This is the mail archive of the gcc-patches@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]

Re: RFA: revert patch prohibiting allocation of global allocnos for -O0

Vladimir Makarov wrote: 
 Hi, I'd like to reverse the patch which I've submitted a few days
ago to fix rethrow6.C degradation with -O0.  I prohibited allocation
of global allocnos for -O0 because I found some hard-registers are
missed from LR_IN and LIVE_IN.  That is not a problem for the old
register allocator because it builds conflicts on reverse traverse of
BBs.

Now after IRA finds conflicting hard registers for allocnos also on
reverse traverse of BBs, there is no necessity for the patch anymore.
Permitting allocation of global allocnos for -O0 improves code for free
(the same compile time).

The patch was successfully bootstrapped with -O0 on x86_64 and tortured
on gcc regression testsuite.

2008-09-04 Vladimir Makarov <vmakarov@redhat.com>

   * ira-color.c (ira_fast_allocation): Permit global allocno
   allocation.

OK. However, I'm a little concerned in that we seem to be bouncing back and forth on a couple issues -- that tends to make me wonder if we're rushing things a little. While it's good to be responsive, let's make sure we're dealing with the core issues the first time around, even if that means there's a longer delay in submitting fixes.

Jeff

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