portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-11-04 05:54:32 +00:00
Code Releases Activity

Moving operator stuff into separate file so that we can switch on/off replacement with

Browse Source 
expression templates
This commit is contained in:
Peter Boyle
2015-05-10 15:23:49 +01:00
parent 8299bc39ea
commit 6bb17502f9
1 changed files with 0 additions and 113 deletions
Download Patch File Download Diff File Expand all files Collapse all files

113
lib/lattice/Grid_lattice_arith.h
View File

@@ -3,19 +3,6 @@
namespace Grid {
//////////////////////////////////////////////////////////////////////////////////////////////////////
// unary negation
//////////////////////////////////////////////////////////////////////////////////////////////////////
template<class vobj>
inline Lattice<vobj> operator -(const Lattice<vobj> &r)
{
Lattice<vobj> ret(r._grid);
#pragma omp parallel for
for(int ss=0;ss<r._grid->oSites();ss++){
vstream(ret._odata[ss], -r._odata[ss]);
}
return ret;
}
//////////////////////////////////////////////////////////////////////////////////////////////////////
// avoid copy back routines for mult, mac, sub, add
@@ -155,106 +142,6 @@ namespace Grid {
}
}
/////////////////////////////////////////////////////////////////////////////////////
// Lattice BinOp Lattice,
//NB mult performs conformable check. Do not reapply here for performance.
/////////////////////////////////////////////////////////////////////////////////////
template<class left,class right>
inline auto operator * (const Lattice<left> &lhs,const Lattice<right> &rhs)-> Lattice<decltype(lhs._odata[0]*rhs._odata[0])>
{
Lattice<decltype(lhs._odata[0]*rhs._odata[0])> ret(rhs._grid);
mult(ret,lhs,rhs);
return ret;
}
template<class left,class right>
inline auto operator + (const Lattice<left> &lhs,const Lattice<right> &rhs)-> Lattice<decltype(lhs._odata[0]+rhs._odata[0])>
{
Lattice<decltype(lhs._odata[0]+rhs._odata[0])> ret(rhs._grid);
add(ret,lhs,rhs);
return ret;
}
template<class left,class right>
inline auto operator - (const Lattice<left> &lhs,const Lattice<right> &rhs)-> Lattice<decltype(lhs._odata[0]-rhs._odata[0])>
{
Lattice<decltype(lhs._odata[0]-rhs._odata[0])> ret(rhs._grid);
sub(ret,lhs,rhs);
return ret;
}
// Scalar BinOp Lattice ;generate return type
template<class left,class right>
inline auto operator * (const left &lhs,const Lattice<right> &rhs) -> Lattice<decltype(lhs*rhs._odata[0])>
{
Lattice<decltype(lhs*rhs._odata[0])> ret(rhs._grid);
#pragma omp parallel for
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs*rhs._odata[0]) tmp=lhs*rhs._odata[ss];
vstream(ret._odata[ss],tmp);
// ret._odata[ss]=lhs*rhs._odata[ss];
}
return ret;
}
template<class left,class right>
inline auto operator + (const left &lhs,const Lattice<right> &rhs) -> Lattice<decltype(lhs+rhs._odata[0])>
{
Lattice<decltype(lhs+rhs._odata[0])> ret(rhs._grid);
#pragma omp parallel for
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs+rhs._odata[0]) tmp =lhs-rhs._odata[ss];
vstream(ret._odata[ss],tmp);
// ret._odata[ss]=lhs+rhs._odata[ss];
}
return ret;
}
template<class left,class right>
inline auto operator - (const left &lhs,const Lattice<right> &rhs) -> Lattice<decltype(lhs-rhs._odata[0])>
{
Lattice<decltype(lhs-rhs._odata[0])> ret(rhs._grid);
#pragma omp parallel for
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs-rhs._odata[0]) tmp=lhs-rhs._odata[ss];
vstream(ret._odata[ss],tmp);
// ret._odata[ss]=lhs-rhs._odata[ss];
}
return ret;
}
template<class left,class right>
inline auto operator * (const Lattice<left> &lhs,const right &rhs) -> Lattice<decltype(lhs._odata[0]*rhs)>
{
Lattice<decltype(lhs._odata[0]*rhs)> ret(lhs._grid);
#pragma omp parallel for
for(int ss=0;ss<lhs._grid->oSites(); ss++){
decltype(lhs._odata[0]*rhs) tmp =lhs._odata[ss]*rhs;
vstream(ret._odata[ss],tmp);
// ret._odata[ss]=lhs._odata[ss]*rhs;
}
return ret;
}
template<class left,class right>
inline auto operator + (const Lattice<left> &lhs,const right &rhs) -> Lattice<decltype(lhs._odata[0]+rhs)>
{
Lattice<decltype(lhs._odata[0]+rhs)> ret(lhs._grid);
#pragma omp parallel for
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs._odata[0]+rhs) tmp=lhs._odata[ss]+rhs;
vstream(ret._odata[ss],tmp);
// ret._odata[ss]=lhs._odata[ss]+rhs;
}
return ret;
}
template<class left,class right>
inline auto operator - (const Lattice<left> &lhs,const right &rhs) -> Lattice<decltype(lhs._odata[0]-rhs)>
{
Lattice<decltype(lhs._odata[0]-rhs)> ret(lhs._grid);
#pragma omp parallel for
for(int ss=0;ss<rhs._grid->oSites(); ss++){
decltype(lhs._odata[0]-rhs) tmp=lhs._odata[ss]-rhs;
vstream(ret._odata[ss],tmp);
// ret._odata[ss]=lhs._odata[ss]-rhs;
}
return ret;
}
template<class sobj,class vobj>
inline void axpy(Lattice<vobj> &ret,sobj a,const Lattice<vobj> &lhs,const Lattice<vobj> &rhs){
conformable(lhs,rhs);