portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-18 15:57:05 +01:00
Code Releases Activity

Global edit with change to View usage. autoView() creates a wrapper object that closes the view when scope closes.

Browse Source
This commit is contained in:
Peter Boyle
2020-06-05 18:52:35 -04:00
parent f39c2a240b
commit 1a4c8c3387
78 changed files with 773 additions and 778 deletions
Download Patch File Download Diff File Expand all files Collapse all files

68
Grid/lattice/Lattice_arith.h
View File

@ -36,9 +36,9 @@ NAMESPACE_BEGIN(Grid);
template<class obj1,class obj2,class obj3> inline
void mult(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
ret.Checkerboard() = lhs.Checkerboard();
auto ret_v = ret.View(AcceleratorWrite);
auto lhs_v = lhs.View(AcceleratorRead);
auto rhs_v = rhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( lhs_v , lhs, AcceleratorRead);
autoView( rhs_v , rhs, AcceleratorRead);
conformable(ret,rhs);
conformable(lhs,rhs);
accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
@ -55,9 +55,9 @@ void mac(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
ret.Checkerboard() = lhs.Checkerboard();
conformable(ret,rhs);
conformable(lhs,rhs);
auto ret_v = ret.View(AcceleratorWrite);
auto lhs_v = lhs.View(AcceleratorRead);
auto rhs_v = rhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( lhs_v , lhs, AcceleratorRead);
autoView( rhs_v , rhs, AcceleratorRead);
accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto lhs_t=lhs_v(ss);
@ -72,9 +72,9 @@ void sub(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
ret.Checkerboard() = lhs.Checkerboard();
conformable(ret,rhs);
conformable(lhs,rhs);
auto ret_v = ret.View(AcceleratorWrite);
auto lhs_v = lhs.View(AcceleratorRead);
auto rhs_v = rhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( lhs_v , lhs, AcceleratorRead);
autoView( rhs_v , rhs, AcceleratorRead);
accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto lhs_t=lhs_v(ss);
@ -88,9 +88,9 @@ void add(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const Lattice<obj3> &rhs){
ret.Checkerboard() = lhs.Checkerboard();
conformable(ret,rhs);
conformable(lhs,rhs);
auto ret_v = ret.View(AcceleratorWrite);
auto lhs_v = lhs.View(AcceleratorRead);
auto rhs_v = rhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( lhs_v , lhs, AcceleratorRead);
autoView( rhs_v , rhs, AcceleratorRead);
accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto lhs_t=lhs_v(ss);
@ -107,8 +107,8 @@ template<class obj1,class obj2,class obj3> inline
void mult(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
ret.Checkerboard() = lhs.Checkerboard();
conformable(lhs,ret);
auto ret_v = ret.View(AcceleratorWrite);
auto lhs_v = lhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( lhs_v , lhs, AcceleratorRead);
accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
mult(&tmp,&lhs_v(ss),&rhs);
@ -120,8 +120,8 @@ template<class obj1,class obj2,class obj3> inline
void mac(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
ret.Checkerboard() = lhs.Checkerboard();
conformable(ret,lhs);
auto ret_v = ret.View(AcceleratorWrite);
auto lhs_v = lhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( lhs_v , lhs, AcceleratorRead);
accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto lhs_t=lhs_v(ss);
@ -134,8 +134,8 @@ template<class obj1,class obj2,class obj3> inline
void sub(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
ret.Checkerboard() = lhs.Checkerboard();
conformable(ret,lhs);
auto ret_v = ret.View(AcceleratorWrite);
auto lhs_v = lhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( lhs_v , lhs, AcceleratorRead);
accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto lhs_t=lhs_v(ss);
@ -147,8 +147,8 @@ template<class obj1,class obj2,class obj3> inline
void add(Lattice<obj1> &ret,const Lattice<obj2> &lhs,const obj3 &rhs){
ret.Checkerboard() = lhs.Checkerboard();
conformable(lhs,ret);
auto ret_v = ret.View(AcceleratorWrite);
auto lhs_v = lhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( lhs_v , lhs, AcceleratorRead);
accelerator_for(ss,lhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto lhs_t=lhs_v(ss);
@ -164,8 +164,8 @@ template<class obj1,class obj2,class obj3> inline
void mult(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
ret.Checkerboard() = rhs.Checkerboard();
conformable(ret,rhs);
auto ret_v = ret.View(AcceleratorWrite);
auto rhs_v = lhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( rhs_v , lhs, AcceleratorRead);
accelerator_for(ss,rhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto rhs_t=rhs_v(ss);
@ -178,8 +178,8 @@ template<class obj1,class obj2,class obj3> inline
void mac(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
ret.Checkerboard() = rhs.Checkerboard();
conformable(ret,rhs);
auto ret_v = ret.View(AcceleratorWrite);
auto rhs_v = lhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( rhs_v , lhs, AcceleratorRead);
accelerator_for(ss,rhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto rhs_t=rhs_v(ss);
@ -192,8 +192,8 @@ template<class obj1,class obj2,class obj3> inline
void sub(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
ret.Checkerboard() = rhs.Checkerboard();
conformable(ret,rhs);
auto ret_v = ret.View(AcceleratorWrite);
auto rhs_v = lhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( rhs_v , lhs, AcceleratorRead);
accelerator_for(ss,rhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto rhs_t=rhs_v(ss);
@ -205,8 +205,8 @@ template<class obj1,class obj2,class obj3> inline
void add(Lattice<obj1> &ret,const obj2 &lhs,const Lattice<obj3> &rhs){
ret.Checkerboard() = rhs.Checkerboard();
conformable(ret,rhs);
auto ret_v = ret.View(AcceleratorWrite);
auto rhs_v = lhs.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( rhs_v , lhs, AcceleratorRead);
accelerator_for(ss,rhs_v.size(),obj1::Nsimd(),{
decltype(coalescedRead(obj1())) tmp;
auto rhs_t=rhs_v(ss);
@ -220,9 +220,9 @@ void axpy(Lattice<vobj> &ret,sobj a,const Lattice<vobj> &x,const Lattice<vobj> &
ret.Checkerboard() = x.Checkerboard();
conformable(ret,x);
conformable(x,y);
auto ret_v = ret.View(AcceleratorWrite);
auto x_v = x.View(AcceleratorRead);
auto y_v = y.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( x_v , x, AcceleratorRead);
autoView( y_v , y, AcceleratorRead);
accelerator_for(ss,x_v.size(),vobj::Nsimd(),{
auto tmp = a*x_v(ss)+y_v(ss);
coalescedWrite(ret_v[ss],tmp);
@ -233,9 +233,9 @@ void axpby(Lattice<vobj> &ret,sobj a,sobj b,const Lattice<vobj> &x,const Lattice
ret.Checkerboard() = x.Checkerboard();
conformable(ret,x);
conformable(x,y);
auto ret_v = ret.View(AcceleratorWrite);
auto x_v = x.View(AcceleratorRead);
auto y_v = y.View(AcceleratorRead);
autoView( ret_v , ret, AcceleratorWrite);
autoView( x_v , x, AcceleratorRead);
autoView( y_v , y, AcceleratorRead);
accelerator_for(ss,x_v.size(),vobj::Nsimd(),{
auto tmp = a*x_v(ss)+b*y_v(ss);
coalescedWrite(ret_v[ss],tmp);