https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63184
--- Comment #18 from Richard Biener <rguenth at gcc dot gnu.org> ---
That works but it doesn't perform the folding in the end. Still combining
&a + b + c + d + e + 2 into (&a + 2) + b + c + d + e is better since most
targets can compute symbol+offset in a single instruction.
We then end up with
<bb 2> [local count: 1073741824]:
i.0_1 = i;
_2 = i.0_1 * 4;
_3 = (sizetype) _2;
_5 = &MEM[(void *)&a + 8B] + _3;
_13 = _3 + 8;
_7 = &a + _13;
if (_5 != _7)
goto <bb 3>; [53.47%]
else
goto <bb 4>; [46.53%]
where the issue is that the 2nd opportunity for &a + _3 + 8 only appears
after SLSR which transforms
_6 = i.0_1 + 2;
_12 = (sizetype) _6;
_13 = _12 * 4;
_7 = &a + _13;
to
_6 = i.0_1 + 2;
_12 = (sizetype) _6;
_13 = _3 + 8;
_7 = &a + _13;
and we have
NEXT_PASS (pass_reassoc, false /* insert_powi_p */);
NEXT_PASS (pass_strength_reduction);
and I'm not sure it's a good idea to swap those two...
If I do we fold things (albeit quite late).
diff --git a/gcc/tree-ssa-reassoc.c b/gcc/tree-ssa-reassoc.c
index a9f45bfd891..fb1f8014633 100644
--- a/gcc/tree-ssa-reassoc.c
+++ b/gcc/tree-ssa-reassoc.c
@@ -5988,6 +5988,31 @@ reassociate_bb (basic_block bb)
}
}
+ /* If the association chain is used in a single
+ POINTER_PLUS_EXPR with an invariant first operand
+ then combine a constant element with the invariant
+ address. */
+ use_operand_p use_p;
+ gimple *use_stmt;
+ if (ops.length () > 1
+ && rhs_code == PLUS_EXPR
+ && TREE_CODE (ops.last ()->op) == INTEGER_CST
+ && single_imm_use (lhs, &use_p, &use_stmt)
+ && is_gimple_assign (use_stmt)
+ && gimple_assign_rhs_code (use_stmt) == POINTER_PLUS_EXPR
+ && TREE_CODE (gimple_assign_rhs1 (use_stmt)) == ADDR_EXPR)
+ {
+ last = ops.pop ();
+ tree addr = gimple_assign_rhs1 (use_stmt);
+ addr = build1 (ADDR_EXPR, TREE_TYPE (addr),
+ fold_build2 (MEM_REF,
+ TREE_TYPE (TREE_TYPE (addr)),
+ addr,
+ fold_convert (ptr_type_node,
+ last->op)));
+ gimple_assign_set_rhs1 (use_stmt, addr);
+ }
+
tree new_lhs = lhs;
/* If the operand vector is now empty, all operands were
consumed by the __builtin_powi optimization. */
