SMP vs MP version
Hi All, would like to know if anyone is using SMP and MP version of Flow-3D? would like to know which one is better and typically how many CPU core is used for weir flow or turbulent free surface flow?
Re: SMP vs MP version
what is the typical size of the mesh that you use? For MP bigger the better. For SMP usually a quad core system is used , so 4 processors on one core. For MP you will need a cluster, so basically more than one quad cores say up to 8 nodes. 1 node=1 core. SMP is based on multithreading so basically do loops are split across multiple processors. With MP the domain is decomposed (using multiple mesh blocks). Then each mesh block can be solved on different core. Since with SMP the RAM is shared is by the processors the speed up is limited. With MP the speed up will be more especially for problems where domain can be decomposed in multiple mesh blocks in such a way that each mesh block has some fluid, which i think is the case for hydraulic problems which is what you are interested in.
Really the best way to decide which version is best for your case to go for is to try running some test cases on both versions. You can send email to flow science at firstname.lastname@example.org regarding some trials. Regards, AC
|All times are GMT -4. The time now is 02:56.|