2012/4/4 Ronald S. Bultje <[email protected]>:
> Yeah, I'm just wondering if it makes sense to describe the data, not
> its use, since the same data could be used for multiple things.

Indeed, this hides it from a grep when one is looking for an already
existing declaration.
On the other hand, having a regular name and a comment just above its
declaration makes it more difficult to grasp its use.
I would think the later is more costly, in the sense of the most
costly resource, developer's time.

> Anyway, this is extremely minor, let's ignore it for now. Rest of the
> patch is obviously OK.

In fact, a bigger highlight of the patch is that there are several
places allowing such substitutions. But then again, once you start
digging, you see that, more importantly, dsputil has lots of functions
with at best mmx2 optimizations.

> I just had this amazing idea of converting all these pw_... tables in
> dsputil_mmx.c to yasm. So pretty. But so much effort...

In a single, unconditionally compiled file then, because if you have
to factor in conditional presence/declaration of said symbols... A
nice goal to reach anyway, but knowledgeable developer's time would be
better spent writing speedier functions IMHO.

Christophe
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to