Optimizing MPI communication within large multicore nodes with kernel assistance.
Stéphanie MoreaudBrice GoglinRaymond NamystDavid GoodellPublished in: IPDPS Workshops (2010)
Keyphrases
- shared memory
- parallel programming
- general purpose
- parallel algorithm
- kernel function
- interconnection networks
- directed graph
- message passing
- feature space
- high performance computing
- fully distributed
- shortest path
- network connectivity
- parallel computing
- parallel implementation
- massively parallel
- support vector
- communication channels
- kernel matrix
- multiple kernel learning
- communication systems
- communication cost
- communication networks
- local area network
- network structure
- computer systems
- communication links
- parallelization strategy