Conclusion

The main thrust of this work was to port OpenFOAM to HECToR and to give users directions on how to run their simulations efficiently. To this end, we have benchmarked a number of cases to give an idea of how many cores should be employed for a given number of mesh points and a given solver.

The results of the benchmarking is summarised in tabe 6.

Table 6: Benchmarking summary

Case	name	solver	No. of cells	Optimum no. of cores
3D cavity	`100'	icoFoam	1,000,000	256
3D cavity	`200.40'	icoFoam	8,000,000	1024
3D dam break	`dambreak'	interFoam	6,230,000	128
3D jet spray	`jet_coarse'	pisoFoam	300,000	64
3D jet spray	`jet_medium'	pisoFoam	2,500,000	256
3D jet spray	`jet_fine'	pisoFoam	19,000,000	512

It is clear from table 6 that there is no simply relation between the optimum number of cores to be used for a given solver and a given number of cells. However, we can state that OpenFOAM scales well on HECToR for both simple tutorial cases and for complex industrial cases.

OpenFOAM is proving to be a very popular CFD package. This is, in part, due to its open-source nature, wherein, unlike commercial CFD packages, users can examine the code and alter it as required (although code written by users is typically not included in future OpenFOAM releases). Further, the code scales well and does not require a per-core license one finds with commercial codes which can be financially prohibitive.