portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-09 21:50:45 +01:00
Code Releases Activity

GPU friendly

Browse Source
This commit is contained in:
Peter Boyle 2019-07-31 01:22:54 +01:00
parent 9dad7a0094
commit 9117f61109
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
Grid/lattice/Lattice_reduction.h
View File

@ -128,7 +128,6 @@ axpy_norm_fast(Lattice<vobj> &z,sobj a,const Lattice<vobj> &x,const Lattice<vobj
template<class sobj,class vobj> strong_inline RealD template<class sobj,class vobj> strong_inline RealD
axpby_norm_fast(Lattice<vobj> &z,sobj a,sobj b,const Lattice<vobj> &x,const Lattice<vobj> &y) axpby_norm_fast(Lattice<vobj> &z,sobj a,sobj b,const Lattice<vobj> &x,const Lattice<vobj> &y)
{ {
const int pad = 8;
z.Checkerboard() = x.Checkerboard(); z.Checkerboard() = x.Checkerboard();
conformable(z,x); conformable(z,x);
conformable(x,y); conformable(x,y);
@ -148,14 +147,15 @@ axpby_norm_fast(Lattice<vobj> &z,sobj a,sobj b,const Lattice<vobj> &x,const Latt
typedef decltype(innerProduct(x_v[0],y_v[0])) inner_t; typedef decltype(innerProduct(x_v[0],y_v[0])) inner_t;
Lattice<inner_t> inner_tmp(grid); Lattice<inner_t> inner_tmp(grid);
auto inner_tmp_v = inner_tmp.View();
accelerator_for( ss, sites, nsimd,{ accelerator_for( ss, sites, nsimd,{
auto tmp = a*x_v(ss)+b*y_v(ss); auto tmp = a*x_v(ss)+b*y_v(ss);
coalescedWrite(inner_tmp[ss],innerProduct(tmp,tmp)); coalescedWrite(inner_tmp_v[ss],innerProduct(tmp,tmp));
coalescedWrite(z_v[ss],tmp); coalescedWrite(z_v[ss],tmp);
}) })
nrm = TensorRemove(sum(inner_tmp)); nrm = real(TensorRemove(sum(inner_tmp)));
z.Grid()->GlobalSum(nrm); z.Grid()->GlobalSum(nrm);
return nrm; return nrm;