Hi, On 28 September 2017 at 09:10, Dave Airlie <airl...@gmail.com> wrote: > If we get a nouveauv ever we would have to have this, so not really wanting > this.
Thanks for taking a look. The reason I did this wasn't just to get it out of the WSI, but because we need it in the driver anyway. Without modifiers, it's easy: we need a linear copy when we have a 'different' GPU as a consumer, and not otherwise. With modifiers, that becomes harder. Intel and AMD _do_ support common tiling formats (the DX-mandated ones: Intel calls the family Yf), so whenever AMD gets modifier support, you could skip the blit-to-linear in that case. But the reverse holds true as well: the WSI can advertise a set of modifiers the driver can't / doesn't want to render directly to, where a blit would be needed. The WSI can't actually know that, only the driver. I can imagine something like an out param for the driver's alloc callback telling WSI that it needs a linear surface allocated to blit to as well, but it didn't seem like that was actually saving any code. Any ideas? Cheers, Daniel _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
