Skip to content
This repository was archived by the owner on Jan 24, 2024. It is now read-only.

Conversation

@BiynXu
Copy link
Collaborator

@BiynXu BiynXu commented Jun 20, 2023

This pr comb the code of op lowering.
Organize the generalized Lower process into four stages:
1.Call Compute and perform narrow Lower (i.e. translate graph to CINN IR).
2. Do Op level schedule.
3.Do group level schedule.
4.Post process, including processing of LoweredFunc arguments and temporary variables, application of low level pass, etc

Then, all types of group lower processes call these processes, abstracting only a small amount of different logic and using different callback functions for processing, to achieve the maximum uniformity of the generalized Lower process.
Specifically, due to significant differences in the Lower logic of CustomCall, it needs to be handled separately by LowerCustomCall().

@paddle-bot
Copy link

paddle-bot bot commented Jun 20, 2023

Thanks for your contribution!

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant