https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87169
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEW |ASSIGNED
Assignee|unassigned at gcc dot gnu.org |rguenth at gcc dot
gnu.org
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Mine as well then. Confirmed even with the last fix.
We're doing
Replaced redundant PHI node defining d_14 with d_20(D)
...
Removing dead stmt d_14 = PHI <d_22(6), d_20(D)(2)>
but the LHS d_14 is still used...
Block 19: BB6 found not executable
that's the block with said stmt - so we fail to eliminate it. That's because
the block is reachable via the irreducible region but we marked the block
as not need iterating.
Ah, this is caused by the loop depth limiting, adjusting --param
rpo-vn-max-loop-depth helps.
Will investigate early next week.
