portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-10 19:36:56 +01:00
Code Releases Activity

Merge branch 'develop' into feature/hmc_generalise

Browse Source
This commit is contained in:
Guido Cossu
2017-05-01 12:13:56 +01:00
parent 8c540333d5 99220f6531
commit 3344788fa1
69 changed files with 3971 additions and 3179 deletions
Download Patch File Download Diff File Expand all files Collapse all files

101
lib/tensors/Tensor_class.h
View File

@ -1,9 +1,6 @@
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./lib/tensors/Tensor_class.h
Copyright (C) 2015
Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
@ -13,16 +10,13 @@ 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
*************************************************************************************/
@ -56,18 +50,18 @@ class iScalar {
typedef vtype element;
typedef typename GridTypeMapper<vtype>::scalar_type scalar_type;
typedef typename GridTypeMapper<vtype>::vector_type vector_type;
typedef typename GridTypeMapper<vtype>::vector_typeD vector_typeD;
typedef typename GridTypeMapper<vtype>::tensor_reduced tensor_reduced_v;
typedef iScalar<tensor_reduced_v> tensor_reduced;
typedef typename GridTypeMapper<vtype>::scalar_object recurse_scalar_object;
typedef iScalar<tensor_reduced_v> tensor_reduced;
typedef iScalar<recurse_scalar_object> scalar_object;
// substitutes a real or complex version with same tensor structure
typedef iScalar<typename GridTypeMapper<vtype>::Complexified> Complexified;
typedef iScalar<typename GridTypeMapper<vtype>::Realified> Realified;
// get double precision version
typedef iScalar<typename GridTypeMapper<vtype>::DoublePrecision> DoublePrecision;
enum { TensorLevel = GridTypeMapper<vtype>::TensorLevel + 1 };
// Scalar no action
@ -80,29 +74,18 @@ class iScalar {
iScalar<vtype> & operator= (const iScalar<vtype> &copyme) = default;
iScalar<vtype> & operator= (iScalar<vtype> &&copyme) = default;
*/
iScalar(scalar_type s)
: _internal(s){}; // recurse down and hit the constructor for vector_type
// template<int N=0>
// iScalar(EnableIf<isSIMDvectorized<vector_type>, vector_type> s) : _internal(s){}; // recurse down and hit the constructor for vector_type
iScalar(scalar_type s) : _internal(s){}; // recurse down and hit the constructor for vector_type
iScalar(const Zero &z) { *this = zero; };
iScalar<vtype> &operator=(const Zero &hero) {
zeroit(*this);
return *this;
}
// managing the internal vector structure
strong_inline scalar_object getlane(int lane){
scalar_object ret;
ret._internal = _internal.getlane(lane);
return ret;
}
strong_inline void putlane(scalar_object &s, int lane){
_internal.putlane(s._internal,lane);
}
friend strong_inline void vstream(iScalar<vtype> &out,
const iScalar<vtype> &in) {
vstream(out._internal, in._internal);
@ -152,42 +135,38 @@ class iScalar {
strong_inline const vtype &operator()(void) const { return _internal; }
// Type casts meta programmed, must be pure scalar to match TensorRemove
template <class U = vtype, class V = scalar_type, IfComplex<V> = 0,
IfNotSimd<U> = 0>
template <class U = vtype, class V = scalar_type, IfComplex<V> = 0, IfNotSimd<U> = 0>
operator ComplexF() const {
return (TensorRemove(_internal));
};
template <class U = vtype, class V = scalar_type, IfComplex<V> = 0,
IfNotSimd<U> = 0>
template <class U = vtype, class V = scalar_type, IfComplex<V> = 0, IfNotSimd<U> = 0>
operator ComplexD() const {
return (TensorRemove(_internal));
};
// template<class U=vtype,class V=scalar_type,IfComplex<V> = 0,IfNotSimd<U> =
// 0> operator RealD () const { return(real(TensorRemove(_internal))); }
template <class U = vtype, class V = scalar_type, IfReal<V> = 0,
IfNotSimd<U> = 0>
template <class U = vtype, class V = scalar_type, IfReal<V> = 0,IfNotSimd<U> = 0>
operator RealD() const {
return TensorRemove(_internal);
}
template <class U = vtype, class V = scalar_type, IfInteger<V> = 0,
IfNotSimd<U> = 0>
template <class U = vtype, class V = scalar_type, IfInteger<V> = 0, IfNotSimd<U> = 0>
operator Integer() const {
return Integer(TensorRemove(_internal));
}
// convert from a something to a scalar via constructor of something arg
template <class T, typename std::enable_if<!isGridTensor<T>::value, T>::type
* = nullptr>
strong_inline iScalar<vtype> operator=(T arg) {
template <class T, typename std::enable_if<!isGridTensor<T>::value, T>::type * = nullptr>
strong_inline iScalar<vtype> operator=(T arg) {
_internal = arg;
return *this;
}
friend std::ostream &operator<<(std::ostream &stream,
const iScalar<vtype> &o) {
friend std::ostream &operator<<(std::ostream &stream,const iScalar<vtype> &o) {
stream << "S {" << o._internal << "}";
return stream;
};
};
///////////////////////////////////////////////////////////
// Allows to turn scalar<scalar<scalar<double>>>> back to double.
