diff --git a/Grid/qcd/action/fermion/WilsonFermion5D.h b/Grid/qcd/action/fermion/WilsonFermion5D.h
index dd83f269..40c1871f 100644
--- a/Grid/qcd/action/fermion/WilsonFermion5D.h
+++ b/Grid/qcd/action/fermion/WilsonFermion5D.h
@@ -119,6 +119,9 @@ public:
   void DhopOE(const FermionField &in, FermionField &out,int dag);
   void DhopEO(const FermionField &in, FermionField &out,int dag);
 
+  void DhopComms  (const FermionField &in, FermionField &out);
+  void DhopCalc   (const FermionField &in, FermionField &out,uint64_t *ids);
+  
   // add a DhopComm
   // -- suboptimal interface will presently trigger multiple comms.
   void DhopDir(const FermionField &in, FermionField &out,int dir,int disp);
diff --git a/Grid/qcd/action/fermion/WilsonKernels.h b/Grid/qcd/action/fermion/WilsonKernels.h
index 2d868c27..ad077dd3 100644
--- a/Grid/qcd/action/fermion/WilsonKernels.h
+++ b/Grid/qcd/action/fermion/WilsonKernels.h
@@ -57,6 +57,10 @@ public:
 			 int Ls, int Nsite, const FermionField &in, FermionField &out,
 			 int interior=1,int exterior=1) ;
 
+  static void DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
+			 int Ls, int Nsite, const FermionField &in, FermionField &out,
+			 uint64_t *ids);
+  
   static void DhopDagKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
 			    int Ls, int Nsite, const FermionField &in, FermionField &out,
 			    int interior=1,int exterior=1) ;
diff --git a/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
index ec3bd94a..92de5a40 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonFermion5DImplementation.h
@@ -438,6 +438,29 @@ void WilsonFermion5D<Impl>::DhopEO(const FermionField &in, FermionField &out,int
 
   DhopInternal(StencilOdd,UmuEven,in,out,dag);
 }
+template<class Impl>
+void WilsonFermion5D<Impl>::DhopComms(const FermionField &in, FermionField &out)
+{
+  int dag =0 ;
+  conformable(in.Grid(),FermionGrid()); // verifies full grid
+  conformable(in.Grid(),out.Grid());
+  out.Checkerboard() = in.Checkerboard();
+  Compressor compressor(dag);
+  Stencil.HaloExchangeOpt(in,compressor);
+}
+template<class Impl>
+void WilsonFermion5D<Impl>::DhopCalc(const FermionField &in, FermionField &out,uint64_t *ids)
+{
+  conformable(in.Grid(),FermionGrid()); // verifies full grid
+  conformable(in.Grid(),out.Grid());
+
+  out.Checkerboard() = in.Checkerboard();
+
+  int LLs = in.Grid()->_rdimensions[0];
+  int Opt = WilsonKernelsStatic::Opt;
+  Kernels::DhopKernel(Opt,Stencil,Umu,Stencil.CommBuf(),LLs,Umu.oSites(),in,out,ids);
+}
+
 template<class Impl>
 void WilsonFermion5D<Impl>::Dhop(const FermionField &in, FermionField &out,int dag)
 {
diff --git a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
index 43662b9c..1d0dfb61 100644
--- a/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
+++ b/Grid/qcd/action/fermion/implementation/WilsonKernelsImplementation.h
@@ -411,6 +411,46 @@ void WilsonKernels<Impl>::DhopDirKernel( StencilImpl &st, DoubledGaugeField &U,S
 #undef LoopBody
 }
 
