02:04 HdkR: karolherbst: Even Nvidia killed Fermi+Vulkan support didn't it?
02:07 karolherbst: HdkR: sure, but I got annoyed that I can't use the copy class and that compute is totally different 🙃
02:08 HdkR: Sounds reasonable enough to me
02:08 karolherbst: yeah.. but we could use m2mf instead, it's just a lot more work than making Kepler work from a Turing codebase
02:09 karolherbst: it's also mainly for zink
02:10 karolherbst: we might not get something vk 1.0 compliant but maybe something enough to use zink
02:10 HdkR: Interesting idea
02:10 HdkR: But also Fermi is old AF and will likely hit other issues :D
02:11 karolherbst: right, but we have this fermi+ GL driver covering everything
02:11 karolherbst: would be nice if we could just replace that completely with zink and then write something from scratch
02:11 karolherbst: or something
02:12 HdkR: If it covers everything, no need for the heartache of Vulkan
02:12 karolherbst: the driver sucks though