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]

Re: [PATCH] Change default max-inline-insns to 100


Mark Mitchell <mark@codesourcery.com> writes:

> --On Friday, July 20, 2001 02:00:25 PM -0400 Daniel Berlin
> <dan@cgsoftware.com> wrote:
> 
>>
>> 2001-07-20  Daniel Berlin  <dan@cgsoftware.com>
>>
>> 	* params.def: Change default max inline insns to 100.
> 
> OK, for branch and mainline.  On the mainline, we will probably want
> to do something more sophisticated, eventually.

Using some stuff Diego's working on, i've got a call graph going, with
frequencies of calls and all that (including proper frequencies for
calls inside loops, arguments to calls, etc).

Using that to inline/procedure clone like the HP paper suggests seems
to improve performance  quite a bit (we also serialize the call graph
so we can use it between modules in order to get better stats on what
to inline, the arguments aren't serialized, i'm waiting for the
precompiled header stuff that will make that easier).

I also started a high order loop transformation infrastructure, and
have loop invariant code motion going at the tree level.

But all of this is secondary fun stuff i work on when not working on
various pieces of the new register allocator, so if someone wants to
kick any of that into high gear, to see results faster, they'd have to
take it over.

--Dan
> 
> Thanks,
> 
> -- 
> Mark Mitchell                mark@codesourcery.com
> CodeSourcery, LLC            http://www.codesourcery.com

-- 
"Every so often, I like to stick my head out the window, look up,
and smile for a satellite picture.
"-Steven Wright


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