This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug middle-end/32044] [4.3/4.4 regression] udivdi3 counterproductive, unwarranted use



------- Comment #47 from steven at gcc dot gnu dot org  2008-12-10 11:42 -------

Re. comment #37:
Mark, bug 38453 has a simple test case that shows the poor optimization choice
for ARM-linux.  Also, there are now 4 bugs closed as duplicates of this one, so
many users run into this and consider it important enough an issue to file a
bug report about it.

Re. comment #16:
Zdenek, do you remember which revision / patch removed the cost check?  And do
you recall (or can you recover) some of the missed-optimization bug report
numbers?  I tried to find them with a Bugzilla query, but failed.

With the removal of the cost check, we've gone from missed-optimization bugs to
too-aggressive-optimization bugs that even require hacks/workarounds from our
users.  To me, it seems we have made the wrong trade-off, then.

In my opinion, this *is* an optimizer bug, and, actually, a much more important
bug than some of the regressions that are P2/P3 now for gcc 4.3 and gcc 4.4.  


-- 

steven at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2007-05-23 15:13:15         |2008-12-10 11:42:50
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=32044


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]