public inbox for gcc-patches@gcc.gnu.org
 help / color / mirror / Atom feed
From: Richard Biener <rguenther@suse.de>
To: gcc-patches@gcc.gnu.org
Subject: [PATCH] tree-optimization/113707 - ICE with VN elimination
Date: Mon, 5 Feb 2024 11:05:52 +0100 (CET)	[thread overview]
Message-ID: <20240205100552.xSyD2W5RBrN0j9nYDLZKO6slIVclWni32E78LYbUXek@z> (raw)

The following avoids different avail answers depending on how the
iteration progressed.

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

	PR tree-optimization/113707
	* tree-ssa-sccvn.cc (rpo_elim::eliminate_avail): After
	checking the avail set treat out-of-region defines as
	available.

	* gcc.dg/torture/pr113707-1.c: New testcase.
	* gcc.dg/torture/pr113707-2.c: Likewise.
---
 gcc/testsuite/gcc.dg/torture/pr113707-1.c | 45 +++++++++++++++++++++++
 gcc/testsuite/gcc.dg/torture/pr113707-2.c | 26 +++++++++++++
 gcc/tree-ssa-sccvn.cc                     |  5 +++
 3 files changed, 76 insertions(+)
 create mode 100644 gcc/testsuite/gcc.dg/torture/pr113707-1.c
 create mode 100644 gcc/testsuite/gcc.dg/torture/pr113707-2.c

diff --git a/gcc/testsuite/gcc.dg/torture/pr113707-1.c b/gcc/testsuite/gcc.dg/torture/pr113707-1.c
new file mode 100644
index 00000000000..c1a50b31025
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/pr113707-1.c
@@ -0,0 +1,45 @@
+/* { dg-do compile } */
+
+int printf(const char *, ...);
+struct a {
+  int b;
+} n;
+int a, c, d, e, f = 1, g, h, j = 1, k, l, m, o;
+int main() {
+  struct a p;
+  int i;
+  p.b = 1;
+  if (!j)
+    goto q;
+  p.b = i = 0;
+  for (; i < 1; i++)
+    if (k)
+      while (m)
+      r:
+      q:
+        if (p.b)
+          g = 1;
+  while (1) {
+    i = 0;
+    for (; i < 5; i++)
+      ;
+    if (l) {
+      while (h)
+        ;
+      if (o) {
+        d = 0;
+        for (; d < 8; d++)
+          ;
+      }
+    }
+    for (; e; e--)
+      while (a)
+        p = n;
+    if (c)
+      goto r;
+    printf("0");
+    if (f)
+      break;
+  }
+  return 0;
+}
diff --git a/gcc/testsuite/gcc.dg/torture/pr113707-2.c b/gcc/testsuite/gcc.dg/torture/pr113707-2.c
new file mode 100644
index 00000000000..957e6f1b534
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/torture/pr113707-2.c
@@ -0,0 +1,26 @@
+/* { dg-do compile } */
+
+int a, b, c, d, e, f, g, h, j, k, l;
+void n() {
+  while (c)
+    if (1) {
+      for (h = 5; h; h--) {
+        int m = e % 2;
+        d = ~g || h ^ m / -1;
+        if (h > 5)
+          e = k;
+      }
+      return;
+    }
+}
+int main() {
+  if (a)
+    for (int i = 0; i < 2; i++) {
+      for (f = 1; f < 6; f++)
+        for (c = 7; c >= 0; c--)
+          if (l)
+            b = 0;
+      n();
+    }
+  return 0;
+}
diff --git a/gcc/tree-ssa-sccvn.cc b/gcc/tree-ssa-sccvn.cc
index bbcf86588f9..8792cd07901 100644
--- a/gcc/tree-ssa-sccvn.cc
+++ b/gcc/tree-ssa-sccvn.cc
@@ -7776,6 +7776,11 @@ rpo_elim::eliminate_avail (basic_block bb, tree op)
 	  av = av->next;
 	}
       while (av);
+      /* While we prefer avail we have to fallback to using the value
+	 directly if defined outside of the region when none of the
+	 available defs suit.  */
+      if (!valnum_info->visited)
+	return valnum;
     }
   else if (valnum != VN_TOP)
     /* valnum is is_gimple_min_invariant.  */
-- 
2.35.3

                 reply	other threads:[~2024-02-05 10:05 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240205100552.xSyD2W5RBrN0j9nYDLZKO6slIVclWni32E78LYbUXek@z \
    --to=rguenther@suse.de \
    --cc=gcc-patches@gcc.gnu.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
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).