On 11/19/21 21:48, Jakub Jelinek wrote:
On Wed, Nov 10, 2021 at 12:31:28AM +0530, Siddhesh Poyarekar wrote:
Put all accesses to object_sizes behind functions so that we can add
dynamic capability more easily.
gcc/ChangeLog:
* tree-object-size.c (object_sizes_grow, object_sizes_release,
object_sizes_unknown_p, object_sizes_get, object_size_set_force,
object_sizes_set): New functions.
(addr_object_size, compute_builtin_object_size,
expr_object_size, call_object_size, unknown_object_size,
merge_object_sizes, plus_stmt_object_size,
cond_expr_object_size, collect_object_sizes_for,
check_for_plus_in_loops_1, init_object_sizes,
fini_object_sizes): Adjust.
@@ -975,8 +994,9 @@ collect_object_sizes_for (struct object_size_info *osi,
tree var)
{
if (bitmap_set_bit (osi->visited, varno))
{
- object_sizes[object_size_type][varno]
- = (object_size_type & 2) ? -1 : 0;
+ /* Initialize to 0 for maximum size and M1U for minimum size so that
+ it gets immediately overridden. */
object_sizes_set_force (osi, varno, unknown (object_size_type ^ 2));
Shouldn't that be unknown (object_size_type
^ OST_MINIMUM)
?
Or, if you redo the series, update the first patch so that it doesn't leave
any & 2 or & 1 etc. around and instead uses the enumerators and redo this
patch?
Thanks, I'll update 1/10 and redo this on top of it.
Siddhesh
