Small simplification and optimization in tick start/end #148
Conversation
Change the EnumSet equivalent of "A &= ~ new(~B)" to "A &= B". This eliminates the need for one temporary object in a frequently called loop, and reduces the number of methods that need to be invoked.
|
Interesting change. Have you tested it? I am reluctant to futz with ticking code unnecessarily. |
|
I've had this patch running on a large server and apparently working fine for about a month - not that that's very good evidence that it's correct, of course. |
|
Yeah, given the ticking code generally hits modloader mods much harder, I would want observations from modloader mod users.. (Therefore, client mods, generally) |
|
Do we have any data on this? |
|
sampling wise, i've mostly got ssp (and some smp)data. It is an insignificant amount of cpu spent here - complementOf rounds to 0.00%, and remove all doesn't appear on the list at all. I agree with the change though - simplifying logic, and having less mem turnover is always a bonus, unless the jvm is optimizing the intermediate object away anyway.... |
|
OK. I'll pull it into the 1.5 snapshot.. |
Change the EnumSet equivalent of "A &= ~ new(~B)" to "A &= B".
This eliminates the need for one temporary object and one static method invocation in a frequently called loop.
Further optimization is possible, but would not necessarily simplify the code, so is not included in this pull request.