Re: LuaJIT2 performance for number crunching

[T T <t34www@...>]

On 18 February 2011 00:50, Leo Razoumov <slonik.az@gmail.com> wrote:
> On Thu, Feb 17, 2011 at 17:42, T T <t34www@googlemail.com> wrote:
>> I ported your
>> rk4-unroll.lua to MATLAB and performance was terrible.  But for an ODE
>> of large dimension it will perform OK, because this overhead won't
>> matter that much.
>>
>
> In more complex cases like heavy event processing with lots of
> branching MATLAB JIT is easily defeated (LuaJIT possibly would not do
> great either). Couple of years ago we did a WiMAX system level
> simulator in MATLAB. Or boy, it was slow.
> Finally most of the heavy lifting was done in C-code and MATLAB just
> became an IO and configuration shell.

Your mileage may vary, as they say.  I don't know what you mean by
event processing, but I agree that certain algorithms do not map well
to MATLAB.  But discussing virtues of MATLAB (or lack thereof) is OT
here.

> P.S. And you pay through the noise as you go: for each CPU core, for
> each toolbox, for updates (via contract) etc.

I did mention that it cost you an arm and both legs ;)  That's one of
the reasons I'm looking now at Python as a replacement for it.

Cheers,

Tomek