0

Let's say I have some time critical code that's run many times from a certain class. However, this code needs to process data encapsulated in another class, so my two options are either get a reference to the data and perform the computation in my original class, or delegate the computation to the class that stores it and return the results. My question is, does calling a method on another class and getting a result come with a performance penalty compared to just getting a reference of data and using a method within the same class ? And if so would this be of practical significance or is it just like 1-2 milliseconds ?

Improve this question
3
  • 3
    1-2ms is eons of time for a computer. If that is your frame of reference, you won't have to think about optimizations at all. Commented Jan 17, 2015 at 23:50
  • profile it, there is no other meaningful answer. Commented Jan 17, 2015 at 23:53
  • My guess is that it would be less than a millisecond penalty which may be reduced to no penalty at all overtime by JVM optimizations. However I'm not an expert. JMV is very complex beast, processors are complex beasts and it may even drill down to cache line alignment. Maybe in some rare cases it may make noticeable difference. Commented Jan 17, 2015 at 23:54

2 Answers 2

Reset to default
2

I think you should not even think about this. This is all up to JVM and jit compiling optimizations. And they are getting better.

I believe short methods can even be compiled and executed inline (that is JVM may not even generate a call instruction)

Better focus on writing optimal code in your critical functions.

Improve this answer
Sign up to request clarification or add additional context in comments.

3 Comments

"I believe short methods can even be compiled and executed inline" <-- that is quite unlikely, to be really honest; the motto of the Java/JVM guys is to leave most of the work to the JVM to do. But anyway, the "problem" will certainly be solved at runtime (by the JVM...)
Sorry for not being clear. I meant jit-compiled (that is runtime - not byte-code compiled)
Thanks, I selected this as best answer, because the advice to better focus on writing optimal code really made me think.
1

My question is, does calling a method on another class and getting a result come with a performance penalty compared to just getting a reference of data and using a method within the same class ?

It won't be in the "long run".

If this piece of code is criticial, it means the JIT will eagerly look for ways to optimize it at runtime, and one of the first optimizations it will perform is inlining.

Of course, this is if the given piece of code is not a 60+k bytecode-long monster...

FWIW, HotSpot, (Oracle's implementation) will start and optimize as soon/late as 10 000 executions of a same piece of code. Other JVM implementations may have their JIT kick in sooner or later than that.

Improve this answer

1 Comment

"Of course if the code is huge" ... it won't be inlined, but the chances are that the relative cost of the call sequence versus what the method body actually does will be tiny.