This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Warning regression with COMPARE fixes
>>>>> "Joe" == Joe Buck <jbuck@racerx.synopsys.com> writes:
Joe> Well, OK, but this suggests that the optimizer is now
Joe> stupider than it used to be.
Definitely.
There's no question that my patch (which avoids generating wrong code)
also pessimizes some code. I'm definitely more concerned about that
problem than the warning.
Of course, generating fast code is easy if you don't mind it being
wrong. :-)
Anyhow, I've been trying to ask (beg?) for help from anyone who is an
expert on this part of the compiler. I asked yesterday about the
problem when it first surfaced, and again when I posted my patch.
It would be terrific if someone knows either a) that my patch is
wrong, or b) that there's a way to avoid the pessimization using some
other technique.
Thanks,
--
Mark Mitchell mark@codesourcery.com
CodeSourcery, LLC http://www.codesourcery.com