Rearranged the order of real-space and kspace part of ufield0c(), delayed device-host transfer from umutual2b() to overlap with kspace part

2022-07-08 14:45:31 -05:00
parent 78d6df5ba9
commit 0c44bd1086
9 changed files with 161 additions and 6 deletions
--- a/lib/gpu/lal_base_amoeba.cpp
+++ b/lib/gpu/lal_base_amoeba.cpp
@ -528,8 +528,8 @@ void BaseAmoebaT::compute_umutual2b(int *host_amtype, int *host_amgroup, double
  const int red_blocks=umutual2b(_eflag,_vflag);

  // copy field and fieldp from device to host (_fieldp store both arrays, one after another)
-
-  _fieldp.update_host(_max_fieldp_size*8,false);
+  // NOTE: move this step to update_fieldp() to delay device-host transfer
+  //_fieldp.update_host(_max_fieldp_size*8,false);
 }

 // ---------------------------------------------------------------------------