portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-14 09:45:36 +00:00
Code Releases Activity
546513861f
Grid/lib/qcd/action/Actions.h

147 lines
7.0 KiB
C
Raw Normal View History

Large scale change to support 5d fermion formulations. Have 5d replicated wilson with 4d gauge working and matrix regressing to Ls copies of wilson.
2015-05-31 15:09:02 +01:00
#ifndef GRID_QCD_ACTIONS_H
#define GRID_QCD_ACTIONS_H
Domain wall fermions now invert ; have the basis set up for Tanh/Zolo * (Cayley/PartFrac/ContFrac) * (Mobius/Shamir/Wilson) Approx Representation Kernel. All are done with space-time taking part in checkerboarding, Ls uncheckerboarded Have only so far tested the Domain Wall limit of mobius, and at that only checked that it i) Inverts ii) 5dim DW == Ls copies of 4dim D2 iii) MeeInv Mee == 1 iv) Meo+Mee+Moe+Moo == M unprec. v) MpcDagMpc is hermitan vi) Mdag is the adjoint of M between stochastic vectors. That said, the RB schur solve, RB MpcDagMpc solve, Unprec solve all converge and the true residual becomes small; so pretty good tests.
2015-06-02 16:57:12 +01:00
// Some reorganisation likely required as both Chroma and IroIro
// are separating the concept of the operator from that of action.
//
// The FermAction contains methods to create
//
// * Linear operators (Hermitian and non-hermitian) .. my LinearOperator
// * System solvers (Hermitian and non-hermitian) .. my OperatorFunction
// * MultiShift System solvers (Hermitian and non-hermitian) .. my OperatorFunction
////////////////////////////////////////////
// Abstract base interface
////////////////////////////////////////////
More progress in the HMC construction
2015-07-03 18:43:14 +01:00
#include <qcd/action/ActionBase.h>
////////////////////////////////////////////
// Gauge Actions
////////////////////////////////////////////
#include <qcd/action/gauge/WilsonGaugeAction.h>
Domain wall fermions now invert ; have the basis set up for Tanh/Zolo * (Cayley/PartFrac/ContFrac) * (Mobius/Shamir/Wilson) Approx Representation Kernel. All are done with space-time taking part in checkerboarding, Ls uncheckerboarded Have only so far tested the Domain Wall limit of mobius, and at that only checked that it i) Inverts ii) 5dim DW == Ls copies of 4dim D2 iii) MeeInv Mee == 1 iv) Meo+Mee+Moe+Moo == M unprec. v) MpcDagMpc is hermitan vi) Mdag is the adjoint of M between stochastic vectors. That said, the RB schur solve, RB MpcDagMpc solve, Unprec solve all converge and the true residual becomes small; so pretty good tests.
2015-06-02 16:57:12 +01:00
////////////////////////////////////////////
// Utility functions
////////////////////////////////////////////
#include <qcd/action/fermion/WilsonCompressor.h> //used by all wilson type fermions
#include <qcd/action/fermion/WilsonKernels.h> //used by all wilson type fermions
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
////////////////////////////////////////////////////////////////////////////////////////////////////
// Explicit explicit template instantiation is still required in the .cc files
//
// - CayleyFermion5D.cc
// - PartialFractionFermion5D.cc
// - WilsonFermion5D.cc
// - WilsonKernelsHand.cc
// - ContinuedFractionFermion5D.cc
// - WilsonFermion.cc
// - WilsonKernels.cc
//
// The explicit instantiation is only avoidable if we move this source to headers and end up with include/parse/recompile
// for EVERY .cc file. This define centralises the list and restores global push of impl cases
////////////////////////////////////////////////////////////////////////////////////////////////////
#define FermOpTemplateInstantiate(A) \
template class A<WilsonImplF>; \
template class A<WilsonImplD>;
Domain wall fermions now invert ; have the basis set up for Tanh/Zolo * (Cayley/PartFrac/ContFrac) * (Mobius/Shamir/Wilson) Approx Representation Kernel. All are done with space-time taking part in checkerboarding, Ls uncheckerboarded Have only so far tested the Domain Wall limit of mobius, and at that only checked that it i) Inverts ii) 5dim DW == Ls copies of 4dim D2 iii) MeeInv Mee == 1 iv) Meo+Mee+Moe+Moo == M unprec. v) MpcDagMpc is hermitan vi) Mdag is the adjoint of M between stochastic vectors. That said, the RB schur solve, RB MpcDagMpc solve, Unprec solve all converge and the true residual becomes small; so pretty good tests.
