min-guk opened a new pull request, #2182: URL: https://github.com/apache/systemds/pull/2182
## Summary An initial version of the enumerator for finding the optimal plan in a `DMLProgram` has been implemented. This design is based on `FederatedPlannerFedAll`. This PR addresses a different topic from the previous one (https://github.com/apache/systemds/pull/2175), despite being related, so it has been created separately. The currently implemented functionalities are as follows: - Recursive enumeration of `StatementBlock` and `Statement` within the program. - Identification of the optimal federated plan (`FedPlan`) for each `Statement` (limited to individual statements, not the entire program). However, the current implementation does not yet support: - Integration and identification of the optimal `FedPlan` and `MemoTable` at the `StatementBlock` or program level. --- ## Questions for Discussion ### 1. Recursive Integration of Optimal FedPlans and MemoTables Across Statements and StatementBlocks - Should the same dynamic programming approach be used for integration? - Or would simply summing up the minimal plans of each block/statement suffice? ### 2. MemoTable Sharing - Is there a need to share the MemoTable? - Are there data/hop dependencies between statements? ### 3. Prediction of Iteration Counts for Loops (For and While) How should the iteration counts for loops be determined? - If from/to/increment are constants: Calculations can be done at compile time. - If they are variables: Use default values at compile time, adjust at runtime, or predict using ML models. --- -- 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: dev-unsubscr...@systemds.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