+#ifdef GRID_SYCL
+extern "C" {
+    ulong SYCL_EXTERNAL __attribute__((overloadable)) intel_get_cycle_counter( void );
+    uint  SYCL_EXTERNAL __attribute__((overloadable)) intel_get_active_channel_mask( void );
+    uint  SYCL_EXTERNAL __attribute__((overloadable)) intel_get_grf_register( uint reg );
+    uint  SYCL_EXTERNAL __attribute__((overloadable)) intel_get_flag_register( uint flag );
+    uint  SYCL_EXTERNAL __attribute__((overloadable)) intel_get_control_register( uint reg );
+    uint  SYCL_EXTERNAL __attribute__((overloadable)) intel_get_hw_thread_id( void );
+    uint  SYCL_EXTERNAL __attribute__((overloadable)) intel_get_slice_id( void );
+    uint  SYCL_EXTERNAL __attribute__((overloadable)) intel_get_subslice_id( void );
+    uint  SYCL_EXTERNAL __attribute__((overloadable)) intel_get_eu_id( void );
+    uint  SYCL_EXTERNAL __attribute__((overloadable)) intel_get_eu_thread_id( void );
+    void  SYCL_EXTERNAL __attribute__((overloadable)) intel_eu_thread_pause( uint value );
+}
+#ifdef GRID_SIMT
+#define MAKE_ID(A) (intel_get_eu_id()<<16)|(intel_get_slice_id()<<8)|(intel_get_subslice_id())
+#else
+#define MAKE_ID(A) (0)
+#endif
+
+#else
+
+#define MAKE_ID(A) (0)
+
+#endif
+
+
+#define KERNEL_CALL_ID(A)						\
+  const uint64_t    NN = Nsite*Ls;					\
+  accelerator_forNB( ss, NN, Simd::Nsimd(), {				\
+      int sF = ss;							\
+      int sU = ss/Ls;							\
+      WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,in_v,out_v);		\
+      const int Nsimd = SiteHalfSpinor::Nsimd();			\
+      const int lane=acceleratorSIMTlane(Nsimd);                        \
+      int idx=sF*Nsimd+lane;						\
+      uint64_t id = MAKE_ID();						\
+      ids[idx]=id;							\
+    });									\
+  accelerator_barrier();
 
 #define KERNEL_CALLNB(A)						\
   const uint64_t    NN = Nsite*Ls;					\
@@ -418,7 +458,7 @@ void WilsonKernels<Impl>::DhopDirKernel( StencilImpl &st, DoubledGaugeField &U,S
       int sF = ss;							\
       int sU = ss/Ls;							\
       WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,in_v,out_v);		\
-  });
+    });
 
 #define KERNEL_CALL(A) KERNEL_CALLNB(A); accelerator_barrier();
 
@@ -451,6 +491,8 @@ void WilsonKernels<Impl>::DhopDirKernel( StencilImpl &st, DoubledGaugeField &U,S
     WilsonKernels<Impl>::A(st_v,U_v,buf,sF,sU,Ls,1,in_v,out_v);		\
     });}
 
