portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-04 19:25:56 +01:00
Code Releases Activity

Part-way through release tidy-up

Browse Source
This commit is contained in:
Michael Marshall 2019-04-29 18:40:38 +01:00
parent ac19c0e04f
commit 4203105104
3 changed files with 189 additions and 193 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

32
Hadrons/Distil.hpp
View File

@ -47,6 +47,9 @@
/******************************************************************************
A consistent set of cross-platform methods for big endian <-> host byte ordering
I imagine this exists already?
RELEASE NOTE:
******************************************************************************/
#if defined(__linux__)
@ -589,6 +592,20 @@ void NamedTensor<Scalar_, NumIndices_, Endian_Scalar_Size>::read(const char * fi
template<typename Scalar_, int NumIndices_, uint16_t Endian_Scalar_Size = sizeof(Scalar_)>
using Perambulator = NamedTensor<Scalar_, NumIndices_, Endian_Scalar_Size>;
struct DistilParameters: Serializable {
GRID_SERIALIZABLE_CLASS_MEMBERS(DistilParameters,
int, TI,
int, LI,
int, nnoise,
int, tsrc,
int, SI,
int, Ns,
int, Nt,
int, Nt_inv)
DistilParameters() = default;
template <class ReaderClass> DistilParameters(Reader<ReaderClass>& Reader){read(Reader,"Distil",*this);}
};
/*************************************************************************************
Rotate eigenvectors into our phase convention
@ -624,21 +641,6 @@ inline void RotateEigen(std::vector<LatticeColourVector> & evec)
}
}
struct DistilParameters: Serializable {
GRID_SERIALIZABLE_CLASS_MEMBERS(DistilParameters,
int, TI,
int, LI,
int, nnoise,
int, tsrc,
int, SI,
int, Ns,
int, Nt,
int, Nt_inv)
DistilParameters() = default;
template <class ReaderClass> DistilParameters(Reader<ReaderClass>& Reader){read(Reader,"Distil",*this);}
};
END_MODULE_NAMESPACE
END_HADRONS_NAMESPACE

255
Hadrons/Modules/MDistil/DistilVectors.hpp
View File

@ -43,54 +43,66 @@
BEGIN_HADRONS_NAMESPACE
/******************************************************************************
* DistilVectors *
* DistilVectors *
* (Create rho and/or phi vectors) *
******************************************************************************/
BEGIN_MODULE_NAMESPACE(MDistil)
class DistilVectorsPar: Serializable
{
public:
GRID_SERIALIZABLE_CLASS_MEMBERS(DistilVectorsPar,
std::string, noise,
std::string, perambulator,
std::string, eigenPack,
bool, multiFile,
int, tsrc,
int, nnoise,
int, LI,
int, SI,
int, TI,
int, nvec,
int, Ns,
int, Nt,
int, Nt_inv);
GRID_SERIALIZABLE_CLASS_MEMBERS(DistilVectorsPar,
std::string, noise,
std::string, perambulator,
std::string, lapevec,
std::string, source,
std::string, sink,
bool, multiFile,
int, tsrc,
//int, nnoise,
int, LI,
int, SI,
int, TI,
int, nvec,
int, Ns,
int, Nt,
int, Nt_inv);
};
template <typename FImpl>
class TDistilVectors: public Module<DistilVectorsPar>
{
public:
FERM_TYPE_ALIASES(FImpl,);
FERM_TYPE_ALIASES(FImpl,);
// This is the type of perambulator I expect
using DistilPeramb = Perambulator<SpinVector, 6, sizeof(Real)>;
// constructor
TDistilVectors(const std::string name);
// destructor
virtual ~TDistilVectors(void);
// dependency relation
virtual std::vector<std::string> getInput(void);
virtual std::vector<std::string> getOutput(void);
// setup
virtual void setup(void);
