> Patrik Stridvall <[EMAIL PROTECTED]> writes:
>
> > Infrastructure for better handling of
> > logical to/from device translation.
>
> I'm not going to apply any patch of this kind unless you demonstrate
> with benchmarks, and preferably real-life applications too, that
> logical-to-device translation is a major bottleneck.
One of the advantages of this patch is that the speed critical
case (MM_TEXT, no offset) bypasses the translation layer
completely. As to how much I don't know. I will see what I can do.
However...
> There is no point
> in adding such a complex infrastructure if it doesn't offer large
> gains in performance.
... please note that the biggest advantage of this patch is not nessarily
speed, it is that maintainance will be easier when most of the device to
logical translation is moved from the drivers and the common code to the
translation layer.
The patch is IMHO not really complex. It might look complex but it really
isn't. It is large, but fairly straight forward.
It makes it easier for the common code to do its work in logical coordinates
without "thinking" about device coordinates and for the device code to
work with device coordinates without thinking about logical coordinates.
Sure when I started with this patch I had speed I mind but then I
realized that it really improved maintainance as well.
So please read the RFC and the patch again and try to see what I see.
Some extra speed is just a bonus of the patch not the main advantage.
