主题：一个问题请教下 -- 小愚

Auto parallelization is a tall billing for any compiler.

I haven't heard of any parallelizing compilers that can

deal with general purpose programs written in C-like

languages using threads.

The closest thing to what you ask is Itanium, which

is based on the VLIW idea. Theoretically, if there is

no dependency between the loops, the compiler will

issue instructions for muliple loops in one shot. Note

the word theoretically. The early product wasn't very

impressive. It might have improved but I don't know.

In addition, languages like HighPerformanceFortran

include additional tags added by the programmer to

guide the compiler's parallelization effort. But those

compilers were originally developed for Vector

computers. I'm not sure if they have been ported

to systems with general-purpose CPUs.

Just my 2c. A compiler or architecture major will

give you a better explanation.