2015-06-02 16:57:12 +01:00
////////////////////////////////////////////
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
// Fermion operators / actions
Domain wall fermions now invert ; have the basis set up for Tanh/Zolo * (Cayley/PartFrac/ContFrac) * (Mobius/Shamir/Wilson) Approx Representation Kernel. All are done with space-time taking part in checkerboarding, Ls uncheckerboarded Have only so far tested the Domain Wall limit of mobius, and at that only checked that it i) Inverts ii) 5dim DW == Ls copies of 4dim D2 iii) MeeInv Mee == 1 iv) Meo+Mee+Moe+Moo == M unprec. v) MpcDagMpc is hermitan vi) Mdag is the adjoint of M between stochastic vectors. That said, the RB schur solve, RB MpcDagMpc solve, Unprec solve all converge and the true residual becomes small; so pretty good tests.
2015-06-02 16:57:12 +01:00
////////////////////////////////////////////
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
#include <qcd/action/fermion/FermionOperator.h>
Overlap Wilson Cayley tanh & zolo
2015-06-03 11:26:54 +01:00
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
#include <qcd/action/fermion/WilsonFermion.h> // 4d wilson like
//#include <qcd/action/fermion/CloverFermion.h>
Overlap Wilson Cayley tanh & zolo
2015-06-03 11:26:54 +01:00
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
#include <qcd/action/fermion/WilsonFermion5D.h> // 5d base used by all 5d overlap types
Domain wall fermions now invert ; have the basis set up for Tanh/Zolo * (Cayley/PartFrac/ContFrac) * (Mobius/Shamir/Wilson) Approx Representation Kernel. All are done with space-time taking part in checkerboarding, Ls uncheckerboarded Have only so far tested the Domain Wall limit of mobius, and at that only checked that it i) Inverts ii) 5dim DW == Ls copies of 4dim D2 iii) MeeInv Mee == 1 iv) Meo+Mee+Moe+Moo == M unprec. v) MpcDagMpc is hermitan vi) Mdag is the adjoint of M between stochastic vectors. That said, the RB schur solve, RB MpcDagMpc solve, Unprec solve all converge and the true residual becomes small; so pretty good tests.
2015-06-02 16:57:12 +01:00
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
#include <qcd/action/fermion/CayleyFermion5D.h> // Cayley types
Domain wall fermions now invert ; have the basis set up for Tanh/Zolo * (Cayley/PartFrac/ContFrac) * (Mobius/Shamir/Wilson) Approx Representation Kernel. All are done with space-time taking part in checkerboarding, Ls uncheckerboarded Have only so far tested the Domain Wall limit of mobius, and at that only checked that it i) Inverts ii) 5dim DW == Ls copies of 4dim D2 iii) MeeInv Mee == 1 iv) Meo+Mee+Moe+Moo == M unprec. v) MpcDagMpc is hermitan vi) Mdag is the adjoint of M between stochastic vectors. That said, the RB schur solve, RB MpcDagMpc solve, Unprec solve all converge and the true residual becomes small; so pretty good tests.
2015-06-02 16:57:12 +01:00
#include <qcd/action/fermion/DomainWallFermion.h>
Mobius Caley form, Mobius Zolotarev operators. Pass Even Odd vs unprec test and hermiticity checks in tests/Grid_any_evenodd.cc; will work on inversion tests shortly.
2015-06-03 09:36:26 +01:00
#include <qcd/action/fermion/DomainWallFermion.h>
#include <qcd/action/fermion/MobiusFermion.h>
Scaled Shamir and Scaled Shamir Zolotarev aliases for special cases of Mobius.
