store-merging: Avoid ICEs on roughly ~0ULL/8 sized stores [PR105094]

On the following testcase on 64-bit targets, store-merging sees
a MEM_REF store from {} ctor with "negative" bitsize where bitoff + bitsize
wraps around to very small end offset.  This later confuses the code
so that it allocates just a few bytes of memory but fills in huge amounts of
it.  Later on there is a param_store_merging_max_size size check but due to
the wrap-around we pass that.

The following patch punts on such large bitsizes.

2022-03-30  Jakub Jelinek  <jakub@redhat.com>

	PR tree-optimization/105094
	* gimple-ssa-store-merging.cc (mem_valid_for_store_merging): Punt if
	bitsize <= 0 rather than just == 0.

	* gcc.dg/pr105094.c: New test.

diff --git a/gcc/gimple-ssa-store-merging.cc b/gcc/gimple-ssa-store-merging.cc
index e2e2157..b952ce5 100644 (file)
--- a/gcc/gimple-ssa-store-merging.cc
+++ b/gcc/gimple-ssa-store-merging.cc
@@ -4940,7 +4940,7 @@ mem_valid_for_store_merging (tree mem, poly_uint64 *pbitsize,
   tree base_addr = get_inner_reference (mem, &bitsize, &bitpos, &offset, &mode,
                                        &unsignedp, &reversep, &volatilep);
   *pbitsize = bitsize;
-  if (known_eq (bitsize, 0))
+  if (known_le (bitsize, 0))
     return NULL_TREE;
 
   if (TREE_CODE (mem) == COMPONENT_REF

diff --git a/gcc/testsuite/gcc.dg/pr105094.c b/gcc/testsuite/gcc.dg/pr105094.c
new file mode 100644 (file)
index 0000000..da6dc17
--- /dev/null
+++ b/gcc/testsuite/gcc.dg/pr105094.c
@@ -0,0 +1,13 @@
+/* PR tree-optimization/105094 */
+/* { dg-do compile } */
+/* { dg-options "-O2" } */
+
+struct S { short a; char b[~(__SIZE_TYPE__)0 / __CHAR_BIT__ - 1]; };
+void bar (struct S *);
+
+void
+foo (void)
+{
+  struct S s = { 5 };
+  bar (&s);
+}