Hey everyone, These were previously sent and there was an extension discussion about performance, finding that there was some slight regression on very large extent workloads and significant performance improvement on smaller extents. This is a *huge* performance increase on my workload, especially the behavior enabled in the #3 patch that allows dirty data to be trimmed quickly during idle periods. Reviewed-by tags would be appreciated so that these can be staged for 4.16. I have additional changes coming in the next couple of days-- changes to the closure code picked from kmo's bcachefs tree that both improve performance and enhance correctness. Just need a little more test and to work on the commit comments some. Thanks, Mike
