Threading support rework.

Placed parallel pragmas as macros; implemented deterministic thread reduction in style of
BFM.

lib/simd/Grid_vComplexD.h

@@ -13,6 +13,9 @@ namespace Grid {
             vzero(*this);
             return (*this);
         }
+        vComplexD( Zero & z){
+            vzero(*this);
+        }
         vComplexD()=default;
         vComplexD(ComplexD a){
 	  vsplat(*this,a);
@@ -286,8 +289,8 @@ friend inline void vstore(const vComplexD &ret, ComplexD *a){
             return ret;
         }
 
-        friend inline vComplexD timesMinusI(const vComplexD &in){
-	  vComplexD ret; vzero(ret);
+        friend inline void timesMinusI(vComplexD &ret,const vComplexD &in){
+	  vzero(ret);
 	  vComplexD tmp;
 #if defined (AVX1)|| defined (AVX2)
 	  tmp.v    =_mm256_addsub_pd(ret.v,in.v); // r,-i
@@ -304,11 +307,10 @@ friend inline void vstore(const vComplexD &ret, ComplexD *a){
 #ifdef QPX
             assert(0);
 #endif
-	  return ret;
 	}
 
-        friend inline vComplexD timesI(const vComplexD &in){
-	  vComplexD ret; vzero(ret);
+        friend inline void timesI(vComplexD &ret, const vComplexD &in){
+	  vzero(ret);
 	  vComplexD tmp;
 #if defined (AVX1)|| defined (AVX2)
 	  tmp.v    =_mm256_shuffle_pd(in.v,in.v,0x5);
@@ -325,9 +327,21 @@ friend inline void vstore(const vComplexD &ret, ComplexD *a){
 #ifdef QPX
             assert(0);
 #endif
+	}
+
+        friend inline vComplexD timesMinusI(const vComplexD &in){
+	  vComplexD ret; 
+	  timesMinusI(ret,in);
 	  return ret;
 	}
 
+        friend inline vComplexD timesI(const vComplexD &in){
+	  vComplexD ret; 
+	  timesI(ret,in);
+	  return ret;
+	}
+
+
 // REDUCE FIXME must be a cleaner implementation
        friend inline ComplexD Reduce(const vComplexD & in)
        { 

lib/simd/Grid_vComplexF.h

@@ -28,6 +28,9 @@ namespace Grid {
             vzero(*this);
             return (*this);
         }
+        vComplexF( Zero & z){
+            vzero(*this);
+        }
         vComplexF()=default;
         vComplexF(ComplexF a){
 	  vsplat(*this,a);
@@ -363,8 +366,7 @@ namespace Grid {
 #endif
             return ret;
         }
-        friend inline vComplexF timesMinusI(const vComplexF &in){
-	  vComplexF ret; 
+        friend inline void timesMinusI( vComplexF &ret,const vComplexF &in){
 	  vzero(ret);
 #if defined (AVX1)|| defined (AVX2)
 	  cvec tmp =_mm256_addsub_ps(ret.v,in.v); // r,-i
@@ -381,10 +383,9 @@ namespace Grid {
 #ifdef QPX
             assert(0);
 #endif
-	  return ret;
 	}
-        friend inline vComplexF timesI(const vComplexF &in){
-	  vComplexF ret; vzero(ret);
+        friend inline void timesI(vComplexF &ret,const vComplexF &in){
+	  vzero(ret);
 #if defined (AVX1)|| defined (AVX2)
 	  cvec tmp =_mm256_shuffle_ps(in.v,in.v,_MM_SHUFFLE(2,3,0,1));//i,r
           ret.v    =_mm256_addsub_ps(ret.v,tmp);     //i,-r
@@ -400,8 +401,18 @@ namespace Grid {
 #ifdef QPX
             assert(0);
 #endif
+	}
+        friend inline vComplexF timesMinusI(const vComplexF &in){
+	  vComplexF ret; 
+	  timesMinusI(ret,in);
 	  return ret;
 	}
+        friend inline vComplexF timesI(const vComplexF &in){
+	  vComplexF ret; 
+	  timesI(ret,in);
+	  return ret;
+	}
+
         
         // Unary negation
         friend inline vComplexF operator -(const vComplexF &r) {

lib/simd/Grid_vRealD.h

@@ -17,6 +17,10 @@ namespace Grid {
         vRealD(Zero &zero){
 	  zeroit(*this);
 	}
+        vRealD & operator = ( Zero & z){
+	  vzero(*this);
+	  return (*this);
+        }
 
         friend inline void mult(vRealD * __restrict__ y,const vRealD * __restrict__ l,const vRealD *__restrict__ r) {*y = (*l) * (*r);}
         friend inline void sub (vRealD * __restrict__ y,const vRealD * __restrict__ l,const vRealD *__restrict__ r) {*y = (*l) - (*r);}

lib/simd/Grid_vRealF.h

@@ -18,6 +18,10 @@ namespace Grid {
         vRealF(Zero &zero){
 	  zeroit(*this);
 	}
+        vRealF & operator = ( Zero & z){
+	  vzero(*this);
+	  return (*this);
+        }
         ////////////////////////////////////
         // Arithmetic operator overloads +,-,*
         ////////////////////////////////////