Skip to content

nvk: switch to Kepler style copies instead of Fermi style copies

What does this MR do and why?

nvk: switch to Kepler style copies instead of Fermi style copies

By "Kepler style" I mean stop the wrapping after (1 >> 17) bytes
and starting new copies to copy the whole buffer. Kepler+ doesnt
*seem* to need this, NVE4 on gallium doesnt do this either.

We push _randomly_ so repeated copies for buffers above (1 >> 17)
bytes could have some real performance implications.

Signed-off-by: Yusuf Khan <yusisamerican@gmail.com>

My Vulkan testing might be flawed as the testing I did was really sketch.

Merge request reports