Threadring

Here is some work to improve performance of c++ threadring implementation.

Why did I do it?

• For fun!
• To learn something new about threading
• To find out if there is way to "cheat" in C++ to beat Haskell

I did archive first and second motivations. But 3rd one is still not yet archived. It is possible to write very fast userspace threading using posix or GNU pth but they both are missing scheduler (posix) or scheduler is bad (GNU pth).

I think that best approach to thread ring problem is non-preemptive M:1 threading in core. Start up initialisation could use kernel threads to speed up settup (unless there is static initializers for everything).

Branches

Master: g++ 4.4 with -std=c++0x. It doesn't have all the latest fancy code

cpp98_combapility: Latest code that should compile with any recent g++

minimal: Provement that C++ code size can be small too. (Too bad it is pick 2 from small code, small memory foot print and lighting fast)

License

Thanks for interest.
Pauli