Re: [ANN] LuaJIT-2.0.0-beta1

[Geoff Leyland <geoff_leyland@...>]

(please excuse me replying to myself!)

On 1/11/2009, at 9:31 PM, Geoff Leyland wrote:
> I'm glad it was useful nonetheless!  I'll run some kdtree stuff  
> soon, and I'll be more careful about isolating relevant results.

Here are some more results (this time making it clear that LJ2 is much  
better!):

Read the DIMACS New York map, preprocess it, and write out:
    Total    Compute only
lua  60s       22s
LJ1  45s       17s
LJ2  35s       13s

Read the preprocessed map, build a kdtree and write it:
    Total    Compute only
lua  547s       526s
LJ1  145s       128s
LJ2  121s       108s

Times are measured with os.time(), so not much resolution.
I tried -jv, but it didn't give me a breakdown of what was interpreted  
and what was JIT compiled - how did you get that?

The kdtree code is in the same place as the heap stuff, but there's no  
test harness or data.  If you'd like that, let me know and I'll put  
something together.  The map preprocessing stuff will be harder (but  
possible) to isolate.

Cheers,
Geoff