0) * dElementRefArea
* (dZtop - dZs[i][j]) / static_cast<double>(nLev);
}
}
}
// Put total energy into file
varEnergy->set_cur(t);
varEnergy->put(&dTotalEnergy, 1);
}
AnnounceEndBlock("Done");
}
AnnounceEndBlock("Done");
} catch(Exception & e) {
Announce(e.ToString().c_str());
}
// Finalize MPI
MPI_Finalize();
}
bool runTestType(cl::Context context, cl::CommandQueue queue)
{
cl_uint size = 64 * 256;
std::vector<T> input(size);
for(size_t i = 0; i < input.size(); ++i)
input[i] = input.size() - i - 1;
// create input buffer using pinned memory
cl::Buffer bufferIn(context, CL_MEM_ALLOC_HOST_PTR |
CL_MEM_COPY_HOST_PTR | CL_MEM_READ_WRITE,
sizeof(T) * input.size(), &input[0])
;
magnet::CL::sort<T> sortFunctor;
sortFunctor.build(queue, context);
sortFunctor(bufferIn);
std::cout << "##Testing generic sort (";
switch(sortFunctor.getMode())
{
case magnet::CL::sort<T>::CPU:
std::cout << "HeapSort";
break;
case magnet::CL::sort<T>::NVIDIA:
std::cout << "radixNVIDIA";
break;
case magnet::CL::sort<T>::AMD:
std::cout << "radixAMD";
break;
default:
M_throw() << "Could not determine which sorting algorithm is being used";
}
std::cout << ") for " << input.size() << " elements and type "
<< magnet::CL::detail::traits<T>::kernel_type();
std::vector<T> output(size);
queue.enqueueReadBuffer(bufferIn, CL_TRUE, 0, input.size() *
sizeof(T), &output[0]);
bool failed = !testOutput(input, output);
std::cout << " key(only) " << (failed ? "FAILED" : "PASSED") << ", ";
//Now test with some data!
//Refresh the input array
queue.enqueueWriteBuffer(bufferIn, CL_TRUE, 0, input.size() *
sizeof(T), &input[0]);
//Write a data array
std::vector<cl_uint> data(size);
for(size_t i = 0; i < input.size(); ++i)
data[i] = i;
cl::Buffer dataIn(context, CL_MEM_ALLOC_HOST_PTR |
CL_MEM_COPY_HOST_PTR | CL_MEM_READ_WRITE,
sizeof(cl_uint) * data.size(), &data[0])
;
sortFunctor(bufferIn, dataIn);
queue.enqueueReadBuffer(dataIn, CL_TRUE, 0, data.size() *
sizeof(cl_uint), &data[0]);
bool keyfail = false;//!testOutput(input, output);
std::cout << " key " << (keyfail ? "FAILED" : "PASSED");
bool datafail = false;
for(size_t i = 0; i < input.size(); ++i)
if (data[i] != input.size() - 1 - i)
datafail = true;
std::cout << " data " << (datafail ? "FAILED" : "PASSED") << std::endl;
return failed || keyfail || datafail;
}