Gabriel39 opened a new pull request, #65921: URL: https://github.com/apache/doris/pull/65921
## What problem does this PR solve? Adds a reproducible local microbenchmark foundation for the native Parquet reader described in the [benchmark design](https://selectdb.feishu.cn/docx/W9qbdFXpaojbivx9I2jcjMbanic). It separates page decoder cost from end-to-end local reader cost and keeps fixture construction and encoding validation outside measured regions. ## What is changed? - Add 152 native decoder cases covering PLAIN, dictionary, byte-stream-split, and DELTA encodings across fixed-width and binary physical types, with clustered and alternating sparse selections. - Add 137 local reader cases covering open-to-first-block, full scan, predicate scan, LIMIT shapes, null density/pattern, predicate selectivity, projection mode, schema width, predicate position, and four physical encodings. - Generate deterministic fixtures lazily and validate every row group and column encoding from Parquet footer metadata before measuring. - Report rows/s, bytes/s, raw/selected rows, fixture size, ns/raw row, and ns/selected row. - Add matrix constraint tests and local build/run documentation. ## Check List - [x] Release benchmark target builds with `ninja -C be/build_RELEASE -j128 benchmark_test` - [x] Scenario matrix tests pass (4/4) - [x] All 152 decoder benchmarks pass - [x] All 137 reader benchmarks pass - [x] clang-format 16 and `git diff --check` pass -- 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]
