diff --git a/lib/qcd/action/fermion/WilsonFermion5D.cc b/lib/qcd/action/fermion/WilsonFermion5D.cc index daddb605..cacc7a9f 100644 --- a/lib/qcd/action/fermion/WilsonFermion5D.cc +++ b/lib/qcd/action/fermion/WilsonFermion5D.cc @@ -406,7 +406,7 @@ void WilsonFermion5D::DhopInternalOverlappedComms(StencilImpl & st, Lebesg // Load imbalance alert. Should use dynamic schedule OMP for loop // Perhaps create a list of only those sites with face work, and // load balance process the list. -#if 1 +#if 0 #pragma omp parallel { int nthreads = omp_get_num_threads(); @@ -425,13 +425,15 @@ void WilsonFermion5D::DhopInternalOverlappedComms(StencilImpl & st, Lebesg #else DhopComputeTime2-=usecond(); if (dag == DaggerYes) { - parallel_for (int ss = 0; ss < U._grid->oSites(); ss++) { +#pragma omp parallel for schedule(static,4) + for (int ss = 0; ss < U._grid->oSites(); ss++) { int sU = ss; int sF = LLs * sU; Kernels::DhopSiteDag(st,lo,U,st.CommBuf(),sF,sU,LLs,1,in,out,0,1); } } else { - parallel_for (int ss = 0; ss < U._grid->oSites(); ss++) { +#pragma omp parallel for schedule(static,1) + for (int ss = 0; ss < U._grid->oSites(); ss++) { int sU = ss; int sF = LLs * sU; Kernels::DhopSite(st,lo,U,st.CommBuf(),sF,sU,LLs,1,in,out,0,1); diff --git a/lib/qcd/action/fermion/WilsonKernels.cc b/lib/qcd/action/fermion/WilsonKernels.cc index 6ed350cf..03c066b0 100644 --- a/lib/qcd/action/fermion/WilsonKernels.cc +++ b/lib/qcd/action/fermion/WilsonKernels.cc @@ -33,8 +33,8 @@ directory namespace Grid { namespace QCD { - int WilsonKernelsStatic::Opt = WilsonKernelsStatic::OptGeneric; - int WilsonKernelsStatic::Comms = WilsonKernelsStatic::CommsAndCompute; +int WilsonKernelsStatic::Opt = WilsonKernelsStatic::OptGeneric; +int WilsonKernelsStatic::Comms = WilsonKernelsStatic::CommsAndCompute; template WilsonKernels::WilsonKernels(const ImplParams &p) : Base(p){};