This is the mail archive of the
mailing list for the GCC project.
Re: Performance regression
- From: David Edelsohn <dje at watson dot ibm dot com>
- To: Richard Henderson <rth at redhat dot com>, Dale Johannesen <dalej at apple dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Fri, 27 Sep 2002 13:34:00 -0400
- Subject: Re: Performance regression
>>>>> Richard Henderson writes:
Richard> It's rs6000_binds_local_p. Since AIX forces flag_pic true, it is
Richard> effectively prevented from auto-inlining any non-static function.
Does Darwin set flag_pic non-zero by default?
I do not see where Darwin changes the value by default, so I do
not understand why Darwin produced different code with the commandline
posted. With the inlining, the testcase should fail with the recent
Richard> I believe David said that AIX normally binds locally in shared
Richard> libraries, but there is a linker switch to change that default.
Richard> He also said that the IBM compiler doesn't change code generation
Richard> wrt that flag; it's up to David to decide what's best here for gcc.
The behavior is correct for AIX.