Challenges of GPU-aware Communication in MPI.
Nathan HanfordRamesh PankajakshanEdgar A. LeónIan KarlinPublished in: ExaMPI@SC (2020)
Keyphrases
- parallel implementation
- parallel computing
- lessons learned
- parallel algorithm
- high performance computing
- parallel programming
- message passing
- communication overhead
- technical challenges
- communication systems
- communication cost
- key issues
- parallel computation
- graphics hardware
- real world
- real time
- parallelization strategy
- communication protocol
- multithreading
- message passing interface
- shared memory
- communication networks
- information systems
- neural network
- communication channels
- computer networks
- fine grained
- graphical models
- open issues
- general purpose
- data sets