mirror of
https://github.com/paboyle/Grid.git
synced 2024-11-10 07:55:35 +00:00
315 lines
10 KiB
C++
315 lines
10 KiB
C++
#ifndef GRID_CARTESIAN_H
|
|
#define GRID_CARTESIAN_H
|
|
|
|
#include <Grid.h>
|
|
#include <Grid_Communicator.h>
|
|
namespace dpo{
|
|
|
|
/////////////////////////////////////////////////////////////////////////////////////////
|
|
// Grid Support. Following will go into Grid.h.
|
|
/////////////////////////////////////////////////////////////////////////////////////////
|
|
// Cartesian grids
|
|
// dpo::Grid
|
|
// dpo::GridCartesian
|
|
// dpo::GridCartesianRedBlack
|
|
|
|
|
|
class SimdGrid : public CartesianCommunicator {
|
|
public:
|
|
|
|
|
|
SimdGrid(std::vector<int> & processor_grid) : CartesianCommunicator(processor_grid) {};
|
|
|
|
// Give Lattice access
|
|
template<class object> friend class Lattice;
|
|
|
|
//protected:
|
|
|
|
// Lattice wide random support. not yet fully implemented. Need seed strategy
|
|
// and one generator per site.
|
|
//std::default_random_engine generator;
|
|
// static std::mt19937 generator( 9 );
|
|
|
|
// Grid information.
|
|
|
|
// Commicator provides
|
|
// unsigned long _ndimension;
|
|
// std::vector<int> _processors; // processor grid
|
|
// int _processor; // linear processor rank
|
|
// std::vector<int> _processor_coor; // linear processor rank
|
|
|
|
std::vector<int> _simd_layout; // Which dimensions get relayed out over simd lanes.
|
|
|
|
|
|
std::vector<int> _fdimensions;// Global dimensions of array with cb removed
|
|
std::vector<int> _gdimensions;// Global dimensions of array
|
|
std::vector<int> _ldimensions;// local dimensions of array with processor images removed
|
|
std::vector<int> _rdimensions;// Reduced local dimensions with simd lane images and processor images removed
|
|
|
|
// std::vector<int> _lstart; // local start of array in gcoors. _processor_coor[d]*_ldimensions[d]
|
|
// std::vector<int> _lend; // local end of array in gcoors _processor_coor[d]*_ldimensions[d]+_ldimensions_[d]-1
|
|
|
|
|
|
std::vector<int> _ostride; // Outer stride for each dimension
|
|
std::vector<int> _istride; // Inner stride i.e. within simd lane
|
|
|
|
int _osites; // _isites*_osites = product(dimensions).
|
|
int _isites;
|
|
|
|
// subslice information
|
|
std::vector<int> _slice_block;
|
|
std::vector<int> _slice_stride;
|
|
std::vector<int> _slice_nblock;
|
|
public:
|
|
|
|
// These routines are key. Subdivide the linearised cartesian index into
|
|
// "inner" index identifying which simd lane of object<vFcomplex> is associated with coord
|
|
// "outer" index identifying which element of _odata in class "Lattice" is associated with coord.
|
|
// Compared to, say, Blitz++ we simply need to store BOTH an inner stride and an outer
|
|
// stride per dimension. The cost of evaluating the indexing information is doubled for an n-dimensional
|
|
// coordinate. Note, however, for data parallel operations the "inner" indexing cost is not paid and all
|
|
// lanes are operated upon simultaneously.
