jaylisde commented on PR #12067: URL: https://github.com/apache/gluten/pull/12067#issuecomment-4625155746
Thanks @marin-ma. Column-buffer hashing isn't retry-deterministic as @wForget mentioned — Gluten splices per-partition column buffers across batches, so identical rows across retries can produce different buffer bytes. UnsafeRow is the data-semantics layer Spark's upstream RowBasedChecksum already uses, so Gluten and vanilla Spark checksums stay comparable. On perf: opt-in (spark.sql.shuffle.orderIndependentChecksum.enabled / ...checksum.mismatchFullRetry.enabled, both default false), only paid when debugging non-deterministic stages. Can add a microbench if useful. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
