https://gcc.gnu.org/bugzilla/show_bug.cgi?id=89150
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|UNCONFIRMED |NEW
Last reconfirmed| |2019-02-01
Ever confirmed|0 |1
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
The marking looks inefficient to me anyhow...
I guess best would be to have user markers for bitmap_head instead and
GTY skip bitmap_element alltogether. That would make it invalid to
hold on bitmap_elements but not its bitmap_head.
void
gt_ggc_mx_bitmap_head (void *x_p)
{
struct bitmap_head * const x = (struct bitmap_head *)x_p;
if (ggc_test_and_set_mark (x))
{
gt_ggc_m_14bitmap_element ((*x).first);
gt_ggc_m_14bitmap_obstack ((*x).obstack);
}
}
void
gt_ggc_mx_bitmap_element (void *x_p)
{
struct bitmap_element * x = (struct bitmap_element *)x_p;
struct bitmap_element * xlimit = x;
while (ggc_test_and_set_mark (xlimit))
xlimit = ((*xlimit).next);
if (x != xlimit)
for (;;)
{
struct bitmap_element * const xprev = ((*x).prev);
if (xprev == NULL) break;
x = xprev;
(void) ggc_test_and_set_mark (xprev);
}
while (x != xlimit)
{
gt_ggc_m_14bitmap_element ((*x).next);
gt_ggc_m_14bitmap_element ((*x).prev);
x = ((*x).next);
}
}
bitmap_obstack is also somewhat weird in that the obstack itself is
skipped but elements/heads is not!? GC allocated bitmaps have a NULL
obstack pointer and free elements are cached in the global
bitmap_ggc_free root.
Anyhow... lots of improvement possibilties I guess. Not sure if
one can skip bitmap_element and instruct GC to walk head->first->next/prev
as list w/o custom walkers.
