MINOR: Make CachingCallSite more JIT friendly

[original-brownbear]

Some improvements I found when profiling the red black benchmark ( + some trivialities along the way :)):

• split org.jruby.RubyModule#searchWithCache(java.lang.String, boolean) into hot and cold path methods to decrease the size of its caller org.jruby.runtime.callsite.CachingCallSite#cacheAndCall(org.jruby.runtime.builtin.IRubyObject, org.jruby.RubyClass, org.jruby.runtime.Block, org.jruby.runtime.ThreadContext, org.jruby.runtime.builtin.IRubyObject, org.jruby.runtime.builtin.IRubyObject, org.jruby.runtime.builtin.IRubyObject, org.jruby.runtime.builtin.IRubyObject)
  • Also removed redundant getter use in it
• Reduced the indirection in org.jruby.runtime.callsite.CacheEntry#typeOk and removed its static version (the static version really made no sense as it required having an instance to feed into it anyhow + the instance version compiles a little smaller)
• Removed the redundant setter updateCache(entry); to save a byte
• Made boolean cacheAndGetBuiltin(RubyClass selfType, String methodName) static since it contains no reference to this

Trivial:

• Added missing @override, where missing to improve readability
• Made a few methods around the CachingCallingSite final to improve readability (I make JIT converge to the final state a little faster I guess :D)
• Made a few methods private that were needlessly protected

Results:

• Redblack benchmark becomes 2.2% faster for me, measured over 1k iterations
• Codecache size for the red black benchmark converges to 6.5M instead of 6.8M

[kares]

you need to careful when removing methods on places such as these.
might be called from JIT-ed byte-code (so you're not seeing it show up as used)

[original-brownbear]

@kares yea, I learnt that today in some other spot already :) Questions (if you have a sec):

1. Can I do anything besides running tests + full-text search to make sure I don't kill such a method accidentally?
2. Am I correct to understand that the @JIT annotation should be put on methods which could be called dynamically like you described?

Thanks!

[kares]

Can I do anything besides running tests + full-text search to make sure I don't kill such a method accidentally?

usually just grep for the method-name string -> "typeOf" should be safe to find whether its used :)

Am I correct to understand that the @jit annotation should be put on methods which could be called dynamically like you described?

well yeah - have run into that as well ... lately.
but what I dislike is that now some places might have a dependency on the IR package (org.jruby.ir.JIT) -> should maybe get relocated somewhere else I guess ... not sure
haven't raised that concern yet -> you should chat with @enebo or @headius on IRC if it was meant to be somehow IR specific or not.

[original-brownbear]

@kares thanks!

[kares]

... merging this piece so I do not have a conflict with my fast-opts call-site changes later