[PATCH v2] Add math-inline benchmark

[Carlos O'Donell]

On 07/21/2015 01:35 PM, Ondřej Bílka wrote:
> On Tue, Jul 21, 2015 at 12:47:05PM -0400, Carlos O'Donell wrote:
>> On 07/21/2015 07:35 AM, Ondřej Bílka wrote:
>>>> One nit, OK to commit with that fixed.
>>>>
>>> No Carlos, this isn't ok. You need to do better review as this patch has
>>> still some issues. 
>>
>> It is incremental progress for code that has no immediate API or ABI impact.
>>
>> Therefore I judge it to be OK.
>>
>> I expect Wilco to improve it incrementally, and he has already agreed to
>> remove the test that has little value.
>>
> Then you should say it so. There is additional problem that it doesn't
> measure isnormal inline at all so somebody could pick that and find that
> builtin doesn't improve isnormal just because of typo here. This is
> relatively harmless here but don't have to be.

My apologies, I will endeavour to be more clear next time I give consensus
for a patch.

If I or Wilco missed something the next step is to respond to the original
email with a detailed response about what is missing. Wilco has responded
to the question about isnormal, please respond to his answer and take it
from there to see if we need to change the test.

If at all possible please provide actionable patches that show how you
think the implementation should be.

Cheers,
Carlos.