diff --git a/benchmarks/Benchmark_dwf_sweep.cc b/benchmarks/Benchmark_dwf_sweep.cc
index 302059a4..94a00903 100644
--- a/benchmarks/Benchmark_dwf_sweep.cc
+++ b/benchmarks/Benchmark_dwf_sweep.cc
@@ -68,10 +68,12 @@ int main (int argc, char ** argv)
   std::cout<<GridLogMessage << "=========================================================================="<<std::endl;
 
   int Lmax=32;
+  int dmin=0;
   if ( getenv("LMAX") ) Lmax=atoi(getenv("LMAX"));
-  for (int L=8;L<Lmax;L*=2){
+  if ( getenv("DMIN") ) dmin=atoi(getenv("DMIN"));
+  for (int L=8;L<=Lmax;L*=2){
     std::vector<int> latt4(4,L);
-    for(int d=4;d>0;d--){
+    for(int d=4;d>dmin;d--){
       if ( d<=3 ) latt4[d]*=2;
       std::cout << GridLogMessage <<"\t";
       for(int d=0;d<Nd;d++){
@@ -170,7 +172,11 @@ void benchDw(std::vector<int> & latt4, int Ls, int threads,int report )
   Dw.Dhop(src,result,0);
   double t1=usecond();
 
+#ifdef TIMERS_OFF
+    int ncall =10;
+#else
   int ncall =1+(int) ((5.0*1000*1000)/(t1-t0));
+#endif
 
   if (ncall < 5 ) exit(0);
 
@@ -297,7 +303,11 @@ void benchsDw(std::vector<int> & latt4, int Ls, int threads, int report )
     sDw.Dhop(ssrc,sresult,0);
     double t1=usecond();
 
+#ifdef TIMERS_OFF
+    int ncall =10;
+#else 
     int ncall =1+(int) ((5.0*1000*1000)/(t1-t0));
+#endif
 
     PerformanceCounter Counter(8);
     Counter.Start();
@@ -340,7 +350,9 @@ void benchsDw(std::vector<int> & latt4, int Ls, int threads, int report )
     CounterSdw.Start();
     t0=usecond();
     for(int i=0;i<ncall;i++){
+      __SSC_START;
       sDw.DhopEO(ssrc_o,sr_e,DaggerNo);
+      __SSC_STOP;
     }
     t1=usecond();
     CounterSdw.Stop();
diff --git a/lib/qcd/action/fermion/WilsonKernelsAsm.cc b/lib/qcd/action/fermion/WilsonKernelsAsm.cc
index cdbe2c8e..33c464ac 100644
--- a/lib/qcd/action/fermion/WilsonKernelsAsm.cc
+++ b/lib/qcd/action/fermion/WilsonKernelsAsm.cc
@@ -67,9 +67,12 @@ int setupSigns(void ){
 }
 static int signInit = setupSigns();
 
+#define label(A)  ilabel(A)
+#define ilabel(A) ".globl\n"  #A ":\n" 
+
 #define MAYBEPERM(A,perm) if (perm) { A ; }
 #define MULT_2SPIN(ptr,pf) MULT_ADDSUB_2SPIN(ptr,pf)
-
+#define FX(A) WILSONASM_ ##A
 template<>
 void WilsonKernels<WilsonImplF>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrder & lo,DoubledGaugeField &U,
 						     std::vector<SiteHalfSpinor,alignedAllocator<SiteHalfSpinor> >  &buf,
@@ -80,6 +83,8 @@ void WilsonKernels<WilsonImplF>::DiracOptAsmDhopSite(StencilImpl &st,LebesgueOrd
 #undef VMOVRDUP
 #undef MAYBEPERM
 #undef MULT_2SPIN
+#undef FX 
+#define FX(A) DWFASM_ ## A
 #define MAYBEPERM(A,B) 
 #define VMOVIDUP(A,B,C)                                  VBCASTIDUPf(A,B,C)
 #define VMOVRDUP(A,B,C)                                  VBCASTRDUPf(A,B,C)
diff --git a/lib/qcd/action/fermion/WilsonKernelsAsmBody.h b/lib/qcd/action/fermion/WilsonKernelsAsmBody.h
index bd96b7d5..d3e86276 100644
--- a/lib/qcd/action/fermion/WilsonKernelsAsmBody.h
+++ b/lib/qcd/action/fermion/WilsonKernelsAsmBody.h
@@ -1,8 +1,7 @@
 {
   int locala,perma, ptypea;
   int localb,permb, ptypeb;
-  int localc,permc, ptypec;
-  uint64_t basea, baseb, basec;
+  uint64_t basea, baseb;
   uint64_t basex;
   const uint64_t plocal =(uint64_t) & in._odata[0];
 
