Relaxations for High-Performance Message Passing on Massively Parallel SIMT Processors.
Benjamin KlenkHolger FröningHans EberleLarry DennisonPublished in: IPDPS (2017)
Keyphrases
- message passing
- massively parallel
- shared memory
- parallel computing
- parallel computers
- distributed memory
- message passing interface
- parallel machines
- belief propagation
- high performance computing
- mesh connected
- processing elements
- graphics processing units
- lp relaxation
- distributed systems
- parallel architectures
- parallel computation
- parallel programming
- factor graphs
- multithreading
- fine grained
- markov random field
- sum product algorithm
- parallel execution
- matrix multiplication
- lower bound
- sum product
- np hard
- tree reweighted
- linear programming
- inference in graphical models
- high quality
- computer architecture
- computing systems
- parallel processing
- graphical models