+
+
 template <class Impl>
 void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
 				     int Ls, int Nsite, const FermionField &in, FermionField &out,
@@ -485,6 +527,18 @@ void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField
    }
    assert(0 && " Kernel optimisation case not covered ");
   }
+
+template <class Impl>
+void WilsonKernels<Impl>::DhopKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
+				     int Ls, int Nsite, const FermionField &in, FermionField &out,
+				     uint64_t *ids)
+{
+    autoView(U_v  ,  U,AcceleratorRead);
+    autoView(in_v , in,AcceleratorRead);
+    autoView(out_v,out,AcceleratorWrite);
+    autoView(st_v , st,AcceleratorRead);
+    KERNEL_CALL_ID(GenericDhopSite);
+}
   template <class Impl>
   void WilsonKernels<Impl>::DhopDagKernel(int Opt,StencilImpl &st,  DoubledGaugeField &U, SiteHalfSpinor * buf,
 					  int Ls, int Nsite, const FermionField &in, FermionField &out,
diff --git a/tests/Test_dwf_dslash_repro.cc b/tests/Test_dwf_dslash_repro.cc
new file mode 100644
index 00000000..1bf813d9
--- /dev/null
+++ b/tests/Test_dwf_dslash_repro.cc
@@ -0,0 +1,239 @@
+    /*************************************************************************************
+
+    Grid physics library, www.github.com/paboyle/Grid 
+
+    Source file: ./tests/Test_dwf_cg_prec.cc
+
+    Copyright (C) 2015
+
+Author: Peter Boyle <paboyle@ph.ed.ac.uk>
+
+    This program is free software; you can redistribute it and/or modify
+    it under the terms of the GNU General Public License as published by
+    the Free Software Foundation; either version 2 of the License, or
+    (at your option) any later version.
+
+    This program is distributed in the hope that it will be useful,
+    but WITHOUT ANY WARRANTY; without even the implied warranty of
+    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+    GNU General Public License for more details.
+
+    You should have received a copy of the GNU General Public License along
+    with this program; if not, write to the Free Software Foundation, Inc.,
+    51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
+
+    See the full license in the file "LICENSE" in the top level distribution directory
+    *************************************************************************************/
+    /*  END LEGAL */
+#include <Grid/Grid.h>
+
+using namespace std;
+using namespace Grid;
+
+#ifndef HOST_NAME_MAX
+#define HOST_NAME_MAX _POSIX_HOST_NAME_MAX
+#endif
+
+typedef LatticeFermionD FermionField;
+
+int VerifyOnDevice(const FermionField &res, FermionField &ref)
+{
+  deviceVector<int> Fails(1);
+  int * Fail = &Fails[0];
+  int FailHost=0;
+  
+  typedef typename FermionField::vector_object vobj;
+  typedef typename vobj::scalar_type scalar_type;
+  typedef typename vobj::vector_type vector_type;
+  
+  const uint64_t NN = res.Grid()->oSites();
+
+  acceleratorPut(*Fail,FailHost);
+
+  accelerator_barrier();
+  // Inject an error
+
+  int injection=0;
+  if(getenv("GRID_ERROR_INJECT")) injection=1;
+  autoView(res_v,res,AcceleratorWrite);
+  autoView(ref_v,ref,AcceleratorRead);
+  if ( res.Grid()->ThisRank()== 0 )
+  {
+    if (((random()&0xF)==0)&&injection) {
+      uint64_t sF = random()%(NN);
+      int lane=0;
+      printf("Error injection site %ld on rank %d\n",sF,res.Grid()->ThisRank());
+      auto vv = acceleratorGet(res_v[sF]);
+      double *dd = (double *)&vv;
+      *dd=M_PI;
+      acceleratorPut(res_v[sF],vv);
+    }
+  }
+
+  accelerator_for( sF, NN, vobj::Nsimd(), {
+#ifdef GRID_SIMT
+      {
+        int blane = acceleratorSIMTlane(vobj::Nsimd());
+#else
+      for(int blane;blane<vobj::Nsimd();blane++){
+#endif
+	vector_type *vtrr = (vector_type *)&res_v[sF];
+	vector_type *vtrf = (vector_type *)&ref_v[sF];
+	int words = sizeof(vobj)/sizeof(vector_type);
+	
+	for(int w=0;w<words;w++){
+	  scalar_type rrtmp = getlane(vtrr[w], blane);
+	  scalar_type rftmp = getlane(vtrf[w], blane);
+	  if ( rrtmp != rftmp) {
+	      *Fail=1;
+	  }
+	}
+      }
+  });
+
+  FailHost = acceleratorGet(*Fail);
+
+  return FailHost;
+}
+void PrintFails(const FermionField &res, FermionField &ref,uint64_t *ids)
+{
+  typedef typename FermionField::vector_object vobj;
+
+  const int Nsimd=vobj::Nsimd();
+  const uint64_t NN = res.Grid()->oSites();
+
+  ///////////////////////////////
