tang-hi commented on issue #12477: URL: https://github.com/apache/lucene/issues/12477#issuecomment-1660786424
> don't we already have the scalar code today (our current gen'd FOR implementation that Hotspot autovectorizes well) that we could fallback to? This is because in the current version, we have implemented some tricks to enable automatic vectorization in the JVM, which makes the existing file format not very friendly for SIMD usage. Therefore, we have two options: 1. Vectorize the existing format, but it won't result in significant improvement. 2. Use a new format that is more SIMD-friendly (can achieve greater improvement), but performance will significantly decrease when using scalar operations. -- 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]
