portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Code Releases Activity

Adding files for multiple implementations (cache opt) and Ls vectorisation

Browse Source 
of the 5D cayley form chiral fermions for the 5d matrix. With Ls entirely
in the vector direction, s-hopping terms involve rotations.

The serial dependence of the LDU inversion for Mobius and 4d even odd
checkerboarding is removed by simply applying Ls^2 operations (vectorised
many ways) as a dense matrix operation.

This should give similar throughput but high flops (non-compulsory flops)
but enable use of the KNL cache friendly kernels throughout the code.

Ls is still constrained to be a multiple of Nsimd, which is as much as 8 for AVX512
with single precision.
This commit is contained in:
paboyle 2016-07-14 22:59:21 +01:00
parent 62601bb649
commit adbc7c1188
6 changed files with 1395 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

206
lib/qcd/action/fermion/CayleyFermion5Dcache.cc Normal file
View File

@ -0,0 +1,206 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./lib/qcd/action/fermion/CayleyFermion5D.cc
Copyright (C) 2015
Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
Author: paboyle <paboyle@ph.ed.ac.uk>
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
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
#include <Grid.h>
namespace Grid {
namespace QCD {
// FIXME -- make a version of these routines with site loop outermost for cache reuse.
// Pminus fowards
// Pplus backwards..
template<class Impl>
void CayleyFermion5D<Impl>::M5D(const FermionField &psi,
const FermionField &phi,
FermionField &chi,
std::vector<RealD> &lower,
std::vector<RealD> &diag,
std::vector<RealD> &upper)
{
int Ls =this->Ls;
GridBase *grid=psi._grid;
assert(phi.checkerboard == psi.checkerboard);
chi.checkerboard=psi.checkerboard;
PARALLEL_FOR_LOOP
for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
for(int s=0;s<Ls;s++){
auto tmp = psi._odata[0];
if ( s==0 ) {
spProj5m(tmp,psi._odata[ss+s+1]);
chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
spProj5p(tmp,psi._odata[ss+Ls-1]);
chi[ss+s]=chi[ss+s]+lower[s]*tmp;
} else if ( s==(Ls-1)) {
spProj5m(tmp,psi._odata[ss+0]);
chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
spProj5p(tmp,psi._odata[ss+s-1]);
chi[ss+s]=chi[ss+s]+lower[s]*tmp;
} else {
spProj5m(tmp,psi._odata[ss+s+1]);
chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
spProj5p(tmp,psi._odata[ss+s-1]);
chi[ss+s]=chi[ss+s]+lower[s]*tmp;
}
}
}
}
template<class Impl>
void CayleyFermion5D<Impl>::M5Ddag(const FermionField &psi,
const FermionField &phi,
FermionField &chi,
std::vector<RealD> &lower,
std::vector<RealD> &diag,
std::vector<RealD> &upper)
{
int Ls =this->Ls;
GridBase *grid=psi._grid;
assert(phi.checkerboard == psi.checkerboard);
chi.checkerboard=psi.checkerboard;
PARALLEL_FOR_LOOP
for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
auto tmp = psi._odata[0];
for(int s=0;s<Ls;s++){
if ( s==0 ) {
spProj5p(tmp,psi._odata[ss+s+1]);
chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
spProj5m(tmp,psi._odata[ss+Ls-1]);
chi[ss+s]=chi[ss+s]+lower[s]*tmp;
} else if ( s==(Ls-1)) {
spProj5p(tmp,psi._odata[ss+0]);
chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
spProj5m(tmp,psi._odata[ss+s-1]);
chi[ss+s]=chi[ss+s]+lower[s]*tmp;
} else {
spProj5p(tmp,psi._odata[ss+s+1]);
chi[ss+s]=diag[s]*phi[ss+s]+upper[s]*tmp;
spProj5m(tmp,psi._odata[ss+s-1]);
chi[ss+s]=chi[ss+s]+lower[s]*tmp;
}
}
}
}
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInv (const FermionField &psi, FermionField &chi)
{
GridBase *grid=psi._grid;
int Ls=this->Ls;
chi.checkerboard=psi.checkerboard;
PARALLEL_FOR_LOOP
for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
auto tmp = psi._odata[0];
// Apply (L^{\prime})^{-1}
chi[ss]=psi[ss]; // chi[0]=psi[0]
for(int s=1;s<Ls;s++){
spProj5p(tmp,chi[ss+s-1]);
chi[ss+s] = psi[ss+s]-lee[s-1]*tmp;
}
// L_m^{-1}
for (int s=0;s<Ls-1;s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
spProj5m(tmp,chi[ss+s]);
chi[ss+Ls-1] = chi[ss+Ls-1] - leem[s]*tmp;
}
// U_m^{-1} D^{-1}
for (int s=0;s<Ls-1;s++){
// Chi[s] + 1/d chi[s]
spProj5p(tmp,chi[ss+Ls-1]);
chi[ss+s] = (1.0/dee[s])*chi[ss+s]-(ueem[s]/dee[Ls-1])*tmp;
}
chi[ss+Ls-1]= (1.0/dee[Ls-1])*chi[ss+Ls-1];
// Apply U^{-1}
for (int s=Ls-2;s>=0;s--){
spProj5m(tmp,chi[ss+s+1]);
chi[ss+s] = chi[ss+s] - uee[s]*tmp;
}
}
}
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &chi)
{
GridBase *grid=psi._grid;
int Ls=this->Ls;
assert(psi.checkerboard == psi.checkerboard);
chi.checkerboard=psi.checkerboard;
PARALLEL_FOR_LOOP
for(int ss=0;ss<grid->oSites();ss+=Ls){ // adds Ls
auto tmp = psi._odata[0];
// Apply (U^{\prime})^{-dagger}
chi[ss]=psi[ss];
for (int s=1;s<Ls;s++){
spProj5m(tmp,chi[ss+s-1]);
chi[ss+s] = psi[ss+s]-uee[s-1]*tmp;
}
// U_m^{-\dagger}
for (int s=0;s<Ls-1;s++){
spProj5p(tmp,chi[ss+s]);
chi[ss+Ls-1] = chi[ss+Ls-1] - ueem[s]*tmp;
}
// L_m^{-\dagger} D^{-dagger}
for (int s=0;s<Ls-1;s++){
spProj5m(tmp,chi[ss+Ls-1]);
chi[ss+s] = (1.0/dee[s])*chi[ss+s]-(leem[s]/dee[Ls-1])*tmp;
}
chi[ss+Ls-1]= (1.0/dee[Ls-1])*chi[ss+Ls-1];
// Apply L^{-dagger}
for (int s=Ls-2;s>=0;s--){
spProj5p(tmp,chi[ss+s+1]);
chi[ss+s] = chi[ss+s] - lee[s]*tmp;
}
}
}
FermOp4dVecTemplateInstantiate(CayleyFermion5D);
GparityFermOpTemplateInstantiate(CayleyFermion5D);
}}