@ -211,6 +190,7 @@ class iVector {
typedef vtype element;
typedef typename GridTypeMapper<vtype>::scalar_type scalar_type;
typedef typename GridTypeMapper<vtype>::vector_type vector_type;
typedef typename GridTypeMapper<vtype>::vector_typeD vector_typeD;
typedef typename GridTypeMapper<vtype>::tensor_reduced tensor_reduced_v;
typedef typename GridTypeMapper<vtype>::scalar_object recurse_scalar_object;
typedef iScalar<tensor_reduced_v> tensor_reduced;
@ -222,8 +202,7 @@ class iVector {
// get double precision version
typedef iVector<typename GridTypeMapper<vtype>::DoublePrecision, N> DoublePrecision;
template <class T, typename std::enable_if<!isGridTensor<T>::value, T>::type
* = nullptr>
strong_inline auto operator=(T arg) -> iVector<vtype, N> {
@ -246,20 +225,6 @@ class iVector {
zeroit(*this);
return *this;
}
strong_inline scalar_object getlane(int lane){
scalar_object ret;
for (int i = 0; i < N; i++) ret._internal[i] = _internal[i].getlane(lane);
return ret;
}
strong_inline void putlane(scalar_object &s, int lane){
for (int i = 0; i < N; i++) _internal[i].putlane(s._internal[i],lane);
}
friend strong_inline void zeroit(iVector<vtype, N> &that) {
for (int i = 0; i < N; i++) {
zeroit(that._internal[i]);
@ -341,6 +306,7 @@ class iMatrix {
typedef vtype element;
typedef typename GridTypeMapper<vtype>::scalar_type scalar_type;
typedef typename GridTypeMapper<vtype>::vector_type vector_type;
typedef typename GridTypeMapper<vtype>::vector_typeD vector_typeD;
typedef typename GridTypeMapper<vtype>::tensor_reduced tensor_reduced_v;
typedef typename GridTypeMapper<vtype>::scalar_object recurse_scalar_object;
@ -350,8 +316,7 @@ class iMatrix {
// get double precision version
typedef iMatrix<typename GridTypeMapper<vtype>::DoublePrecision, N> DoublePrecision;
// Tensor removal
typedef iScalar<tensor_reduced_v> tensor_reduced;
typedef iMatrix<recurse_scalar_object, N> scalar_object;
@ -390,25 +355,6 @@ class iMatrix {
return *this;
}
strong_inline scalar_object getlane(int lane){
scalar_object ret;
for (int i = 0; i < N; i++) {
for (int j = 0; j < N; j++) {
ret._internal[i][j] = _internal[i][j].getlane(lane);
}
}
return ret;
}
strong_inline void putlane(scalar_object &s, int lane){
for (int i = 0; i < N; i++)
for (int j = 0; j < N; j++) _internal[i][j].putlane(s._internal[i][j],lane);
}
friend strong_inline void zeroit(iMatrix<vtype,N> &that){
for(int i=0;i<N;i++){
for(int j=0;j<N;j++){
@ -527,3 +473,6 @@ void vprefetch(const iMatrix<v, N> &vv) {
}
}
#endif

142
lib/tensors/Tensor_inner.h
View File

@ -29,51 +29,109 @@ Author: Peter Boyle <paboyle@ph.ed.ac.uk>
#ifndef GRID_MATH_INNER_H
#define GRID_MATH_INNER_H
namespace Grid {
///////////////////////////////////////////////////////////////////////////////////////
// innerProduct Scalar x Scalar -> Scalar
// innerProduct Vector x Vector -> Scalar
// innerProduct Matrix x Matrix -> Scalar
///////////////////////////////////////////////////////////////////////////////////////
template<class sobj> inline RealD norm2(const sobj &arg){
typedef typename sobj::scalar_type scalar;
decltype(innerProduct(arg,arg)) nrm;
nrm = innerProduct(arg,arg);
RealD ret = real(nrm);
return ret;
}
///////////////////////////////////////////////////////////////////////////////////////
// innerProduct Scalar x Scalar -> Scalar
