This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/87985] Compile-time and memory hog w/ -O1 -ftree-slp-vectorize
- From: "rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Mon, 12 Nov 2018 13:23:49 +0000
- Subject: [Bug tree-optimization/87985] Compile-time and memory hog w/ -O1 -ftree-slp-vectorize
- Auto-submitted: auto-generated
- References: <bug-87985-4@http.gcc.gnu.org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87985
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
diff --git a/gcc/tree-data-ref.c b/gcc/tree-data-ref.c
index 6019c6168bf..d60d389fa0a 100644
--- a/gcc/tree-data-ref.c
+++ b/gcc/tree-data-ref.c
@@ -682,7 +684,8 @@ split_constant_offset_1 (tree type, tree op0, enum
tree_code code, tree op1,
case SSA_NAME:
{
- if (SSA_NAME_OCCURS_IN_ABNORMAL_PHI (op0))
+ if (SSA_NAME_OCCURS_IN_ABNORMAL_PHI (op0)
+ || !has_single_use (op0))
return false;
gimple *def_stmt = SSA_NAME_DEF_STMT (op0);
fixes this but that's very likely too restrictive. We are merely interested
to limit multiple uses in the expression itself as we'd otherwise un-CSE
those. A full-blown hash-map would do the trick here I guess. But at
what expense?