Hi!
As discussed in the PR, the problem here is that the routines changed in
this patch sign extend the difference of index and low_bound from the
precision of the index, so e.g. when index is unsigned int and contains
value -2U, we treat it as index -2 rather than 0x00000000fffffffeU on 64-bit
arches.
On the other hand, get_inner_reference which is used during expansion, does:
if (! integer_zerop (low_bound))
index = fold_build2 (MINUS_EXPR, TREE_TYPE (index),
index, low_bound);
offset = size_binop (PLUS_EXPR, offset,
size_binop (MULT_EXPR,
fold_convert (sizetype, index),
unit_size));
which effectively requires that either low_bound is constant 0 and then
index in ARRAY_REFs can be arbitrary type which is then sign or zero
extended to sizetype, or low_bound is something else and then index and
low_bound must have compatible types and it is still converted afterwards to
sizetype and from there then a few lines later:
expr.c- if (poly_int_tree_p (offset))
expr.c- {
expr.c: poly_offset_int tem = wi::sext (wi::to_poly_offset (offset),
expr.c- TYPE_PRECISION (sizetype));
The following patch makes those routines match what get_inner_reference is
doing.
Bootstrapped/regtested on x86_64-linux (as usually, including Ada) and
i686-linux, ok for trunk?
2021-01-22 Jakub Jelinek <ja...@redhat.com>
PR tree-optimization/98255
* tree-dfa.c (get_ref_base_and_extent): For ARRAY_REFs, sign
extend index - low_bound from sizetype's precision rather than index
precision.
(get_addr_base_and_unit_offset_1): Likewise.
* tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Likewise.
* gimple-fold.c (fold_const_aggregate_ref_1): Likewise.
* gcc.dg/pr98255.c: New test.
--- gcc/tree-dfa.c.jj 2021-01-04 10:25:37.220252089 +0100
+++ gcc/tree-dfa.c 2021-01-21 16:43:33.302962080 +0100
@@ -503,7 +503,7 @@ get_ref_base_and_extent (tree exp, poly_
poly_offset_int woffset
= wi::sext (wi::to_poly_offset (index)
- wi::to_poly_offset (low_bound),
- TYPE_PRECISION (TREE_TYPE (index)));
+ TYPE_PRECISION (sizetype));
woffset *= wi::to_offset (unit_size);
woffset <<= LOG2_BITS_PER_UNIT;
bit_offset += woffset;
@@ -564,7 +564,7 @@ get_ref_base_and_extent (tree exp, poly_
{
poly_offset_int woffset
= wi::sext (omin - lbound,
- TYPE_PRECISION (TREE_TYPE (index)));
+ TYPE_PRECISION (sizetype));
woffset *= wi::to_offset (unit_size);
woffset <<= LOG2_BITS_PER_UNIT;
bit_offset += woffset;
@@ -822,7 +822,7 @@ get_addr_base_and_unit_offset_1 (tree ex
poly_offset_int woffset
= wi::sext (wi::to_poly_offset (index)
- wi::to_poly_offset (low_bound),
- TYPE_PRECISION (TREE_TYPE (index)));
+ TYPE_PRECISION (sizetype));
woffset *= wi::to_offset (unit_size);
byte_offset += woffset.force_shwi ();
}
--- gcc/tree-ssa-sccvn.c.jj 2021-01-13 11:22:30.807345423 +0100
+++ gcc/tree-ssa-sccvn.c 2021-01-21 16:46:52.780699466 +0100
@@ -1108,7 +1108,7 @@ ao_ref_init_from_vn_reference (ao_ref *r
poly_offset_int woffset
= wi::sext (wi::to_poly_offset (op->op0)
- wi::to_poly_offset (op->op1),
- TYPE_PRECISION (TREE_TYPE (op->op0)));
+ TYPE_PRECISION (sizetype));
woffset *= wi::to_offset (op->op2) * vn_ref_op_align_unit (op);
woffset <<= LOG2_BITS_PER_UNIT;
offset += woffset;
--- gcc/gimple-fold.c.jj 2021-01-05 09:17:24.754566798 +0100
+++ gcc/gimple-fold.c 2021-01-21 16:45:58.960309941 +0100
@@ -8007,7 +8007,7 @@ fold_const_aggregate_ref_1 (tree t, tree
poly_offset_int woffset
= wi::sext (wi::to_poly_offset (idx)
- wi::to_poly_offset (low_bound),
- TYPE_PRECISION (TREE_TYPE (idx)));
+ TYPE_PRECISION (sizetype));
woffset *= tree_to_uhwi (unit_size);
woffset *= BITS_PER_UNIT;
if (woffset.to_shwi (&offset))
--- gcc/testsuite/gcc.dg/pr98255.c.jj 2021-01-21 16:54:39.258416413 +0100
+++ gcc/testsuite/gcc.dg/pr98255.c 2021-01-21 16:54:11.531730111 +0100
@@ -0,0 +1,49 @@
+/* PR tree-optimization/98255 */
+/* { dg-do run } */
+/* { dg-options "-Os" } */
+/* { dg-additional-options "-fPIC" { target fpic } } */
+
+struct A { volatile unsigned b; unsigned c; };
+int d, *e, h, k, l;
+static struct A f;
+long g;
+static unsigned i = -2U;
+volatile int j;
+
+long
+foo (void)
+{
+ char n[4][4][3]
+ = { { {9, 2, 8}, {9, 2, 8}, {9, 2, 8}, {9} }, { {8} }, { {8} }, { {2} } };
+ while (d)
+ {
+ for (; f.c < 4; f.c++)
+ {
+ *e = 0;
+ h = n[f.c + 4][0][d];
+ }
+ while (g)
+ return n[0][3][i];
+ while (1)
+ {
+ if (k)
+ {
+ j = 0;
+ if (j)
+ continue;
+ }
+ if (l)
+ break;
+ }
+ }
+ return 0;
+}
+
+int
+main ()
+{
+ asm volatile ("" : "+g" (d), "+g" (g), "+g" (f.c));
+ asm volatile ("" : "+g" (e), "+g" (k), "+g" (l));
+ foo ();
+ return 0;
+}
Jakub
