Re: LuaJIT2 performance for number crunching

[T T <t34www@...>]

On 18 February 2011 00:33, Leo Razoumov <slonik.az@gmail.com> wrote:
> On Thu, Feb 17, 2011 at 14:54, T T <t34www@googlemail.com> wrote:
>> On 17 February 2011 17:57, Leo Razoumov <slonik.az@gmail.com> wrote:
>>>
>>> The performance ratio is 40 times in favor of Lua-5.1.4 *interpreter*.
>>> I guess LuaJIT-2 will be even better.
>>> MATLAB sucks on loops!
>>
>> You do realize that you succeeded only in testing MATLAB's interpreter
>> and not jit, don't you?!? MATLAB doesn't jit scripts.  Slap a
>> 'function foo()' at the top and watch it run almost 80 times faster.
>> For N=1e8 I get 0.91sec with MATLAB and 0.68sec with LuaJIT-2.  Pretty
>> close I would say.
>>
>> Cheers,
>>
>
> If you had read my email carefully you would have noticed that I was
> comparing Lua-5.1.4 *interpreter* to MATLAB *interpreter* only. And
> yes, I am well aware that MATLAB JIT does not do scripts. I do not
> have LuaJIT on that machine and I had to make sure that I am comparing
> apples-to-apples.

Sir, your post was in direct reply to my comparison of MATLAB JIT to
LuaJIT, so please forgive me if I got confused as to what you were
trying to show.  MATLAB interpreter is slow, just as almost every
other interpreter out there (Python, Perl, Ruby, R, you name it), and
nobody disputes that.

And not to be nasty but you posted the same exact benchmark in the
past saying (emphasis mine) [1]:

  "It seems that MATLAB Windows port is better optimized but it is still
  about 100 times slower than LuaJIT-1. The joke is that MATLAB has its
  own JIT since version 6.5 and JIT *was enabled* for these tests."

[1] http://lua-users.org/lists/lua-l/2009-11/msg00357.html

So I wasn't exactly sure how much you know about jit in MATLAB.

Cheers,

Tomek