@@ -12,22 +11,15 @@
   MASK_REGS;
 
   for(int site=0;site<Ns;site++) {
-  int sU=lo.Reorder(ssU);
-
+  int sU=lo.Reorder(ssU);  
   for(int s=0;s<Ls;s++) {
-  ss     =sU*Ls+s;
-
+  ss=sU*Ls+s;
   ////////////////////////////////
   // Xp
   ////////////////////////////////
   int ent=ss*8;// 2*Ndim
   basea = st.GetInfo(ptypea,locala,perma,Xp,ent,plocal); ent++;
-  PREFETCH_CHIMU(basea);
   baseb = st.GetInfo(ptypeb,localb,permb,Yp,ent,plocal); ent++;
-  PREFETCH_CHIMU(baseb);
-  basec = st.GetInfo(ptypec,localc,permc,Zp,ent,plocal); ent++;
-  PREFETCH_CHIMU(basec);
-
   basex = basea;
 
   if ( locala ) {
@@ -47,7 +39,6 @@
   // Yp
   ////////////////////////////////
   basea = st.GetInfo(ptypea,locala,perma,Xp,ent,plocal); ent++;
-  PREFETCH_CHIMU(basea);
   if ( localb ) {
     LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
     YM_PROJMEM(baseb);
@@ -56,7 +47,7 @@
     LOAD_CHI(baseb);
   }
   {
-    MULT_2SPIN_DIR_PFYP(Yp,basec);
+    MULT_2SPIN_DIR_PFYP(Yp,basea);
   }
   LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
   YM_RECON_ACCUM;
@@ -65,16 +56,15 @@
   // Zp
   ////////////////////////////////
   baseb = st.GetInfo(ptypeb,localb,permb,Yp,ent,plocal); ent++;
-  PREFETCH_CHIMU(baseb);
-  if ( localc ) {
+  if ( locala ) {
     LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
-    ZM_PROJMEM(basec);
-    MAYBEPERM(PERMUTE_DIR1,permc);
+    ZM_PROJMEM(basea);
+    MAYBEPERM(PERMUTE_DIR1,perma);
   } else { 
-    LOAD_CHI(basec);
+    LOAD_CHI(basea);
   }
   {
-    MULT_2SPIN_DIR_PFZP(Zp,basea);
+    MULT_2SPIN_DIR_PFZP(Zp,baseb);
   }
   LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
   ZM_RECON_ACCUM;
@@ -82,17 +72,16 @@
   ////////////////////////////////
   // Tp
   ////////////////////////////////
-  basec = st.GetInfo(ptypec,localc,permc,Xp,ent,plocal); ent++;
-  PREFETCH_CHIMU(basec);
-  if ( locala ) {
+  basea = st.GetInfo(ptypea,locala,perma,Xp,ent,plocal); ent++;
+  if ( localb ) {
     LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
-    TM_PROJMEM(basea);
-    MAYBEPERM(PERMUTE_DIR0,perma);
+    TM_PROJMEM(baseb);
+    MAYBEPERM(PERMUTE_DIR0,permb);
   } else { 
-    LOAD_CHI(basea);
+    LOAD_CHI(baseb);
   }
   {
-    MULT_2SPIN_DIR_PFTP(Tp,baseb);
+    MULT_2SPIN_DIR_PFTP(Tp,basea);
   }
   LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
   TM_RECON_ACCUM;
@@ -100,17 +89,16 @@
   ////////////////////////////////
   // Xm
   ////////////////////////////////
-  basea = st.GetInfo(ptypea,locala,perma,Yp,ent,plocal); ent++;
-  PREFETCH_CHIMU(basea);
-  if ( localb ) {
+  baseb = st.GetInfo(ptypeb,localb,permb,Yp,ent,plocal); ent++;
+  if ( locala ) {
     LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
-    XP_PROJMEM(baseb);
-    MAYBEPERM(PERMUTE_DIR3,permb);
+    XP_PROJMEM(basea);
+    MAYBEPERM(PERMUTE_DIR3,perma);
   } else { 
-    LOAD_CHI(baseb);
+    LOAD_CHI(basea);
   }
   {
-    MULT_2SPIN_DIR_PFXM(Xm,basec);
+    MULT_2SPIN_DIR_PFXM(Xm,baseb);
   }
   LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
   XP_RECON_ACCUM;
@@ -118,14 +106,13 @@
   ////////////////////////////////
   // Ym
   ////////////////////////////////
-  baseb = st.GetInfo(ptypeb,localb,permb,Xp,ent,plocal); ent++;
-  PREFETCH_CHIMU(baseb);
-  if ( localc ) {
+  basea = st.GetInfo(ptypea,locala,perma,Xp,ent,plocal); ent++;
+  if ( localb ) {
     LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
-    YP_PROJMEM(basec);
-    MAYBEPERM(PERMUTE_DIR2,permc);
+    YP_PROJMEM(baseb);
+    MAYBEPERM(PERMUTE_DIR2,permb);
   } else { 
-    LOAD_CHI(basec);
+    LOAD_CHI(baseb);
   }
   {
     MULT_2SPIN_DIR_PFYM(Ym,basea);
@@ -136,8 +123,7 @@
   ////////////////////////////////
   // Zm
   ////////////////////////////////
-  basec = st.GetInfo(ptypec,localc,permc,Yp,ent,plocal); ent++;
-  PREFETCH_CHIMU(basec);
+  baseb = st.GetInfo(ptypeb,localb,permb,Yp,ent,plocal); ent++;
   if ( locala ) {
     LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
     ZP_PROJMEM(basea);
@@ -155,7 +141,6 @@
   // Tm
   ////////////////////////////////
   basea = (uint64_t)&out._odata[ss];
-  PREFETCH_CHIMU(basea);
   if ( localb ) {
     LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
     TP_PROJMEM(baseb);
@@ -163,16 +148,16 @@
   } else { 
     LOAD_CHI(baseb);
   }
+  baseb = st.GetInfo(ptypeb,localb,permb,Yp,ent,plocal);
   {
-    MULT_2SPIN_DIR_PFTM(Tm,basec);
+    MULT_2SPIN_DIR_PFTM(Tm,basea);
   }
   LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
   TP_RECON_ACCUM;
 
