Re: .machine "altivec"

[Buday Gergely <gergoe@math.bme.hu>]

Nick Clifton wrote:

>> I'm using openembedded to create a powerpc executable on my x86 linux box. 
>> oe complains about
>
> What was the configure command line that you used ?

/opt/openembedded/powerpc-tmp/work/glibc-initial-2.3.2+cvs20040726-r15/libc/configure 
--host=powerpc-linux --build=i686-linux --without-cvs --disable-sanity-checks 
--with-headers=/opt/openembedded/powerpc-tmp/cross/powerpc-linux/include
  --enable-hacker-mode

Though it was bitbake and not me.

>> | checking for .machine support... no
>> | configure: error:
>> | *** A binutils version which can handle
>> | ***    .machine "altivec"
>> | *** is needed.
>
> I could not find this error messages in the sources anywhere.  Are you using 
> a customised version of the configure scripts by any chance ?

Well, openembedded people might changed them.

> The 2.15.94 sources should support the .machine "altivec" directive. Have a 
> look at the function parse_cpu() in gas/config/tc-pcc.c for confirmation of 
> this.

tc-ppc.c I suppose you meant. It supports, as far as I can judge:

=========
   else if (strcmp (arg, "altivec") == 0)
     {
       if (ppc_cpu == 0)
         ppc_cpu = PPC_OPCODE_PPC | PPC_OPCODE_CLASSIC | 
PPC_OPCODE_ALTIVEC;
       else
         ppc_cpu |= PPC_OPCODE_ALTIVEC;
     }

====

How ./configure checks whether it supports?

Regards

- Gergely