westonpace commented on pull request #9656: URL: https://github.com/apache/arrow/pull/9656#issuecomment-812074834
+1 for Antoine's approach. Everything to the "left" of the background generator should be simple reads. Immediately after the background generator it should transfer to the CPU thread. My theory (I will hopefully prove someday :) ) is that anytime you start processing data (e.g. ReadMessage) on I/O threads you are 1) not reading as fast as you can from the input stream and 2) forcing it to load data into the CPU caches twice. -- 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
