Re: I need some help on some ideas

To be honest, I have to code it first then I can see the performance improvement.

There may be a possibility that the speed performance improvement is negative.

But if implemented well, users should be able to forget about the memory management totally.

BTW, circular referencing problem is the road block when we use RC most of time.

The luarc-51 project solved this problem by mixing GC and RC. GC works at the time when RC can not deal with

circular references.

In my algorithm, circular referencing problem is solved directly by RC. Therefore, GC is not needed.

I don't know if this is the improvement, but will see soon.

BR,

Yilin

From: Ranier Vilela <ranier.vf@gmail.com>
Sent: Tuesday, October 6, 2020 3:54 PM
To: Lua mailing list <lua-l@lists.lua.org>
Subject: Re: I need some help on some ideas

Em ter., 6 de out. de 2020 às 12:01, 逸霖杨 <yyl_20050115@hotmail.com> escreveu:

Yes, by "reference counting" is really whatever I've in mind.

And if it's OK for others, I'm quite happy to make contributions to Lua 5.4 directly.

Since GC is already widely used and accepted, I just don't think I have the right to force others to get rid of the way of thinking in GC.

So whatever I do now is: provide a switch for users to use GC or disable it (then use reference counting, and default is GC enabled).

Would it be good, show how reference counting can improve the actual garbage collector?

I'm not an expert in GC and this is a harder arena.

I need help to get a better understanding on where to apply reference counting (where to increase reference count and where to decrease reference count).

Your idea is applicable to add reference counting for the userdata objects?

regards,

Ranier Vilela