- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
Hi everybody,
I'd like to share results of stress testing ofIntel OpenMP library for a 32-bit platform:
libiomp5md.dll
Intel OMP Performance Library
version 5.0.20111219 for IA-32 architecture
built on 2011-12-19 17:12:07 UTC
A purpose of the stress test was as follows:
Create as many as possible OpenMP threads in a parallel region on aone CPU system.
Details:
It isstated by Intel that up to 16,384 OpenMP threads could be created in a parallel region. This is not true,
because a simpletest application easily created more than 18,600 OpenMP threads on a 32-bit Windows XP
operating system ( 4GB of Virtual Memory ).
How would I compare Intel OpenMP library with Microsoft OpenMP library?
Intel's library is far more advancedand demonstrated outstanding capabilities (!).Microsoft's library
doesn't allow to create more than 64 OpenMP threads in a parallel region. Intel's OpenMP library could be used
forsophisticated scientific projects that require a veryhigh level of parallelization.
Fourscreenshootswill beenclosed.
Best regards,
Sergey
I'd like to share results of stress testing ofIntel OpenMP library for a 32-bit platform:
libiomp5md.dll
Intel OMP Performance Library
version 5.0.20111219 for IA-32 architecture
built on 2011-12-19 17:12:07 UTC
A purpose of the stress test was as follows:
Create as many as possible OpenMP threads in a parallel region on aone CPU system.
Details:
It isstated by Intel that up to 16,384 OpenMP threads could be created in a parallel region. This is not true,
because a simpletest application easily created more than 18,600 OpenMP threads on a 32-bit Windows XP
operating system ( 4GB of Virtual Memory ).
How would I compare Intel OpenMP library with Microsoft OpenMP library?
Intel's library is far more advancedand demonstrated outstanding capabilities (!).Microsoft's library
doesn't allow to create more than 64 OpenMP threads in a parallel region. Intel's OpenMP library could be used
forsophisticated scientific projects that require a veryhigh level of parallelization.
Fourscreenshootswill beenclosed.
Best regards,
Sergey
コピーされたリンク
10 返答(返信)
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
I found a very interesting article regarding limits of Windows platform(s):
http://blogs.technet.com/b/markrussinovich/archive/2009/07/08/3261309.aspx
http://blogs.technet.com/b/markrussinovich/archive/2009/07/08/3261309.aspx
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
I've done some review and unfortunately a Post #4 is a spam-post. It is not related to the subject of the thread.
Best regards,
Sergey
Best regards,
Sergey
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
Quoting billywesley
I am little bit comfused in terms of How it actually works? Can you explain it in brief?
I've created a smalltest-case in C++ that uses some OpenMP functionalityand compiled it with Intel C++ compiler.
I could post the source codes if you need and please confirm me.
Best regards,
Sergey
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
It means that theunion 'my_union_t' can not be splitted in two cache lines. Sorry for a little deviation from the main subject.
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
Quoting sureshgupta22
It means that theunion 'my_union_t' can not be splitted in two cache lines. Sorry for a little deviation from the main subject.
Sorry, but it looks like that statementshould beaddressedto a different thread.
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
It is not related to the subject of the thread.
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
>>>I found a very interesting article regarding limits of Windows platform(s):>>>
Thanks very insightful information.
