Decoupled Approach for CDT3D
- Take advantage of CDT3D performance and mesh quality capability and create a ‘coarse’ mesh
- Decompose the mesh into N subdomains (N > P)
- Distribute and refine each subdomain in parallel with no communication.
Due to the absence of appropriate software simple heuristics have been employed
Preliminary Results (May 2017)
|empty||User Handler threads||Total Time||Subdomain Migrations|
64 processes, 128 cores without ILB
64 processes, 128 cores with ILB.
42 processes, 126 cores with ILB.
25 processes, 125 cores with ILB.