This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] SEE optimization removal
- From: Mark Mitchell <mark at codesourcery dot com>
- To: Sergei Dyshel <SERGEID at il dot ibm dot com>
- Cc: gcc-patches at gcc dot gnu dot org, Mircea Namolaru <NAMOLARU at il dot ibm dot com>, Ayal Zaks <ZAKS at il dot ibm dot com>
- Date: Mon, 06 Apr 2009 09:55:40 -0700
- Subject: Re: [PATCH] SEE optimization removal
- References: <OF44CF0ED6.D13C27F3-ONC225757E.004F15BE-C225758B.0055F378@il.ibm.com>
Sergei Dyshel wrote:
> In my opinion it's best to pull SEE out of mainline, unless anyone objects.
Sadly, I agree. At present, this is badly broken, and anybody trying to
use it will suffer. We don't serve our users well by inviting them to
use broken optimizations.
The patch would be OK, except that we want to continue to accept old
optimization options (just ignoring them) to avoid breaking existing
Makefiles. Please look at the code for examples of how this has been
handled for other options, and resubmit after changing that. If you CC:
me on the patch, I will review it for you.
> It isn't enabled by default thus making it hard to maintain.
That's not the problem. It's perfectly possible to write test cases for
a pass that isn't enabled by default, and it's equally possible to run
the testsuite with another option turned on. We should not take the
attitude that things are either enabled by default or broken. (Note
that "enabled by default" and "tested" are not the same; thus I am not
calling into question the conventional wisdom that things are either
tested or broken.) The core problem here is that nobody cared enough to
test and maintain this optimization pass.
Thanks,
--
Mark Mitchell
CodeSourcery
mark@codesourcery.com
(650) 331-3385 x713