2015-06-03 09:51:06 +01:00
#include <qcd/action/fermion/ScaledShamirFermion.h>
Overlap Wilson Cayley tanh & zolo
2015-06-03 11:26:54 +01:00
#include <qcd/action/fermion/MobiusZolotarevFermion.h>
#include <qcd/action/fermion/ShamirZolotarevFermion.h>
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
#include <qcd/action/fermion/OverlapWilsonCayleyTanhFermion.h>
Overlap Wilson Cayley tanh & zolo
2015-06-03 11:26:54 +01:00
#include <qcd/action/fermion/OverlapWilsonCayleyZolotarevFermion.h>
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
#include <qcd/action/fermion/ContinuedFractionFermion5D.h> // Continued fraction
#include <qcd/action/fermion/OverlapWilsonContFracTanhFermion.h>
#include <qcd/action/fermion/OverlapWilsonContFracZolotarevFermion.h>
Overlap Wilson Cayley tanh & zolo
2015-06-03 11:26:54 +01:00
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
#include <qcd/action/fermion/PartialFractionFermion5D.h> // Partial fraction
PartialFraction Hw with Zolo and Tanh approx converged under CG and passed EO breakdown and hermiticity tests.
2015-06-04 13:28:37 +01:00
#include <qcd/action/fermion/OverlapWilsonPartialFractionTanhFermion.h>
#include <qcd/action/fermion/OverlapWilsonPartialFractionZolotarevFermion.h>
Domain wall fermions now invert ; have the basis set up for Tanh/Zolo * (Cayley/PartFrac/ContFrac) * (Mobius/Shamir/Wilson) Approx Representation Kernel. All are done with space-time taking part in checkerboarding, Ls uncheckerboarded Have only so far tested the Domain Wall limit of mobius, and at that only checked that it i) Inverts ii) 5dim DW == Ls copies of 4dim D2 iii) MeeInv Mee == 1 iv) Meo+Mee+Moe+Moo == M unprec. v) MpcDagMpc is hermitan vi) Mdag is the adjoint of M between stochastic vectors. That said, the RB schur solve, RB MpcDagMpc solve, Unprec solve all converge and the true residual becomes small; so pretty good tests.
2015-06-02 16:57:12 +01:00
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
////////////////////////////////////////////////////////////////////////////////////////////////////
// More maintainable to maintain the following typedef list centrally, as more "impl" targets
// are added, (e.g. extension for gparity, half precision project in comms etc..)
////////////////////////////////////////////////////////////////////////////////////////////////////
// Cayley 5d
namespace Grid {
namespace QCD {
typedef WilsonFermion<WilsonImplR> WilsonFermionR;
typedef WilsonFermion<WilsonImplF> WilsonFermionF;
typedef WilsonFermion<WilsonImplD> WilsonFermionD;
typedef DomainWallFermion<WilsonImplR> DomainWallFermionR;
typedef DomainWallFermion<WilsonImplF> DomainWallFermionF;
typedef DomainWallFermion<WilsonImplD> DomainWallFermionD;
typedef MobiusFermion<WilsonImplR> MobiusFermionR;
typedef MobiusFermion<WilsonImplF> MobiusFermionF;
typedef MobiusFermion<WilsonImplD> MobiusFermionD;
typedef ScaledShamirFermion<WilsonImplR> ScaledShamirFermionR;
typedef ScaledShamirFermion<WilsonImplF> ScaledShamirFermionF;
typedef ScaledShamirFermion<WilsonImplD> ScaledShamirFermionD;
typedef MobiusZolotarevFermion<WilsonImplR> MobiusZolotarevFermionR;
typedef MobiusZolotarevFermion<WilsonImplF> MobiusZolotarevFermionF;
typedef MobiusZolotarevFermion<WilsonImplD> MobiusZolotarevFermionD;
typedef ShamirZolotarevFermion<WilsonImplR> ShamirZolotarevFermionR;
typedef ShamirZolotarevFermion<WilsonImplF> ShamirZolotarevFermionF;
typedef ShamirZolotarevFermion<WilsonImplD> ShamirZolotarevFermionD;
typedef OverlapWilsonCayleyTanhFermion<WilsonImplR> OverlapWilsonCayleyTanhFermionR;
typedef OverlapWilsonCayleyTanhFermion<WilsonImplF> OverlapWilsonCayleyTanhFermionF;
typedef OverlapWilsonCayleyTanhFermion<WilsonImplD> OverlapWilsonCayleyTanhFermionD;
typedef OverlapWilsonCayleyZolotarevFermion<WilsonImplR> OverlapWilsonCayleyZolotarevFermionR;