129
lib/qcd/action/fermion/CayleyFermion5Ddense.cc Normal file
View File

@ -0,0 +1,129 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./lib/qcd/action/fermion/CayleyFermion5D.cc
Copyright (C) 2015
Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
Author: paboyle <paboyle@ph.ed.ac.uk>
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
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
#include <Grid/Eigen/Dense>
#include <Grid.h>
namespace Grid {
namespace QCD {
/*
* Dense matrix versions of routines
*/
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &chi)
{
this->MooeeInternal(psi,chi,DaggerYes,InverseYes);
}
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInv(const FermionField &psi, FermionField &chi)
{
this->MooeeInternal(psi,chi,DaggerNo,InverseYes);
}
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv)
{
int Ls=this->Ls;
int LLs = psi._grid->_rdimensions[0];
int vol = psi._grid->oSites()/LLs;
chi.checkerboard=psi.checkerboard;
assert(Ls==LLs);
Eigen::MatrixXd Pplus = Eigen::MatrixXd::Zero(Ls,Ls);
Eigen::MatrixXd Pminus = Eigen::MatrixXd::Zero(Ls,Ls);
for(int s=0;s<Ls;s++){
Pplus(s,s) = bee[s];
Pminus(s,s)= bee[s];
}
for(int s=0;s<Ls-1;s++){
Pminus(s,s+1) = -cee[s];
}
for(int s=0;s<Ls-1;s++){
Pplus(s+1,s) = -cee[s+1];
}
Pplus (0,Ls-1) = mass*cee[0];
Pminus(Ls-1,0) = mass*cee[Ls-1];
Eigen::MatrixXd PplusMat ;
Eigen::MatrixXd PminusMat;
if ( inv ) {
PplusMat =Pplus.inverse();
PminusMat=Pminus.inverse();
} else {
PplusMat =Pplus;
PminusMat=Pminus;
}
if(dag){
PplusMat.adjointInPlace();
PminusMat.adjointInPlace();
}
// For the non-vectorised s-direction this is simple
for(auto site=0;site<vol;site++){
SiteSpinor SiteChi;
SiteHalfSpinor SitePplus;
SiteHalfSpinor SitePminus;
for(int s1=0;s1<Ls;s1++){
SiteChi =zero;
for(int s2=0;s2<Ls;s2++){
int lex2 = s2+Ls*site;
if ( PplusMat(s1,s2) != 0.0 ) {
spProj5p(SitePplus,psi[lex2]);
accumRecon5p(SiteChi,PplusMat (s1,s2)*SitePplus);
}
if ( PminusMat(s1,s2) != 0.0 ) {
spProj5m(SitePminus,psi[lex2]);
accumRecon5m(SiteChi,PminusMat(s1,s2)*SitePminus);
}
}
chi[s1+Ls*site] = SiteChi*0.5;
}
}
}
FermOp4dVecTemplateInstantiate(CayleyFermion5D);
}}

