public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc r13-4317] tree-optimization/107876 - unswitching of switch
@ 2022-11-28  9:05 Richard Biener
  0 siblings, 0 replies; only message in thread
From: Richard Biener @ 2022-11-28  9:05 UTC (permalink / raw)
  To: gcc-cvs

https://gcc.gnu.org/g:238cf114de16518c6569f0cdd2c4d6211ac98a74

commit r13-4317-g238cf114de16518c6569f0cdd2c4d6211ac98a74
Author: Richard Biener <rguenther@suse.de>
Date:   Mon Nov 28 09:19:33 2022 +0100

    tree-optimization/107876 - unswitching of switch
    
    The following shows a missed update of dominators when unswitching
    removes unreachable edges from switch stmts it unswitches.  Fixed
    by wiping dominator info in that case.
    
            PR tree-optimization/107876
            * tree-ssa-loop-unswitch.cc (clean_up_after_unswitching): Wipe
            dominator info if we removed an edge.
    
            * g++.dg/tree-ssa/pr107876.C: New testcase.

Diff:
---
 gcc/testsuite/g++.dg/tree-ssa/pr107876.C | 38 ++++++++++++++++++++++++++++++++
 gcc/tree-ssa-loop-unswitch.cc            | 10 ++++++++-
 2 files changed, 47 insertions(+), 1 deletion(-)

diff --git a/gcc/testsuite/g++.dg/tree-ssa/pr107876.C b/gcc/testsuite/g++.dg/tree-ssa/pr107876.C
new file mode 100644
index 00000000000..6cff2f1b166
--- /dev/null
+++ b/gcc/testsuite/g++.dg/tree-ssa/pr107876.C
@@ -0,0 +1,38 @@
+// { dg-do compile }
+// { dg-require-effective-target c++11 }
+// { dg-options "-O2 -funswitch-loops --param max-unswitch-insns=5 -fdump-tree-unswitch-details" }
+
+class X {
+public:
+  X();
+  X(const X&);
+  X(const volatile X &);
+  ~X();
+};
+
+X test17(int i) {
+  if (false) {
+  impossible:
+    if (i == 3)
+      return X();
+  }
+
+  while (true) {
+    X x;
+    if (i == 0)
+      return x;
+    if (i == 1)
+      break;
+    if (i == 2)
+      continue;
+    if (i == 3)
+      goto impossible;
+    if (i == 4)
+      __builtin_exit(1);
+    if (i == 5)
+      return x;
+  }
+  return X();
+}
+
+// { dg-final { scan-tree-dump "unswitching loop 1 on .switch. with condition: i_\[0-9\]+\\(D\\) == 2" "unswitch" } }
diff --git a/gcc/tree-ssa-loop-unswitch.cc b/gcc/tree-ssa-loop-unswitch.cc
index 186ae953f04..e8c9bd6812a 100644
--- a/gcc/tree-ssa-loop-unswitch.cc
+++ b/gcc/tree-ssa-loop-unswitch.cc
@@ -1631,6 +1631,7 @@ clean_up_after_unswitching (int ignored_edge_flag)
   basic_block bb;
   edge e;
   edge_iterator ei;
+  bool removed_edge = false;
 
   FOR_EACH_BB_FN (bb, cfun)
     {
@@ -1655,7 +1656,10 @@ clean_up_after_unswitching (int ignored_edge_flag)
 		     to preserve its edge.  But we can remove the
 		     non-default CASE sharing the edge.  */
 		  if (e != default_e)
-		    remove_edge (e);
+		    {
+		      remove_edge (e);
+		      removed_edge = true;
+		    }
 		}
 	      else
 		{
@@ -1672,6 +1676,10 @@ clean_up_after_unswitching (int ignored_edge_flag)
       FOR_EACH_EDGE (e, ei, bb->succs)
 	e->flags &= ~ignored_edge_flag;
     }
+
+  /* If we removed an edge we possibly have to recompute dominators.  */
+  if (removed_edge)
+    free_dominance_info (CDI_DOMINATORS);
 }
 
 /* Loop unswitching pass.  */

^ permalink raw reply	[flat|nested] only message in thread

only message in thread, other threads:[~2022-11-28  9:05 UTC | newest]

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-11-28  9:05 [gcc r13-4317] tree-optimization/107876 - unswitching of switch Richard Biener

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).