From f574c2011851a178af597fcf36f528515ca433e4 Mon Sep 17 00:00:00 2001
From: paboyle <paboyle@ph.ed.ac.uk>
Date: Sat, 27 Jan 2018 23:50:17 +0000
Subject: [PATCH] Zero changes, __VA_ARGS__ and swap

---
 lib/lattice/Lattice_base.h      | 11 +++++++++++
 lib/lattice/Lattice_reduction.h | 18 +++++++++---------
 lib/lattice/Lattice_transfer.h  |  8 ++++----
 lib/lattice/Lattice_where.h     |  4 ++--
 4 files changed, 26 insertions(+), 15 deletions(-)

diff --git a/lib/lattice/Lattice_base.h b/lib/lattice/Lattice_base.h
index 89e0bbcb..26eabde4 100644
--- a/lib/lattice/Lattice_base.h
+++ b/lib/lattice/Lattice_base.h
@@ -76,6 +76,7 @@ public:
   accelerator_inline uint64_t size(void)   const { return _odata_size; };
   accelerator_inline vobj & operator[](size_t i)             { return _odata[i]; };
   accelerator_inline const vobj & operator[](size_t i) const { return _odata[i]; };
+
 };
     
 class LatticeExpressionBase {};
@@ -398,6 +399,16 @@ public:
     *this = (*this)+r;
     return *this;
   }
+
+  friend inline void swap(Lattice &l, Lattice &r) { 
+    conformable(l,r);
+    LatticeAccelerator<vobj> tmp;
+    LatticeAccelerator<vobj> *lp = (LatticeAccelerator<vobj> *)&l;
+    LatticeAccelerator<vobj> *rp = (LatticeAccelerator<vobj> *)&r;
+    tmp = *lp;    *lp=*rp;    *rp=tmp;
+  }
+
+
 }; // class Lattice
 
 template<class vobj> std::ostream& operator<< (std::ostream& stream, const Lattice<vobj> &o){
diff --git a/lib/lattice/Lattice_reduction.h b/lib/lattice/Lattice_reduction.h
index 86d27aa6..b39f43fb 100644
--- a/lib/lattice/Lattice_reduction.h
+++ b/lib/lattice/Lattice_reduction.h
@@ -52,14 +52,14 @@ inline ComplexD innerProduct(const Lattice<vobj> &left,const Lattice<vobj> &righ
     int mywork, myoff;
     GridThread::GetWork(left.Grid()->oSites(),thr,mywork,myoff);
     
-    decltype(innerProductD(left[0],right[0])) vnrm=zero; // private to thread; sub summation
+    decltype(innerProductD(left[0],right[0])) vnrm=Zero(); // private to thread; sub summation
     for(int ss=myoff;ss<mywork+myoff; ss++){
       vnrm = vnrm + innerProductD(left[ss],right[ss]);
     }
     sumarray[thr]=TensorRemove(vnrm) ;
   }
   
-  vector_type vvnrm; vvnrm=zero;  // sum across threads
+  vector_type vvnrm; vvnrm=Zero();  // sum across threads
   for(int i=0;i<grid->SumArraySize();i++){
     vvnrm = vvnrm+sumarray[i];
   } 
@@ -101,21 +101,21 @@ inline typename vobj::scalar_object sum(const Lattice<vobj> &arg)
   
   std::vector<vobj,alignedAllocator<vobj> > sumarray(grid->SumArraySize());
   for(int i=0;i<grid->SumArraySize();i++){
-    sumarray[i]=zero;
+    sumarray[i]=Zero();
   }
   
   parallel_for(int thr=0;thr<grid->SumArraySize();thr++){
     int mywork, myoff;
     GridThread::GetWork(grid->oSites(),thr,mywork,myoff);
     
-    vobj vvsum=zero;
+    vobj vvsum=Zero();
     for(int ss=myoff;ss<mywork+myoff; ss++){
       vvsum = vvsum + arg[ss];
     }
     sumarray[thr]=vvsum;
   }
   
