https://hgpu.org/?p=13363
Parallel Implementation of the Finite Element Method on Graphics Processors for the Solution of Incompressible Flows