[PATCH,picochip] Removed profiling support in picochip

[Hariharan Sandanagobalane]

Hi Paolo,
Thanks for your suggestion.

Two things made me remove profiling support in picochip
1. I wasnt sure about how much performance is to be had by just having 
fprofile-arcs and fbranch-probabilities. Seongbae seems to think that it 
is very small (and none if i didnt want a code bloat).

2. Generating profile information from a simulator and feeding it back 
to gcc seems to be rarely tried. My message 
http://gcc.gnu.org/ml/gcc/2009-01/msg00305.html didnt get any response.

If the way forward is to add more profiling under fprofile-generate and 
fprofile-use, it will be very difficult for me to keep up my hack and it 
wouldnt be production-worthy.

Let me know if you think otherwise.

Cheers
Hari


Paolo Bonzini wrote:
> Hariharan wrote:
>> Hi,
>> Further to my discussions at
>> http://gcc.gnu.org/ml/gcc/2009-01/msg00111.html, it doesnt make sense to
>> keep profiling support in picochip port anymore. This change (commited)
>> removes profiling support.
> 
> Just add something like
> 
>         flag_profile_values = 0;
>         flag_unroll_loops = 0;
>         flag_peel_loops = 0;
>         flag_tracer = 0;
>         flag_value_profile_transformations = 0;
>         flag_inline_functions = 0;
>         flag_ipa_cp = 0;
>         flag_ipa_cp_clone = 0;
>         flag_predictive_commoning = 0;
>         flag_unswitch_loops = 0;
>         flag_gcse_after_reload = 0;
> 
> to your target's OPTIMIZATION_OPTIONS instead.
> 
> Paolo