Delete useless .split2 dump file at -O0 on non-x86 targets

[Eric Botcazou]

Doing -fdump-rtl-all at -O0 on a non-x86 target yields a .split2 file which is 
mightily confusing since nothing is split, so you can easily burn a dozen of 
minutes trying to pinpoint the first split point after reload...

Tested on x86-64/Linux and SPARC/Solaris, applied on the mainline.


2016-09-16  Eric Botcazou  <ebotca...@adacore.com>

        * recog.c (rest_of_handle_split_after_reload): Delete.
        (pass_split_after_reload::gate): New method.
        (pass_split_after_reload::execute): Call split_all_insns directly.

-- 
Eric Botcazou
Index: recog.c
===================================================================
--- recog.c	(revision 240142)
+++ recog.c	(working copy)
@@ -3901,17 +3901,6 @@ make_pass_split_all_insns (gcc::context
   return new pass_split_all_insns (ctxt);
 }
 
-static unsigned int
-rest_of_handle_split_after_reload (void)
-{
-  /* If optimizing, then go ahead and split insns now.  */
-#ifndef STACK_REGS
-  if (optimize > 0)
-#endif
-    split_all_insns ();
-  return 0;
-}
-
 namespace {
 
 const pass_data pass_data_split_after_reload =
@@ -3935,9 +3924,23 @@ public:
   {}
 
   /* opt_pass methods: */
+  virtual bool gate (function *)
+    {
+      /* If optimizing, then go ahead and split insns now.  */
+      if (optimize > 0)
+	return true;
+
+#ifdef STACK_REGS
+      return true;
+#else
+      return false;
+#endif
+    }
+
   virtual unsigned int execute (function *)
     {
-      return rest_of_handle_split_after_reload ();
+      split_all_insns ();
+      return 0;
     }
 
 }; // class pass_split_after_reload