The following plugs possible overflow issues in vn_reference_compute_hash
and possibly in vn_reference_eq. The inchash "integer" adds are a bit
of a mess, but I know overloads with different integer types can get
messy, so not this time. For hashing simply truncate to 64bits.
Bootstrap and regtest running on x86_64-unknown-linux-gnu.
PR tree-optimization/90242
* tree-ssa-sccvn.cc (vn_reference_compute_hash): Use
poly_offset_int for offset accumulation. For hashing
truncate to 64 bits and also hash 64 bits.
(vn_reference_eq): Likewise.
---
gcc/tree-ssa-sccvn.cc | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/gcc/tree-ssa-sccvn.cc b/gcc/tree-ssa-sccvn.cc
index a3117dacb96..3974c4d4583 100644
--- a/gcc/tree-ssa-sccvn.cc
+++ b/gcc/tree-ssa-sccvn.cc
@@ -717,7 +717,7 @@ vn_reference_compute_hash (const vn_reference_t vr1)
hashval_t result;
int i;
vn_reference_op_t vro;
- poly_int64 off = -1;
+ poly_offset_int off = -1;
bool deref = false;
FOR_EACH_VEC_ELT (vr1->operands, i, vro)
@@ -736,7 +736,7 @@ vn_reference_compute_hash (const vn_reference_t vr1)
{
if (maybe_ne (off, -1)
&& maybe_ne (off, 0))
- hstate.add_poly_int (off);
+ hstate.add_poly_hwi (off.force_shwi ());
off = -1;
if (deref
&& vro->opcode == ADDR_EXPR)
@@ -850,7 +850,7 @@ vn_reference_eq (const_vn_reference_t const vr1,
const_vn_reference_t const vr2)
j = 0;
do
{
- poly_int64 off1 = 0, off2 = 0;
+ poly_offset_int off1 = 0, off2 = 0;
vn_reference_op_t vro1, vro2;
vn_reference_op_s tem1, tem2;
bool deref1 = false, deref2 = false;
--
2.43.0