|
|
|
|
inline int oIndexReduced(std::vector<int> &rcoor)
|
|
{
|
|
int idx=0;
|
|
for(int d=0;d<_ndimension;d++) idx+=_ostride[d]*rcoor[d];
|
|
return idx;
|
|
}
|
|
virtual int oIndex(std::vector<int> &coor)
|
|
{
|
|
int idx=0;
|
|
for(int d=0;d<_ndimension;d++) idx+=_ostride[d]*(coor[d]%_rdimensions[d]);
|
|
return idx;
|
|
}
|
|
inline int iIndex(std::vector<int> &rcoor)
|
|
{
|
|
int idx=0;
|
|
for(int d=0;d<_ndimension;d++) idx+=_istride[d]*(rcoor[d]/_rdimensions[d]);
|
|
return idx;
|
|
}
|
|
|
|
inline int oSites(void) { return _osites; };
|
|
inline int iSites(void) { return _isites; };
|
|
|
|
inline int CheckerboardFromOsite (int Osite){
|
|
std::vector<int> ocoor;
|
|
CoordFromOsite(ocoor,Osite);
|
|
int ss=0;
|
|
for(int d=0;d<_ndimension;d++){
|
|
ss=ss+ocoor[d];
|
|
}
|
|
return ss&0x1;
|
|
}
|
|
inline void CoordFromOsite (std::vector<int>& coor,int Osite){
|
|
coor.resize(_ndimension);
|
|
for(int d=0;d<_ndimension;d++){
|
|
coor[d] = Osite % _rdimensions[d];
|
|
Osite = Osite / _rdimensions[d];
|
|
}
|
|
}
|
|
|
|
virtual int CheckerBoard(std::vector<int> site)=0;
|
|
virtual int CheckerBoardDestination(int source_cb,int shift)=0;
|
|
virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite)=0;
|
|
};
|
|
|
|
class GridCartesian: public SimdGrid {
|
|
public:
|
|
virtual int CheckerBoard(std::vector<int> site){
|
|
return 0;
|
|
}
|
|
virtual int CheckerBoardDestination(int cb,int shift){
|
|
return 0;
|
|
}
|
|
virtual int CheckerBoardShift(int source_cb,int dim,int shift, int osite){
|
|
return shift;
|
|
}
|
|
GridCartesian(std::vector<int> &dimensions,
|
|
std::vector<int> &simd_layout,
|
|
std::vector<int> &processor_grid
|
|
) : SimdGrid(processor_grid)
|
|
{
|
|
///////////////////////
|
|
// Grid information
|
|
///////////////////////
|
|
_ndimension = dimensions.size();
|
|
|
|
_fdimensions.resize(_ndimension);
|
|
_gdimensions.resize(_ndimension);
|
|
_ldimensions.resize(_ndimension);
|
|
_rdimensions.resize(_ndimension);
|
|
_simd_layout.resize(_ndimension);
|
|
|
|
_ostride.resize(_ndimension);
|
|
_istride.resize(_ndimension);
|
|
|
|
_osites = 1;
|
|
_isites = 1;
|
|
for(int d=0;d<_ndimension;d++){
|
|
_fdimensions[d] = dimensions[d]; // Global dimensions
|
|
_gdimensions[d] = _fdimensions[d]; // Global dimensions
|
|
_simd_layout[d] = simd_layout[d];
|
|
|
|
//FIXME check for exact division
|
|
|
|
// Use a reduced simd grid
|
|
_ldimensions[d]= _gdimensions[d]/_processors[d]; //local dimensions
|
|
_rdimensions[d]= _ldimensions[d]/_simd_layout[d]; //overdecomposition
|
|
_osites *= _rdimensions[d];
|
|
_isites *= _simd_layout[d];
|
|
|
|
// Addressing support
|
|
if ( d==0 ) {
|
|
_ostride[d] = 1;
|
|
_istride[d] = 1;
|
|
} else {
|
|
_ostride[d] = _ostride[d-1]*_rdimensions[d-1];
|
|
_istride[d] = _istride[d-1]*_simd_layout[d-1];
|
|
}
|
|
}
|
|
|
|
///////////////////////
|
|
// subplane information
|
|
///////////////////////
|
|
_slice_block.resize(_ndimension);
|
|
_slice_stride.resize(_ndimension);
|
|
_slice_nblock.resize(_ndimension);
|
|
|
|
int block =1;
|
|
int nblock=1;
|
|
for(int d=0;d<_ndimension;d++) nblock*=_rdimensions[d];
|
|
|
|
for(int d=0;d<_ndimension;d++){
|
|
nblock/=_rdimensions[d];
|
|
_slice_block[d] =block;
|
|
_slice_stride[d]=_ostride[d]*_rdimensions[d];
|
|
_slice_nblock[d]=nblock;
|
|
block = block*_rdimensions[d];
|
|
}
|
|
|
|
if ( _isites != vComplex::Nsimd()) {
|
|
printf("bad layout for grid isites %d Nsimd %d\n",_isites,vComplex::Nsimd());
|
|
exit(0);
|
|
}
|
|
};
|
|
};
|
|
|
|
// Specialise this for red black grids storing half the data like a chess board.
|
|
class GridRedBlackCartesian : public SimdGrid
|
|
{
|
|
public:
|
|
virtual int CheckerBoard(std::vector<int> site){
|
|
return (site[0]+site[1]+site[2]+site[3])&0x1;
|
|
}
|
|
|
|
// Depending on the cb of site, we toggle source cb.
