Thanks @nihui, @tqchen, @jwfromm. 

I generalized this approach to handle devices that don't support the push 
descriptor (see the Stream::LaunchDeferred APIs), and devices that don't 
support dedicated allocation APIs, and removed the previous runtime 
implementation as @tqchen suggested. I also added some more tests which (under 
Vulkan API validation) revealed some thread safety bugs.

I think this should be closer to something that is acceptable, so I'll remove 
the RFC and WIP tags, this should be ready for review. 

File/class structure is certainly up for debate. I'm fine with whatever folks 
propose.

-- 
You are receiving this because you are subscribed to this thread.
Reply to this email directly or view it on GitHub:
https://github.com/dmlc/tvm/pull/3849#issuecomment-528190493

Reply via email to