FullPack: Full Vector Utilization for Sub-Byte Quantized Inference on General Purpose CPUs.
Hossein KatebiNavidreza AsadiMaziar GoudarziPublished in: CoRR (2022)
Keyphrases
- general purpose
- special purpose
- graphics processing units
- parallel processing
- domain specific
- inference process
- probabilistic inference
- dynamic bayesian networks
- application specific
- bayesian model
- bayesian inference
- vector space
- inference mechanism
- sparse matrix
- neural network
- knowledge base
- tightly coupled
- error correcting codes
- random fields
- programming language
- multiresolution
- data structure
- similarity measure