I don't think defer is the main bottleneck here. I tested the benchmark with Go tip and with removing defers. While it becomes faster it is still slower than Java.
I believe that the reason the Java version is faster is that it's lock implementation behaves better under contention. While I am not 100% sure here are a few observations that support this statement. If the benchmark is executed with GOMAXPROCS=1 it runs much faster (same speed as Java or faster). This proves that the lock contention plays the main role in this benchmark. Starting 1000 gorountines is not the same as starting 1000 threads. The benchmark shows better results when started with GOMAXPROCS=1000. I think this happens because the time between different threads acquiring the lock is longer and each thread has more time to do useful work instead of waiting for the lock. On Oct 3, 2016 09:14, "Aliaksandr Valialkin" <valy...@gmail.com> wrote: > I bet the bottleneck is exactly at defer. Good news - it looks like go 1.8 > will have much faster defer and, as a bonus, faster cgo calls: > - https://github.com/golang/go/commit/f8b2314c563be4366f645536e8031a > 132cfdf3dd > - https://github.com/golang/go/commit/441502154fa5f78e93c9c7985fbea7 > 8a02c21f4f > > -- > You received this message because you are subscribed to a topic in the > Google Groups "golang-nuts" group. > To unsubscribe from this topic, visit https://groups.google.com/d/ > topic/golang-nuts/I-p5vmyln9c/unsubscribe. > To unsubscribe from this group and all its topics, send an email to > golang-nuts+unsubscr...@googlegroups.com. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "golang-nuts" group. To unsubscribe from this group and stop receiving emails from it, send an email to golang-nuts+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
