Garbage Collection for Systems Programmers

[Sergey]

On Saturday, 6 April 2024 at 15:52:21 UTC, Carl Sturtivant wrote:
>> Very similar test..
>> D with LDC showed more FPS than Rust and C++ :P
>> https://github.com/NCrashed/asteroids-arena
>
> This is all very well, but this side thread is conflating the 
> existing GC with the one we are postulating for D that would be 
> the game changer, with performance as per the article linked 
> when this thread was started.

I've just shared with Ogi some other tests showing that 
everything not that bad as in tests Ogi mentioned)

That article is not new and was discussed previously. Moreover at 
D Discord several discussions about improvements of GC were taken.
But I can't see any real points from topic starter or you. Maybe 
you can help me with that.

Let me briefly summarize some things to be on the same page:

* have "better" GC will be good for D (kinda obviously, but even 
"better" could mean different things for different people)
* but it is very hard and expensive research (Google made some 
for Go)
* GC overall is not simple thing, and you need someone who will 
be able to spend significant resources on it
* Rikki has some ideas for improvements, that related to 
barriers, but afaic there is no consensus about it
* Walter previously on Dconf was mentioning, that some metaprog 
features of D are available only because we have current GC 
approach, and they could not working with another architecture
* even C# (if we trust to Mono's creater Miguel 
https://www.youtube.com/watch?v=tzt36EGKEZo) has not very good GC
* One of the state of the art Azul's "pauseless" GC is a result 
of hard work the whole company, which is impossible for the 
current state of the D

So what is the point then you wanted to discuss in this thread?