On 10/21/11 15:42, Bernd Schmidt wrote:
> On 10/14/11 17:35, Vladimir Makarov wrote:
>> The scheduler part of the patch is ok for me (other part changes are
>> obvious). Could you only commit it at the beginning of the next week.
>
> I've committed this variant. It's updated for some recent changes in trunk:
And this fixlet prevents ports that don't expect REG_DEP_CONTROL from
seeing this type of dependency. Committed as obvious.
Bernd
Index: gcc/ChangeLog
===================================================================
--- gcc/ChangeLog (revision 180309)
+++ gcc/ChangeLog (working copy)
@@ -1,3 +1,9 @@
+2011-10-21 Bernd Schmidt <ber...@codesourcery.com>
+
+ PR bootstrap/50825
+ * sched-deps.c (add_dependence): If not doing predication, promote
+ REG_DEP_CONTROL to REG_DEP_ANTI.
+
2011-10-21 Georg-Johann Lay <a...@gjlay.de>
* config/avr/avr.h (LEGITIMIZE_RELOAD_ADDRESS): Pass address of X
Index: gcc/sched-deps.c
===================================================================
--- gcc/sched-deps.c (revision 180302)
+++ gcc/sched-deps.c (working copy)
@@ -1505,6 +1505,10 @@ sd_debug_lists (rtx insn, sd_list_types_
void
add_dependence (rtx con, rtx pro, enum reg_note dep_type)
{
+ if (dep_type == REG_DEP_CONTROL
+ && !(current_sched_info->flags & DO_PREDICATION))
+ dep_type = REG_DEP_ANTI;
+
/* A REG_DEP_CONTROL dependence may be eliminated through predication,
so we must also make the insn dependent on the setter of the
condition. */
