portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-09-19 16:55:37 +01:00
Code Releases Activity

Coarse space is expensive, but gives a speed up in fine matrix multiplies now.

Browse Source 
Down to optimisation
This commit is contained in:
Peter Boyle 2023-09-25 17:24:19 -04:00
parent 9246e653cd
commit 37884d369f
1 changed files with 171 additions and 102 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

273
tests/debug/Test_general_coarse_hdcg.cc
View File

@ -2,11 +2,11 @@
Grid physics library, www.github.com/paboyle/Grid
Source file: ./tests/Test_padded_cell.cc
Source file: ./tests/Test_general_coarse_hdcg.cc
Copyright (C) 2023
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
Author: Peter Boyle <pboyle@bnl.gov>
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
@ -29,10 +29,22 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
#include <Grid/lattice/PaddedCell.h>
#include <Grid/stencil/GeneralLocalStencil.h>
#include <Grid/algorithms/GeneralCoarsenedMatrix.h>
#include <Grid/algorithms/iterative/AdefGeneric.h>
using namespace std;
using namespace Grid;
template<class Field> class TestSolver : public LinearFunction<Field> {
public:
TestSolver() {};
void operator() (const Field &in, Field &out){ out = Zero(); }
};
RealD InverseApproximation(RealD x){
return 1.0/x;
}
// Want Op in CoarsenOp to call MatPcDagMatPc
template<class Field>
class HermOpAdaptor : public LinearOperatorBase<Field>
@ -40,20 +52,33 @@ class HermOpAdaptor : public LinearOperatorBase<Field>
LinearOperatorBase<Field> & wrapped;
public:
HermOpAdaptor(LinearOperatorBase<Field> &wrapme) : wrapped(wrapme) {};
void OpDiag (const Field &in, Field &out) { assert(0); }
void Op (const Field &in, Field &out) { wrapped.HermOp(in,out); }
void HermOp(const Field &in, Field &out) { wrapped.HermOp(in,out); }
void AdjOp (const Field &in, Field &out){ wrapped.HermOp(in,out); }
void OpDiag (const Field &in, Field &out) { assert(0); }
void OpDir (const Field &in, Field &out,int dir,int disp) { assert(0); }
void OpDirAll (const Field &in, std::vector<Field> &out){ assert(0); };
void Op (const Field &in, Field &out){
wrapped.HermOp(in,out);
}
void AdjOp (const Field &in, Field &out){
wrapped.HermOp(in,out);
}
void OpDirAll (const Field &in, std::vector<Field> &out) { assert(0); };
void HermOpAndNorm(const Field &in, Field &out,RealD &n1,RealD &n2){ assert(0); }
void HermOp(const Field &in, Field &out){
wrapped.HermOp(in,out);
};
template<class Field,class Matrix> class ChebyshevSmoother : public LinearFunction<Field>
{
public:
using LinearFunction<Field>::operator();
typedef LinearOperatorBase<Field> FineOperator;
FineOperator & _SmootherOperator;
Chebyshev<Field> Cheby;
ChebyshevSmoother(RealD _lo,RealD _hi,int _ord, FineOperator &SmootherOperator) :
_SmootherOperator(SmootherOperator),
Cheby(_lo,_hi,_ord,InverseApproximation)
{
std::cout << GridLogMessage<<" Chebyshev smoother order "<<_ord<<" ["<<_lo<<","<<_hi<<"]"<<std::endl;
};
void operator() (const Field &in, Field &out)
{
Field tmp(in.Grid());
tmp = in;
Cheby(_SmootherOperator,tmp,out);
}
};
int main (int argc, char ** argv)
@ -62,69 +87,68 @@ int main (int argc, char ** argv)
const int Ls=16;
GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(),
GridDefaultSimd(Nd,vComplex::Nsimd()),
GridDefaultMpi());
GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
GridCartesian * FGrid = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
// Construct a coarsened grid
// 4^4 cell
// Construct a coarsened grid with 4^4 cell
Coordinate clatt = GridDefaultLatt();
for(int d=0;d<clatt.size();d++){
clatt[d] = clatt[d]/4;
}
GridCartesian *Coarse4d = SpaceTimeGrid::makeFourDimGrid(clatt, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());;
GridCartesian *Coarse4d = SpaceTimeGrid::makeFourDimGrid(clatt,
GridDefaultSimd(Nd,vComplex::Nsimd()),
GridDefaultMpi());;
GridCartesian *Coarse5d = SpaceTimeGrid::makeFiveDimGrid(1,Coarse4d);
///////////////////////// RNGs /////////////////////////////////
std::vector<int> seeds4({1,2,3,4});
std::vector<int> seeds5({5,6,7,8});
std::vector<int> cseeds({5,6,7,8});
GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
GridParallelRNG CRNG(Coarse5d);CRNG.SeedFixedIntegers(cseeds);