144
lib/qcd/action/fermion/CayleyFermion5Dssp.cc Normal file
View File

@ -0,0 +1,144 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./lib/qcd/action/fermion/CayleyFermion5D.cc
Copyright (C) 2015
Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
Author: paboyle <paboyle@ph.ed.ac.uk>
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
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
#include <Grid.h>
namespace Grid {
namespace QCD {
// FIXME -- make a version of these routines with site loop outermost for cache reuse.
// Pminus fowards
// Pplus backwards
template<class Impl>
void CayleyFermion5D<Impl>::M5D(const FermionField &psi,
const FermionField &phi,
FermionField &chi,
std::vector<RealD> &lower,
std::vector<RealD> &diag,
std::vector<RealD> &upper)
{
int Ls=this->Ls;
for(int s=0;s<Ls;s++){
if ( s==0 ) {
axpby_ssp_pminus(chi,diag[s],phi,upper[s],psi,s,s+1);
axpby_ssp_pplus (chi,1.0,chi,lower[s],psi,s,Ls-1);
} else if ( s==(Ls-1)) {
axpby_ssp_pminus(chi,diag[s],phi,upper[s],psi,s,0);
axpby_ssp_pplus (chi,1.0,chi,lower[s],psi,s,s-1);
} else {
axpby_ssp_pminus(chi,diag[s],phi,upper[s],psi,s,s+1);
axpby_ssp_pplus(chi,1.0,chi,lower[s],psi,s,s-1);
}
}
}
template<class Impl>
void CayleyFermion5D<Impl>::M5Ddag(const FermionField &psi,
const FermionField &phi,
FermionField &chi,
std::vector<RealD> &lower,
std::vector<RealD> &diag,
std::vector<RealD> &upper)
{
int Ls=this->Ls;
for(int s=0;s<Ls;s++){
if ( s==0 ) {
axpby_ssp_pplus (chi,diag[s],phi,upper[s],psi,s,s+1);
axpby_ssp_pminus(chi,1.0,chi,lower[s],psi,s,Ls-1);
} else if ( s==(Ls-1)) {
axpby_ssp_pplus (chi,diag[s],phi,upper[s],psi,s,0);
axpby_ssp_pminus(chi,1.0,chi,lower[s],psi,s,s-1);
} else {
axpby_ssp_pplus (chi,diag[s],phi,upper[s],psi,s,s+1);
axpby_ssp_pminus(chi,1.0,chi,lower[s],psi,s,s-1);
}
}
}
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInv (const FermionField &psi, FermionField &chi)
{
chi.checkerboard=psi.checkerboard;
int Ls=this->Ls;
// Apply (L^{\prime})^{-1}
axpby_ssp (chi,1.0,psi, 0.0,psi,0,0); // chi[0]=psi[0]
for (int s=1;s<Ls;s++){
axpby_ssp_pplus(chi,1.0,psi,-lee[s-1],chi,s,s-1);// recursion Psi[s] -lee P_+ chi[s-1]
}
// L_m^{-1}
for (int s=0;s<Ls-1;s++){ // Chi[ee] = 1 - sum[s<Ls-1] -leem[s]P_- chi
axpby_ssp_pminus(chi,1.0,chi,-leem[s],chi,Ls-1,s);
}
// U_m^{-1} D^{-1}
for (int s=0;s<Ls-1;s++){
// Chi[s] + 1/d chi[s]
axpby_ssp_pplus(chi,1.0/dee[s],chi,-ueem[s]/dee[Ls-1],chi,s,Ls-1);
}
axpby_ssp(chi,1.0/dee[Ls-1],chi,0.0,chi,Ls-1,Ls-1); // Modest avoidable
// Apply U^{-1}
for (int s=Ls-2;s>=0;s--){
axpby_ssp_pminus (chi,1.0,chi,-uee[s],chi,s,s+1); // chi[Ls]
}
}
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &chi)
{
chi.checkerboard=psi.checkerboard;
int Ls=this->Ls;
// Apply (U^{\prime})^{-dagger}
axpby_ssp (chi,1.0,psi, 0.0,psi,0,0); // chi[0]=psi[0]
for (int s=1;s<Ls;s++){
axpby_ssp_pminus(chi,1.0,psi,-uee[s-1],chi,s,s-1);
}
// U_m^{-\dagger}
for (int s=0;s<Ls-1;s++){
axpby_ssp_pplus(chi,1.0,chi,-ueem[s],chi,Ls-1,s);
}
// L_m^{-\dagger} D^{-dagger}
for (int s=0;s<Ls-1;s++){
axpby_ssp_pminus(chi,1.0/dee[s],chi,-leem[s]/dee[Ls-1],chi,s,Ls-1);
}
axpby_ssp(chi,1.0/dee[Ls-1],chi,0.0,chi,Ls-1,Ls-1); // Modest avoidable
// Apply L^{-dagger}
for (int s=Ls-2;s>=0;s--){
axpby_ssp_pplus (chi,1.0,chi,-lee[s],chi,s,s+1); // chi[Ls]
}
}
FermOp4dVecTemplateInstantiate(CayleyFermion5D);
GparityFermOpTemplateInstantiate(CayleyFermion5D);
}
}

