Re: Options for signed 64bit ints

[Andrew Starks <andrew.starks@...>]

On Thursday, August 1, 2013, Justin Cormack  wrote:
> On Thu, Aug 1, 2013 at 6:10 PM, Andrew Starks <andrew.starks@trms.com> wrote:
>
> >
> >
> > On Thursday, August 1, 2013, Lorenzo Donati wrote:
> >>
> >> On 01/08/2013 9.32, Chodera, Ian wrote:
> >>>
> >>> nano = 10^-9
> >>>
> >>> see: http://physics.nist.gov/cuu/Units/prefixes.html
> >>>
> >>>
> >>>
> >>>
> >>>
> >>
> >> I think the OP meant he has to work in the nanoseconds *range*,
> >> using basic time units of 100ns:
> >>
> >> 10^7 "units" == 1s
> >> => 1 unit = 10^-7s = 100 * 10^-9s = 100ns
> >>
> >> -- Lorenzo
> >>
> >> --
> >> ()  ascii ribbon campaign - against html e-mail
> >> /\  www.asciiribbon.org   - against proprietary attachments
> >>
> >
> > This is correct.
> >
> > I'm on 5.2 and limiting dependancies is important.  I'd also prefer an
> > approach that avoided patching code in individual Lua files. Libraries, even
> > ones replacing core functionality, are okay, provided they're stable.
> >
> > Is grafting portions of LuaJIT possible, given the above preferences?
>
> No but luaffi, linked above has a compatible interface, and you really
> don't need much of it (have been meaning to rip out its type handling
> parts but havent had time).
>
>
> > I'm still temped by 5.3 work 1...
>
> Sounds a reasonable option...
>
> One more option to look at is if long doubles work for you - depends
> on your compiler though. This would migrate to 5.3 fairly smoothly
> (you could test on both).
>
> Justin

Thanks for that suggestion, Justin. Long doubles might be an interesting experiment. 

Another is to take the hint that Lua is providing me and to make this part of the library in C. If I'm honest, this might be the correct (simple) approach.  

- Andrew