portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-20 08:46:55 +01:00
Code Releases Activity

Accelerator mark up of entire tensore space for offload

Browse Source
This commit is contained in:
paboyle
2018-01-24 13:27:30 +00:00
parent 69327db9a9
commit 8e99264f40
18 changed files with 614 additions and 661 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
lib/tensors/Tensor_arith_mac.h
View File

@ -47,12 +47,12 @@ NAMESPACE_BEGIN(Grid);
// scal x vec = vec
///////////////////////////
template<class rtype,class vtype,class mtype>
strong_inline void mac(iScalar<rtype> * __restrict__ ret,const iScalar<vtype> * __restrict__ lhs,const iScalar<mtype> * __restrict__ rhs)
accelerator_inline void mac(iScalar<rtype> * __restrict__ ret,const iScalar<vtype> * __restrict__ lhs,const iScalar<mtype> * __restrict__ rhs)
{
mac(&ret->_internal,&lhs->_internal,&rhs->_internal);
}
template<class rrtype,class ltype,class rtype,int N>
strong_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
accelerator_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
for(int c3=0;c3<N;c3++){
for(int c1=0;c1<N;c1++){
for(int c2=0;c2<N;c2++){
@ -62,7 +62,7 @@ strong_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,
}
template<class rrtype,class ltype,class rtype,int N>
strong_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs){
accelerator_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs){
for(int c1=0;c1<N;c1++){
for(int c2=0;c2<N;c2++){
mac(&ret->_internal[c1][c2],&lhs->_internal[c1][c2],&rhs->_internal);
@ -70,7 +70,7 @@ strong_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iMatrix<ltype,
return;
}
template<class rrtype,class ltype,class rtype,int N>
strong_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iScalar<ltype> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
accelerator_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iScalar<ltype> * __restrict__ lhs,const iMatrix<rtype,N> * __restrict__ rhs){
for(int c1=0;c1<N;c1++){
for(int c2=0;c2<N;c2++){
mac(&ret->_internal[c1][c2],&lhs->_internal,&rhs->_internal[c1][c2]);
@ -78,7 +78,7 @@ strong_inline void mac(iMatrix<rrtype,N> * __restrict__ ret,const iScalar<ltype>
return;
}
template<class rrtype,class ltype,class rtype,int N>
strong_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iVector<rtype,N> * __restrict__ rhs)
accelerator_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iMatrix<ltype,N> * __restrict__ lhs,const iVector<rtype,N> * __restrict__ rhs)
{
for(int c1=0;c1<N;c1++){
for(int c2=0;c2<N;c2++){
@ -87,7 +87,7 @@ strong_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iMatrix<ltype,
return;
}
template<class rrtype,class ltype,class rtype,int N>
strong_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iScalar<ltype> * __restrict__ lhs,const iVector<rtype,N> * __restrict__ rhs)
accelerator_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iScalar<ltype> * __restrict__ lhs,const iVector<rtype,N> * __restrict__ rhs)
{
for(int c1=0;c1<N;c1++){
mac(&ret->_internal[c1],&lhs->_internal,&rhs->_internal[c1]);
@ -95,7 +95,7 @@ strong_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iScalar<ltype>
return;
}
template<class rrtype,class ltype,class rtype,int N>
strong_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iVector<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs)
accelerator_inline void mac(iVector<rrtype,N> * __restrict__ ret,const iVector<ltype,N> * __restrict__ lhs,const iScalar<rtype> * __restrict__ rhs)
{
for(int c1=0;c1<N;c1++){
mac(&ret->_internal[c1],&lhs->_internal[c1],&rhs->_internal);