///////////////////////// Configuration /////////////////////////////////
LatticeGaugeField Umu(UGrid);
FieldMetaData header;
std::string file("ckpoint_lat.4000");
NerscIO::readConfiguration(Umu,header,file);
//////////////////////// Fermion action //////////////////////////////////
RealD mass=0.01;
RealD M5=1.8;
RealD b=1.5;
RealD c=0.5;
MobiusFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
MobiusFermionD Dpv(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,1.0,M5,b,c);
const int nbasis = 4;
SchurDiagMooeeOperator<MobiusFermionD, LatticeFermion> HermOpEO(Ddwf);
typedef HermOpAdaptor<LatticeFermionD> HermFineMatrix;
HermFineMatrix FineHermOp(HermOpEO);
LatticeFermion result(FrbGrid); result=Zero();
LatticeFermion src(FrbGrid); random(RNG5,src);
// Run power method on FineHermOp
PowerMethod<LatticeFermion> PM; PM(HermOpEO,src);
////////////////////////////////////////////////////////////
///////////// Coarse basis and Little Dirac Operator ///////
////////////////////////////////////////////////////////////
const int nbasis = 40;
const int cb = 0 ;
LatticeFermion prom(FrbGrid);
typedef GeneralCoarsenedMatrix<vSpinColourVector,vTComplex,nbasis> LittleDiracOperator;
typedef LittleDiracOperator::CoarseVector CoarseVector;
NextToNextToNextToNearestStencilGeometry5D geom;
std::cout<<GridLogMessage<<std::endl;
std::cout<<GridLogMessage<<"*******************************************"<<std::endl;
std::cout<<GridLogMessage<<std::endl;
SchurDiagMooeeOperator<MobiusFermionD, LatticeFermion> HermOpEO(Ddwf);
HermOpAdaptor<LatticeFermionD> HOA(HermOpEO);
// Run power method on HOA??
LatticeFermion result(FrbGrid); result=Zero();
LatticeFermion ref(FrbGrid); ref=Zero();
LatticeFermion tmp(FrbGrid);
LatticeFermion err(FrbGrid);
{
LatticeFermion src(FrbGrid); random(RNG5,src);
PowerMethod<LatticeFermion> PM; PM(HermOpEO,src);
}
// exit(0);
NextToNextToNextToNearestStencilGeometry5D geom(Coarse5d);
// Warning: This routine calls PVdagM.Op, not PVdagM.HermOp
typedef Aggregation<vSpinColourVector,vTComplex,nbasis> Subspace;
@ -132,72 +156,117 @@ int main (int argc, char ** argv)
Aggregates.CreateSubspaceChebyshev(RNG5,
HermOpEO,
nbasis,
90.0,
0.1,
500,
500,
100,
0.0);
// 100.0,
// 0.1, // Low pass is pretty high still -- 311 iters
// 250.0,
// 0.01, // subspace too low filter power wrong
// 250.0,
// 0.2, // slower
95.0,
// 0.05, // nbasis 12 - 311 -- wrong coarse inv
// 0.05, // nbasis 12 - 154 -- right filt
// 0.1, // nbasis 12 - 169 oops
// 0.05, // nbasis 16 -- 127 iters
// 0.03, // nbasis 16 -- 13-
// 0.1, // nbasis 16 -- 142; sloppy solve
0.1, // nbasis 24
300);
////////////////////////////////////////////////////////////
// Need to check about red-black grid coarsening
////////////////////////////////////////////////////////////
LittleDiracOperator LittleDiracOp(geom,FrbGrid,Coarse5d);
LittleDiracOp.CoarsenOperator(HOA,Aggregates);
LittleDiracOp.CoarsenOperatorColoured(FineHermOp,Aggregates);
std::cout<<GridLogMessage<<std::endl;
std::cout<<GridLogMessage<<"*******************************************"<<std::endl;
typedef HermitianLinearOperator<LittleDiracOperator,CoarseVector> HermMatrix;
HermMatrix CoarseOp (LittleDiracOp);
std::cout<<GridLogMessage<<std::endl;
std::cout<<GridLogMessage<<"Testing coarsened operator "<<std::endl;
std::cout<<GridLogMessage<<"*******************************************"<<std::endl;
//////////////////////////////////////////
// Build a coarse lanczos
//////////////////////////////////////////
Chebyshev<CoarseVector> IRLCheby(0.02,50.0,71); // 1 iter
FunctionHermOp<CoarseVector> IRLOpCheby(IRLCheby,CoarseOp);
PlainHermOp<CoarseVector> IRLOp (CoarseOp);
int Nk=64;
int Nm=128;
int Nstop=Nk;
ImplicitlyRestartedLanczos<CoarseVector> IRL(IRLOpCheby,IRLOp,Nstop,Nk,Nm,1.0e-5,20);
int Nconv;
std::vector<RealD> eval(Nm);
std::vector<CoarseVector> evec(Nm,Coarse5d);
CoarseVector c_src(Coarse5d); c_src=1.0;
IRL.calc(eval,evec,c_src,Nconv);
DeflatedGuesser<CoarseVector> DeflCoarseGuesser(evec,eval);
CoarseVector c_src (Coarse5d);
CoarseVector c_res (Coarse5d);
CoarseVector c_proj(Coarse5d);
std::vector<LatticeFermion> subspace(nbasis,FrbGrid);
subspace=Aggregates.subspace;
Complex one(1.0);
c_src = one; // 1 in every element for vector 1.
