Index: modulo-sched.c
===================================================================
--- modulo-sched.c	(revision 130895)
+++ modulo-sched.c	(working copy)
@@ -688,7 +688,22 @@
 	  emit_insn (copy_rtx (PATTERN (reg_move)));
 	if (SCHED_STAGE (u_node) >= from_stage
 	    && SCHED_STAGE (u_node) <= to_stage)
-	  duplicate_insn_chain (u_node->first_note, u_node->insn);
+        {
+            /* The note will get removed from insn stream in fixup.  */
+            emit_note (NOTE_INSN_DELETED);
+            /* Avoid coping reg-moves that were created for the body of
+               the loop.  */
+            if (u_node->first_note != u_node->insn)
+              {
+                rtx insn;
+
+                for (insn = u_node->first_note; insn != u_node->insn;
+                     insn = NEXT_INSN (insn))
+                  if (NOTE_P (insn))
+                    emit_note_copy (insn);
+              }
+            emit_copy_of_insn_after (u_node->insn, get_last_insn ()); 
+         }
       }
 }
 
@@ -794,7 +809,11 @@
 {
 
   if (loop->inner || !loop_outer (loop))
+  {
+    if (dump_file)
+      fprintf (dump_file, "SMS loop inner or !loop_outer\n");
     return false;
+  }
 
   if (!single_exit (loop))
     {
@@ -910,6 +929,12 @@
      We use loop->num as index into this array.  */
   g_arr = XCNEWVEC (ddg_ptr, number_of_loops ());
 
+  if (dump_file)
+  {
+    fprintf (dump_file, "\n\nSMS analysis phase\n");
+    fprintf (dump_file, "===================\n\n");
+  }
+
   /* Build DDGs for all the relevant loops and hold them in G_ARR
      indexed by the loop index.  */
   FOR_EACH_LOOP (li, loop, 0)
@@ -926,11 +951,24 @@
           break;
         }
 
+      if (dump_file)
+      {
+         rtx insn = BB_END (loop->header);
+
+         fprintf (dump_file, "SMS loop num: %d, file: %s, line: %d\n",
+                  loop->num, insn_file (insn), insn_line (insn));
+
+      }
+
       if (! loop_canon_p (loop))
         continue;
 
       if (! loop_single_full_bb_p (loop))
+      {
+        if (dump_file)
+          fprintf (dump_file, "SMS not loop_single_full_bb_p\n");
 	continue;
+      }
 
       bb = loop->header;
 
@@ -971,7 +1009,11 @@
 
       /* Make sure this is a doloop.  */
       if ( !(count_reg = doloop_register_get (head, tail)))
+      {
+        if (dump_file)
+          fprintf (dump_file, "SMS doloop_register_get failed\n");
 	continue;
+      }
 
       /* Don't handle BBs with calls or barriers, or !single_set insns,
          or auto-increment insns (to avoid creating invalid reg-moves
@@ -1021,7 +1063,15 @@
         }
 
       g_arr[loop->num] = g;
+      if (dump_file)
+        fprintf (dump_file, "...OK\n");
+
     }
+  if (dump_file)
+  {
+    fprintf (dump_file, "\nSMS transformation phase\n");
+    fprintf (dump_file, "=========================\n\n");
+  }
 
   /* We don't want to perform SMS on new loops - created by versioning.  */
   FOR_EACH_LOOP (li, loop, 0)
@@ -1036,8 +1086,15 @@
         continue;
 
       if (dump_file)
-	print_ddg (dump_file, g);
+      {
+         rtx insn = BB_END (loop->header);
 
+         fprintf (dump_file, "SMS loop num: %d, file: %s, line: %d\n",
+                  loop->num, insn_file (insn), insn_line (insn));
+
+         print_ddg (dump_file, g);
+      }
+
       get_ebb_head_tail (loop->header, loop->header, &head, &tail);
 
       latch_edge = loop_latch_edge (loop);