Re: [PATCH 3/12] phiopt: Fix phiopt ICE on vops [PR102989]

[Jakub Jelinek <jakub@redhat.com>]

On Wed, Aug 09, 2023 at 01:06:22PM -0700, Andrew Pinski wrote:
> That was my next thought, returning NULL from
> single_non_singleton_phi_for_edges if it would return a virtual OP
> might be better even.
> Either version of these patches are ok with me (though I am not the
> maintainer here).

In patch form that would be (but so far untested):

2023-08-09  <jakub@redhat.com>

	PR c/102989
	* tree-ssa-phiopt.cc (single_non_singleton_phi_for_edges): Never
	return virtual phis and return NULL if there is a virtual phi
	where the arguments from E0 and E1 edges aren't equal.

--- gcc/tree-ssa-phiopt.cc.jj	2023-08-09 22:08:07.974563266 +0200
+++ gcc/tree-ssa-phiopt.cc	2023-08-09 22:11:37.291517911 +0200
@@ -63,7 +63,13 @@ single_non_singleton_phi_for_edges (gimp
   gimple_stmt_iterator i;
   gphi *phi = NULL;
   if (gimple_seq_singleton_p (seq))
-    return as_a <gphi *> (gsi_stmt (gsi_start (seq)));
+    {
+      phi = as_a <gphi *> (gsi_stmt (gsi_start (seq)));
+      /* Never return virtual phis.  */
+      if (virtual_operand_p (gimple_phi_result (phi)))
+	return NULL;
+      return phi;
+    }
   for (i = gsi_start (seq); !gsi_end_p (i); gsi_next (&i))
     {
       gphi *p = as_a <gphi *> (gsi_stmt (i));
@@ -72,6 +78,10 @@ single_non_singleton_phi_for_edges (gimp
 				       gimple_phi_arg_def (p, e1->dest_idx)))
 	continue;
 
+      /* Punt on virtual phis with different arguments from the edges.  */
+      if (virtual_operand_p (gimple_phi_result (p)))
+	return NULL;
+
       /* If we already have a PHI that has the two edge arguments are
 	 different, then return it is not a singleton for these PHIs. */
       if (phi)


	Jakub