This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Tree inlining for the C front end (part 3 of 3)
- To: Alexandre Oliva <aoliva at redhat dot com>
- Subject: Re: Tree inlining for the C front end (part 3 of 3)
- From: Daniel Berlin <dan at cgsoftware dot com>
- Date: Tue, 25 Sep 2001 10:45:46 -0400
- Cc: Gerald Pfeifer <pfeifer at dbai dot tuwien dot ac dot at>, <gcc-patches at gcc dot gnu dot org>
On Tuesday, September 25, 2001, at 09:15 AM, Alexandre Oliva wrote:
> On Sep 25, 2001, Daniel Berlin <dan@cgsoftware.com> wrote:
>
>> Not doing performance testing on patches specifically meant to improve
>> performance (Otherwise, why the heck are we doing tree inlining at
>> all?)
>
> See the message I've just posted.
Okey.
>
>> Had performance testing been done, it would have been noticed that it
>> made 0% difference.
>
> I guess performance testing was done at the time store motion was
> implemented. I suspect it ended up broken because of countless
> merges, possibly with errors, before it was contributed.
In this case, this is not possible. It couldn't have ever worked
properly. In fact, I have a tree from the day it was introduced into
red hat's source tree (months before it was contributed, IIRC), and it
was broken in the exact same way. If that piece had been fixed, in
actuality, it would have been noticed that the rest of it was broken in
other ways, too.
> This is
> exactly what I'm trying to avoid by contributing this chunk of code as
> early as possible.
I understand that, and don't get me wrong, i'm grateful for the work. I
was thinking of doing it myself on the ast-optimizer branch a few months
ago.
As far as i'm concerned, a 3% difference is acceptable (you did disable
RTL inlining, right?).
--Dan