Re: Patch to use colored NaNs on LP64

[Ivo Beltchev <ivo@...>]

This is a bit OT.

I have a special NaN to represent a NIL value - see my other thread 
"Comparing userdata with other types". My experiments show that it gets 
preserved by the FPU operations:
Special_NaN+5=Special_NaN
-Special_NaN=Special_NaN

In your experience, do you think I can count on that? (I'm only 
interested in 32/64-bit Windows).

Thanks
Ivo

Mike Pall wrote:
> So there's no need to canonicalize NaNs after
> every operation, which is an expensive step in your patch. The FPU
> on its own only generates 0xfff80000_00000000. Canonicalization is
> only needed for the two ingress points (lua_pushnumber and
> string2number).
>