compare gcc vs clang on HORIZONTAL tuple implementation

[Larry Evans]

On 09/22/12 06:51, Christopher Schmidt wrote:
> Larry Evans <cppljevans@suddenlink.net> writes:
>> I thought you'd find the following interesting.
>>
>> The 2 attached tuple.benchmark.time.COMPILER files show
>> clang compiles the HORIZONTAL implementation at about
>> the same speed as the VERTICAL implementation
>> (TUPLE_TEST_IMPL:0 is HORZONTAL, TUPLE_TEST_IMP:1 is VERTICAL).
>>
>>   113.12(HORIZ) vs 114.29(VERT)
>>
>> In contrast, as noted in a previous email, the gcc compiler
>> is *much* slower on the HORIZ implementation:
>>
>>   316.75(HORIZ) vs 91.19(VERT)
> Wow, that's quiet a number.  I think it might be a good idea to submit a
> bug report, including both, pre-processed test files.  IDK.  WDYT?
>
>         Christopher
>
The above was a private email I sent to Christopher.  Since then,
similar results were reported here:

  http://article.gmane.org/gmane.comp.compilers.clang.devel/24145

So, I thought the g++ template implementors might be interested
in figuring out why g++ is so much slower than clang.

-regards,
Larry