// innerProduct Vector x Vector -> Scalar
// innerProduct Matrix x Matrix -> Scalar
///////////////////////////////////////////////////////////////////////////////////////
template<class sobj> inline RealD norm2(const sobj &arg){
auto nrm = innerProductD(arg,arg);
RealD ret = real(nrm);
return ret;
}
//////////////////////////////////////
// If single promote to double and sum 2x
//////////////////////////////////////
template<class l,class r,int N> inline
auto innerProduct (const iVector<l,N>& lhs,const iVector<r,N>& rhs) -> iScalar<decltype(innerProduct(lhs._internal[0],rhs._internal[0]))>
{
typedef decltype(innerProduct(lhs._internal[0],rhs._internal[0])) ret_t;
iScalar<ret_t> ret;
ret=zero;
for(int c1=0;c1<N;c1++){
ret._internal += innerProduct(lhs._internal[c1],rhs._internal[c1]);
}
return ret;
inline ComplexD innerProductD(const ComplexF &l,const ComplexF &r){ return innerProduct(l,r); }
inline ComplexD innerProductD(const ComplexD &l,const ComplexD &r){ return innerProduct(l,r); }
inline RealD innerProductD(const RealD &l,const RealD &r){ return innerProduct(l,r); }
inline RealD innerProductD(const RealF &l,const RealF &r){ return innerProduct(l,r); }
inline vComplexD innerProductD(const vComplexD &l,const vComplexD &r){ return innerProduct(l,r); }
inline vRealD innerProductD(const vRealD &l,const vRealD &r){ return innerProduct(l,r); }
inline vComplexD innerProductD(const vComplexF &l,const vComplexF &r){
vComplexD la,lb;
vComplexD ra,rb;
Optimization::PrecisionChange::StoD(l.v,la.v,lb.v);
Optimization::PrecisionChange::StoD(r.v,ra.v,rb.v);
return innerProduct(la,ra) + innerProduct(lb,rb);
}
inline vRealD innerProductD(const vRealF &l,const vRealF &r){
vRealD la,lb;
vRealD ra,rb;
Optimization::PrecisionChange::StoD(l.v,la.v,lb.v);
Optimization::PrecisionChange::StoD(r.v,ra.v,rb.v);
return innerProduct(la,ra) + innerProduct(lb,rb);
}
template<class l,class r,int N> inline
auto innerProductD (const iVector<l,N>& lhs,const iVector<r,N>& rhs) -> iScalar<decltype(innerProductD(lhs._internal[0],rhs._internal[0]))>
{
typedef decltype(innerProductD(lhs._internal[0],rhs._internal[0])) ret_t;
iScalar<ret_t> ret;
ret=zero;
for(int c1=0;c1<N;c1++){
ret._internal += innerProductD(lhs._internal[c1],rhs._internal[c1]);
}
template<class l,class r,int N> inline
auto innerProduct (const iMatrix<l,N>& lhs,const iMatrix<r,N>& rhs) -> iScalar<decltype(innerProduct(lhs._internal[0][0],rhs._internal[0][0]))>
{
typedef decltype(innerProduct(lhs._internal[0][0],rhs._internal[0][0])) ret_t;
iScalar<ret_t> ret;
iScalar<ret_t> tmp;
ret=zero;
for(int c1=0;c1<N;c1++){
for(int c2=0;c2<N;c2++){
ret._internal+=innerProduct(lhs._internal[c1][c2],rhs._internal[c1][c2]);
}}
return ret;
}
template<class l,class r> inline
auto innerProduct (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(innerProduct(lhs._internal,rhs._internal))>
{
typedef decltype(innerProduct(lhs._internal,rhs._internal)) ret_t;
iScalar<ret_t> ret;
ret._internal = innerProduct(lhs._internal,rhs._internal);
return ret;
return ret;
}
template<class l,class r,int N> inline
auto innerProductD (const iMatrix<l,N>& lhs,const iMatrix<r,N>& rhs) -> iScalar<decltype(innerProductD(lhs._internal[0][0],rhs._internal[0][0]))>
{
typedef decltype(innerProductD(lhs._internal[0][0],rhs._internal[0][0])) ret_t;
iScalar<ret_t> ret;
