[patch]: Add support of new __int128 type for targets having 128-bit integer scalar support

Kai Tietz ktietz70@googlemail.com
Wed May 26 09:56:00 GMT 2010


2010/5/26 Jason Merrill <jason@redhat.com>:
> On 05/25/2010 05:44 PM, Kai Tietz wrote:
>>
>> +__extension__ typedef __SIZE_TYPE__ ssize_t;
>
> Why "ssize_t"?  That looks to me like "signed size type", but __SIZE_TYPE__
> is unsigned.  You also don't need __extension__.

Please note about the #define unsigned before. This trick is a bit
ugly but already in use at many places to get an signed type by
__SIZE_TYPE__ macro. As prior version worked on signed integers, I
kept here this.
Maybe an idea would be to define as builtin macro also the
__SSIZE_TYPE__, but as this isn't present at the moment this kludge
works fine.

> Also still need a mangling test in g++.dg/abi and a demangling test in
> libiberty/testsuite/demangle-expected.
>
>> Here the updated patch for __int128. It doesn't use for __int128
>> keyword the D_EXT any more. Not sure if this is that good, as it means
>> that the token __int128 is interpreted as keyword for
>> non-gnu-extension, but well.
>
> That's the normal rule; keywords that start with __ are available in
> non-gnu-extension mode.
>
> Jason
>

Kai

-- 
|  (\_/) This is Bunny. Copy and paste
| (='.'=) Bunny into your signature to help
| (")_(") him gain world domination



More information about the Gcc-patches mailing list