what's the reason for low speedup on windows operating system?

[GoogleCodeExporter]

I spawn the code on windows platform and using MPICH2 to develop.
My hardware platform is on two workstation with multicore(intel Xeon 2CPU 
with E5435).So I have 16 cores to use.The communication between two 
workstaion is using high speed lan. But I got low speedup on it. Could you 
tell me the reason? Thanks.
The speedup figures can be found in following figs. The communication time 
sometimes are longer than the computation time.

Original issue reported on code.google.com by polygan...@gmail.com on 5 May 2008 at 11:24

[GoogleCodeExporter]

Original comment by polygan...@gmail.com on 5 May 2008 at 11:25

Attachments:

• [two workstation computation core16.zip](https://storage.googleapis.com/google-code-attachments/psvm/issue-1/comment-1/two workstation computation core16.zip)

[GoogleCodeExporter]

Hi William,

For you problem, the problem size is too small.
If the dataset could be larger, most time will be spent on the step E = I + H^T 
* D *
H and this step could archive good speedup. For small dataset, time is mostly 
spent
on other steps (such as CF: cholesky factorization) and in this case, the 
speedup is low.

Original comment by baihong...@gmail.com on 5 May 2008 at 12:17

[GoogleCodeExporter]

Thanks. Let me have a try on larger dataset.
Rcv1 dataset is selected to test it.

Original comment by polygan...@gmail.com on 6 May 2008 at 9:46

[GoogleCodeExporter]

Hi, 
After do some large dataset, i found the psvm is a large memory-consuming code 
during the ICF 
period.The RCV1 and ijcnn1 dataset are tested on my cluster. The result is not 
so well. 

Original comment by polygan...@gmail.com on 9 May 2008 at 10:30