It is strange and so interesting (related with OpenMP).

fender0107401

Hello everyone:

As I didn't install any input method, I have to write this in English and sorry about that.

Here is the thing. I am learning OpenMP right now. I found that seems OpenMP only slow down my program.

For instance, I take the following codes to speed up something.

1. 
   
2. #pragma omp parallel for num_threads(x)
   

复制代码

But I found that "x = 1" will only need 40 second and:
x = 2 needs 50
x = 3 needs 60
x = 4 needs 70

Why?

When x = 4, the top is so fun!

fender0107401

Why it slow things down?

It is about my Intel I3 chip set?

bruceteen

没代码没真相
只能说，不是多线程就一定快的，尤其是互相影响时，人多则乱

fender0107401

I3 only contains two physical cores, but it show OS four cores and "omp_get_num_procs()" also return 4 for me. Maybe that is the reason or some part of the reason.

Anyway, I still hope OpenMP will improve the performance.

linux_c_py_php

代码是什么样的

cdtits

无码无真相，不如看看空空去...

fender0107401

I can show that codes, but I am sure the codes are fine to run in parallel.

One more interesting thing is I found my AMD Athlon 2.7 CPU (desktop) can complete the computation task in 40 seconds with "x = 1" and my Intel I3 2.4 CPU (laptop) also need 40 seconds to do that with "x = 1". :wink:

Also, the kernel version and compiler version is higher then my desktop.