301
lib/qcd/action/fermion/CayleyFermion5Dvec.cc Normal file
View File

@ -0,0 +1,301 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./lib/qcd/action/fermion/CayleyFermion5D.cc
Copyright (C) 2015
Author: Peter Boyle <pabobyle@ph.ed.ac.uk>
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
Author: Peter Boyle <peterboyle@Peters-MacBook-Pro-2.local>
Author: paboyle <paboyle@ph.ed.ac.uk>
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
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
#include <Grid/Eigen/Dense>
#include <Grid.h>
namespace Grid {
namespace QCD {
/*
* Dense matrix versions of routines
*/
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInvDag (const FermionField &psi, FermionField &chi)
{
this->MooeeInternal(psi,chi,DaggerYes,InverseYes);
}
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInv(const FermionField &psi, FermionField &chi)
{
this->MooeeInternal(psi,chi,DaggerNo,InverseYes);
}
template<class Impl>
void CayleyFermion5D<Impl>::M5D(const FermionField &psi,
const FermionField &phi,
FermionField &chi,
std::vector<RealD> &lower,
std::vector<RealD> &diag,
std::vector<RealD> &upper)
{
GridBase *grid=psi._grid;
int Ls = this->Ls;
int LLs = grid->_rdimensions[0];
int nsimd= Simd::Nsimd();
Vector<iSinglet<Simd> > u(LLs);
Vector<iSinglet<Simd> > l(LLs);
Vector<iSinglet<Simd> > d(LLs);
assert(Ls/LLs==nsimd);
assert(phi.checkerboard == psi.checkerboard);
chi.checkerboard=psi.checkerboard;
// just directly address via type pun
typedef typename Simd::scalar_type scalar_type;
scalar_type * u_p = (scalar_type *)&u[0];
scalar_type * l_p = (scalar_type *)&l[0];
scalar_type * d_p = (scalar_type *)&d[0];
for(int o=0;o<LLs;o++){ // outer
for(int i=0;i<nsimd;i++){ //inner
int s = o+i*LLs;
int ss = o*nsimd+i;
u_p[ss] = upper[s];
l_p[ss] = lower[s];
d_p[ss] = diag[s];
}}
PARALLEL_FOR_LOOP
for(int ss=0;ss<grid->oSites();ss+=LLs){ // adds LLs
alignas(64) SiteHalfSpinor hp;
alignas(64) SiteHalfSpinor hm;
alignas(64) SiteSpinor fp;
alignas(64) SiteSpinor fm;
for(int v=0;v<LLs;v++){
int vp=(v+1)%LLs;
int vm=(v+LLs-1)%LLs;
spProj5m(hp,psi[ss+vp]);
spProj5p(hm,psi[ss+vm]);
if ( vp<=v ) rotate(hp,hp,1);
if ( vm>=v ) rotate(hm,hm,nsimd-1);
hp=hp*0.5;
hm=hm*0.5;
spRecon5m(fp,hp);
spRecon5p(fm,hm);
chi[ss+v] = d[v]*phi[ss+v]+u[v]*fp;
chi[ss+v] = chi[ss+v] +l[v]*fm;
}
}
}
template<class Impl>
void CayleyFermion5D<Impl>::M5Ddag(const FermionField &psi,
const FermionField &phi,
FermionField &chi,
std::vector<RealD> &lower,
std::vector<RealD> &diag,
std::vector<RealD> &upper)
{
GridBase *grid=psi._grid;
int Ls = this->Ls;
int LLs = grid->_rdimensions[0];
int nsimd= Simd::Nsimd();
Vector<iSinglet<Simd> > u(LLs);
Vector<iSinglet<Simd> > l(LLs);
Vector<iSinglet<Simd> > d(LLs);
assert(Ls/LLs==nsimd);
assert(phi.checkerboard == psi.checkerboard);
chi.checkerboard=psi.checkerboard;
// just directly address via type pun
typedef typename Simd::scalar_type scalar_type;
scalar_type * u_p = (scalar_type *)&u[0];
scalar_type * l_p = (scalar_type *)&l[0];
scalar_type * d_p = (scalar_type *)&d[0];
for(int o=0;o<LLs;o++){ // outer
for(int i=0;i<nsimd;i++){ //inner
int s = o+i*LLs;
int ss = o*nsimd+i;
u_p[ss] = upper[s];
l_p[ss] = lower[s];
d_p[ss] = diag[s];
}}
PARALLEL_FOR_LOOP
for(int ss=0;ss<grid->oSites();ss+=LLs){ // adds LLs
alignas(64) SiteHalfSpinor hp;
alignas(64) SiteHalfSpinor hm;
alignas(64) SiteSpinor fp;
alignas(64) SiteSpinor fm;
for(int v=0;v<LLs;v++){
int vp=(v+1)%LLs;
int vm=(v+LLs-1)%LLs;
spProj5p(hp,psi[ss+vp]);
spProj5m(hm,psi[ss+vm]);
if ( vp<=v ) rotate(hp,hp,1);
if ( vm>=v ) rotate(hm,hm,nsimd-1);
hp=hp*0.5;
hm=hm*0.5;
spRecon5p(fp,hp);
spRecon5m(fm,hm);
chi[ss+v] = d[v]*phi[ss+v]+u[v]*fp;
chi[ss+v] = chi[ss+v] +l[v]*fm;
}
}
}
template<class Impl>
void CayleyFermion5D<Impl>::MooeeInternal(const FermionField &psi, FermionField &chi,int dag, int inv)
{
int Ls=this->Ls;
int LLs = psi._grid->_rdimensions[0];
int vol = psi._grid->oSites()/LLs;
chi.checkerboard=psi.checkerboard;
Eigen::MatrixXd Pplus = Eigen::MatrixXd::Zero(Ls,Ls);
Eigen::MatrixXd Pminus = Eigen::MatrixXd::Zero(Ls,Ls);
for(int s=0;s<Ls;s++){
Pplus(s,s) = bee[s];
Pminus(s,s)= bee[s];
}
for(int s=0;s<Ls-1;s++){
Pminus(s,s+1) = -cee[s];
}
for(int s=0;s<Ls-1;s++){
Pplus(s+1,s) = -cee[s+1];
}
Pplus (0,Ls-1) = mass*cee[0];
Pminus(Ls-1,0) = mass*cee[Ls-1];
Eigen::MatrixXd PplusMat ;
Eigen::MatrixXd PminusMat;
if ( inv ) {
PplusMat =Pplus.inverse();
PminusMat=Pminus.inverse();
} else {
PplusMat =Pplus;
PminusMat=Pminus;
}
if(dag){
PplusMat.adjointInPlace();
PminusMat.adjointInPlace();
}
typedef typename SiteHalfSpinor::scalar_type scalar_type;
const int Nsimd=Simd::Nsimd();
Vector<iSinglet<Simd> > Matp(Ls*LLs);
Vector<iSinglet<Simd> > Matm(Ls*LLs);
for(int s2=0;s2<Ls;s2++){
for(int s1=0;s1<LLs;s1++){
int istride = LLs;
int ostride = 1;
Simd Vp;
Simd Vm;
scalar_type *sp = (scalar_type *)&Vp;
scalar_type *sm = (scalar_type *)&Vm;
for(int l=0;l<Nsimd;l++){
sp[l] = PplusMat (l*istride+s1*ostride ,s2);
sm[l] = PminusMat(l*istride+s1*ostride,s2);
}
Matp[LLs*s2+s1] = Vp;
Matm[LLs*s2+s1] = Vm;
}
}
// Dynamic allocate on stack to get per thread without serialised heap acces
PARALLEL_FOR_LOOP
for(auto site=0;site<vol;site++){
// SiteHalfSpinor *SitePplus =(SiteHalfSpinor *) alloca(LLs*sizeof(SiteHalfSpinor));
// SiteHalfSpinor *SitePminus=(SiteHalfSpinor *) alloca(LLs*sizeof(SiteHalfSpinor));
// SiteSpinor *SiteChi =(SiteSpinor *) alloca(LLs*sizeof(SiteSpinor));
Vector<SiteHalfSpinor> SitePplus(LLs);
Vector<SiteHalfSpinor> SitePminus(LLs);
Vector<SiteHalfSpinor> SiteChiP(LLs);
Vector<SiteHalfSpinor> SiteChiM(LLs);
Vector<SiteSpinor> SiteChi(LLs);
SiteHalfSpinor BcastP;
SiteHalfSpinor BcastM;
for(int s=0;s<LLs;s++){
int lex = s+LLs*site;
spProj5p(SitePplus[s] ,psi[lex]);
spProj5m(SitePminus[s],psi[lex]);
SiteChiP[s]=zero;
SiteChiM[s]=zero;
}
int s=0;
for(int l=0; l<Simd::Nsimd();l++){ // simd lane
for(int s2=0;s2<LLs;s2++){ // Column loop of right hand side
vbroadcast(BcastP,SitePplus [s2],l);
vbroadcast(BcastM,SitePminus[s2],l);
for(int s1=0;s1<LLs;s1++){ // Column loop of reduction variables
SiteChiP[s1]=SiteChiP[s1]+Matp[LLs*s+s1]*BcastP;
SiteChiM[s1]=SiteChiM[s1]+Matm[LLs*s+s1]*BcastM;
}
s++;
}}
for(int s=0;s<LLs;s++){
int lex = s+LLs*site;
spRecon5p(SiteChi[s],SiteChiP[s]);
accumRecon5m(SiteChi[s],SiteChiM[s]);
chi[lex] = SiteChi[s]*0.5;
}
}
}
FermOp5dVecTemplateInstantiate(CayleyFermion5D);
}}