// execution
virtual void execute(void);
protected:
// These variables are created in setup() and freed in Cleanup()
GridCartesian * grid3d; // Owned by me, so I must delete it
GridCartesian * grid4d; // Owned by environment (so I won't delete it)
virtual void Cleanup(void);
public:
// constructor
TDistilVectors(const std::string name);
// destructor
virtual ~TDistilVectors(void);
// dependency relation
virtual std::vector<std::string> getInput(void);
virtual std::vector<std::string> getOutput(void);
// setup
virtual void setup(void);
// execution
virtual void execute(void);
protected:
// These variables are created in setup() and freed in Cleanup()
GridCartesian * grid3d; // Owned by me, so I must delete it
GridCartesian * grid4d; // Owned by environment (so I won't delete it)
protected:
virtual void Cleanup(void);
// These variables contain parameters
std::string PerambulatorName;
std::string NoiseVectorName;
std::string LapEvecName;
bool bMakeSource;
bool bMakeSink;
std::string SourceName;
std::string SinkName;
int nnoise;
};
MODULE_REGISTER_TMP(DistilVectors, TDistilVectors<FIMPL>, MDistil);
@ -114,40 +126,70 @@ TDistilVectors<FImpl>::~TDistilVectors(void)
template <typename FImpl>
std::vector<std::string> TDistilVectors<FImpl>::getInput(void)
{
std::vector<std::string> in;
in.push_back(par().noise);
in.push_back(par().perambulator);
in.push_back(par().eigenPack);
return in;
PerambulatorName = par().perambulator;
if( PerambulatorName.size() == 0 ) {
PerambulatorName = getName();
PerambulatorName.append( "_peramb" );
}
NoiseVectorName = par().noise;
if( NoiseVectorName.size() == 0 ) {
NoiseVectorName = PerambulatorName;
NoiseVectorName.append( "_noise" );
}
LapEvecName = par().lapevec;
if( LapEvecName.size() == 0 ) {
LapEvecName = PerambulatorName;
LapEvecName.append( "_lapevec" );
}
return { PerambulatorName, NoiseVectorName, LapEvecName };
}
template <typename FImpl>
std::vector<std::string> TDistilVectors<FImpl>::getOutput(void)
{
std::vector<std::string> out = {getName() + "_rho", getName() + "_phi"};
return out;
SourceName = par().source;
SinkName = par().sink;
bMakeSource = ( SourceName.size() > 0 );
bMakeSink = ( SinkName.size() > 0 );
if( !bMakeSource && !bMakeSink ) {
SourceName = getName();
SinkName = SourceName;
SourceName.append( "_rho" );
SinkName.append( "_phi" );
bMakeSource = true;
bMakeSink = true;
}
std::vector<std::string> out;
if( bMakeSource )
out.push_back( SourceName );
if( bMakeSink )
out.push_back( SinkName );
return out;
}
// setup ///////////////////////////////////////////////////////////////////////
template <typename FImpl>
void TDistilVectors<FImpl>::setup(void)
{
Cleanup();
auto &noise = envGet(std::vector<Complex>, par().noise);
Cleanup();
auto &noise = envGet(std::vector<Complex>, NoiseVectorName);
auto &perambulator = envGet(DistilPeramb, PerambulatorName);
int nnoise=par().nnoise;
int LI=par().LI;
int Ns=par().Ns;
int SI=par().SI;
int Nt_inv=par().Nt_inv;
// We expect the perambulator to have been created with these indices
std::array<std::string,6> sIndexNames{"Nt", "nvec", "LI", "nnoise", "Nt_inv", "SI"};
for(int i = 0; i < DistilPeramb::NumIndices; i++ )
assert( sIndexNames[i] == perambulator.IndexNames[i] && "Perambulator indices bad" );
envCreate(std::vector<FermionField>, getName() + "_rho", 1,
