New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[jit] replace method invocation counter with jit.time.delta #5887
base: master
Are you sure you want to change the base?
Conversation
…been reached BUT it has taken longer than -Xjit.time.delta ns since the last time saved when the counter was set to 0. This is totally untuned in that I made up a number and it I suspect is too dramatically not JITting methods. In running an oj bench and gem list (with and without --dev) and starting up rails the times do not appear to be much different. After tuning the value (which I guess would be for this machine -- a follow up to this work will be whether one magic constant can work here) which would be figuring out proper time delta and also determing whether any important methods which should have JITted had not there is a followup change I want to try: change threshold to be more than just a method call counter. I would like to also add threadpolls as an increment value. Then larger methods which are not called as much will end up as JIT candidates.
and also a way to disable System.nanoTime checking when < 0
e0b270e
to
8083987
Compare
should be ready now, it could use an com.headius:options release to get rid of the LongOption class |
I believe this is safe but without pushback I am skipping for somewhat imminent 9.2.10.0. |
I merged latest master into this to see where we stand on it, but I don't see us making this change right before 9.3. |
we did not do a lot of experiments with |
Untargeting until we have a specific release in mind. @kares Let us know what you think we should do here. |
this is @enebo's work from https://github.com/jruby/jruby/tree/counter_nuke
rebased + with a few touches for usability (able to specify jit.time.delta as long)
might need more work as we still do not have real-world numbers (hopefully soon).
... or we can add it into next 9.2 but having jit.time.delta = -1 (disabled) by default?