blockPromote(c_src,err,subspace);
prom=Zero();
for(int b=0;b<nbasis;b++){
prom=prom+subspace[b];
}
err=err-prom;
std::cout<<GridLogMessage<<"Promoted back from subspace: err "<<norm2(err)<<std::endl;
std::cout<<GridLogMessage<<"c_src "<<norm2(c_src)<<std::endl;
std::cout<<GridLogMessage<<"prom "<<norm2(prom)<<std::endl;
HermOpEO.HermOp(prom,tmp);
blockProject(c_proj,tmp,subspace);
std::cout<<GridLogMessage<<" Called Big Dirac Op "<<norm2(tmp)<<std::endl;
LittleDiracOp.M(c_src,c_res);
std::cout<<GridLogMessage<<" Called Little Dirac Op c_src "<< norm2(c_src) << " c_res "<< norm2(c_res) <<std::endl;
std::cout<<GridLogMessage<<"Little dop : "<<norm2(c_res)<<std::endl;
std::cout<<GridLogMessage<<"Big dop in subspace : "<<norm2(c_proj)<<std::endl;
c_proj = c_proj - c_res;
std::cout<<GridLogMessage<<" ldop error: "<<norm2(c_proj)<<std::endl;
std::cout<<GridLogMessage<<std::endl;
std::cout<<GridLogMessage<<"*******************************************"<<std::endl;
std::cout<<GridLogMessage << "Done "<< std::endl;
//////////////////////////////////////////
// Build a coarse space solver
random(CRNG,c_src);
c_res=Zero();
// ZeroGuesser<CoarseVector> Guess;
RealD tol = 1.0e-8;
int maxit=2000;
ConjugateGradient<CoarseVector> CG(tol,maxit,false);
HermitianLinearOperator<LittleDiracOperator,CoarseVector> Hop (LittleDiracOp);
CG(Hop, c_src, c_res);
//////////////////////////////////////////
int maxit=20000;
ConjugateGradient<CoarseVector> CG(1.0e-8,maxit,false);
ConjugateGradient<LatticeFermionD> CGfine(1.0e-8,10000,false);
ZeroGuesser<CoarseVector> CoarseZeroGuesser;
// HPDSolver<CoarseVector> HPDSolve(CoarseOp,CG,CoarseZeroGuesser);
HPDSolver<CoarseVector> HPDSolve(CoarseOp,CG,DeflCoarseGuesser);
//////////////////////////////////////////
// Build a smoother
//////////////////////////////////////////
// ChebyshevSmoother<LatticeFermionD,HermFineMatrix > Smoother(10.0,100.0,10,FineHermOp); //499
// ChebyshevSmoother<LatticeFermionD,HermFineMatrix > Smoother(3.0,100.0,10,FineHermOp); //383
// ChebyshevSmoother<LatticeFermionD,HermFineMatrix > Smoother(1.0,100.0,10,FineHermOp); //328
// std::vector<RealD> los({0.5,1.0,3.0}); // 147/142/146 nbasis 1
// std::vector<RealD> los({1.0,2.0}); // Nbasis 24: 88,86 iterations
// std::vector<RealD> los({2.0,4.0}); // Nbasis 32 == 52, iters
// std::vector<RealD> los({2.0,4.0}); // Nbasis 40 == 36,36 iters
//
// Turns approx 2700 iterations into 340 fine multiplies with Nbasis 40
// Need to measure cost of coarse space.
//
// -- i) Reduce coarse residual -- 0.04
// -- ii) Lanczos on coarse space -- done
// -- iii) Possible 1 hop project and/or preconditioning it - easy - PrecCG it and
// use a limited stencil. Reread BFM code to check on evecs / deflation strategy with prec
//
std::vector<RealD> los({3.0}); // Nbasis 40 == 36,36 iters
std::vector<int> ords({7,8,10}); // Nbasis 40 == 40,38,36 iters (320,342,396 mults)
// Standard CG
// result=Zero();
// CGfine(HermOpEO, src, result);
for(int l=0;l<los.size();l++){
RealD lo = los[l];
for(int o=0;o<ords.size();o++){
ConjugateGradient<CoarseVector> CGsloppy(4.0e-2,maxit,false);
HPDSolver<CoarseVector> HPDSolveSloppy(CoarseOp,CGsloppy,DeflCoarseGuesser);
// ChebyshevSmoother<LatticeFermionD,HermFineMatrix > Smoother(lo,92,10,FineHermOp); // 36 best case
ChebyshevSmoother<LatticeFermionD,HermFineMatrix > Smoother(lo,92,ords[o],FineHermOp); // 311
//////////////////////////////////////////
// Build a HDCG solver
//////////////////////////////////////////
TwoLevelFlexiblePcg<LatticeFermion,CoarseVector,Subspace>
HDCG(1.0e-8, 3000,
FineHermOp,
Smoother,
HPDSolveSloppy,
HPDSolve,
Aggregates);
// result=Zero();
// HDCG(src,result);
result=Zero();
HDCG.Inflexible(src,result);
}
}
Grid_finalize();
return 0;
}