zhiics opened a new pull request #6068: URL: https://github.com/apache/incubator-tvm/pull/6068
This PR enables the invocation of accelerator specific optimization flow to make BYOC consistent to TVM compilation flow. For example, we separate the optimization and codegen. Optimization is now invoked when a function is partitioned, and codegen only focuses on generating the needed runtime module. This is needed by libraries such as ARM compute library in #5915 Note that we now have explicitly registered various type of APIs for the BYOC flow and implicitly invoke them to accomplish the end-to-end flow. It would require users/vendors to change multiple places for registration. @jroesch also has a concern for this. @comaniac Lets think of a way to centralize the APIs and make an RFC. @masahi @lhutton1 @mbaret @trevor-m ---------------------------------------------------------------- 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: [email protected]
