MPI is used in several small and large clusters and supercomputers and supports both multiprocessor environments and heterogeneous architectures with GPUs and other accelerators. Several Message Passing Interfaces (MPIs) today offer support for transport plugins.
This master project is available for one or two students, and the tasks would be to:
- Select one open source MPI library.
- Create basic transport using standard PIO and RDMA functionality offered by Dolphin's SISCI API.
- Optimize collective operations by using PCI Express multicast functionality.
- Integrate and test with CUDA (nVidia GPUDirect) using PCI Express peer-to-peer functionality.
- Collaborate with the MPI open source development team to submit results to open source projects.
Dolphin will provide the required guidance on using SISCI and PCI Express interconnect cards, as well as the required hardware for testing and benchmarking.
Qualifications
Good low-level computer systems understanding. The student should have completed IN3000 or equivalent.