mei-ye commented on pull request #5645:
URL: https://github.com/apache/incubator-tvm/pull/5645#issuecomment-632434003


   Tianqi and Thierry,
   
   Thanks for our advise.  I can see your points,  especially the benefit of 
supporting both vulkan and opencl.  But I also have some concerns:
   1. We want to get rid of data movements between CPU and GPU.  Do we need a 
new runtime?  If yes, having a new target will be more convenient.
   2.  It looks like target-dependent condition checking codes are heavily 
relying on string match for target names.   Without adding a new target, we 
will need to check both target name and device_name.   Is device_name available 
throughout TVM?
   
   
   
   


----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
us...@infra.apache.org


Reply via email to