Systematic Approach in Optimizing Numerical Memory-Bound Kernels on GPU.
Ahmad AbdelfattahDavid E. KeyesHatem LtaiefPublished in: Euro-Par Workshops (2012)
Keyphrases
- real time
- upper bound
- memory usage
- data transfer
- qualitative and quantitative
- lower bound
- graphics hardware
- kernel function
- memory requirements
- parallel computing
- sensitivity analysis
- error bounds
- gpu accelerated
- parallel computation
- parallel implementation
- graphics processors
- limited memory
- memory space
- numerical data
- memory bandwidth
- intel xeon
- kernel methods
- computational power
- gaussian processes
- computing power
- graphics processing units
- associative memory
- main memory
- linear combination
- gpu implementation
- worst case
- support vector
- level parallelism
- machine learning
- data sets