Hacker News Reader

Comment by foldr | Hacker News Reader

It's worth noting that Go, unlike Java, can pass objects by value. That makes it less crucial for the GC to be able to deal efficiently with small allocations, since a typical Go program will have considerably fewer of these than a typical Java program.

View on HN

Same with .NET (C#), where the generational hypothesis certainly holds and therefore .NET has a generational garbage collector.

It's also pretty bad from an ergonomics (not to mention performance) perspective if the programmer has to manually switch to by-value copying of structs in order to keep escape analysis happy and avoid allocation.

Nonetheless, having copy-by-value semantics available for small objects reduces the need for a generational collector.

I'm not sure what you're referring to w.r.t ergonomics. Passing small structs by value should probably be the default choice in Go, and it is not difficult to switch from passing via a pointer to passing by value.

It reduces the need for a collector, any collector, as there will be less garbage. You have not made any arguments that indicate or even hint that the actual share of young objects that will be generated will be reduced.

I believe there is generally thought to be a rough correlation between object size and object lifetime in most programs. So reducing the number of small objects that are allocated in the GCed heap should, on average, reduce the proportion of young objects. Indeed, if an object can be passed by value and/or allocated on the stack following escape analysis, that suggests that it is an object that would have had a short lifetime in the equivalent Java program.

But I don't disagree with what you are saying. It may be that Go would perform better with a generational GC. My claim is not that Go is better off with a non-generational GC (I don't know if it is or it isn't), but simply that it is less in need of one than e.g. Java.

I don't believe that's true, because the Java VM will promote heap objects to stack objects via escape analysis. It even can do this post-speculative-devirtualization.

I'm not seeing the connection between that and anything that I said. Could you expand a bit?

edit: Do you mean that adding pass-object-by-value semantics to Java wouldn't decrease the number of short-lived object allocations because Java already does escape analysis? That may well be true. My claim was that the availability of pass-object-by-value in Go probably reduces the number of short-lived object allocations in Go, compared to what it would be if objects could not be passed by value and everything else remained the same.

> It's also pretty bad from an ergonomics (not to mention performance) perspective if the programmer has to manually switch to by-value copying of structs in order to keep escape analysis happy and avoid allocation.

I kind of like that Go's escape analysis is simple enough that I can reason about it reliably without profiling. I also tend to pass and return values unless I have a compelling reason to pass pointers (like mutation or rarely performance in the case of large structs), so I don't think I experience the sort of ergonomic issues you cite (and incidentally, I think this habit of mine makes writing Rust feel quite unergonomic to me since it seems to prefer egregious use of pointers). And I haven't noticed a significant speedup by switching from value copying to pointer dereferencing (but I've only tested this a handful of times).

You shouldn't have to manually reason about escape analysis. As compiler optimizations of 6g/8g evolve, your ability to reason about it will probably dramatically decrease, due to the effects of inlining, SROA, GVN, LICM, etc. etc. That's why ultimately there's no escaping the need for generational GC.

The beauty of bump allocation in the nursery is that heap allocation is as cheap as stack allocation, at least in the common case.

Even if heap allocation itself is not more expensive, there must clearly be performance benefits to stack allocation, or the JVM wouldn't bother doing escape analysis in the first place.

Java passes objects by value, not by reference. Try implementing swap(Object a, Object b) in Java, and you will see it is by Val not by ref

Java passes references to objects by value. That is, in Java, a variable for a heap-allocated object is a reference to that object. When you pass that variable to a method, the method gets a copy of that reference. So if the method makes changes to the object itself, that is visible outside of the method, but if it makes changes to its own copy of the reference (say, makes it point to a different object), that's not visible outside of the method.

I still consider this behavior "passing by reference" since it passes around references/pointers to objects, and does not incur an object copy.

Expanding on the sibling, Java is a pass by value language which is commonly mistakenly thought to be pass by reference. (That's probably the mistake you think I'm making.) It is not, however, possible to pass objects by value in Java.

Reply on news.ycombinator.com