nnoise*LI*SI*Nt_inv, envGetGrid(FermionField));
envCreate(std::vector<FermionField>, getName() + "_phi", 1,
nnoise*LI*SI*Nt_inv, envGetGrid(FermionField));
nnoise = static_cast<int>( noise.size() );
int LI=par().LI;
int Ns=par().Ns;
int SI=par().SI;
int Nt_inv=par().Nt_inv;
if( bMakeSource )
envCreate(std::vector<FermionField>, SourceName, 1, nnoise*LI*SI*Nt_inv, envGetGrid(FermionField));
if( bMakeSink )
envCreate(std::vector<FermionField>, SinkName, 1, nnoise*LI*SI*Nt_inv, envGetGrid(FermionField));
grid4d = env().getGrid();
std::vector<int> latt_size = GridDefaultLatt();
@ -158,10 +200,10 @@ void TDistilVectors<FImpl>::setup(void)
simd_layout_3.push_back( 1 );
mpi_layout[Nd-1] = 1;
grid3d = MakeLowerDimGrid(grid4d);
envTmp(LatticeSpinColourVector, "tmp2",1,LatticeSpinColourVector(grid4d));
envTmp(LatticeColourVector, "tmp_nospin",1,LatticeColourVector(grid4d));
//envTmp(LatticeColourVector, "tmp_nospin",1,LatticeColourVector(grid4d));
envTmp(LatticeSpinColourVector, "tmp3d",1,LatticeSpinColourVector(grid3d));
envTmp(LatticeColourVector, "tmp3d_nospin",1,LatticeColourVector(grid3d));
envTmp(LatticeSpinColourVector, "sink_tslice",1,LatticeSpinColourVector(grid3d));
@ -183,26 +225,22 @@ void TDistilVectors<FImpl>::Cleanup(void)
template <typename FImpl>
void TDistilVectors<FImpl>::execute(void)
{
auto &noise = envGet(std::vector<Complex>, par().noise);
auto &perambulator = envGet(Perambulator<SpinVector COMMA 6 COMMA sizeof(Real)>, par().perambulator);
auto &epack = envGet(Grid::Hadrons::EigenPack<LatticeColourVector>, par().eigenPack);
auto &rho = envGet(std::vector<FermionField>, getName() + "_rho");
auto &phi = envGet(std::vector<FermionField>, getName() + "_phi");
auto &noise = envGet(std::vector<Complex>, NoiseVectorName);
auto &perambulator = envGet(DistilPeramb, PerambulatorName);
auto &epack = envGet(Grid::Hadrons::EigenPack<LatticeColourVector>, LapEvecName);
envGetTmp(LatticeSpinColourVector, tmp2);
envGetTmp(LatticeColourVector, tmp_nospin);
//envGetTmp(LatticeColourVector, tmp_nospin);
envGetTmp(LatticeSpinColourVector, tmp3d);
envGetTmp(LatticeColourVector, tmp3d_nospin);
envGetTmp(LatticeSpinColourVector, sink_tslice);
envGetTmp(LatticeColourVector, evec3d);
int Ntlocal = grid4d->LocalDimensions()[3];
int Ntfirst = grid4d->LocalStarts()[3];
int tsrc=par().tsrc;
int nnoise=par().nnoise;
int LI=par().LI;
int Ns=par().Ns;
int Nt_inv=par().Nt_inv; // TODO: No input, but define through Nt, TI
@ -212,28 +250,31 @@ void TDistilVectors<FImpl>::execute(void)
int SI=par().SI;
bool full_tdil=(TI==Nt);
int vecindex;
int t_inv;
for (int inoise = 0; inoise < nnoise; inoise++) {
for (int dk = 0; dk < LI; dk++) {
for (int dt = 0; dt < Nt_inv; dt++) {
for (int ds = 0; ds < SI; ds++) {
vecindex = inoise + nnoise * dk + nnoise * LI * ds + nnoise *LI * SI*dt;
rho[vecindex] = zero;
tmp3d_nospin = zero;
for (int it = dt; it < Nt; it += TI){
if (full_tdil) t_inv = tsrc; else t_inv = it;
if( t_inv >= Ntfirst && t_inv < Ntfirst + Ntlocal ) {
for (int ik = dk; ik < nvec; ik += LI){
for (int is = ds; is < Ns; is += SI){
ExtractSliceLocal(evec3d,epack.evec[ik],0,t_inv,3);
tmp3d_nospin = evec3d * noise[inoise + nnoise*(t_inv + Nt*(ik+nvec*is))];
tmp3d=zero;
pokeSpin(tmp3d,tmp3d_nospin,is);
tmp2=zero;
