etseidl commented on PR #8587: URL: https://github.com/apache/arrow-rs/pull/8587#issuecomment-3391726888
This PR shows the value of reducing vector allocations. Two major offenders during decoding are the `encodings` and `page_encoding_stats` vectors in `ColumnChunkMetaData`. This PR gets rid of the former. The biggest impact is in the wide schema test. ``` default features group encodings_default main ----- ----------------- ------- decode parquet metadata 1.00 16.2±0.28µs ? ?/sec 1.08 17.4±0.31µs ? ?/sec decode parquet metadata (wide) 1.00 62.0±1.92ms ? ?/sec 1.20 74.2±1.58ms ? ?/sec open(default) 1.00 16.9±0.25µs ? ?/sec 1.07 18.1±0.41µs ? ?/sec open(page index) 1.01 255.8±6.08µs ? ?/sec 1.00 252.9±4.98µs ? ?/sec all features group encodings_encr main ----- -------------- ------------ decode parquet metadata 1.00 16.9±0.29µs ? ?/sec 1.07 18.0±0.32µs ? ?/sec decode parquet metadata (wide) 1.00 67.6±1.37ms ? ?/sec 1.16 78.5±1.69ms ? ?/sec open(default) 1.00 17.5±0.35µs ? ?/sec 1.07 18.8±0.39µs ? ?/sec open(page index) 1.00 254.1±4.33µs ? ?/sec 1.00 254.6±5.31µs ? ?/sec ``` -- 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]