-  //  PREFETCH_CHIMU(basex);
-  SAVE_RESULT(&out._odata[ss]);
-  
-  }
+  SAVE_RESULT(&out._odata[ss],baseb);
+
+  } 
   ssU++;
   }
 }
diff --git a/lib/qcd/action/fermion/WilsonKernelsAsmBody.h.abc b/lib/qcd/action/fermion/WilsonKernelsAsmBody.h.abc
new file mode 100644
index 00000000..5a3e01f7
--- /dev/null
+++ b/lib/qcd/action/fermion/WilsonKernelsAsmBody.h.abc
@@ -0,0 +1,187 @@
+{
+  int locala,perma, ptypea;
+  int localb,permb, ptypeb;
+  int localc,permc, ptypec;
+  uint64_t basea, baseb, basec;
+  uint64_t basex;
+  const uint64_t plocal =(uint64_t) & in._odata[0];
+
+  //  vComplexF isigns[2] = { signs[0], signs[1] };
+  vComplexF *isigns = &signs[0];
+
+  MASK_REGS;
+
+  for(int site=0;site<Ns;site++) {
+  int sU=lo.Reorder(ssU);
+
+  for(int s=0;s<Ls;s++) {
+  ss     =sU*Ls+s;
+
+  ////////////////////////////////
+  // Xp
+  ////////////////////////////////
+  int ent=ss*8;// 2*Ndim
+  basea = st.GetInfo(ptypea,locala,perma,Xp,ent,plocal); ent++;
+  PREFETCH_CHIMU(basea);
+  baseb = st.GetInfo(ptypeb,localb,permb,Yp,ent,plocal); ent++;
+  PREFETCH_CHIMU(baseb);
+  basec = st.GetInfo(ptypec,localc,permc,Zp,ent,plocal); ent++;
+  PREFETCH_CHIMU(basec);
+
+  basex = basea;
+
+  label(FX(XP) );
+  if ( locala ) {
+    LOAD64(%r10,isigns);
+    XM_PROJMEM(basea);
+    MAYBEPERM(PERMUTE_DIR3,perma);
+  } else { 
+    LOAD_CHI(basea);
+  }
+  {
+    MULT_2SPIN_DIR_PFXP(Xp,baseb);
+  }
+  LOAD64(%r10,isigns);
+  XM_RECON;
+
+  ////////////////////////////////
+  // Yp
+  ////////////////////////////////
+  basea = st.GetInfo(ptypea,locala,perma,Xp,ent,plocal); ent++;
+  PREFETCH_CHIMU(basea);
+  label(FX(YP) );
+  if ( localb ) {
+    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+    YM_PROJMEM(baseb);
+    MAYBEPERM(PERMUTE_DIR2,permb);
+  } else { 
+    LOAD_CHI(baseb);
+  }
+  {
+    MULT_2SPIN_DIR_PFYP(Yp,basec);
+  }
+  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+  YM_RECON_ACCUM;
+
+  ////////////////////////////////
+  // Zp
+  ////////////////////////////////
+  baseb = st.GetInfo(ptypeb,localb,permb,Yp,ent,plocal); ent++;
+  PREFETCH_CHIMU(baseb);
+  label(FX(ZP) );
+  if ( localc ) {
+    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+    ZM_PROJMEM(basec);
+    MAYBEPERM(PERMUTE_DIR1,permc);
+  } else { 
+    LOAD_CHI(basec);
+  }
+  {
+    MULT_2SPIN_DIR_PFZP(Zp,basea);
+  }
+  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+  ZM_RECON_ACCUM;
+
+  ////////////////////////////////
+  // Tp
+  ////////////////////////////////
+  basec = st.GetInfo(ptypec,localc,permc,Xp,ent,plocal); ent++;
+  PREFETCH_CHIMU(basec);
+  label(FX(TP) );
+  if ( locala ) {
+    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+    TM_PROJMEM(basea);
+    MAYBEPERM(PERMUTE_DIR0,perma);
+  } else { 
+    LOAD_CHI(basea);
+  }
+  {
+    MULT_2SPIN_DIR_PFTP(Tp,baseb);
+  }