iScalar<ret_t> tmp;
ret=zero;
for(int c1=0;c1<N;c1++){
for(int c2=0;c2<N;c2++){
ret._internal+=innerProductD(lhs._internal[c1][c2],rhs._internal[c1][c2]);
}}
return ret;
}
template<class l,class r> inline
auto innerProductD (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(innerProductD(lhs._internal,rhs._internal))>
{
typedef decltype(innerProductD(lhs._internal,rhs._internal)) ret_t;
iScalar<ret_t> ret;
ret._internal = innerProductD(lhs._internal,rhs._internal);
return ret;
}
//////////////////////
// Keep same precison
//////////////////////
template<class l,class r,int N> inline
auto innerProduct (const iVector<l,N>& lhs,const iVector<r,N>& rhs) -> iScalar<decltype(innerProduct(lhs._internal[0],rhs._internal[0]))>
{
typedef decltype(innerProduct(lhs._internal[0],rhs._internal[0])) ret_t;
iScalar<ret_t> ret;
ret=zero;
for(int c1=0;c1<N;c1++){
ret._internal += innerProduct(lhs._internal[c1],rhs._internal[c1]);
}
return ret;
}
template<class l,class r,int N> inline
auto innerProduct (const iMatrix<l,N>& lhs,const iMatrix<r,N>& rhs) -> iScalar<decltype(innerProduct(lhs._internal[0][0],rhs._internal[0][0]))>
{
typedef decltype(innerProduct(lhs._internal[0][0],rhs._internal[0][0])) ret_t;
iScalar<ret_t> ret;
iScalar<ret_t> tmp;
ret=zero;
for(int c1=0;c1<N;c1++){
for(int c2=0;c2<N;c2++){
ret._internal+=innerProduct(lhs._internal[c1][c2],rhs._internal[c1][c2]);
}}
return ret;
}
template<class l,class r> inline
auto innerProduct (const iScalar<l>& lhs,const iScalar<r>& rhs) -> iScalar<decltype(innerProduct(lhs._internal,rhs._internal))>
{
typedef decltype(innerProduct(lhs._internal,rhs._internal)) ret_t;
iScalar<ret_t> ret;
ret._internal = innerProduct(lhs._internal,rhs._internal);
return ret;
}
}
#endif

40
lib/tensors/Tensor_traits.h
View File

@ -1,29 +1,21 @@
/*************************************************************************************
/*************************************************************************************
Grid physics library, www.github.com/paboyle/Grid
Source file: ./lib/tensors/Tensor_traits.h
Copyright (C) 2015
Author: Azusa Yamaguchi <ayamaguc@staffmail.ed.ac.uk>
Author: Peter Boyle <paboyle@ph.ed.ac.uk>
Author: Christopher Kelly <ckelly@phys.columbia.edu>
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 */
@ -53,6 +45,7 @@ namespace Grid {
public:
typedef typename T::scalar_type scalar_type;
typedef typename T::vector_type vector_type;
typedef typename T::vector_typeD vector_typeD;
typedef typename T::tensor_reduced tensor_reduced;
typedef typename T::scalar_object scalar_object;
typedef typename T::Complexified Complexified;
@ -68,6 +61,7 @@ namespace Grid {
public:
typedef RealF scalar_type;
typedef RealF vector_type;
typedef RealD vector_typeD;
typedef RealF tensor_reduced ;
typedef RealF scalar_object;
typedef ComplexF Complexified;
@ -79,6 +73,7 @@ namespace Grid {
public:
typedef RealD scalar_type;
typedef RealD vector_type;
typedef RealD vector_typeD;
typedef RealD tensor_reduced;
typedef RealD scalar_object;
typedef ComplexD Complexified;
@ -90,6 +85,7 @@ namespace Grid {
public:
typedef ComplexF scalar_type;
typedef ComplexF vector_type;
typedef ComplexD vector_typeD;
typedef ComplexF tensor_reduced;
typedef ComplexF scalar_object;
typedef ComplexF Complexified;
@ -101,6 +97,7 @@ namespace Grid {
