Re: Introducing Lua nodes

[Marc Balmer <marc@...>]

Am 06.06.2014 um 09:08 schrieb Michal Kolodziejczyk <miko@wp.pl>:

> 
> On 05.06.2014 22:34, Marc Balmer wrote:
>> Lua nodes is a "different" way to execute Lua code.  It combines three
>> existing technologies:  Lua, POSIX threads, and, 0MQ (zeromq.orq). 
> 
> Looks great. Does it work with luajit? Is it released yet?

I don't see how it should not work with LuaJIT, since it is a C binding.  Unfortunately I did not yet find the time to release it, and since I will be travelling the next weeks, it will likely stay that way for a few more days/weeks.

In the meantime, I will work on the code and make some more experiments with it, since I plan to use it in production, I expect it to be evry solid.

FWIW, the big difference to existing Lua ZeroMQ bindings is that it creates the ZeroMQ context once in the "node" process and then runs n Lua states in parallel.  Existing zmq Lua bindings allow you to use ZeroMQ in a Lua state, but they do, of course, not provide the POSIX threading stuff.

> 
> +1 for name change, like "paranodes - parallel lua nodes".
> Or "nodulo" - "node" in Spanish.

When I chose "node" I did not think to much about clashes, it merely describes what it is "a computing node".  I will think of a better name over the next days, luamq, mqlua, zmqlua, plua, tlua come to my mind.  Maybe mqlua would be an not-too-bad name for the binary (I prefer it over zmqlua in case one day the underlying MQ framework would be changed).

> 
>> Comments, suggestions, and, ideas welcome.
> 
> One idea taken from julialang.org - make parallel computing transparent
> (independent on number of cores/machines):
> 
> http://docs.julialang.org/en/release-0.2/manual/parallel-computing/

Thanks for that link!

> 
> Regards,
> miko
> 


- Marc