155
tests/core/Test_checker.cc Normal file
View File

@ -0,0 +1,155 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./tests/Test_dwf_cg_prec.cc
Copyright (C) 2015
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
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
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
#include <Grid/Grid.h>
using namespace std;
using namespace Grid;
using namespace Grid::QCD;
template<class d>
struct scal {
d internal;
};
Gamma::GammaMatrix Gmu [] = {
Gamma::GammaX,
Gamma::GammaY,
Gamma::GammaZ,
Gamma::GammaT
};
int toint(const char* str){
std::stringstream os; os << str;
int out; os >> out;
return out;
}
int main (int argc, char ** argv)
{
Grid_init(&argc,&argv);
assert(argc >= 5);
std::vector<int> latt(4,0);
latt[0] = toint(argv[1]);
latt[1] = toint(argv[2]);
latt[2] = toint(argv[3]);
latt[3] = toint(argv[4]);
const int Ls= toint(argv[5]);
std::cout << "Lattice size (" << latt[0] << "," << latt[1] << "," << latt[2] << "," << latt[3] << ") Ls=" << Ls << std::endl;
std::vector<int> simd_layout = GridDefaultSimd(Nd,vComplexD::Nsimd());
std::cout << "SIMD layout (" << simd_layout[0] << "," << simd_layout[1] << "," << simd_layout[2] << "," << simd_layout[3] << ")" << std::endl;
GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(latt, simd_layout,GridDefaultMpi());
GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
GridCartesian * FGrid = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
std::vector<int> seeds4({1,2,3,4});
std::vector<int> seeds5({5,6,7,8});
GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
//typedef Lattice<iGparitySpinColourVector<vComplexD> > LatticeType;
typedef LatticeFermionD LatticeType;
LatticeType src(FGrid); random(RNG5,src);
LatticeType src_o(FrbGrid);
pickCheckerboard(Odd,src_o,src);
std::vector<int> site(5);
std::vector<int> cbsite(5);
typedef typename GridTypeMapper<LatticeType::vector_object>::scalar_object sobj;
// std::cout << "sizeof(vobj) " << sizeof(LatticeType::vector_object) << std::endl;
// std::cout << "sizeof(sobj) " << sizeof(sobj) << std::endl;
std::cout << "v1 from uncheckerboarded field, v2 from odd-parity red-black field\n";
for(site[0]=0;site[0]<Ls;site[0]++){
for(site[4]=0;site[4]<latt[3];site[4]++){
for(site[3]=0;site[3]<latt[2];site[3]++){
for(site[2]=0;site[2]<latt[1];site[2]++){
for(site[1]=0;site[1]<latt[0];site[1]++){
if(src_o._grid->CheckerBoard(site) != src_o.checkerboard)
continue;
std::cout << "Site (" << site[0] << "," << site[1] << "," << site[2] << "," << site[3] << "," << site[4] << ")" << std::endl;
sobj v1, v2;
peekLocalSite(v1,src,site);
peekLocalSite(v2,src_o,site);
RealD v1_norm = norm2(v1);
RealD v2_norm = norm2(v2);
RealD diff = v2_norm - v1_norm;
std::cout << v1_norm << " " << v2_norm << " " << diff << '\n';
if(fabs(diff) > 1e-12){
std::cout << "ERROR!\n";
exit(-1);
}
}
}
}
}
}
// LatticeFermion result(FGrid); result=zero;
// LatticeGaugeField Umu(UGrid);
// SU3::HotConfiguration(RNG4,Umu);
// std::vector<LatticeColourMatrix> U(4,UGrid);
// for(int mu=0;mu<Nd;mu++){
// U[mu] = PeekIndex<LorentzIndex>(Umu,mu);
// }
// RealD mass=0.1;
// RealD M5=1.8;
// DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
// LatticeFermion src_o(FrbGrid);
// LatticeFermion result_o(FrbGrid);
// pickCheckerboard(Odd,src_o,src);
// result_o=zero;
// SchurDiagMooeeOperator<DomainWallFermionR,LatticeFermion> HermOpEO(Ddwf);
// ConjugateGradient<LatticeFermion> CG(1.0e-8,10000);
// CG(HermOpEO,src_o,result_o);
Grid_finalize();
}

