bb-reorder.c (find_traces_1_round): Don't connect easy to copy successors with multiple predecessors.

* bb-reorder.c (find_traces_1_round): Don't connect easy to copy
        successors with multiple predecessors.
        (connect_traces): Try harder to copy traces of length 1.

        * function.h (struct function): Add computed_goto_common_label,
        computed_goto_common_reg.
        * function.c (free_after_compilation): Zap them.
        * stmt.c (expand_computed_goto): Use them to produce one
        indirect branch per function.

From-SVN: r62944

gcc/ChangeLog

@@ -38,6 +38,18 @@
 	* config/rs6000/rs64.md: New file.
 	[Some DFA descriptions based on work by Michael Hayes]
 
+2003-02-15  Richard Henderson  <rth@redhat.com>
+
+	* bb-reorder.c (find_traces_1_round): Don't connect easy to copy
+	successors with multiple predecessors.
+	(connect_traces): Try harder to copy traces of length 1.
+
+	* function.h (struct function): Add computed_goto_common_label,
+	computed_goto_common_reg.
+	* function.c (free_after_compilation): Zap them.
+	* stmt.c (expand_computed_goto): Use them to produce one 
+	indirect branch per function.
+
 2003-02-15  Richard Henderson  <rth@redhat.com>
 
 	* cfgcleanup.c: Include params.h.

gcc/bb-reorder.c

@@ -448,6 +448,12 @@ find_traces_1_round (branch_th, exec_th, count_th, traces, n_traces, round,
 		  || prob < branch_th || freq < exec_th || e->count < count_th)
 		continue;
 
+	      /* If the destination has multiple precessesors, and can be
+		 duplicated cheaper than a jump, don't allow it to be added
+		 to a trace.  We'll duplicate it when connecting traces.  */
+	      if (e->dest->pred->pred_next && copy_bb_p (e->dest, 0))
+		continue;
+
 	      if (better_edge_p (bb, e, prob, freq, best_prob, best_freq))
 		{
 		  best_edge = e;
@@ -905,19 +911,27 @@ connect_traces (n_traces, traces)
 	      /* Try to connect the traces by duplication of 1 block.  */
 	      edge e2;
 	      basic_block next_bb = NULL;
+	      bool try_copy = false;
 
 	      for (e = traces[t].last->succ; e; e = e->succ_next)
 		if (e->dest != EXIT_BLOCK_PTR
 		    && (e->flags & EDGE_CAN_FALLTHRU)
 		    && !(e->flags & EDGE_COMPLEX)
-		    && (EDGE_FREQUENCY (e) >= freq_threshold)
-		    && (e->count >= count_threshold)
-		    && (!best
-			|| e->probability > best->probability))
+		    && (!best || e->probability > best->probability))
 		  {
 		    edge best2 = NULL;
 		    int best2_len = 0;
 
+		    /* If the destination trace is only one block
+		       long, then no need to search the successor
+		       blocks of the trace.  Accept it.  */
+		   if (traces[bbd[e->dest->index].start_of_trace].length == 1)
+		     {
+		       best = e;
+		       try_copy = true;
+		       continue;
+		     }
+
 		    for (e2 = e->dest->succ; e2; e2 = e2->succ_next)
 		      {
 			int di = e2->dest->index;
@@ -942,10 +956,18 @@ connect_traces (n_traces, traces)
 			    else
 			      best2_len = INT_MAX;
 			    next_bb = e2->dest;
+			    try_copy = true;
 			  }
 		      }
 		  }
-	      if (best && next_bb && copy_bb_p (best->dest, !optimize_size))
+
+	      /* Copy tiny blocks always; copy larger blocks only when the
+		 edge is traversed frequently enough.  */
+	      if (try_copy
+		  && copy_bb_p (best->dest,
+				!optimize_size
+				&& EDGE_FREQUENCY (best) >= freq_threshold
+				&& best->count >= count_threshold))
 		{
 		  basic_block new_bb;
 
@@ -953,7 +975,9 @@ connect_traces (n_traces, traces)
 		    {
 		      fprintf (rtl_dump_file, "Connection: %d %d ",
 			       traces[t].last->index, best->dest->index);
-		      if (next_bb == EXIT_BLOCK_PTR)
+		      if (!next_bb)
+			fputc ('\n', rtl_dump_file);
+		      else if (next_bb == EXIT_BLOCK_PTR)
 			fprintf (rtl_dump_file, "exit\n");
 		      else
 			fprintf (rtl_dump_file, "%d\n", next_bb->index);
@@ -961,7 +985,7 @@ connect_traces (n_traces, traces)
 
 		  new_bb = copy_bb (best->dest, best, traces[t].last, t);
 		  traces[t].last = new_bb;
-		  if (next_bb != EXIT_BLOCK_PTR)
+		  if (next_bb && next_bb != EXIT_BLOCK_PTR)
 		    {
 		      t = bbd[next_bb->index].start_of_trace;
 		      RBI (traces[last_trace].last)->next = traces[t].first;

gcc/function.c

@@ -452,6 +452,8 @@ free_after_compilation (f)
   f->x_nonlocal_goto_stack_level = NULL;
   f->x_cleanup_label = NULL;
   f->x_return_label = NULL;
+  f->computed_goto_common_label = NULL;
+  f->computed_goto_common_reg = NULL;
   f->x_save_expr_regs = NULL;
   f->x_stack_slot_list = NULL;
   f->x_rtl_expr_chain = NULL;

gcc/function.h

@@ -268,6 +268,10 @@ struct function GTY(())
      on machines which require execution of the epilogue on all returns.  */
   rtx x_return_label;
 
+  /* Label and register for unswitching computed gotos.  */
+  rtx computed_goto_common_label;
+  rtx computed_goto_common_reg;
+
   /* List (chain of EXPR_LISTs) of pseudo-regs of SAVE_EXPRs.
      So we can mark them all live at the end of the function, if nonopt.  */
   rtx x_save_expr_regs;

gcc/stmt.c

@@ -539,10 +539,23 @@ expand_computed_goto (exp)
 #endif
 
   emit_queue ();
-  do_pending_stack_adjust ();
-  emit_indirect_jump (x);
 
-  current_function_has_computed_jump = 1;
+  if (! cfun->computed_goto_common_label)
+    {
+      cfun->computed_goto_common_reg = copy_to_mode_reg (Pmode, x);
+      cfun->computed_goto_common_label = gen_label_rtx ();
+      emit_label (cfun->computed_goto_common_label);
+  
+      do_pending_stack_adjust ();
+      emit_indirect_jump (cfun->computed_goto_common_reg);
+
+      current_function_has_computed_jump = 1;
+    }
+  else
+    {
+      emit_move_insn (cfun->computed_goto_common_reg, x);
+      emit_jump (cfun->computed_goto_common_label);
+    }
 }
 
 /* Handle goto statements and the labels that they can go to.  */