This project focus on the multi-GPU parallel implementation of LBM(Lattice Boltzmann Method). We use OpenACC to accelerate codes on single GPU and MPI for inter-GPU communication.

Bold for completed, italic for coded but not documented. Bold and italic means we are now working on this.

• MPI (Pure) MPI implementation for multi processor, Including communication overlap(non-blocked version), and scalable decomposition of sub-domains. (sub-domains arranged in 3 dimensions.) contains:

  • Laplace Well-know jacobi iteration for testing
  • Lid_driven_cavity Lid-driven cavity flow.
  • Thermal_flow Thermal flow.
  • Particle_flow Particle flow.
  • Multi_Phase_flow Multiphase flow.

• OpenACC OpenACC accelerated codes for single GPU.

• MPI + OpenACC Multi-GPU solver.