460
tests/debug/Test_cayley_even_odd_vec.cc Normal file
View File

@ -0,0 +1,460 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./tests/Test_cayley_even_odd.cc
Copyright (C) 2015
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
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
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License along
with this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
See the full license in the file "LICENSE" in the top level distribution directory
*************************************************************************************/
/* END LEGAL */
#include <Grid/Grid.h>
using namespace std;
using namespace Grid;
using namespace Grid::QCD;
template<class d>
struct scal {
d internal;
};
Gamma::GammaMatrix Gmu [] = {
Gamma::GammaX,
Gamma::GammaY,
Gamma::GammaZ,
Gamma::GammaT
};
typedef DomainWallFermion<DomainWallVec5dImplR> DomainWallVecFermionR;
typedef MobiusFermion<DomainWallVec5dImplR> MobiusVecFermionR;
typedef MobiusZolotarevFermion<DomainWallVec5dImplR> MobiusZolotarevVecFermionR;
typedef ScaledShamirFermion<DomainWallVec5dImplR> ScaledShamirVecFermionR;
typedef ShamirZolotarevFermion<DomainWallVec5dImplR> ShamirZolotarevVecFermionR;
typedef OverlapWilsonCayleyTanhFermion<DomainWallVec5dImplR> OverlapWilsonCayleyTanhVecFermionR;
typedef OverlapWilsonCayleyZolotarevFermion<DomainWallVec5dImplR> OverlapWilsonCayleyZolotarevVecFermionR;
template<class What>
void TestWhat(What & Ddwf,
GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
GridCartesian * UGrid,
RealD mass, RealD M5,
GridParallelRNG *RNG4, GridParallelRNG *RNG5);
template<class This,class That>
void TestMoo(This & Dw, That &sDw);
int main (int argc, char ** argv)
{
Grid_init(&argc,&argv);
int threads = GridThread::GetThreads();
std::cout<<GridLogMessage << "Grid is setup to use "<<threads<<" threads"<<std::endl;
const int Ls=16;
std::vector<int> latt4 =GridDefaultLatt();
GridCartesian * UGrid = SpaceTimeGrid::makeFourDimGrid(latt4, GridDefaultSimd(Nd,vComplex::Nsimd()),GridDefaultMpi());
GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
GridCartesian * FGrid = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
GridCartesian * sUGrid = SpaceTimeGrid::makeFourDimDWFGrid(latt4,GridDefaultMpi());
GridRedBlackCartesian * sUrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(sUGrid);
GridCartesian * sFGrid = SpaceTimeGrid::makeFiveDimDWFGrid(Ls,UGrid);
GridRedBlackCartesian * sFrbGrid = SpaceTimeGrid::makeFiveDimDWFRedBlackGrid(Ls,UGrid);
std::vector<int> seeds4({1,2,3,4});
std::vector<int> seeds5({5,6,7,8});
GridParallelRNG RNG4(UGrid); RNG4.SeedFixedIntegers(seeds4);
GridParallelRNG RNG5(FGrid); RNG5.SeedFixedIntegers(seeds5);
GridParallelRNG sRNG4(sUGrid); sRNG4.SeedFixedIntegers(seeds4);
GridParallelRNG sRNG5(sFGrid); sRNG5.SeedFixedIntegers(seeds5);
LatticeGaugeField Umu(UGrid); random(RNG4,Umu);
RealD mass=0.1;
RealD M5 =1.8;
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
std::cout<<GridLogMessage <<"DomainWallFermion vectorised test"<<std::endl;
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
DomainWallFermionR Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
DomainWallVecFermionR sDdwf(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5);
TestMoo(Ddwf,sDdwf);
TestWhat<DomainWallFermionR>(Ddwf,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
TestWhat<DomainWallVecFermionR>(sDdwf,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
RealD b=1.5;// Scale factor b+c=2, b-c=1
RealD c=0.5;
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
std::cout<<GridLogMessage <<"MobiusFermion test"<<std::endl;
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
MobiusFermionR Dmob(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c);
MobiusVecFermionR sDmob(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,b,c);
TestMoo(Dmob,sDmob);
TestWhat<MobiusFermionR>(Dmob,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
TestWhat<MobiusVecFermionR>(sDmob,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
std::cout<<GridLogMessage <<"MobiusZolotarevFermion test"<<std::endl;
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
MobiusZolotarevFermionR Dzolo(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,b,c,0.1,2.0);
MobiusZolotarevVecFermionR sDzolo(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,b,c,0.1,2.0);
TestMoo(Dzolo,sDzolo);
TestWhat<MobiusZolotarevFermionR>(Dzolo,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
TestWhat<MobiusZolotarevVecFermionR>(sDzolo,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
std::cout<<GridLogMessage <<"ScaledShamirFermion test"<<std::endl;
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
ScaledShamirFermionR Dsham(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,2.0);
ScaledShamirVecFermionR sDsham(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,2.0);
