Trying to use opengl hardware acceleration on the Weston RDP backend I found
that the “has_llc” limitation on the “intel_readpixels_tiled_memcpy” fast path
was avoiding a fast intelPixelRead behavior. I’m using an Intel Bay Trail GEN7,
so a byt device family with 0x0F31 id (no llc, I gues). What I want to say is
that this limitation seems not be necessary in this case, as if I bypass it I
get this fast path running succesfully and the intelReadPixels() function works
very well and fast.
mesa-dev mailing list