A more serious question, while you’re around, is to ask you why C# is generally reckoned to be slower than C++, “because it has garbage collection”. In another thread you opined that garbage collectors don’t necessarily make code slow. Thoughts, please.

[image] FrancisKing: A more serious question, while you’re around, is to ask you why C# is generally reckoned to be slower than C++, “because it has garbage collection”. In another thread you opined that garbage collectors don’t necessarily make code slow. Thoughts, please. I have no idea. I…

Prefacing my comment, garbage collection (GC) performance should probably be split off to its own thread, possibly tagged off-topic if it’s not about Julia’s implementation specifically. There’s really no definitive opinion on whether manual malloc/free or GC is faster, there are many confounding f…

[image] Benny: The only clear thing, and something Julia users know well: reduce garbage. I think this is actually the main difference between GC and non-GC languages: languages with GC tend to make it easy to allocate a lot and often, and, as in the case of Java (and Python), some even make …

[image] StefanKarpinski: ref counting has a lot of overhead (it has more small pauses though). I think you mean that as a pro of ref-counting (non-GC languages). But I believe it can mean arbitrarily long pauses (for a thread at least), think of a linked list. While a GC can be real-time, in …

It’s the opposite: ref counting is low latency, high throughout overhead.

Ref counting can actually have pretty high worst case latency if you aren’t very careful about the implementation (since you naively can free arbitrarily many objects at the same time). Average latency is much better than GC though.

There is a “downside” though, Julia is in a rare position of being one of the GCed languages that abstract away the pointers/stack/heap yet be a language where users have an indirect control by understanding the version-specific compiler. Repeatedly I see people struggle with the semantics of variab…

I don’t think that’s a unique position. Every compiler tries to eliminate allocations and optimize things away. That’s true of Java, C#, C and C++. When you upgrade the compiler (or even change options), you sometimes get free performance. On the other hand, sometimes something that used to be fast …

[image] StefanKarpinski: The only languages with truly predictable performance are the ones with predictably bad performance. I would say that you also get more predictable performance if the language allows you to be more expressive about describing the desired optimizations to the compiler.…

imo explicit could be clearer for very small quick programs or specific hardware, but it gets miserable for general use. I’d almost always prefer an optimizing compiler to have a say on what goes on the stack or heap or what is copied or referenced. I’ve heard arguments that abstractions mean I have…

Do Garbage Collectors necessarily make code slow?

Offtopic

Topic		Replies	Views
What exactly is "allocation" in Julia? Performance question , memory-allocation	45	6595	November 4, 2022
Julia position in the Debian Benchmark Game can be improved, and categorization of some Julia there is unfair Performance	29	2204	December 12, 2024
Is my understanding of Julia correct? New to Julia question	38	4663	March 8, 2022
Julia written in Julia? Internals & Design question	37	6472	July 6, 2019
Programming Language Benchmark 2 Performance	25	4004	April 8, 2024

Do Garbage Collectors necessarily make code slow?

Related topics