Re: Lua/LuaJIT2 performance on Android

[Adam Strzelecki <ono@...>]

> The sponsorship currently only entails an ARMv5 soft-float port.
> It may be extended later to add VFP support or new features of
> ARMv6/ARMv7. But at this time I do not have a covenant for this.

Fair enough.

> Note that NEON isn't very useful here, because it only helps with
> single-precision FP arithmetic. But Lua numbers are doubles.

Thanks for the valuable info, this explains why iOS CGFloats are mapped single precision while on Mac they are doubles.

> The ARM port uses the dual-number mode of LuaJIT. All Lua numbers
> are represented either as integers or doubles internally. This is
> transparent to the user of the VM. Except for the performance
> effects, of course: e.g. 'for i=1,1e7 do end' runs much faster
> than 'for i=1.5,1e7 do end'.

This sounds great. It is amazing that so many things are already anticipated in LuaJIT architecture.

I must confess I thought once about using NaNs as a technique of handling dynamic types in scripting languages, then I found LuaJIT and surprisingly it was already there implemented and running.
Looking at all the innovative techniques used in LuaJIT I think all your work deserve great respect from world-wide community and spreading the news so also other languages such as Ruby (which I use everyday) or Python adopt these inventions.

Regards,
-- 
Adam Strzelecki