Re: [ANN] Lua 5.4.1 now available

[Roberto Ierusalimschy <roberto@...>]

> >>>>> "Roberto" == Roberto Ierusalimschy <roberto@inf.puc-rio.br> writes:
> 
>  Roberto> Apropos, version 5.4.2 will roll back to the previous
>  Roberto> stackless implementation. This C-stack issue is causing too
>  Roberto> much noise.
> 
> Also, out of curiosity, has the performance impact of stack-based vs.
> stackless implementation been measured, and how big is it?

The C-stack version was a little faster (that is why we changed in
the first place), but the difference seems to be smaller now, due to
some other changes. For a few examples, the stackless implementation
seems faster.  Anyway, in programs that do stuff besides only calling
functions (that is, almost any program except benchmarks like recursive
fibonacci and Ackerman), the difference is near zero.

Logically (that is, trying to count the number of jumps, memory
reads, memory writes, etc. as written in C), I think the stackless
implementation does one extra memory read per function call.

-- Roberto