Accelerating ML Recommendation with over a Thousand RISC-V/Tensor Processors on Esperanto's ET-SoC-1 Chip.
David R. DitzelRoger EspasaNivard AymerichAllen BaumTom BergJim BurrEric HaoJayesh IyerMiquel IzquierdoShankar JayaratnamDarren JonesChris KlingnerJin KimStephen LeeMarc LuponGrigorios MagklisBojan MaricRajib NathMike NeillyJ. Duane NorthcuttBill OrnerJose RenauGerard RevesXavier RevesTom RiordanPedro SanchezSridhar SamudralaGuillem SoleRaymond TangTommy ThornFrancisco TorresSebastia TortellaDaniel YauPublished in: HCS (2021)
Keyphrases
- low power consumption
- low power
- instruction set
- signal processor
- low cost
- memory subsystem
- power consumption
- high speed
- multithreading
- maximum likelihood
- recommender systems
- level parallelism
- single chip
- embedded systems
- parallel algorithm
- collaborative filtering
- parallel processing
- processor core
- memory access
- recommendation systems
- high order
- hardware and software
- application specific
- signal processing
- higher order
- high density
- user preferences
- real time
- cmos technology
- floating point
- application specific integrated circuits
- analog vlsi
- high end
- personalized recommendation
- parallel computing
- operating system
- tensor factorization
- single processor
- tensor decomposition
- hardware architecture
- parallel processors
- gray scale