Using hardware performance counters to speed up autotuning convergence on GPUs.
Jiri FilipovicJana HozzováAmin NezaratJaroslav OlhaFilip PetrovicPublished in: CoRR (2021)
Keyphrases
- graphics hardware
- graphics processing units
- digital signal processing
- low cost
- graphics cards
- computational power
- hardware and software
- parallel architectures
- real time
- general purpose
- commodity hardware
- convergence rate
- hardware implementation
- floating point
- gpu implementation
- highly parallel
- graphics processors
- processing units
- high end
- massively parallel
- parallel processing
- parallel programming
- vlsi implementation
- convergence speed
- fixed point
- parallel hardware
- heterogeneous computing