diff --git a/lib/lattice/Lattice_local.h b/lib/lattice/Lattice_local.h
index 97e118f0..95b0dd2c 100644
--- a/lib/lattice/Lattice_local.h
+++ b/lib/lattice/Lattice_local.h
@@ -43,9 +43,9 @@ template<class vobj>
 inline auto localNorm2 (const Lattice<vobj> &rhs)-> Lattice<typename vobj::tensor_reduced>
 {
   Lattice<typename vobj::tensor_reduced> ret(rhs._grid);
-  parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
+  accelerator_loop(ss,rhs,{
     ret._odata[ss]=innerProduct(rhs._odata[ss],rhs._odata[ss]);
-  }
+  });
   return ret;
 }
   
@@ -54,9 +54,9 @@ template<class vobj>
 inline auto localInnerProduct (const Lattice<vobj> &lhs,const Lattice<vobj> &rhs) -> Lattice<typename vobj::tensor_reduced>
 {
   Lattice<typename vobj::tensor_reduced> ret(rhs._grid);
-  parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
+  accelerator_loop(ss,rhs,{
     ret._odata[ss]=innerProduct(lhs._odata[ss],rhs._odata[ss]);
-  }
+  });
   return ret;
 }
   
@@ -66,9 +66,9 @@ template<class ll,class rr>
 inline auto outerProduct (const Lattice<ll> &lhs,const Lattice<rr> &rhs) -> Lattice<decltype(outerProduct(lhs._odata[0],rhs._odata[0]))>
 {
   Lattice<decltype(outerProduct(lhs._odata[0],rhs._odata[0]))> ret(rhs._grid);
-  parallel_for(int ss=0;ss<rhs._grid->oSites(); ss++){
+  accelerator_loop(ss,rhs,{
     ret._odata[ss]=outerProduct(lhs._odata[ss],rhs._odata[ss]);
-  }
+  });
   return ret;
 }
 NAMESPACE_END(Grid);