While I was looking into the code generation of PR 122727,
I noticed that TODO_cleanup_cfg could be set from the call to
execute_fixup_cfg even though cleanupcfg did nothing afterwards.
This means the forwarder blocks that were just created with
make_forwarders_with_degenerate_phis are being removed.
Instead of conditionally unsetting TODO_cleanup_cfg,
unconditionally unset TODO_cleanup_cfg after the call
to make_forwarders_with_degenerate_phis. Since we already
did the cleanup (maybe twice).
Bootstrapped and tested on x86_64-linux-gnu.
PR tree-optimization/46555
gcc/ChangeLog:
* tree-cfgcleanup.cc (execute_cleanup_cfg_post_optimizing):
Unconditionally set TODO_cleanup_cfg.
Signed-off-by: Andrew Pinski <[email protected]>
---
gcc/tree-cfgcleanup.cc | 10 ++++++----
1 file changed, 6 insertions(+), 4 deletions(-)
diff --git a/gcc/tree-cfgcleanup.cc b/gcc/tree-cfgcleanup.cc
index 093fde93da6..2521e6d09fd 100644
--- a/gcc/tree-cfgcleanup.cc
+++ b/gcc/tree-cfgcleanup.cc
@@ -1409,10 +1409,7 @@ execute_cleanup_cfg_post_optimizing (void)
{
unsigned int todo = execute_fixup_cfg ();
if (cleanup_tree_cfg ())
- {
- todo &= ~TODO_cleanup_cfg;
- todo |= TODO_update_ssa;
- }
+ todo |= TODO_update_ssa;
maybe_remove_unreachable_handlers ();
cleanup_dead_labels ();
if (group_case_labels () && cleanup_tree_cfg ())
@@ -1423,6 +1420,11 @@ execute_cleanup_cfg_post_optimizing (void)
if (optimize)
make_forwarders_with_degenerate_phis (cfun);
+ /* Make sure todo does not have cleanup cfg as we don't want
+ remove the forwarder blocks we just created. cleanup cfg
+ has already happened. */
+ todo &= ~TODO_cleanup_cfg;
+
basic_block bb = single_succ (ENTRY_BLOCK_PTR_FOR_FN (cfun));
gimple_stmt_iterator gsi = gsi_start_nondebug_after_labels_bb (bb);
/* If the first (and only) bb and the only non debug
--
2.43.0