+  // Pull back to host
+  ///////////////////////////////
+  autoView(res_v,res,CpuRead);
+  autoView(ref_v,ref,CpuRead);
+  
+  std::vector<uint64_t> ids_host(NN*Nsimd);
+  
+  acceleratorCopyFromDevice(ids,&ids_host[0],NN*Nsimd*sizeof(uint64_t));
+
+  //////////////////////////////////////////////////////////////
+  // Redo check on host and print IDs
+  //////////////////////////////////////////////////////////////
+  
+  for(int ss=0;ss< NN; ss++){				
+      int sF = ss;
+      for(int lane=0;lane<Nsimd;lane++){
+	
+	auto rr = extractLane(lane,res_v[sF]);
+	auto rf = extractLane(lane,ref_v[sF]);
+	uint64_t id = ids_host[lane+Nsimd*sF];
+	//	std::cout << GridHostname()<<" id["<<sF<<"] lane "<<lane<<" id "<<id<<std::endl;
+	for(int s=0;s<4;s++){
+	  for(int c=0;c<3;c++){
+	    if ( rr()(s)(c)!=rf()(s)(c) ) {
+	      int subslice=(id>>0 )&0xFF;
+	      int slice   =(id>>8 )&0xFF;
+	      int eu      =(id>>16)&0xFF;
+	      std::cout << GridHostname()<<" miscompare site "<<sF<<" "<<rr()(s)(c)<<" "<<rf()(s)(c)<<" EU "<<eu<<" slice "<<slice<<" subslice "<<subslice<<std::endl;
+	    }
+	  }
+	}
+      }
+  };
+  return;
+}
+
+
+
+int main (int argc, char ** argv)
+{
+  char hostname[HOST_NAME_MAX+1];
+  gethostname(hostname, HOST_NAME_MAX+1);
+  std::string host(hostname);
+  
+  Grid_init(&argc,&argv);
+
+  const int Ls=12;
+
+  GridCartesian         * UGrid   = SpaceTimeGrid::makeFourDimGrid(GridDefaultLatt(), GridDefaultSimd(Nd,vComplexD::Nsimd()),GridDefaultMpi());
+  GridRedBlackCartesian * UrbGrid = SpaceTimeGrid::makeFourDimRedBlackGrid(UGrid);
+  GridCartesian         * FGrid   = SpaceTimeGrid::makeFiveDimGrid(Ls,UGrid);
+  GridRedBlackCartesian * FrbGrid = SpaceTimeGrid::makeFiveDimRedBlackGrid(Ls,UGrid);
+
+  std::vector<int> seeds4({1,2,3,4});
+  std::vector<int> seeds5({5,6,7,8});
+  GridParallelRNG          RNG5(FGrid);  RNG5.SeedFixedIntegers(seeds5);
+  GridParallelRNG          RNG4(UGrid);  RNG4.SeedFixedIntegers(seeds4);
+
+  LatticeGaugeField Umu(UGrid);
+  LatticeFermionD    src(FGrid); random(RNG5,src);
+  LatticeFermionD   junk(FGrid); random(RNG5,junk);
+
+  LatticeFermionD result(FGrid); result=Zero();
+  LatticeFermionD ref(FGrid); ref=Zero();
+  
+  SU<Nc>::HotConfiguration(RNG4,Umu);
+
+  RealD mass=0.1;
+  RealD M5=1.8;
+
+  DomainWallFermionD Ddwf(Umu,*FGrid,*FrbGrid,*UGrid,*UrbGrid,mass,M5);
+
+  int nsecs=600;
+  if( GridCmdOptionExists(argv,argv+argc,"--seconds") ){
+    std::string arg = GridCmdOptionPayload(argv,argv+argc,"--seconds");
+    GridCmdOptionInt(arg,nsecs);
+  }
+  
+  std::cout << GridLogMessage << "::::::::::::: Job startup Barrier " << std::endl;
+  UGrid->Barrier();
+  std::cout << GridLogMessage << "::::::::::::: Job startup Barrier complete" << std::endl;
+
+  std::cout << GridLogMessage << "::::::::::::: Starting DWF repro for "<<nsecs <<" seconds" << std::endl;
+
+  time_t now;
+  time_t start = time(NULL);
+  UGrid->Broadcast(0,(void *)&start,sizeof(start));
+
+  FlightRecorder::ContinueOnFail = 0;
+  FlightRecorder::PrintEntireLog = 0;
+  FlightRecorder::ChecksumComms  = 0;
+  FlightRecorder::ChecksumCommsSend=0;
+
+  if(char *s=getenv("GRID_PRINT_ENTIRE_LOG"))  FlightRecorder::PrintEntireLog     = atoi(s);
+  if(char *s=getenv("GRID_CHECKSUM_RECV_BUF")) FlightRecorder::ChecksumComms      = atoi(s);
+  if(char *s=getenv("GRID_CHECKSUM_SEND_BUF")) FlightRecorder::ChecksumCommsSend  = atoi(s);
+
+  const uint64_t NN = FGrid->oSites()*vComplexD::Nsimd();
+  
+  deviceVector<uint64_t> ids_device(NN);
+  uint64_t *ids = &ids_device[0];
+  
+
+  Ddwf.DhopComms(src,ref);
+  Ddwf.DhopCalc(src,ref,ids);
+
+  Ddwf.DhopComms(src,result);
+  
+  int iter=0;
+  do {
+    
+    result=junk;
+
+    Ddwf.DhopCalc(src,result,ids);
+
+    if ( VerifyOnDevice(result, ref) ) {
+      printf("Node %s Iter %d detected fails\n",GridHostname(),iter);
+      PrintFails(result,ref,ids);
+      //      std::cout << " Dslash "<<iter<<" is WRONG! "<<std::endl;
+    }
+    //else {
+    //      printf("Node %s Iter %d detected NO fails\n",GridHostname(),iter);
+    //      PrintFails(result,ref,ids);
+    //      std::cout << " Dslash "<<iter<<" is OK! "<<std::endl;
+    //}
+
+
+    iter ++;
+    now = time(NULL); UGrid->Broadcast(0,(void *)&now,sizeof(now));
+  } while (now < (start + nsecs) );
+
+  
+  Grid_finalize();
+}