Multi-dimensional Parallel Training of Winograd Layer on Memory-Centric Architecture.
Byungchul HongYeonju RoJohn KimPublished in: MICRO (2018)
Keyphrases
- multi dimensional
- processing elements
- level parallelism
- multi layer
- hierarchical architecture
- master slave
- multithreading
- memory hierarchy
- parallel processing
- parallel hardware
- multi core processors
- management system
- real time
- associative memory
- distributed processing
- middle layer
- memory usage
- memory footprint
- restricted boltzmann machine
- parallel processors
- parallel computers
- hardware architecture
- multi processor
- multi threaded
- massively parallel
- parallel implementation
- shared memory
- compute intensive
- range queries
- feedforward artificial neural networks
- abstraction layer
- distributed shared memory
- parallel architecture
- training process
- memory requirements
- neural network
- training set
- single instruction multiple data
- operating system
- memory subsystem
- training examples
- index structure
- memory bandwidth
- distributed memory
- memory management
- computational power
- parallel computing
- user centric
- computer architecture
- random access
- parallel computation