Concurrency in Lua's core language is co-operative (coroutines explicitly yield to each other). This is because the language designers don't want to impose any

Concurrency in Lua's core language is co-operative (coroutines explicitly yield to each other). This is because the language designers don't want to impose any particular threading architecture -- Lua is intentded to be as portable and non-opinionated about that sort of thing as possible. Its coroutines are intended for the same sort of issues as generators/callbacks/continuations, not for explicit parallelism. They won't do multi-core mapreduce, but they'll let you cleanly factor code that has multiple main loops. You can get multi-threaded/multi-process concurrency by messaging between separate Lua VMs, using libraries such as Lua Lanes (http://kotisivu.dnainternet.net/askok/bin/lanes/ http://kotisivu.dnainternet.net/askok/bin/lanes/). The Lua VM is implemented without static state, so several VMs can run in the same process without a GIL. There is LuaPi (https://github.com/LuaDist/luapi https://github.com/LuaDist/luapi), an implementation of the Pi calculus (a variant of CSP), but I don't really know much about it. I'm more familiar with Erlang-flavored concurrency models (and there's also ConcurrentLua, https://github.com/lefcha/concurrentlua https://github.com/lefcha/concurrentlua).