https://gcc.gnu.org/bugzilla/show_bug.cgi?id=46590
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Known to fail|16.0 |14.3.0
Target Milestone|14.4 |---
Summary|[14 Regression] long |long compile time with -O2
|compile time with -O2 and |and many loops
|many loops |
Known to work| |4.4.7
--- Comment #62 from Richard Biener <rguenth at gcc dot gnu.org> ---
Meh, instead of making this one a regression in 14 you should have opened a new
bug for this. It's probably a ranger thing that wasn't backported (and IMO
it's a bit late for this as well now).
It's OK to leave this bug open, there's still non-linear growth with the number
of loops in some passes. For the testcase created by the script
GCC 4.4.7 takes 138s (and 11GB peak memory)
GCC 13.4 takes 80s (and 1.3GB peak memory)
GCC 15.2 takes 110s (and 1.4GB peak memory)
so we fixed the original reported regression in GCC 4.6, but there's room
for improvement:
df reaching defs : 40.09 ( 36%) 0 ( 0%)
like possibly re-writing invariant motion to RTL SSA. What we do at -O2
also changes over time of course.
