Closed chrisforbes closed 10 years ago
Presumably this is very similar to the GS or CLIP kernel setups, just with different inputs and outputs.
I have no idea if the Gen7 tesselation engine is fast enough to bother with though.
Not as similar as I thought, but doable.
This is what URB atomics are really for -- implementing semaphores so we know who is the last kernel to finish and so needing to deallocate the input VUEs.
Presumably this is very similar to the GS or CLIP kernel setups, just with different inputs and outputs.
I have no idea if the Gen7 tesselation engine is fast enough to bother with though.