jfboeuf commented on PR #13989: URL: https://github.com/apache/lucene/pull/13989#issuecomment-2471157022
@jpountz [I modified the benchmark to make it more realistic by adding a header to the `IndexOutput` ](https://github.com/apache/lucene/commit/8dc6eac23b3a1158ef4c82860d8574c779bad047) so: * `updateLongs(long[], int, int)` triggers a real flush before processing (the benchmark was biased because the buffer was empty, which is unlikely in real operations where there is at least the header and the table size before). * write accesses to the buffer by `updateLong(long)` are not necessarily aligned (they are always with the new `updateLongs(long[], int, int)`). After testing with different sizes of `long[]` it seems the loop of `updateLong(long)` is only worth it for **really small** arrays and this specific case probably doesn't deserve the additional complexity/if_branch in `updateLongs(long[], int, int)`. ``` Benchmark (size) Mode Cnt Score Error Units BufferedChecksumIndexInputBenchmark.decodeLongArray 8 thrpt 15 5992.000 ± 123.429 ops/ms BufferedChecksumIndexInputBenchmark.decodeLongArray 16 thrpt 15 5729.068 ± 123.423 ops/ms BufferedChecksumIndexInputBenchmark.decodeLongArray 32 thrpt 15 5307.703 ± 149.885 ops/ms BufferedChecksumIndexInputBenchmark.decodeLongArray 64 thrpt 15 4960.178 ± 124.810 ops/ms BufferedChecksumIndexInputBenchmark.decodeSingleLongs 8 thrpt 15 6182.986 ± 253.181 ops/ms BufferedChecksumIndexInputBenchmark.decodeSingleLongs 16 thrpt 15 5472.158 ± 190.560 ops/ms BufferedChecksumIndexInputBenchmark.decodeSingleLongs 32 thrpt 15 4739.634 ± 404.963 ops/ms BufferedChecksumIndexInputBenchmark.decodeSingleLongs 64 thrpt 15 3858.685 ± 81.880 ops/ms ``` @rmuir I'll add unit tests to thoroughly check the method behaves properly. -- 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: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