InsertSliceLocal(tmp3d,tmp2,0,t_inv-Ntfirst,Grid::QCD::Tdir);
rho[vecindex] += tmp2;
if( bMakeSource ) {
auto &rho = envGet(std::vector<FermionField>, SourceName);
for (int inoise = 0; inoise < nnoise; inoise++) {
for (int dk = 0; dk < LI; dk++) {
for (int dt = 0; dt < Nt_inv; dt++) {
for (int ds = 0; ds < SI; ds++) {
vecindex = inoise + nnoise * dk + nnoise * LI * ds + nnoise *LI * SI*dt;
rho[vecindex] = zero;
tmp3d_nospin = zero;
for (int it = dt; it < Nt; it += TI){
if (full_tdil) t_inv = tsrc; else t_inv = it;
if( t_inv >= Ntfirst && t_inv < Ntfirst + Ntlocal ) {
for (int ik = dk; ik < nvec; ik += LI){
for (int is = ds; is < Ns; is += SI){
ExtractSliceLocal(evec3d,epack.evec[ik],0,t_inv,3);
tmp3d_nospin = evec3d * noise[inoise + nnoise*(t_inv + Nt*(ik+nvec*is))];
tmp3d=zero;
pokeSpin(tmp3d,tmp3d_nospin,is);
tmp2=zero;
InsertSliceLocal(tmp3d,tmp2,0,t_inv-Ntfirst,Grid::QCD::Tdir);
rho[vecindex] += tmp2;
}
}
}
}
@ -242,31 +283,27 @@ void TDistilVectors<FImpl>::execute(void)
}
}
}
for (int inoise = 0; inoise < nnoise; inoise++) {
for (int dk = 0; dk < LI; dk++) {
for (int dt = 0; dt < Nt_inv; dt++) {
for (int ds = 0; ds < SI; ds++) {
vecindex = inoise + nnoise * dk + nnoise * LI * ds + nnoise *LI * SI*dt;
phi[vecindex] = zero;
for (int t = Ntfirst; t < Ntfirst + Ntlocal; t++) {
sink_tslice=zero;
for (int ivec = 0; ivec < nvec; ivec++) {
ExtractSliceLocal(evec3d,epack.evec[ivec],0,t,3);
sink_tslice += evec3d * perambulator(t, ivec, dk, inoise,dt,ds);
if( bMakeSink ) {
auto &phi = envGet(std::vector<FermionField>, SinkName);
for (int inoise = 0; inoise < nnoise; inoise++) {
for (int dk = 0; dk < LI; dk++) {
for (int dt = 0; dt < Nt_inv; dt++) {
for (int ds = 0; ds < SI; ds++) {
vecindex = inoise + nnoise * dk + nnoise * LI * ds + nnoise *LI * SI*dt;
phi[vecindex] = zero;
for (int t = Ntfirst; t < Ntfirst + Ntlocal; t++) {
sink_tslice=zero;
for (int ivec = 0; ivec < nvec; ivec++) {
ExtractSliceLocal(evec3d,epack.evec[ivec],0,t,3);
sink_tslice += evec3d * perambulator(t, ivec, dk, inoise,dt,ds);
}
InsertSliceLocal(sink_tslice,phi[vecindex],0,t-Ntfirst,Grid::QCD::Tdir);
}
InsertSliceLocal(sink_tslice,phi[vecindex],0,t-Ntfirst,Grid::QCD::Tdir);
}
}
}
}
}
// TEST TO SEE WHETHER THIS MIGHT BE THE MEMORY LEAK
Cleanup();
std::cout << "size rho" << rho.size() << std::endl;
std::cout << "size phi" << phi.size() << std::endl;
}
END_MODULE_NAMESPACE

95
Hadrons/Modules/MDistil/LapEvec.hpp
View File

@ -74,7 +74,8 @@ struct LanczosParameters: Serializable {
int, Np,
int, MaxIt,
//int, MinRes,
double, resid)
double, resid,
std::string, Log) // Any non-empty string will enable logging
LanczosParameters() = default;
template <class ReaderClass> LanczosParameters(Reader<ReaderClass>& Reader){read(Reader,"Lanczos",*this);}
};
@ -133,14 +134,10 @@ MODULE_REGISTER_TMP(LapEvec, TLapEvec<GIMPL>, MDistil);
TLapEvec implementation
******************************************************************************/
//constexpr char szEigenPackSuffix[] = "_eigenPack";
// constructor /////////////////////////////////////////////////////////////////
template <typename GImpl>
TLapEvec<GImpl>::TLapEvec(const std::string name) : gridLD{nullptr}, Module<LapEvecPar>(name)
{
//LOG(Message) << "TLapEvec constructor : TLapEvec<GImpl>::TLapEvec(const std::string name)" << std::endl;
