PME reduction for CUDA F buffer operations

[alexxy/gromacs.git] / src / gromacs / ewald / pme_only.cpp

diff --git a/src/gromacs/ewald/pme_only.cpp b/src/gromacs/ewald/pme_only.cpp
index 6f8ebe1723e80eb7cca84ddbfeada3a23a7d9064..e3e948a856ac6edac000bf8e4958ac044f8529d2 100644 (file)
--- a/src/gromacs/ewald/pme_only.cpp
+++ b/src/gromacs/ewald/pme_only.cpp
@@ -626,7 +626,7 @@ int gmx_pmeonly(struct gmx_pme_t *pme,
             pme_gpu_prepare_computation(pme, boxChanged, box, wcycle, pmeFlags);
             pme_gpu_launch_spread(pme, as_rvec_array(pme_pp->x.data()), wcycle);
             pme_gpu_launch_complex_transforms(pme, wcycle);
-            pme_gpu_launch_gather(pme, wcycle, PmeForceOutputHandling::Set);
+            pme_gpu_launch_gather(pme, wcycle, PmeForceOutputHandling::Set, false);
             output = pme_gpu_wait_finish_task(pme, pmeFlags, wcycle);
             pme_gpu_reinit_computation(pme, wcycle);
         }