clvk icon indicating copy to clipboard operation
clvk copied to clipboard
verified publisher icon kpet

Better support discrete GPUs

Open kpet opened this issue 6 years ago • 1 comments

clvk is currently designed with Unified Memory Architecture systems in mind. All the device memory allocations are currently HOST_VISIBLE which is inefficient on systems with a discrete GPU. Instead clvk should use DEVICE_LOCAL allocations to back memory objects used on the GPU and transfer to/from HOST_VISIBLE allocations as needed.

kpet avatar Apr 14 '19 10:04 kpet

Can you elaborate on why all allocations are HOST_VISIBLE?

pvelesko avatar Nov 21 '24 08:11 pvelesko