This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Probable GCC Bug
- From: "Vladimir N. Makarov" <vmakarov at redhat dot com>
- To: Pankaj Dev <panks11 at hotmail dot com>
- Cc: gcc at gcc dot gnu dot org
- Date: Fri, 05 Sep 2003 10:08:47 -0400
- Subject: Re: Probable GCC Bug
- References: <Law10-F47jWARKToVR8000285c7@hotmail.com>
Pankaj Dev wrote:
> To my understanding in order to disable the use of DFA Scheduling by the GCC
> i removed the "(1 << PROCESSOR_TSC701)" from the function
> // File : sparc.c
> static int
> sparc_use_dfa_pipeline_interface ()
>
> but left(didnt remove) the above expression i.e. "(1 << PROCESSOR_TSC701)"
> in the function (which i added earlier)
> // File : sparc.c
> static int
> sparc_use_sched_lookahead ()
>
> and then build GCC.. It is giving an error
>
> I just wanted to know if I am wrong in the steps taken to disable the use of
> DFA based instruction scheduling in gcc or is it a bug in gcc..
Currently, to disable the dfa pipeline hazard recognizer the hooks
use_dfa_pipeline_interface and first_cycle_multipass_dfa_lookahead should
return 0. That is what I see. Apparently that is not what people expect when
they read the documentation because the first cycle multipass scheduling works
only with the dfa.
I'll fix this. The patch will be coming soon. Thanks for the message.
Vlad