store-merging: Fix up a -fcompare-debug bug in get_status_for_store_merging [PR104263]
Commit Message
Hi!
As mentioned in the PRthe following testcase fails, because the last
stmt of a bb with -g is a debug stmt and get_status_for_store_merging
uses gimple_seq_last_stmt (bb_seq (bb)) when testing if it is valid
for store merging. The debug stmt isn't valid, while a stmt at that
position with -g0 is valid and so the divergence.
As we walk the whole bb already, this patch just remembers the last
non-debug stmt, so that we don't need to skip backwards debug stmts at the
end of the bb to find last real stmt.
Bootstrapped/regtested on powerpc64le-linux, ok for trunk?
2022-01-28 Jakub Jelinek <jakub@redhat.com>
PR tree-optimization/104263
* gimple-ssa-store-merging.cc (get_status_for_store_merging): For
cfun->can_throw_non_call_exceptions && cfun->eh test whether
last non-debug stmt in the bb is store_valid_for_store_merging_p
rather than last stmt.
* gcc.dg/pr104263.c: New test.
Jakub
Comments
On 1/28/2022 9:31 AM, Jakub Jelinek via Gcc-patches wrote:
> Hi!
>
> As mentioned in the PRthe following testcase fails, because the last
> stmt of a bb with -g is a debug stmt and get_status_for_store_merging
> uses gimple_seq_last_stmt (bb_seq (bb)) when testing if it is valid
> for store merging. The debug stmt isn't valid, while a stmt at that
> position with -g0 is valid and so the divergence.
>
> As we walk the whole bb already, this patch just remembers the last
> non-debug stmt, so that we don't need to skip backwards debug stmts at the
> end of the bb to find last real stmt.
>
> Bootstrapped/regtested on powerpc64le-linux, ok for trunk?
>
> 2022-01-28 Jakub Jelinek <jakub@redhat.com>
>
> PR tree-optimization/104263
> * gimple-ssa-store-merging.cc (get_status_for_store_merging): For
> cfun->can_throw_non_call_exceptions && cfun->eh test whether
> last non-debug stmt in the bb is store_valid_for_store_merging_p
> rather than last stmt.
>
> * gcc.dg/pr104263.c: New test.
OK
jeff
> Am 28.01.2022 um 18:36 schrieb Jeff Law <jeffreyalaw@gmail.com>:
>
>
>
>> On 1/28/2022 9:31 AM, Jakub Jelinek via Gcc-patches wrote:
>> Hi!
>>
>> As mentioned in the PRthe following testcase fails, because the last
>> stmt of a bb with -g is a debug stmt and get_status_for_store_merging
>> uses gimple_seq_last_stmt (bb_seq (bb)) when testing if it is valid
>> for store merging. The debug stmt isn't valid, while a stmt at that
>> position with -g0 is valid and so the divergence.
>>
>> As we walk the whole bb already, this patch just remembers the last
>> non-debug stmt, so that we don't need to skip backwards debug stmts at the
>> end of the bb to find last real stmt.
>>
>> Bootstrapped/regtested on powerpc64le-linux, ok for trunk
Ok
Thanks,
Richard
>> 2022-01-28 Jakub Jelinek <jakub@redhat.com>
>>
>> PR tree-optimization/104263
>> * gimple-ssa-store-merging.cc (get_status_for_store_merging): For
>> cfun->can_throw_non_call_exceptions && cfun->eh test whether
>> last non-debug stmt in the bb is store_valid_for_store_merging_p
>> rather than last stmt.
>>
>> * gcc.dg/pr104263.c: New test.
> OK
> jeff
>
@@ -5364,6 +5364,7 @@ get_status_for_store_merging (basic_bloc
unsigned int num_constructors = 0;
gimple_stmt_iterator gsi;
edge e;
+ gimple *last_stmt = NULL;
for (gsi = gsi_after_labels (bb); !gsi_end_p (gsi); gsi_next (&gsi))
{
@@ -5372,6 +5373,8 @@ get_status_for_store_merging (basic_bloc
if (is_gimple_debug (stmt))
continue;
+ last_stmt = stmt;
+
if (store_valid_for_store_merging_p (stmt) && ++num_statements >= 2)
break;
@@ -5398,7 +5401,7 @@ get_status_for_store_merging (basic_bloc
return BB_INVALID;
if (cfun->can_throw_non_call_exceptions && cfun->eh
- && store_valid_for_store_merging_p (gimple_seq_last_stmt (bb_seq (bb)))
+ && store_valid_for_store_merging_p (last_stmt)
&& (e = find_fallthru_edge (bb->succs))
&& e->dest == bb->next_bb)
return BB_EXTENDED_VALID;
@@ -0,0 +1,25 @@
+/* PR tree-optimization/104263 */
+/* { dg-do compile } */
+/* { dg-options "-O2 -fcompare-debug -fnon-call-exceptions -fno-inline-small-functions" } */
+
+int n;
+
+int
+bar (void)
+{
+ int a;
+
+ n = 0;
+ a = 0;
+
+ return n;
+}
+
+__attribute__ ((pure, returns_twice)) int
+foo (void)
+{
+ n = bar () + 1;
+ foo ();
+
+ return 0;
+}