portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-10 07:55:35 +00:00
Code Releases Activity

Project on group fix on GPU tracked to reciprocal sqrt collision between CUDA and Grid rsqrt

Browse Source
This commit is contained in:
Peter Boyle 2020-10-31 18:12:47 -04:00
parent d10422ded8
commit cc9c993f74
4 changed files with 6 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
Grid/lattice/Lattice_ET.h
View File

@ -350,7 +350,6 @@ GridUnopClass(UnaryTimesI, timesI(a));
GridUnopClass(UnaryTimesMinusI, timesMinusI(a));
GridUnopClass(UnaryAbs, abs(a));
GridUnopClass(UnarySqrt, sqrt(a));
GridUnopClass(UnaryRsqrt, rsqrt(a));
GridUnopClass(UnarySin, sin(a));
GridUnopClass(UnaryCos, cos(a));
GridUnopClass(UnaryAsin, asin(a));
@ -463,7 +462,6 @@ GRID_DEF_UNOP(timesMinusI, UnaryTimesMinusI);
GRID_DEF_UNOP(abs, UnaryAbs); // abs overloaded in cmath C++98; DON'T do the
// abs-fabs-dabs-labs thing
GRID_DEF_UNOP(sqrt, UnarySqrt);
GRID_DEF_UNOP(rsqrt, UnaryRsqrt);
GRID_DEF_UNOP(sin, UnarySin);
GRID_DEF_UNOP(cos, UnaryCos);
GRID_DEF_UNOP(asin, UnaryAsin);

8
Grid/simd/Grid_vector_unops.h
View File

@ -125,14 +125,6 @@ accelerator_inline Grid_simd<S, V> sqrt(const Grid_simd<S, V> &r) {
return SimdApply(SqrtRealFunctor<S>(), r);
}
template <class S, class V>
accelerator_inline Grid_simd<S, V> rsqrt(const Grid_simd<S, V> &r) {
return SimdApply(RSqrtRealFunctor<S>(), r);
}
template <class Scalar>
accelerator_inline Scalar rsqrt(const Scalar &r) {
return (RSqrtRealFunctor<Scalar>(), r);
}
template <class S, class V>
accelerator_inline Grid_simd<S, V> cos(const Grid_simd<S, V> &r) {
return SimdApply(CosRealFunctor<S>(), r);
}

7
Grid/tensors/Tensor_Ta.h
View File

@ -92,17 +92,22 @@ accelerator_inline iMatrix<vtype,N> ProjectOnGroup(const iMatrix<vtype,N> &arg)
{
// need a check for the group type?
iMatrix<vtype,N> ret(arg);
vtype rnrm;
vtype nrm;
vtype inner;
for(int c1=0;c1<N;c1++){
// Normalises row c1
zeroit(inner);
for(int c2=0;c2<N;c2++)
inner += innerProduct(ret._internal[c1][c2],ret._internal[c1][c2]);
nrm = rsqrt(inner);
nrm = sqrt(inner);
nrm = 1.0/nrm;
for(int c2=0;c2<N;c2++)
ret._internal[c1][c2]*= nrm;
// Remove c1 from rows c1+1...N-1
for (int b=c1+1; b<N; ++b){
decltype(ret._internal[b][b]*ret._internal[b][b]) pr;
zeroit(pr);

1
Grid/tensors/Tensor_unary.h
View File

@ -84,7 +84,6 @@ NAMESPACE_BEGIN(Grid);
}
UNARY(sqrt);
UNARY(rsqrt);
UNARY(sin);
UNARY(cos);
UNARY(asin);