On Wed, 24 Mar 2010 at 17:36:49 +0200, Johann Haarhoff wrote:
> 
> > > Removing wrlib will be very disrupting to wmaker stability, but
> > > the end goal seems worth pursuing. And we'll keep the stable
> >
> >could someone give me a quick rundown as to why it is beneficial? i'm
> >not looking to hear "because it renders fractal gradients 5.128%
> >faster", but why _really_ :)
> 
> For me the main advantage is the built-in support for XRender (and
> by extension hardware acceleration). So instead of having to tweak
> MMX/assembly in wrlib you get the performance for free in hardware.

I am sorry for not looking deeply right now, but is that MMX/assembly
stuff being used at all?

I have a modern laptop and all that, and this is what I get in the
end of configure

Use assembly routines for wrlib     : no
Use inline MMX(tm) x86 assembly     : no

So at least in my case there is no tweak at all. Oh, and now I see
this in configure.ac:

# until we fix it, leave it disabled
asm_support=no
mmx_support=no

And once I used 'perf' to make some performance profiles of wmaker,
and by far most of the time was consumed by wrlib stuff.

> That being said, it might be more effort than it is worth, so please
> make the case for sticking with wrlib as well.

For me the idea of getting rid of wrlib and using cairo
seems nice because I think having a graphics library exclusively
for wmaker has smaller chances of getting better/cleaner/faster
than using one which is actively developed.

If we can pass the responsibility for the graphics stuff to
cairo instead of nurturing our own library (which probably
a _lot_ less people review or will be able to fix bugs),
I think it is worth.

But I guess it will be a huge effort to get rid of wrlib.


-- 
To unsubscribe, send mail to [email protected].

Reply via email to