TestMoo(Dsham,sDsham);
TestWhat<ScaledShamirFermionR>(Dsham,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
TestWhat<ScaledShamirVecFermionR>(sDsham,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
std::cout<<GridLogMessage <<"ShamirZolotarevFermion test"<<std::endl;
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
ShamirZolotarevFermionR Dshamz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
ShamirZolotarevVecFermionR sDshamz(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,0.1,2.0);
TestMoo(Dshamz,sDshamz);
TestWhat<ShamirZolotarevFermionR>(Dshamz,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
TestWhat<ShamirZolotarevVecFermionR>(sDshamz,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
std::cout<<GridLogMessage <<"OverlapWilsonCayleyTanhFermion test"<<std::endl;
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
OverlapWilsonCayleyTanhFermionR Dov(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,1.0);
OverlapWilsonCayleyTanhVecFermionR sDov(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,1.0);
TestMoo(Dov,sDov);
TestWhat<OverlapWilsonCayleyTanhFermionR>(Dov,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
TestWhat<OverlapWilsonCayleyTanhVecFermionR>(sDov,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
std::cout<<GridLogMessage <<"OverlapWilsonCayleyZolotarevFermion test"<<std::endl;
std::cout<<GridLogMessage<<"**************************************************************"<<std::endl;
OverlapWilsonCayleyZolotarevFermionR Dovz(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5,0.1,2.0);
OverlapWilsonCayleyZolotarevVecFermionR sDovz(Umu,*sFGrid,*sFrbGrid,*sUGrid,*sUrbGrid,mass,M5,0.1,2.0);
TestMoo(Dovz,sDovz);
TestWhat<OverlapWilsonCayleyZolotarevFermionR>(Dovz,FGrid,FrbGrid,UGrid,mass,M5,&RNG4,&RNG5);
TestWhat<OverlapWilsonCayleyZolotarevVecFermionR>(sDovz,sFGrid,sFrbGrid,sUGrid,mass,M5,&sRNG4,&sRNG5);
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
Grid_finalize();
}
template<class What>
void TestWhat(What & Ddwf,
GridCartesian * FGrid, GridRedBlackCartesian * FrbGrid,
GridCartesian * UGrid,
RealD mass, RealD M5,
GridParallelRNG *RNG4,
GridParallelRNG *RNG5)
{
LatticeFermion src (FGrid); random(*RNG5,src);
LatticeFermion phi (FGrid); random(*RNG5,phi);
LatticeFermion chi (FGrid); random(*RNG5,chi);
LatticeFermion result(FGrid); result=zero;
LatticeFermion ref(FGrid); ref=zero;
LatticeFermion tmp(FGrid); tmp=zero;
LatticeFermion err(FGrid); tmp=zero;
LatticeFermion src_e (FrbGrid);
LatticeFermion src_o (FrbGrid);
LatticeFermion r_e (FrbGrid);
LatticeFermion r_o (FrbGrid);
LatticeFermion r_eo (FGrid);
LatticeFermion r_eeoo(FGrid);
std::cout<<GridLogMessage<<"=========================================================="<<std::endl;
std::cout<<GridLogMessage<<"= Testing that Meo + Moe + Moo + Mee = Munprec "<<std::endl;
std::cout<<GridLogMessage<<"=========================================================="<<std::endl;
pickCheckerboard(Even,src_e,src);
pickCheckerboard(Odd,src_o,src);
Ddwf.Meooe(src_e,r_o); std::cout<<GridLogMessage<<"Applied Meo"<<std::endl;
Ddwf.Meooe(src_o,r_e); std::cout<<GridLogMessage<<"Applied Moe"<<std::endl;
setCheckerboard(r_eo,r_o);
setCheckerboard(r_eo,r_e);
Ddwf.Mooee(src_e,r_e); std::cout<<GridLogMessage<<"Applied Mee"<<std::endl;
Ddwf.Mooee(src_o,r_o); std::cout<<GridLogMessage<<"Applied Moo"<<std::endl;
setCheckerboard(r_eeoo,r_e);
setCheckerboard(r_eeoo,r_o);
r_eo=r_eo+r_eeoo;
Ddwf.M(src,ref);
// std::cout<<GridLogMessage << r_eo<<std::endl;
// std::cout<<GridLogMessage << ref <<std::endl;
err= ref - r_eo;
std::cout<<GridLogMessage << "EO norm diff "<< norm2(err)<< " "<<norm2(ref)<< " " << norm2(r_eo) <<std::endl;
LatticeComplex cerr(FGrid);
cerr = localInnerProduct(err,err);
// std::cout<<GridLogMessage << cerr<<std::endl;
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
std::cout<<GridLogMessage<<"= Test Ddagger is the dagger of D by requiring "<<std::endl;
std::cout<<GridLogMessage<<"= < phi | Deo | chi > * = < chi | Deo^dag| phi> "<<std::endl;
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
LatticeFermion chi_e (FrbGrid);
LatticeFermion chi_o (FrbGrid);
LatticeFermion dchi_e (FrbGrid);
LatticeFermion dchi_o (FrbGrid);
LatticeFermion phi_e (FrbGrid);
LatticeFermion phi_o (FrbGrid);
LatticeFermion dphi_e (FrbGrid);
LatticeFermion dphi_o (FrbGrid);
pickCheckerboard(Even,chi_e,chi);
pickCheckerboard(Odd ,chi_o,chi);
pickCheckerboard(Even,phi_e,phi);
pickCheckerboard(Odd ,phi_o,phi);
Ddwf.Meooe(chi_e,dchi_o);
Ddwf.Meooe(chi_o,dchi_e);
Ddwf.MeooeDag(phi_e,dphi_o);
Ddwf.MeooeDag(phi_o,dphi_e);
ComplexD pDce = innerProduct(phi_e,dchi_e);
ComplexD pDco = innerProduct(phi_o,dchi_o);
ComplexD cDpe = innerProduct(chi_e,dphi_e);
ComplexD cDpo = innerProduct(chi_o,dphi_o);
std::cout<<GridLogMessage <<"e "<<pDce<<" "<<cDpe <<std::endl;