+  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+  TM_RECON_ACCUM;
+
+  ////////////////////////////////
+  // Xm
+  ////////////////////////////////
+  basea = st.GetInfo(ptypea,locala,perma,Yp,ent,plocal); ent++;
+  PREFETCH_CHIMU(basea);
+  label(FX(XM) );
+  if ( localb ) {
+    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+    XP_PROJMEM(baseb);
+    MAYBEPERM(PERMUTE_DIR3,permb);
+  } else { 
+    LOAD_CHI(baseb);
+  }
+  {
+    MULT_2SPIN_DIR_PFXM(Xm,basec);
+  }
+  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+  XP_RECON_ACCUM;
+
+  ////////////////////////////////
+  // Ym
+  ////////////////////////////////
+  baseb = st.GetInfo(ptypeb,localb,permb,Xp,ent,plocal); ent++;
+  PREFETCH_CHIMU(baseb);
+  label(FX(YM) );
+  if ( localc ) {
+    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+    YP_PROJMEM(basec);
+    MAYBEPERM(PERMUTE_DIR2,permc);
+  } else { 
+    LOAD_CHI(basec);
+  }
+  {
+    MULT_2SPIN_DIR_PFYM(Ym,basea);
+  }
+  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+  YP_RECON_ACCUM;
+
+  ////////////////////////////////
+  // Zm
+  ////////////////////////////////
+  basec = st.GetInfo(ptypec,localc,permc,Yp,ent,plocal); ent++;
+  PREFETCH_CHIMU(basec);
+  label(FX(ZM) );
+  if ( locala ) {
+    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+    ZP_PROJMEM(basea);
+    MAYBEPERM(PERMUTE_DIR1,perma);
+  } else { 
+    LOAD_CHI(basea);
+  }
+  {
+    MULT_2SPIN_DIR_PFZM(Zm,baseb);
+  }
+  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+  ZP_RECON_ACCUM;
+
+  ////////////////////////////////
+  // Tm
+  ////////////////////////////////
+  basea = (uint64_t)&out._odata[ss];
+  PREFETCH_CHIMU(basea);
+  label(FX(TM) );
+  if ( localb ) {
+    LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+    TP_PROJMEM(baseb);
+    MAYBEPERM(PERMUTE_DIR0,permb);
+  } else { 
+    LOAD_CHI(baseb);
+  }
+  {
+    MULT_2SPIN_DIR_PFTM(Tm,basec);
+  }
+  LOAD64(%r10,isigns);  // times i => shuffle and xor the real part sign bit
+  TP_RECON_ACCUM;
+
+  //  PREFETCH_CHIMU(basex);
+  label(FX(SAV) );
+  SAVE_RESULT(&out._odata[ss]);
+  
+  }
+  ssU++;
+  }
+}
diff --git a/lib/simd/Intel512common.h b/lib/simd/Intel512common.h
index a3cd980d..6878bcfb 100644
--- a/lib/simd/Intel512common.h
+++ b/lib/simd/Intel512common.h
@@ -1,4 +1,4 @@
-    /*************************************************************************************
+   /*************************************************************************************
 
     Grid physics library, www.github.com/paboyle/Grid 
 
@@ -37,6 +37,8 @@ Author: paboyle <paboyle@ph.ed.ac.uk>
            "mov     $0x5555, %%eax \n"\
            "kmovw    %%eax, %%k7 \n" : : : "%eax");
 
+//#define label(B) __asm__ ( __func__ __LINE__ #B ":\n" );
+
 #define VZEROf(A)       "vpxorq " #A ","  #A "," #A ";\n"
 #define VZEROd(A)       "vpxorq " #A ","  #A "," #A ";\n"