portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-04 19:25:56 +01:00
Code Releases Activity

Compile fix, multishift mixed prec support

Browse Source
This commit is contained in:
Peter Boyle 2022-09-23 16:19:27 -04:00
parent 5b128a6f9f
commit e4c117aabf
2 changed files with 20 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
Grid/lattice/Lattice_reduction.h
View File

@ -92,9 +92,7 @@ inline typename vobj::scalar_objectD sumD_cpu(const vobj *arg, Integer osites)
ssum = ssum+sumarray[i];
}
typedef typename vobj::scalar_object ssobj;
ssobj ret = ssum;
return ret;
return ssum;
}
/*
Threaded max, don't use for now

19
Grid/lattice/Lattice_transfer.h
View File

@ -1080,6 +1080,25 @@ vectorizeFromRevLexOrdArray( std::vector<sobj> &in, Lattice<vobj> &out)
});
}
template<class VobjOut, class VobjIn>
void precisionChangeFast(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)
{
typedef typename VobjOut::scalar_object SobjOut;
typedef typename VobjIn::scalar_object SobjIn;
conformable(out.Grid(),in.Grid());
out.Checkerboard() = in.Checkerboard();
int nsimd = out.Grid()->Nsimd();
autoView( out_v , out, AcceleratorWrite);
autoView( in_v , in, AcceleratorRead);
accelerator_for(idx,out.Grid()->oSites(),nsimd,{
auto itmp = coalescedRead(in_v[idx]);
auto otmp = coalescedRead(out_v[idx]);
#ifdef GRID_SIMT
otmp=itmp;
#endif
coalescedWrite(out_v[idx],otmp);
});
}
//Convert a Lattice from one precision to another
template<class VobjOut, class VobjIn>
void precisionChange(Lattice<VobjOut> &out, const Lattice<VobjIn> &in)