This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [dataflow] remove try_auto_increment from regmove


Steven Bosscher wrote:
> On 11/5/06, Kenneth Zadeck <zadeck@naturalbridge.com> wrote:
>> I have no problem with this patch.  I would like to see it go in.
>> However, the ppc only has pre* auto insns and the x86 has none.
>
> But iiuyc, ARM does?
>
>> You need to test this on a machine that has post* auto insns like the
>> arm or the ia-64.  Unfortunately we do not have these ports working
>> yet.  If you want to track down some bugs on these ports, we would be
>> happy for the help.
>
> Today I successfully built&tested an ARM compiler with the dataflow
> branch, see http://gcc.gnu.org/ml/gcc-testresults/2006-11/msg00428.html
>
> I also compared some assembler output (~440000 lines, preprocessed GCC
> 3.1) with and without Paolo's patch.There were no assembler
> differences. A more thorough check ought to be performed but the
> result is very promissing IMO.
>
> Gr.
> Steven
your 89 arm failures compared to the mainline are the reason that I
still want to hold off.
Seongbae found a bug in the post decrement code today that I sent him a
private patch for and he is moving to the next ia-64 bug even as we speak.

I expect that he will post a progress report patch sometime soon with
the several ia-64 fixes he has made and then we can retest the arm.

However, I admit that these results are promising.  I just think that we
should wait until we can test this clean.

kenny


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]