Threading and multi-core systems

Question

Do you think the threadpooling design pattern is the way for a multi-core future?

The thread library, for example, if used widely, makes / forces the application writer

(1) to break the problem down into separate parallel jobs, which contributes to (enforcing :)) parallelism

(2) With abstraction from all low level OS calls, synchronization, etc. simplifies the life of a programmer. (Especially for C programmers :))

I strongly believe that this is the best way (or one of the "best" ways :)) for a multi-core future ...

So my question is what I write, thinking this way, or am I misled :)

Hi,

Microkernel

Answer 1

- , , . , .

: , , - , (Thread ).

, . , , , . , , .

, . . . , ( ) , .

. , ... .

Answer 2

, : "".

http://en.wikipedia.org/wiki/Embarrassingly_parallel

, ,

Answer 3

Windows NT , -. . IOCP , , , IOCP , - IOC ( ) , IOCP. IOCP . , -, IOCP ( -).

IOC -.

IOCP . IOCP , IOCP, . , . IOC , -, , , , , . , " ". , , IOCP, , .

Answer 4

I really like @yaankee's answer, but I would say that thread pooling is almost always the right way. Reason: The thread pool can degenerate into a simple static partitioning model for problems such as matrix matrix. OpenMP is guided by these lines.