[vlc-devel] commit: New C++ binding for libvlc. ( R?mi Duraffort )

[Pierre d'Herbemont]

On Mon, Jan 25, 2010 at 7:20 PM,  <jpd at videolan.org> wrote:
> On Mon, Jan 25, 2010 at 05:25:14PM +0100, Pierre d'Herbemont wrote:
>> I wonder what tool you use to do your performance optimization, and
>> how you proceed. If this becomes a performance issue (ie, takes
>> several percent of execution time), this has to be fixed, not the
>> other way around.
>>
>> This micro optimization makes no sense here. Non virtual call C++ call
>> nearly resolves to a call and a load.
>
> Did you even read what I wrote? I don't think you did more than look at
> the letters and failed to grep agreement, because I quite clearly stated
> what and why, and all you do is merely repeat what was said before.

lol @ grep.

Probably. So far I was trying you to agree that micro optimization is
not something valuable, and wasn't participating at all as an argument
in favor of inlining. Saying this was horrifyingly candid.

> I didn't say it was critical. I did say it was senseless to introduce
> something that a) doesn't gain anything and b) is easily removable.

Sure, agreed.

> I did also mention that I'd be perfectly happy with the calls provided
> they do gain us something, like complete client independence from the
> libvlc C headers. But this doesn't. Therefore it is useless overhead,
> however small.

Yes, if you read the thread again (arf, seems that we can't read each
other, or that we can't express ourselves clearly), you'll see that I
did see your point and even rephrase it. I do agree with you. So you
don't need to explain or try to convince me again.

However the goal I want to pursue is header independence and library
independence. This should explain the rest of about auto-generation.

The reason why I want header independence are better API stability,
and sound layering...

And I would prefer us to have something that match as closely as we
can the direction we want to take, even if it's uncomplete now. (We
probably disagree here, but fine, could be a taste issue.)

> Go read the thread again, for I did explain myself in some detail.

I won't, what you say seems perfectly understandable :-)

Pierre.