//LOG(Message) << "this=" << this << ", gridLD=" << gridLD << std::endl;
}
// destructor /////////////////////////////////////////////////////////////////
@ -155,7 +152,6 @@ template <typename GImpl>
std::vector<std::string> TLapEvec<GImpl>::getInput(void)
{
std::vector<std::string> in = {par().gauge};
return in;
}
@ -163,7 +159,6 @@ template <typename GImpl>
std::vector<std::string> TLapEvec<GImpl>::getOutput(void)
{
std::vector<std::string> out = {getName()}; // This is the higher dimensional eigenpack
return out;
}
@ -209,11 +204,14 @@ void TLapEvec<GImpl>::Cleanup(void)
template <typename GImpl>
void TLapEvec<GImpl>::execute(void)
{
LOG(Message) << "execute() : start for " << getName() << std::endl;
const ChebyshevParameters &ChebPar{par().Cheby};
const LanczosParameters &LPar{par().Lanczos};
const int &nvec{LPar.Nvec};
// Enable IRL logging if requested
if( LPar.Log.size() > 0 )
GridLogIRL.Active(1);
//const bool exact_distillation{TI==Nt && LI==nvec};
//const bool full_tdil{TI==Nt};
//const int &Nt_inv{full_tdil ? 1 : TI};
@ -243,21 +241,10 @@ void TLapEvec<GImpl>::execute(void)
}
LOG(Message) << "Smeared plaquette: " << WilsonLoops<PeriodicGimplR>::avgPlaquette(Umu_smear) << std::endl;
// For debugging only, write logging output to a local file
std::ofstream * ll = nullptr;
const int rank{gridHD->ThisRank()};
if((0)) { // debug to a local log file
std::string filename{"Local_"};
filename.append(std::to_string(rank));
filename.append(".log");
ll = new std::ofstream(filename);
}
////////////////////////////////////////////////////////////////////////
// Invert Peardon Nabla operator separately on each time-slice
////////////////////////////////////////////////////////////////////////
bool bReturnValue = true;
auto & eig4d = envGet(DistilEP, getName() );
envGetTmp(std::vector<DistilEP>, eig); // Eigenpack for each timeslice
envGetTmp(LatticeGaugeField, UmuNoTime); // Gauge field without time dimension
@ -266,68 +253,45 @@ void TLapEvec<GImpl>::execute(void)
const int Ntfirst{gridHD->LocalStarts()[Tdir]};
const char DefaultOperatorXml[] = "<OPERATOR>Michael</OPERATOR>";
const char DefaultsolverXml[] = "<SOLVER>Felix</SOLVER>";
for(int t=Ntfirst;bReturnValue && t<Ntfirst+Ntlocal;t++){
std::cout << GridLogMessage << "------------------------------------------------------------" << std::endl;
std::cout << GridLogMessage << " Compute eigenpack, Timeslice = " << t << std::endl;
std::cout << GridLogMessage << "------------------------------------------------------------" << std::endl;
std::cout << "T: " << t << " / " << Ntfirst + Ntlocal << std::endl;
for(int t = Ntfirst; t < Ntfirst + Ntlocal; t++ ) {
LOG(Message) << "------------------------------------------------------------" << std::endl;
LOG(Message) << " Compute eigenpack, Timeslice = " << t << " / " << Ntfirst + Ntlocal << std::endl;
LOG(Message) << "------------------------------------------------------------" << std::endl;
eig[t].resize(LPar.Nk+LPar.Np,gridLD);
// Construct smearing operator
ExtractSliceLocal(UmuNoTime,Umu_smear,0,t-Ntfirst,Grid::QCD::Tdir); // switch to 3d/4d objects
LinOpPeardonNabla<LatticeColourVector> PeardonNabla(UmuNoTime);
std::cout << "Chebyshev preconditioning to order " << ChebPar.PolyOrder