typedef OverlapWilsonCayleyZolotarevFermion<WilsonImplF> OverlapWilsonCayleyZolotarevFermionF;
typedef OverlapWilsonCayleyZolotarevFermion<WilsonImplD> OverlapWilsonCayleyZolotarevFermionD;
// Continued fraction
typedef OverlapWilsonContFracTanhFermion<WilsonImplR> OverlapWilsonContFracTanhFermionR;
typedef OverlapWilsonContFracTanhFermion<WilsonImplF> OverlapWilsonContFracTanhFermionF;
typedef OverlapWilsonContFracTanhFermion<WilsonImplD> OverlapWilsonContFracTanhFermionD;
typedef OverlapWilsonContFracZolotarevFermion<WilsonImplR> OverlapWilsonContFracZolotarevFermionR;
typedef OverlapWilsonContFracZolotarevFermion<WilsonImplF> OverlapWilsonContFracZolotarevFermionF;
typedef OverlapWilsonContFracZolotarevFermion<WilsonImplD> OverlapWilsonContFracZolotarevFermionD;
// Partial fraction
typedef OverlapWilsonPartialFractionTanhFermion<WilsonImplR> OverlapWilsonPartialFractionTanhFermionR;
typedef OverlapWilsonPartialFractionTanhFermion<WilsonImplF> OverlapWilsonPartialFractionTanhFermionF;
typedef OverlapWilsonPartialFractionTanhFermion<WilsonImplD> OverlapWilsonPartialFractionTanhFermionD;
typedef OverlapWilsonPartialFractionZolotarevFermion<WilsonImplR> OverlapWilsonPartialFractionZolotarevFermionR;
typedef OverlapWilsonPartialFractionZolotarevFermion<WilsonImplF> OverlapWilsonPartialFractionZolotarevFermionF;
typedef OverlapWilsonPartialFractionZolotarevFermion<WilsonImplD> OverlapWilsonPartialFractionZolotarevFermionD;
}}
Conjugate residual added
2015-06-05 18:16:25 +01:00
///////////////////////////////////////////////////////////////////////////////
// G5 herm -- this has to live in QCD since dirac matrix is not in the broader sector of code
///////////////////////////////////////////////////////////////////////////////
#include <qcd/action/fermion/g5HermitianLinop.h>
Domain wall fermions now invert ; have the basis set up for Tanh/Zolo * (Cayley/PartFrac/ContFrac) * (Mobius/Shamir/Wilson) Approx Representation Kernel. All are done with space-time taking part in checkerboarding, Ls uncheckerboarded Have only so far tested the Domain Wall limit of mobius, and at that only checked that it i) Inverts ii) 5dim DW == Ls copies of 4dim D2 iii) MeeInv Mee == 1 iv) Meo+Mee+Moe+Moo == M unprec. v) MpcDagMpc is hermitan vi) Mdag is the adjoint of M between stochastic vectors. That said, the RB schur solve, RB MpcDagMpc solve, Unprec solve all converge and the true residual becomes small; so pretty good tests.
2015-06-02 16:57:12 +01:00
Two flavour HMC for Wilson/Wilson is conserving energy. Still to check plaq and <e(-dH)>, but nevertheless this is progress
2015-07-29 09:53:39 +01:00
////////////////////////////////////////
Rework/global edit to enforce type templating of fermion operators. Allows multi-precision work and paves the way for alternate BC's and such like allowing for example G-parity which is important for K pipi programme. In particular, can drive an extra flavour index into the fermion fields using template types.
2015-08-10 20:47:44 +01:00
// Pseudo fermion combinations for HMC
Two flavour HMC for Wilson/Wilson is conserving energy. Still to check plaq and <e(-dH)>, but nevertheless this is progress
2015-07-29 09:53:39 +01:00
////////////////////////////////////////
#include <qcd/action/pseudofermion/TwoFlavour.h>
Adding components for even odd decomposed determinant in HMC. dH not yet conserved, so something wrong in the eo force code still
2015-08-07 08:37:15 +01:00
#include <qcd/action/pseudofermion/TwoFlavourEvenOdd.h>
Two flavour HMC for Wilson/Wilson is conserving energy. Still to check plaq and <e(-dH)>, but nevertheless this is progress
2015-07-29 09:53:39 +01:00
Large scale change to support 5d fermion formulations. Have 5d replicated wilson with 4d gauge working and matrix regressing to Ls copies of wilson.
2015-05-31 15:09:02 +01:00
#endif
Copy Permalink