std::cout<<GridLogMessage <<"o "<<pDco<<" "<<cDpo <<std::endl;
std::cout<<GridLogMessage <<"pDce - conj(cDpo) "<< pDce-conj(cDpo) <<std::endl;
std::cout<<GridLogMessage <<"pDco - conj(cDpe) "<< pDco-conj(cDpe) <<std::endl;
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
std::cout<<GridLogMessage<<"= Test MeeInv Mee = 1 "<<std::endl;
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
pickCheckerboard(Even,chi_e,chi);
pickCheckerboard(Odd ,chi_o,chi);
Ddwf.Mooee(chi_e,src_e);
Ddwf.MooeeInv(src_e,phi_e);
Ddwf.Mooee(chi_o,src_o);
Ddwf.MooeeInv(src_o,phi_o);
setCheckerboard(phi,phi_e);
setCheckerboard(phi,phi_o);
err = phi-chi;
std::cout<<GridLogMessage << "norm diff "<< norm2(err)<< std::endl;
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
std::cout<<GridLogMessage<<"= Test MeeInvDag MeeDag = 1 "<<std::endl;
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
pickCheckerboard(Even,chi_e,chi);
pickCheckerboard(Odd ,chi_o,chi);
Ddwf.MooeeDag(chi_e,src_e);
Ddwf.MooeeInvDag(src_e,phi_e);
Ddwf.MooeeDag(chi_o,src_o);
Ddwf.MooeeInvDag(src_o,phi_o);
setCheckerboard(phi,phi_e);
setCheckerboard(phi,phi_o);
err = phi-chi;
std::cout<<GridLogMessage << "norm diff "<< norm2(err)<< std::endl;
}
template<class This,class That>
void TestMoo(This & Dw, That &sDw)
{
GridBase *sgrid= sDw.FermionGrid();
GridBase *ngrid= Dw.FermionGrid();
int Ls = Dw.Ls;
LatticeFermion ssrc(sgrid);
LatticeFermion nsrc(ngrid);
LatticeFermion zz(ngrid); zz=zero;
LatticeFermion sres(sgrid);
LatticeFermion nres(ngrid);
LatticeFermion ndiff(ngrid);
LatticeFermion sdiff(sgrid);
Gamma g5( Gamma::Gamma5 );
std::vector<int> seeds({1,2,3,4,5,7,8});
GridParallelRNG RNG5(ngrid);
RNG5.SeedFixedIntegers(seeds);
random(RNG5,nsrc);
// nsrc = nsrc + g5*nsrc;
// Lattice<iScalar<vInteger> > coor(ngrid);
// LatticeCoordinate(coor,0);//scoor
// nsrc = where(coor==(Integer)0,zz,nsrc);
std::vector<int> latt4(4);
for(int d=0;d<4;d++){
latt4[d] = ngrid->_fdimensions[d+1];
}
for(int x=0;x<latt4[0];x++){
for(int y=0;y<latt4[1];y++){
for(int z=0;z<latt4[2];z++){
for(int t=0;t<latt4[3];t++){
for(int s=0;s<Ls;s++){
std::vector<int> site({s,x,y,z,t});
SpinColourVector tmp;
peekSite(tmp,nsrc,site);
pokeSite(tmp,ssrc,site);
}}}}}
sDw.Mooee(ssrc,sres);
Dw.Mooee(nsrc,nres);
sDw.MooeeInternal(ssrc,sdiff,DaggerNo,InverseNo);
for(int x=0;x<latt4[0];x++){
for(int y=0;y<latt4[1];y++){
for(int z=0;z<latt4[2];z++){
for(int t=0;t<latt4[3];t++){
for(int s=0;s<Ls;s++){
std::vector<int> site({s,x,y,z,t});
SpinColourVector stmp;
SpinColourVector itmp;
SpinColourVector dtmp;
peekSite(stmp,sres,site);
peekSite(itmp,sdiff,site);
dtmp=itmp-stmp;
if ( norm2(dtmp)>1.0e-6) {
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
std:: cout << x<<" "<<y<<" "<< z<< " "<<t<<"; s= "<<s<<std::endl;
std:: cout << "stmp "<< stmp <<std::endl;
std:: cout << "itmp "<< itmp <<std::endl;
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
}
}}}}}
sdiff = sdiff -sres;
std::cout<<GridLogMessage<<" norm MooInternal diff "<<norm2(sdiff)<<std::endl;
for(int x=0;x<latt4[0];x++){
for(int y=0;y<latt4[1];y++){
for(int z=0;z<latt4[2];z++){
for(int t=0;t<latt4[3];t++){
for(int s=0;s<Ls;s++){
std::vector<int> site({s,x,y,z,t});
SpinColourVector tmp;
peekSite(tmp,sres,site);
pokeSite(tmp,ndiff,site);
}}}}}
ndiff=ndiff-nres;
std::cout<<GridLogMessage<<" norm Moo diff "<<norm2(ndiff)<<std::endl;
sDw.MooeeDag(ssrc,sres);
Dw.MooeeDag(nsrc,nres);
sDw.MooeeInternal(ssrc,sdiff,DaggerYes,InverseNo);
sdiff = sdiff -sres;
std::cout<<GridLogMessage<<" norm MooInternalDag diff "<<norm2(sdiff)<<std::endl;
for(int x=0;x<latt4[0];x++){
for(int y=0;y<latt4[1];y++){
for(int z=0;z<latt4[2];z++){
for(int t=0;t<latt4[3];t++){
for(int s=0;s<Ls;s++){
std::vector<int> site({s,x,y,z,t});
SpinColourVector tmp;
peekSite(tmp,sres,site);
pokeSite(tmp,ndiff,site);
}}}}}
ndiff=ndiff-nres;
std::cout<<GridLogMessage<<" norm MooeeDag diff "<<norm2(ndiff)<<std::endl;
sDw.MooeeInv(ssrc,sres);
Dw.MooeeInv(nsrc,nres);
for(int x=0;x<latt4[0];x++){
for(int y=0;y<latt4[1];y++){
for(int z=0;z<latt4[2];z++){
for(int t=0;t<latt4[3];t++){
for(int s=0;s<Ls;s++){
std::vector<int> site({s,x,y,z,t});
SpinColourVector tmp;
peekSite(tmp,sres,site);
pokeSite(tmp,ndiff,site);
}}}}}
ndiff=ndiff-nres;
std::cout<<GridLogMessage<<" norm MooeeInv diff "<<norm2(ndiff)<<std::endl;
sDw.MooeeInvDag(ssrc,sres);
Dw.MooeeInvDag(nsrc,nres);
for(int x=0;x<latt4[0];x++){
for(int y=0;y<latt4[1];y++){
for(int z=0;z<latt4[2];z++){
for(int t=0;t<latt4[3];t++){
for(int s=0;s<Ls;s++){
std::vector<int> site({s,x,y,z,t});
SpinColourVector tmp;
peekSite(tmp,sres,site);
pokeSite(tmp,ndiff,site);
}}}}}
ndiff=ndiff-nres;
std::cout<<GridLogMessage<<" norm MooeeInvDag diff "<<norm2(ndiff)<<std::endl;
std::cout<<GridLogMessage<<"=============================================================="<<std::endl;
}