uschindler commented on pull request #2176: URL: https://github.com/apache/lucene-solr/pull/2176#issuecomment-754302399
> No, the floats are read using readBytes() into a ByteBuffer and then converted using a derived FloatBuffer. So I think your intuition about readBytes() is correct? Huh? Really. I thought you are using the code in readLELongs default implementation on IndexInput. So you also copy stuff 2 times! That's even worse, it would have been better to the ByteBuffer directly from IndexInput: `curSegment.slice(offset, length).asByteBuffer()` would return a view on the bytes. Or alternatively do it like with readLELongs(). But nevertheless, for readBytes the wrapping is extensive and may be a slowdown. Can you try this again using a loop of `Float.intBitsToFloat(input.readInt())`? If this is faster I would say that my first expectation is correct: If the number of integers or single bytes to read is small (maybe <50 or <100) then it might really be better to just have a loop calling readByte() or readLong() instead of the array views. ---------------------------------------------------------------- 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. 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