-  vobj vsum=zero;  // sum across threads
+  vobj vsum=Zero();  // sum across threads
   for(int i=0;i<grid->SumArraySize();i++){
     vsum = vsum+sumarray[i];
   } 
@@ -159,12 +159,12 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<
   int rd=grid->_rdimensions[orthogdim];
 
   std::vector<vobj,alignedAllocator<vobj> > lvSum(rd); // will locally sum vectors first
-  std::vector<sobj> lsSum(ld,zero);                    // sum across these down to scalars
+  std::vector<sobj> lsSum(ld,Zero());                    // sum across these down to scalars
   std::vector<sobj> extracted(Nsimd);                  // splitting the SIMD
 
   result.resize(fd); // And then global sum to return the same vector to every node 
   for(int r=0;r<rd;r++){
-    lvSum[r]=zero;
+    lvSum[r]=Zero();
   }
 
   int e1=    grid->_slice_nblock[orthogdim];
@@ -211,7 +211,7 @@ template<class vobj> inline void sliceSum(const Lattice<vobj> &Data,std::vector<
     if ( pt == grid->_processor_coor[orthogdim] ) {
       gsum=lsSum[lt];
     } else {
-      gsum=zero;
+      gsum=Zero();
     }
 
     grid->GlobalSum(gsum);
@@ -245,7 +245,7 @@ static void sliceInnerProductVector( std::vector<ComplexD> & result, const Latti
 
   result.resize(fd); // And then global sum to return the same vector to every node for IO to file
   for(int r=0;r<rd;r++){
-    lvSum[r]=zero;
+    lvSum[r]=Zero();
   }
 
   int e1=    grid->_slice_nblock[orthogdim];
diff --git a/lib/lattice/Lattice_transfer.h b/lib/lattice/Lattice_transfer.h
index ccb5e7e9..1ee91a52 100644
--- a/lib/lattice/Lattice_transfer.h
+++ b/lib/lattice/Lattice_transfer.h
@@ -103,7 +103,7 @@ inline void blockProject(Lattice<iVector<CComplex,nbasis > > &coarseData,
     assert(block_r[d]*coarse->_rdimensions[d] == fine->_rdimensions[d]);
   }
 
-  coarseData=zero;
+  coarseData=Zero();
 
   // Loop over coars parallel, and then loop over fine associated with coarse.
   thread_loop( (int sf=0;sf<fine->oSites();sf++),{
@@ -192,7 +192,7 @@ template<class vobj,class CComplex>
 inline void blockNormalise(Lattice<CComplex> &ip,Lattice<vobj> &fineX)
 {
   GridBase *coarse = ip.Grid();
-  Lattice<vobj> zz(fineX.Grid()); zz=zero; zz.Checkerboard()=fineX.Checkerboard();
+  Lattice<vobj> zz(fineX.Grid()); zz=Zero(); zz.Checkerboard()=fineX.Checkerboard();
   blockInnerProduct(ip,fineX,fineX);
   ip = pow(ip,-0.5);
   blockZAXPY(fineX,ip,fineX,zz);
@@ -217,7 +217,7 @@ inline void blockSum(Lattice<vobj> &coarseData,const Lattice<vobj> &fineData)
 
   // Turn this around to loop threaded over sc and interior loop 
   // over sf would thread better
-  coarseData=zero;
+  coarseData=Zero();
   thread_region {
 
     int sc;
@@ -247,7 +247,7 @@ inline void blockPick(GridBase *coarse,const Lattice<vobj> &unpicked,Lattice<vob
   Lattice<vobj> zz(fine); zz.Checkerboard() = unpicked.Checkerboard();
   Lattice<iScalar<vInteger> > fcoor(fine);
 
-  zz = zero;
+  zz = Zero();
 
   picked = unpicked;
   for(int d=0;d<fine->_ndimension;d++){
diff --git a/lib/lattice/Lattice_where.h b/lib/lattice/Lattice_where.h
index ad524cc6..3aa9098c 100644
--- a/lib/lattice/Lattice_where.h
+++ b/lib/lattice/Lattice_where.h
@@ -58,7 +58,7 @@ inline void whereWolf(Lattice<vobj> &ret,const Lattice<iobj> &predicate,Lattice<
   std::vector<scalar_object> truevals (Nsimd);
   std::vector<scalar_object> falsevals(Nsimd);
 
-  thread_loop( (int ss=iftrue.begin(); ss<iftrue.end();ss++) , COMMA_SAFE({
+  thread_loop( (int ss=iftrue.begin(); ss<iftrue.end();ss++) , {
 
     extract(iftrue[ss]   ,truevals);
     extract(iffalse[ss]  ,falsevals);
@@ -69,7 +69,7 @@ inline void whereWolf(Lattice<vobj> &ret,const Lattice<iobj> &predicate,Lattice<
     }
 
     merge(ret[ss],falsevals);
-  }) 
+  } 
   );
 }