Abhisheklearn12 commented on PR #25793: URL: https://github.com/apache/datafusion/pull/25793#issuecomment-5858023215
thanks for running the benchmarks @kumarUjjawal! about the non-union rows that moved in the bot run, for example `utf8_view` up to 1.18x, this pr doesn't touch that code, so i looked into it locally on x86 with an i7-11700f: * the rows that were noticeably slower in the bot run, like `utf8_view`, `dictionary`, and `list_array`, don't slow down here. instead, other untouched rows move around in both directions. for example, `large_utf8: single, no nulls` is 0.62x (faster), while `dictionary_utf8_int32: single, no nulls` is 1.28x (slower). * building this branch twice with only `strip` toggled moves these rows by a similar amount, even though there are no source changes. for example, the `struct_array` rows get 1.27x to 1.34x slower, and `utf8: single, no nulls` gets 0.78x (faster). * with symbols, the untouched hash functions, including `hash_single_array` and the string, string view, dictionary, struct, and list hashers, are instruction-identical to main. only their alignment differs. so, imo, these movements look like code placement effects from the lto build rather than changes in the hashing itself. -- 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]