public:
typedef ComplexD scalar_type;
typedef ComplexD vector_type;
typedef ComplexD vector_typeD;
typedef ComplexD tensor_reduced;
typedef ComplexD scalar_object;
typedef ComplexD Complexified;
@ -112,6 +109,7 @@ namespace Grid {
public:
typedef Integer scalar_type;
typedef Integer vector_type;
typedef Integer vector_typeD;
typedef Integer tensor_reduced;
typedef Integer scalar_object;
typedef void Complexified;
@ -124,6 +122,7 @@ namespace Grid {
public:
typedef RealF scalar_type;
typedef vRealF vector_type;
typedef vRealD vector_typeD;
typedef vRealF tensor_reduced;
typedef RealF scalar_object;
typedef vComplexF Complexified;
@ -135,6 +134,7 @@ namespace Grid {
public:
typedef RealD scalar_type;
typedef vRealD vector_type;
typedef vRealD vector_typeD;
typedef vRealD tensor_reduced;
typedef RealD scalar_object;
typedef vComplexD Complexified;
@ -142,10 +142,23 @@ namespace Grid {
typedef vRealD DoublePrecision;
enum { TensorLevel = 0 };
};
template<> class GridTypeMapper<vComplexH> {
public:
typedef ComplexF scalar_type;
typedef vComplexH vector_type;
typedef vComplexD vector_typeD;
typedef vComplexH tensor_reduced;
typedef ComplexF scalar_object;
typedef vComplexH Complexified;
typedef vRealH Realified;
typedef vComplexD DoublePrecision;
enum { TensorLevel = 0 };
};
template<> class GridTypeMapper<vComplexF> {
public:
typedef ComplexF scalar_type;
typedef vComplexF vector_type;
typedef vComplexD vector_typeD;
typedef vComplexF tensor_reduced;
typedef ComplexF scalar_object;
typedef vComplexF Complexified;
@ -157,6 +170,7 @@ namespace Grid {
public:
typedef ComplexD scalar_type;
typedef vComplexD vector_type;
typedef vComplexD vector_typeD;
typedef vComplexD tensor_reduced;
typedef ComplexD scalar_object;
typedef vComplexD Complexified;
@ -168,6 +182,7 @@ namespace Grid {
public:
typedef Integer scalar_type;
typedef vInteger vector_type;
typedef vInteger vector_typeD;
typedef vInteger tensor_reduced;
typedef Integer scalar_object;
typedef void Complexified;
@ -252,7 +267,8 @@ namespace Grid {
template<typename T>
class isSIMDvectorized{
template<typename U>
static typename std::enable_if< !std::is_same< typename GridTypeMapper<typename getVectorType<U>::type>::scalar_type, typename GridTypeMapper<typename getVectorType<U>::type>::vector_type>::value, char>::type test(void *);
static typename std::enable_if< !std::is_same< typename GridTypeMapper<typename getVectorType<U>::type>::scalar_type,
typename GridTypeMapper<typename getVectorType<U>::type>::vector_type>::value, char>::type test(void *);
template<typename U>
static double test(...);
@ -264,13 +280,15 @@ namespace Grid {
//Get the precision of a Lattice, tensor or scalar type in units of sizeof(float)
template<typename T>
class getPrecision{
public:
typedef typename getVectorType<T>::type vector_obj; //get the vector_obj (i.e. a grid Tensor) if its a Lattice<vobj>, do nothing otherwise (i.e. if fundamental or grid Tensor)
typedef typename GridTypeMapper<vector_obj>::scalar_type scalar_type; //get the associated scalar type. Works on fundamental and tensor types
public:
typedef typename GridTypeMapper<scalar_type>::Realified real_scalar_type; //remove any std::complex wrapper, should get us to the fundamental type
enum { value = sizeof(real_scalar_type)/sizeof(float) };
};
}
#endif