Starting with r16-4438-ga93f80feeef744, the edges started to be sorted
in ascending order. Thus the most likely branches were deprioritized
instead of optimized to fallthroughs.
Fix by restoring the sorting order prior to r16-4438-ga93f80feeef744.
There are no regressions for C and C++ on x86_64-pc-linux-gnu.
I struggle to write a test case. Suggestions how to check the BB order
are welcome.
Ok for trunk?
PR rtl-optimization/122675
gcc/ChangeLog:
* bb-reorder.cc (edge_order): Fix BB edge ordering to be
descending.
Signed-off-by: Dimitar Dimitrov <[email protected]>
---
gcc/bb-reorder.cc | 7 ++++++-
1 file changed, 6 insertions(+), 1 deletion(-)
diff --git a/gcc/bb-reorder.cc b/gcc/bb-reorder.cc
index e4efdee0b16..43fa019984a 100644
--- a/gcc/bb-reorder.cc
+++ b/gcc/bb-reorder.cc
@@ -2392,7 +2392,12 @@ edge_order (const void *ve1, const void *ve2)
gcov_type gc1 = c1.initialized_p () ? c1.to_gcov_type () : 0;
gcov_type gc2 = c2.initialized_p () ? c2.to_gcov_type () : 0;
gcov_type m = MAX (gc1, gc2);
- return (m == gc1) - (m == gc2);
+ /* While gcc_stablesort sorts in ascending order, the edges should
+ be sorted in descending order of their execution frequency.
+ So return a reversed comparison. Expressed with a spaceship operator:
+ return gc2 <=> gc1;
+ */
+ return (m == gc2) - (m == gc1);
}
/* Reorder basic blocks using the "simple" algorithm. This tries to
--
2.51.1
