gcc -march=nocona -mtune=nocona --help=target

gcc -march=nocona -mtune=nocona --help=target

[Oleg Kravchenko]

Hello!

Why gcc say what mmx, sse, sse2 is disabled?

$ gcc -march=nocona -mtune=nocona --help=target

The following options are target specific:
  -m128bit-long-double                  [disabled]
  -m32                                  [enabled]
  -m3dnow                               [disabled]
  -m3dnowa                              [disabled]
  -m64                                  [disabled]
  -m80387                               [enabled]
  -m96bit-long-double                   [enabled]
  -mabm                                 [disabled]
  -maccumulate-outgoing-args            [disabled]
  -malign-double                        [disabled]
  -malign-functions=
  -malign-jumps=
  -malign-loops=
  -malign-stringops                     [enabled]
  -march=                               nocona
  -masm=
  -mbranch-cost=
  -mcld                                 [disabled]
  -mcmodel=
  -mcx16                                [disabled]
  -mfancy-math-387                      [enabled]
  -mfp-ret-in-387                       [enabled]
  -mfpmath=
  -mfused-madd                          [enabled]
  -mglibc                               [enabled]
  -mhard-float                          [enabled]
  -mieee-fp                             [enabled]
  -minline-all-stringops                [disabled]
  -minline-stringops-dynamically        [disabled]
  -mintel-syntax                        [disabled]
  -mlarge-data-threshold=
  -mmmx                                 [disabled]
  -mms-bitfields                        [disabled]
  -mno-align-stringops                  [disabled]
  -mno-fancy-math-387                   [disabled]
  -mno-push-args                        [disabled]
  -mno-red-zone                         [disabled]
  -mno-sse4                             [enabled]
  -momit-leaf-frame-pointer             [disabled]
  -mpc
  -mpopcnt                              [disabled]
  -mpreferred-stack-boundary=
  -mpush-args                           [enabled]
  -mrecip                               [disabled]
  -mred-zone                            [enabled]
  -mregparm=
  -mrtd                                 [disabled]
  -msahf                                [disabled]
  -msoft-float                          [disabled]
  -msse                                 [disabled]
  -msse2                                [disabled]
  -msse3                                [disabled]
  -msse4                                [disabled]
  -msse4.1                              [disabled]
  -msse4.2                              [disabled]
  -msse4a                               [disabled]
  -msse5                                [disabled]
  -msseregparm                          [disabled]
  -mssse3                               [disabled]
  -mstack-arg-probe                     [disabled]
  -mstackrealign                        [disabled]
  -mstringop-strategy=
  -mtls-dialect=
  -mtls-direct-seg-refs                 [enabled]
  -mtune=                               nocona
  -muclibc                              [disabled]
  -mveclibabi=

Re: gcc -march=nocona -mtune=nocona --help=target

[Ian Lance Taylor]

Oleg Kravchenko <oleg@kaa.org.ua> writes:

> Why gcc say what mmx, sse, sse2 is disabled?
>
> $ gcc -march=nocona -mtune=nocona --help=target

They are disabled because you didn't explicitly request them.

Ian

Re: gcc -march=nocona -mtune=nocona --help=target

[Alexander Monakov]

On Wed, Apr 22, 2009 at 06:30:39AM -0700, Ian Lance Taylor wrote:
> Oleg Kravchenko <oleg@kaa.org.ua> writes:
> 
> > Why gcc say what mmx, sse, sse2 is disabled?
> >
> > $ gcc -march=nocona -mtune=nocona --help=target
> 
> They are disabled because you didn't explicitly request them.
> 
> Ian

Sorry, Ian, but is this behaviour really intended?  These extensions are
enabled by gcc/config/i386/i386.c:override_options() for -march=nocona, and if
one uses -fverbose-asm, they are really listed as enabled there.  IMHO, in
this situation behaviour of -Q --help=target and -fverbose-asm seems
contradictory and a bit confusing to the user.

Perhaps GCC could do better by calling OVERRIDE_OPTIONS (currently in
toplev.c:process_options()) before calling decode_options()?  Is it worth a
bugzilla entry?

Thanks.

-- 
Alexander Monakov

Re: gcc -march=nocona -mtune=nocona --help=target

[Ian Lance Taylor]

Alexander Monakov <amonakov@ispras.ru> writes:

> On Wed, Apr 22, 2009 at 06:30:39AM -0700, Ian Lance Taylor wrote:
>> Oleg Kravchenko <oleg@kaa.org.ua> writes:
>> 
>> > Why gcc say what mmx, sse, sse2 is disabled?
>> >
>> > $ gcc -march=nocona -mtune=nocona --help=target
>> 
>> They are disabled because you didn't explicitly request them.
>> 
>> Ian
>
> Sorry, Ian, but is this behaviour really intended?  These extensions are
> enabled by gcc/config/i386/i386.c:override_options() for -march=nocona, and if
> one uses -fverbose-asm, they are really listed as enabled there.  IMHO, in
> this situation behaviour of -Q --help=target and -fverbose-asm seems
> contradictory and a bit confusing to the user.
>
> Perhaps GCC could do better by calling OVERRIDE_OPTIONS (currently in
> toplev.c:process_options()) before calling decode_options()?  Is it worth a
> bugzilla entry?

As I see it, --help=target reports the state of the command line
options.  I agree that this behaviour can be confusing, but I also think
it would be difficult to fix in a consistent way across targets.  But,
yes, I think this would be worth an enhancement request in bugzilla to
avoid user confusion.  Thanks.

Ian

Re: gcc -march=nocona -mtune=nocona --help=target

[Oleg Kravchenko]

Ian Lance Taylor написав(ла):
> Oleg Kravchenko <oleg@kaa.org.ua> writes:
>
>   
>> Ian Lance Taylor написав(ла):
>>     
>>> Oleg Kravchenko <oleg@kaa.org.ua> writes:
>>>
>>>   
>>>       
>>>> Why gcc say what mmx, sse, sse2 is disabled?
>>>>
>>>> $ gcc -march=nocona -mtune=nocona --help=target
>>>>     
>>>>         
>>> They are disabled because you didn't explicitly request them.
>>>
>>> Ian
>>>   
>>>       
>> It's mean my code will be compiled without optimization?
>>     
>
> Please send e-mail to the list, not just to me.  Thanks.
>
> I assume you are asking whether the code will use SSE instructions if
> appropriate.  Yes, it will, because you specified -march with a
> processor that supports SSE.  The output of --help=target just tells you
> the status of the various command line options; it doesn't tell you
> which architecture will be used.
>
> Ian
>   
Thank you very much! :)

Re: gcc -march=nocona -mtune=nocona --help=target

[Tim Prince]

Alexander Monakov wrote:
> On Wed, Apr 22, 2009 at 06:30:39AM -0700, Ian Lance Taylor wrote:
>> Oleg Kravchenko <oleg@kaa.org.ua> writes:
>>
>>> Why gcc say what mmx, sse, sse2 is disabled?
>>>
>>> $ gcc -march=nocona -mtune=nocona --help=target
>> They are disabled because you didn't explicitly request them.
>>
>> Ian
> 
> Sorry, Ian, but is this behaviour really intended?  These extensions are
> enabled by gcc/config/i386/i386.c:override_options() for -march=nocona, and if
> one uses -fverbose-asm, they are really listed as enabled there.  IMHO, in
> this situation behaviour of -Q --help=target and -fverbose-asm seems
> contradictory and a bit confusing to the user.
> 

Is prescott still preferred as the 32-bit equivalent to 64-bit nocona?  I
couldn't make out which version is under discussion.  There are too many
product names partly supported without useful distinction