ch4/wait: Optimize waitall when all requests are local

Optimize progress for a batch of local communication requests by
skipping netmod progress. Shows a 5-10% improvement in ch4/shm
bandwidth measurements on a single node of Cascade Lake.

Author: raffenet

src/mpid/ch4/src/ch4_wait.h

@@ -22,7 +22,8 @@ MPL_STATIC_INLINE_PREFIX void MPIDI_set_progress_vci(MPIR_Request * req,
 MPL_STATIC_INLINE_PREFIX void MPIDI_set_progress_vci_n(int n, MPIR_Request ** reqs,
                                                        MPID_Progress_state * state)
 {
-    state->flag = MPIDI_PROGRESS_ALL;   /* TODO: check request is_local/anysource */
+    state->flag = MPIDI_PROGRESS_ALL;
+    bool local_only = true;
 
     int idx = 0;
     for (int i = 0; i < n; i++) {
@@ -45,8 +46,15 @@ MPL_STATIC_INLINE_PREFIX void MPIDI_set_progress_vci_n(int n, MPIR_Request ** re
         if (!found) {
             state->vci[idx++] = vci;
         }
+#ifndef MPIDI_CH4_DIRECT_NETMOD
+        local_only = local_only && MPIDI_REQUEST(reqs[i], is_local);
+#endif
     }
     state->vci_count = idx;
+
+    if (local_only) {
+        state->flag &= ~MPIDI_PROGRESS_NM;
+    }
 }
 
 /* MPID_Test, MPID_Testall, MPID_Testany, MPID_Testsome */