[Pixman] [RFC] A couple of minor tweaks for sse2 blitting

[Soeren Sandmann]

Chris Wilson <chris at chris-wilson.co.uk> writes:

> The generic implementation of the BLT routine is to deny the operation.
> So we need to add support to the sse2 implementation in order to perform
> 8bpp. This looked relatively straightforward, so I'm obviously
> overlooking something.

I don't think you are overlooking anything - I added the blt() routine a
long time ago as a quick hack because fbBlt was showing up on RHEL4
profiles, and so pixman_blt() ended up as a not-very-well-though-out API
with a bare minimal implementation.

> Please review kindly. I think adding a fuzzer like blitters-test would
> be useful for validation of these changes. Presumably we would like to
> verify that we can blit between all identical formats from 1bpp to
> 64bpp.

Indeed, such a test would be very useful.

> Should the generic implementation fallback to converting the BLT
> into an image composite operation?

Yes, it would make sense to do that and then never return FALSE. That
does complicate the situation a little if we add support for overlapping
blt at some point, though, but probably nothing that can't be handled.

These patches look basically right to me, but I'll follow up with a
couple of minor comments on each one.


Thanks,
Soren