Re: LuaJIT2 performance for number crunching

[Francesco Abbate <francesco.bbt@...>]

2011/2/23 Mike Pall <mikelu-1102@mike.de>:
> I guess I have to completely redesign the loop unrolling heuristics.
> This will take me quite a bit of time, because I need to verify that
> it still performs well for many different benchmarks. Alas, other
> work has priority right now, so don't hold your breath.

Ok, thank you Mike, this is fine for me.

In the meantime I will keep working on other algorithms, I'm going to
finalize the adaptive integration routine. I'm also planning to
formalize a set of benchmarks to perform a set of tests in a more
systematic way.

And thanks again for your feedback. I'm aware that I'm in the easy
position of throwing algorithms and point with my fingers to problems
with JIT compilation and all the hard work is for you. For the other
side I believe that this kind of work is a good opportunity for
LuaJIT2 to evolve and became more reliable for complex algorithm
implementations. We have already a great tools and I'm sure that it is
going to improve very rapidly over a few problems that a young program
like LuaJIT2 can have.

Best regards,
Francesco