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: patch to fix PR59511


On Wed, Jan 15, 2014 at 9:34 AM, Vladimir Makarov <vmakarov@redhat.com> wrote:
> The following patch fixes:
>
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59511
>
> Register move cost calculations were too conservative for some cases
> and it resulted into overflows.  The patch fixes also some wrong
> frequencies usage I found.
>
> The patch has very small affect on SPEC2000.  Only 7 programs out of
> 26 has a different code for -mtune=corei7.  For -mtune=generic, it is
> only 3 programs. And there is no visible changes in performance for
> given programs.  So I'd say this PR is not worth P1 status.
>
> The patch was successfully bootstrapped and tested on x86/x86-64.
>
> Committed as rev. 206636.
>
>
> 2014-01-15  Vladimir Makarov  <vmakarov@redhat.com>
>
>         PR rtl-optimization/59511
>         * ira.c (ira_init_register_move_cost): Use memory costs for some
>         cases of register move cost calculations.
>         * lra-constraints.c (lra_constraints): Use REG_FREQ_FROM_BB
>         instead of BB frequency.
>         * lra-coalesce.c (move_freq_compare_func, lra_coalesce): Ditto.
>         * lra-assigns.c (find_hard_regno_for): Ditto.

This caused:

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=59835


-- 
H.J.


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