Fix for force copy dependency of GPU force reduction

Fixes issue #3729

diff --git a/src/gromacs/mdlib/sim_util.cpp b/src/gromacs/mdlib/sim_util.cpp
index 47b6baac1bf0ef2fddddae85a59d91230af4e361..e788c6b66b2d2d5237d0f005f4aad8795402a016 100644 (file)
--- a/src/gromacs/mdlib/sim_util.cpp
+++ b/src/gromacs/mdlib/sim_util.cpp
@@ -1089,8 +1089,10 @@ static void setupGpuForceReductions(gmx::MdrunScheduleWorkload* runScheduleWork,
     if ((runScheduleWork->domainWork.haveCpuLocalForceWork || havePPDomainDecomposition(cr))
         && !runScheduleWork->simulationWork.useGpuHaloExchange)
     {
+        auto forcesReadyLocality = havePPDomainDecomposition(cr) ? AtomLocality::Local : AtomLocality::All;
+        const bool useGpuForceBufferOps = true;
         fr->gpuForceReduction[gmx::AtomLocality::Local]->addDependency(
-                stateGpu->getForcesReadyOnDeviceEvent(AtomLocality::Local, true));
+                stateGpu->getForcesReadyOnDeviceEvent(forcesReadyLocality, useGpuForceBufferOps));
     }
 
     if (runScheduleWork->simulationWork.useGpuHaloExchange)