Search the Community
Showing results for tags 'multicore execution'.
I would like to utilize the mulitcore capabilities of GWB v. 8+, and while reading about this feature in the GWB transport manual, I became curious as to how the parallel algorithm actually subdivides the domain. I understand that the domain is broken into threads of m blocks, each containing n nodes. Each core handles a block, m, but I am wondering how the nodes the physical and chemical data are transferred between m1, m2, m3, etc. I think that with the Operator Splitting Method all transport terms are solved first, mobile species are redistributed, then the ensuing chemical reaction terms are evaluated. Obviously the spatial arrangement of the nodes matters, so how will the nodes in block m1 share data with m2, m3, etc? And how can I be sure that mass and charge balance are preserved?