diff mbox series

tree-optimization/106198 - CFG cleanup vs LC SSA

Message ID 20220705140423.4E16C3856080@sourceware.org
State New
Headers show
Series tree-optimization/106198 - CFG cleanup vs LC SSA | expand

Commit Message

Richard Biener July 5, 2022, 2:03 p.m. UTC 
This is another case like PR106182 where for the 2nd testcase in
the bug there are no removed or discovered loops but still changing
loop exits invalidates LC SSA and it is not enough to just scan for
uses in the blocks that changed loop depth.  One might argue that
if we'd include former exit destinations we'd pick up the original
LC SSA use but for virtuals on block merging we'd have propagated
those out (while for regular uses we insert copies).  CFG cleanup
can also be entered with loops needing fixup so any heuristics
based on loop structure are bound to fail.

Boostrapped and tested on x86_64-unknown-linux-gnu, pushed.

	PR tree-optimization/106198
	* tree-cfgcleanup.cc (repair_loop_structures): Always do a
	full LC SSA rewrite but only if any blocks changed loop
	depth.

	* gcc.dg/pr106198.c: New testcase.
---
 gcc/testsuite/gcc.dg/pr106198.c | 22 ++++++++++++++++++++++
 gcc/tree-cfgcleanup.cc          |  6 +++---
 2 files changed, 25 insertions(+), 3 deletions(-)
 create mode 100644 gcc/testsuite/gcc.dg/pr106198.c
diff mbox series

Patch

diff --git a/gcc/testsuite/gcc.dg/pr106198.c b/gcc/testsuite/gcc.dg/pr106198.c
new file mode 100644
index 00000000000..00d2758efa7
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr106198.c
@@ -0,0 +1,22 @@ 
+/* { dg-do compile } */
+/* { dg-options "-O3" } */
+
+int printf(const char *, ...);
+long a;
+int b;
+volatile int c;
+int main() {
+  long e = a;
+  int f = a;
+ L:
+  if (b > 0) {
+    printf("0");
+    goto L;
+  }
+  if (f) {
+    printf("%ld", (long)b);
+    goto L;
+  }
+  e >= b && c;
+  return 0;
+}
diff --git a/gcc/tree-cfgcleanup.cc b/gcc/tree-cfgcleanup.cc
index 3b24e021b6b..b9ff6896ce6 100644
--- a/gcc/tree-cfgcleanup.cc
+++ b/gcc/tree-cfgcleanup.cc
@@ -1183,9 +1183,9 @@  repair_loop_structures (void)
      become unreachable by back edges from latch).  Also a former
      irreducible loop can become reducible - in this case force a full
      rewrite into loop-closed SSA form.  */
-  if (loops_state_satisfies_p (LOOP_CLOSED_SSA))
-    rewrite_into_loop_closed_ssa (n_new_or_deleted_loops ? NULL : changed_bbs,
-				  TODO_update_ssa);
+  if (loops_state_satisfies_p (LOOP_CLOSED_SSA)
+      && (!bitmap_empty_p (changed_bbs) || n_new_or_deleted_loops))
+    rewrite_into_loop_closed_ssa (NULL, TODO_update_ssa);
 
   BITMAP_FREE (changed_bbs);