<< " with parameters (alpha,beta) = (" << ChebPar.alpha << "," << ChebPar.beta << ")" << std::endl;
LOG(Debug) << "Chebyshev preconditioning to order " << ChebPar.PolyOrder
<< " with parameters (alpha,beta) = (" << ChebPar.alpha << "," << ChebPar.beta << ")" << std::endl;
Chebyshev<LatticeColourVector> Cheb(ChebPar.alpha,ChebPar.beta,ChebPar.PolyOrder);
//from Test_Cheby.cc
if ( ((0)) && Ntfirst == 0 && t==0) {
std::ofstream of("cheby_" + std::to_string(ChebPar.alpha) + "_" + std::to_string(ChebPar.beta) + "_" + std::to_string(ChebPar.PolyOrder));
Cheb.csv(of);
}
//if( Ntfirst == 0 && t==0) {
//std::ofstream of("cheby_" + std::to_string(ChebPar.alpha) + "_" + std::to_string(ChebPar.beta) + "_" + std::to_string(ChebPar.PolyOrder));
//Cheb.csv(of);
//}
// Construct source vector according to Test_dwf_compressed_lanczos.cc
src=11.0;
src = 11.0;
RealD nn = norm2(src);
nn = Grid::sqrt(nn);
src = src * (1.0/nn);
GridLogIRL.Active(1);
LinOpPeardonNablaHerm<LatticeColourVector> PeardonNablaCheby(Cheb,PeardonNabla);
ImplicitlyRestartedLanczos<LatticeColourVector> IRL(PeardonNablaCheby,PeardonNabla,LPar.Nvec,LPar.Nk,LPar.Nk+LPar.Np,LPar.resid,LPar.MaxIt);
ImplicitlyRestartedLanczos<LatticeColourVector>
IRL(PeardonNablaCheby,PeardonNabla,LPar.Nvec,LPar.Nk,LPar.Nk+LPar.Np,LPar.resid,LPar.MaxIt);
int Nconv = 0;
if(ll) *ll << t << " : Before IRL.calc()" << std::endl;
IRL.calc(eig[t].eval,eig[t].evec,src,Nconv);
if(ll) *ll << t << " : After IRL.calc()" << std::endl;
if( Nconv < LPar.Nvec ) {
bReturnValue = false;
if(ll) *ll << t << " : Convergence error : Only " << Nconv << " converged!" << std::endl;
} else {
if( Nconv > LPar.Nvec )
eig[t].resize( LPar.Nvec, gridLD );
std::cout << GridLogMessage << "Timeslice " << t << " has " << eig[t].eval.size() << " eigenvalues and " << eig[t].evec.size() << " eigenvectors." << std::endl;
// Now rotate the eigenvectors into our phase convention
RotateEigen( eig[t].evec );
if((0)) { // Debugging only
// Write the eigenvectors and eigenvalues to disk
//std::cout << GridLogMessage << "Writing eigenvalues/vectors to " << pszEigenPack << std::endl;
eig[t].record.operatorXml = DefaultOperatorXml;
eig[t].record.solverXml = DefaultsolverXml;
eig[t].write("DistilEigen",false,t);
//std::cout << GridLogMessage << "Written eigenvectors" << std::endl;
}
}
std::cout << "T: " << t << " / " << Ntfirst + Ntlocal << std::endl;
assert( Nconv >= LPar.Nvec && "MDistil::LapEvec : Error - not enough eigenvectors converged" );
if( Nconv > LPar.Nvec )
eig[t].resize( LPar.Nvec, gridLD );
RotateEigen( eig[t].evec ); // Rotate the eigenvectors into our phase convention
for (int i=0;i<LPar.Nvec;i++){
std::cout << "Inserting Timeslice " << t << " into vector " << i << std::endl;
InsertSliceLocal(eig[t].evec[i],eig4d.evec[i],0,t,3);
// TODO: Discuss: is this needed? Is there a better way?
if(t==0)
eig4d.eval[i] = eig[t].eval[i];
eig4d.eval[i] = eig[t].eval[i]; // TODO: Discuss: is this needed? Is there a better way?
}
}
@ -338,13 +302,6 @@ void TLapEvec<GImpl>::execute(void)
sEigenPackName.append(".");
sEigenPackName.append(std::to_string(vm().getTrajectory()));
eig4d.write(sEigenPackName,false);
// Close the local debugging log file
if( ll ) {
*ll << " Returning " << bReturnValue << std::endl;
delete ll;
}
LOG(Message) << "execute() : end" << std::endl;
}
END_MODULE_NAMESPACE