elvin-n commented on PR #14010: URL: https://github.com/apache/tvm/pull/14010#issuecomment-1442332877
> That definitely looks specific to OpenCL and loading precompiled OpenCL? Are all of these modules generated ahead of TVM compilation? I don't think it necessarily has to be deep in the core compiler, but it would make sense for such options to live near `Target` specific logic. It's a bit tricky to work through this without an example though, it'd be good to raise something outside this PR else I think this will take a lot longer. Probably I missed the initial proposal and don't feel that I fully understand proposal `to live near `Target` specific logic.` But conversion to FP16 and execution in low precision is not specific neither to opencl nor to Adreno. This is generic for many platforms. We have to support such conversion for * any opencl * arm cpu * hexagon * cuda * vulkan * metal * rockm If we add such feature dedicated only to Adreno, we will have to extend it to each above target. That would be very undesirable. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
