Re: PATCH: PATCH: Add builtin floating point types

["H.J. Lu" <hjl.tools@gmail.com>]

On Sun, Feb 7, 2010 at 1:52 PM, Mark Kettenis <mark.kettenis@xs4all.nl> wrote:
>> Date: Thu, 4 Feb 2010 17:14:47 -0800
>> From: "H.J. Lu" <hongjiu.lu@intel.com>
>>
>> I am resending this patch. The motivation is I am working on x86 xml
>> target descriptions.  x86 has i387_ext type. I added
>>
>>   case TDESC_TYPE_I387_EXT:
>>      return arch_float_type (gdbarch, -1, "builtin_type_i387_ext",
>>                              floatformats_i387_ext);
>>
>> to tdesc_gdb_type. I would up 8 i387_ext types at 8 different addresses
>> with the same bits. x86 does
>>
>> if (i386_fp_regnum_p (gdbarch, regnum))
>>    {
>>      /* Floating point registers must be converted unless we are
>>         accessing them in their hardware type.  */
>>      if (type == i387_ext_type (gdbarch))
>>        return 0;
>>      else
>>        return 1;
>>    }
>>
>> It expects 2  i387_ext types should have the same address.  This
>> patch caches ieee_single, ieee_double and i387_ext. OK to install?
>
> Sorry, but I don't like this diff.  At the very least, keep the
> i387_ext_type() function around.  It makes the diff far less invasive.

Is there a way to call i387_ext_type from  target-descriptions.c?
Otherwise,  XML target description will create multiple i386_ext
types and won't work for x86 since x86 requires single i386_ext
type per arch.

Thanks.


-- 
H.J.