|
|
// for block #b, element #e = (b, e)
|
|
// we need
|
|
virtual int CheckerBoardShift(int source_cb,int dim,int shift,int osite){
|
|
|
|
if(dim != 0) return shift;
|
|
|
|
int fulldim =_fdimensions[0];
|
|
shift = (shift+fulldim)%fulldim;
|
|
|
|
// Probably faster with table lookup;
|
|
// or by looping over x,y,z and multiply rather than computing checkerboard.
|
|
int ocb=CheckerboardFromOsite(osite);
|
|
|
|
if ( (source_cb+ocb)&1 ) {
|
|
return (shift)/2;
|
|
} else {
|
|
return (shift+1)/2;
|
|
}
|
|
}
|
|
|
|
virtual int CheckerBoardDestination(int source_cb,int shift){
|
|
if ((shift+_fdimensions[0])&0x1) {
|
|
return 1-source_cb;
|
|
} else {
|
|
return source_cb;
|
|
}
|
|
};
|
|
GridRedBlackCartesian(std::vector<int> &dimensions,
|
|
std::vector<int> &simd_layout,
|
|
std::vector<int> &processor_grid) : SimdGrid(processor_grid)
|
|
{
|
|
///////////////////////
|
|
// Grid information
|
|
///////////////////////
|
|
_ndimension = dimensions.size();
|
|
|
|
_fdimensions.resize(_ndimension);
|
|
_gdimensions.resize(_ndimension);
|
|
_ldimensions.resize(_ndimension);
|
|
_rdimensions.resize(_ndimension);
|
|
_simd_layout.resize(_ndimension);
|
|
|
|
_ostride.resize(_ndimension);
|
|
_istride.resize(_ndimension);
|
|
|
|
_osites = 1;
|
|
_isites = 1;
|
|
for(int d=0;d<_ndimension;d++){
|
|
_fdimensions[d] = dimensions[d];
|
|
_gdimensions[d] = _fdimensions[d];
|
|
if (d==0) _gdimensions[0] = _gdimensions[0]/2; // Remove a checkerboard
|
|
_ldimensions[d] = _gdimensions[d]/_processors[d];
|
|
|
|
// Use a reduced simd grid
|
|
_simd_layout[d] = simd_layout[d];
|
|
_rdimensions[d]= _ldimensions[d]/_simd_layout[d];
|
|
|
|
_osites *= _rdimensions[d];
|
|
_isites *= _simd_layout[d];
|
|
|
|
// Addressing support
|
|
if ( d==0 ) {
|
|
_ostride[d] = 1;
|
|
_istride[d] = 1;
|
|
} else {
|
|
_ostride[d] = _ostride[d-1]*_rdimensions[d-1];
|
|
_istride[d] = _istride[d-1]*_simd_layout[d-1];
|
|
}
|
|
}
|
|
|
|
////////////////////////////////////////////////////////////////////////////////////////////
|
|
// subplane information
|
|
// It may be worth the investment of generating a more general subplane "iterator",
|
|
// and providing support for threads grabbing a unit of allocation.
|
|
////////////////////////////////////////////////////////////////////////////////////////////
|
|
_slice_block.resize(_ndimension);
|
|
_slice_stride.resize(_ndimension);
|
|
_slice_nblock.resize(_ndimension);
|
|
|
|
int block =1;
|
|
int nblock=1;
|
|
for(int d=0;d<_ndimension;d++) nblock*=_rdimensions[d];
|
|
|
|
for(int d=0;d<_ndimension;d++){
|
|
nblock/=_rdimensions[d];
|
|
_slice_block[d] =block;
|
|
_slice_stride[d]=_ostride[d]*_rdimensions[d];
|
|
_slice_nblock[d]=nblock;
|
|
block = block*_rdimensions[d];
|
|
}
|
|
|
|
if ( _isites != vComplex::Nsimd()) {
|
|
printf("bad layout for grid isites %d Nsimd %d\n",_isites,vComplex::Nsimd());
|
|
exit(0);
|
|
}
|
|
};
|
|
protected:
|
|
virtual int oIndex(std::vector<int> &coor)
|
|
{
|
|
int idx=_ostride[0]*((coor[0]/2)%_rdimensions[0]);
|
|
for(int d=1;d<_ndimension;d++) idx+=_ostride[d]*(coor[d]%_rdimensions[d]);
|
|
return idx;
|
|
};
|
|
|
|
};
|
|
|
|
}
|
|
#endif
|