[PATCH 3/3] debug.h: hint a compiler: TRACE is not executed in common usage

[Alexandre Julliard]

Konstantin Kharlamov <hi-angel at yandex.ru> writes:

> On 30.01.2019 13:43, Alexandre Julliard wrote:
>> Konstantin Kharlamov <hi-angel at yandex.ru> writes:
>>
>>> I'll test later if it works under Windows and report a wine bug if it
>>> does; but if anybody is aware of any other free windows benchmarkss
>>> for memory allocation of threads creation that work under wine, I'll
>>> be happy to try them out.
>>>
>>> I guess there's no haste since someone for some reason told Marvin to
>>> reject my patch (let's hope it was just a misclick without any
>>> malicious intention).
>>
>> I rejected the patch, and I explained why. If you can come up with
>> convincing benchmarks numbers, please resend it and we can discuss it
>> further.
>
> We're programmers, please, let's stay technical. Your explanation was
> some abstract "burden", and I asked you to formalize the "burden". You
> ignored me. At the same time on my part I provided you with good
> arguments why the patch is nice to have. What you did next is ignoring
> my reply whatsoever and rejecting the patch. And I must mention, you
> did that in fact twice, even though discussion was ongoing.

You are the only one who's "discussing" it. I asked you to show
convincing numbers, so you have a path forward. Until you come back with
the numbers, there's no reason to continue the argument.

The most precious resource we need to conserve is not branch prediction
cycles, it's developers' cycles. That's why you shouldn't change things
that don't need changing, you shouldn't add complexity that isn't
needed, and you should spend extra effort on your side to save other
developers' time. These are the kind of burdens I'm talking about.

> In other news, I managed to build "osbench" for Windows, and in thread
> creation/memory allocation benches I didn't see improvement beyond
> statistical noise, so let's hold off this patch until we find a
> workload that definitely gets improved it.

Glad to see we agree ;-)

-- 
Alexandre Julliard
julliard at winehq.org