masahi commented on PR #80: URL: https://github.com/apache/tvm-rfcs/pull/80#issuecomment-1164849034
> At the commit/wait abstraction level, what if we rename it from "stage" to "queue"? Yeah, my thinking has been that there is a 1-1 mapping between a stage and a queue. So I have no problem with this change. > To double check, would this kind of proposed usage affect the implementation at all? Oh, what I meant was that we might want to change the terminology "stage" throughout our existing TIR software pipeline implementation first, not just the implementation of this proposal. To align with more standard terminology and avoid potential confusion with this proposal. But reading the discussions more carefully, I'm realizing that your suggestion is not necessarily changing the existing use of "stage" in TIR software pipeline, but rather decoupling asynchrony (commit, wait) from the "stage" in this proposal. As the title of this proposal literally says, the original intention has been to bring asynchrony to the "stage" in TIR software pipeline. So, the current mechanics of commit / wait are naturally tied to the "stage" in the TIR sense and the proposal / implementation are highly influenced by how the current TIR software pipeline works / is implemented. @Lunderberg rightfully hinted when he said: "That the "queue" is produced from the "stage" feels more like an implementation detail of the lowering, rather than something inherent to the commit/wait functionality." I'll go through discussion comments more carefully today and think about how to incorporate the proposed suggestions. Thank you very much for the detailed feedback so far! -- 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]
