This is the mail archive of the
mailing list for the GCC project.
Re: Reduce inline-insns-auto
- From: "H.J. Lu" <hjl dot tools at gmail dot com>
- To: Jan Hubicka <hubicka at ucw dot cz>
- Cc: gcc-patches at gcc dot gnu dot org, Michael Matz <matz at suse dot de>
- Date: Wed, 10 Nov 2010 18:19:53 -0800
- Subject: Re: Reduce inline-insns-auto
- References: <20090914101202.GF1349@kam.mff.cuni.cz> <AANLkTi=HtYFdQO5P22xCFSozrr=4Re91gyfyH798MWZK@mail.gmail.com>
On Fri, Oct 29, 2010 at 12:02 PM, H.J. Lu <firstname.lastname@example.org> wrote:
> On Mon, Sep 14, 2009 at 3:12 AM, Jan Hubicka <email@example.com> wrote:
>> this patch reduces inline-insns-auto from 60 to 50. ?This does not cause
>> performance regressions on our testsuite (SPEC/C++ benchmarks) and
>> significandly reduces size of some C testcases from SPEC since C code
>> tends to have number of functions of this size, unlike C++ benchmarks
>> where majority of functions are smaller.
>> I would like to reduce the limits further, but there is problem with eon
>> benchmark where not inlining the initialization loops cause major
>> regression. ?I will look into if we can handle this via Martin's better
>> cost estimate patch. (in general the bottleneck of inlining seems to
>> have shifted from early inlining limits and call costs to this
>> parameter. It is obvious that we need more informed estimates of
>> function body size after inlining and I hope that handling few most
>> common cases such as parameter becoming known constant will suffice).
>> I bootstrapped/regtested x86_64-linux and will commit the patch tomorrow
>> if there are no objections.
>> ? ? ? ?* doc/invoke.texi (inline-insns-auto): Set to 50 (from 60).
>> ? ? ? ?* params.def (inline-sinsns-auto): Set to 50.
> This caused:
Your fix for PR 46228. which I couldn't find in the gcc-patches archive,