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: PR rtl-optimization/39241: [4.4 Regression] ICE in subreg_get_info, at rtlanal.c:3104


> The original patch, which caused this regression, calls
> subreg_offset_representable_p and subreg_regno_offset.
> Both of them are wrappers for subreg_get_info. I don't
> see a reason to make 2 calls to subreg_get_info
> instead of one call. Why do we unnecessarily make
> gcc to run slower?

As you noted, this was the pre-existing idiom used in jump.c (true_regnum) so 
the patch was consistent in this respect.  We'll revisit that for 4.5.

> BTW, there are no regressions when 2 calls
> to subreg_offset_representable_p are removed.
>
> Can I check in the patch just to remove 2 calls
> to subreg_offset_representable_p?

Yes, the patch that just removes the 2nd and 4th calls is pre-approved for 
mainline, assuming that it does fix the PR.  Thanks for your patience.

-- 
Eric Botcazou


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