ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff errors when…

ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff errors when comparing specialized and unspecialized times.


	* ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
	errors when comparing specialized and unspecialized times.

From-SVN: r247528

gcc/ChangeLog

+2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
+
+	* ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
+	errors when comparing specialized and unspecialized times.
+
 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
 
 	* diagnostic-show-locus.c
@@ -356,6 +361,11 @@
 
 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
 
+	* ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
+	overflow check.
+
+2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
+
 	PR ipa/79224
 	* ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
 	(account_size_time): Use two predicates - exec_pred and

gcc/ipa-inline-analysis.c

@@ -3422,7 +3422,15 @@ estimate_node_size_and_time (struct cgraph_node *node,
   min_size = (*info->entry)[0].size;
   gcc_checking_assert (size >= 0);
   gcc_checking_assert (time >= 0);
-  gcc_checking_assert (nonspecialized_time >= time);
+  /* nonspecialized_time should be always bigger than specialized time.
+     Roundoff issues however may get into the way.  */
+  gcc_checking_assert ((nonspecialized_time - time) >= -1);
+
+  /* Roundoff issues may make specialized time bigger than nonspecialized
+     time.  We do not really want that to happen because some heurstics
+     may get confused by seeing negative speedups.  */
+  if (time > nonspecialized_time)
+    time = nonspecialized_time;
 
   if (info->loop_iterations
       && !evaluate_predicate (info->loop_iterations, possible_truths))