Added more barriers to GMRES

kokkos · Jan 26, 2022 · 64aa095 · 64aa095
1 parent 1abd4e2
commit 64aa095
Show file tree

Hide file tree

Showing 3 changed files with 15 additions and 8 deletions.
diff --git a/src/batched/sparse/KokkosBatched_JacobiPrec.hpp b/src/batched/sparse/KokkosBatched_JacobiPrec.hpp
@@ -146,7 +146,11 @@ class JacobiPrec {
   KOKKOS_INLINE_FUNCTION void apply(const MemberType &member,
                                     const XViewType &X,
                                     const YViewType &Y) const {
-    if (!computed_inverse) this->computeInverse<MemberType, ArgMode>(member);
+    if (!computed_inverse)
+    {
+      this->computeInverse<MemberType, ArgMode>(member);
+      member.team_barrier();  //Finish writing to this->diag_values
+    }
 
     KokkosBatched::HadamardProduct<MemberType, ArgMode>::template invoke<
         ValuesViewType, XViewType, YViewType>(member, diag_values, X, Y);

diff --git a/src/batched/sparse/impl/KokkosBatched_GMRES_TeamVector_Impl.hpp b/src/batched/sparse/impl/KokkosBatched_GMRES_TeamVector_Impl.hpp
@@ -172,15 +172,16 @@ struct TeamVectorGMRES {
       P.template apply<MemberType, ScratchPadVectorViewType,
                        ScratchPadVectorViewType, Trans::NoTranspose,
                        Mode::TeamVector, 1>(member, W, W);
-      member.team_barrier();
 
       for (size_t i = 0; i < j + 1; ++i) {
+        member.team_barrier();  //Finish writing to W
         auto V_i = Kokkos::subview(V, Kokkos::ALL, i, Kokkos::ALL);
         TeamVectorDot<MemberType>::invoke(member, W, V_i, tmp);
         member.team_barrier();
         TeamVectorCopy1D::invoke(member, tmp,
                                  Kokkos::subview(H, Kokkos::ALL, i, j));
 
+        member.team_barrier();  //Don't start modifying tmp until copy above finishes
         Kokkos::parallel_for(
             Kokkos::TeamVectorRange(member, 0, numMatrices),
             [&](const OrdinalType& ii) { tmp(ii) = -tmp(ii); });
@@ -276,11 +277,12 @@ struct TeamVectorGMRES {
     member.team_barrier();  // Finish writing to G
 
     for (size_t j = 0; j < maximum_iteration; ++j)
+    {
       TeamVectorAxpy<MemberType>::invoke(
           member, Kokkos::subview(G, Kokkos::ALL, j),
           Kokkos::subview(V, Kokkos::ALL, j, Kokkos::ALL), X);
-
-    member.team_barrier();  // Finish writing to X
+      member.team_barrier();  //Finish writing to X
+    }
 
     TeamVectorCopy<MemberType>::invoke(member, X, _X);
     return status;

diff --git a/src/batched/sparse/impl/KokkosBatched_GMRES_Team_Impl.hpp b/src/batched/sparse/impl/KokkosBatched_GMRES_Team_Impl.hpp
@@ -171,14 +171,14 @@ struct TeamGMRES {
       P.template apply<MemberType, ScratchPadVectorViewType,
                        ScratchPadVectorViewType, Trans::NoTranspose, Mode::Team,
                        1>(member, W, W);
-      member.team_barrier();
 
       for (size_t i = 0; i < j + 1; ++i) {
+        member.team_barrier();  //Finish writing to W
         auto V_i = Kokkos::subview(V, Kokkos::ALL, i, Kokkos::ALL);
         TeamDot<MemberType>::invoke(member, W, V_i, tmp);
         member.team_barrier();
         TeamCopy1D::invoke(member, tmp, Kokkos::subview(H, Kokkos::ALL, i, j));
-
+        member.team_barrier();  //Don't start modifying tmp until copy above finishes
         Kokkos::parallel_for(
             Kokkos::TeamThreadRange(member, 0, numMatrices),
             [&](const OrdinalType& ii) { tmp(ii) = -tmp(ii); });
@@ -274,11 +274,12 @@ struct TeamGMRES {
     member.team_barrier();  // Finish writing to G
 
     for (size_t j = 0; j < maximum_iteration; ++j)
+    {
       TeamAxpy<MemberType>::invoke(
           member, Kokkos::subview(G, Kokkos::ALL, j),
           Kokkos::subview(V, Kokkos::ALL, j, Kokkos::ALL), X);
-
-    member.team_barrier();  // Finish writing to X
+      member.team_barrier();  //Finish writing to X
+    }
 
     TeamCopy<MemberType>::invoke(member, X, _X);
     return status;