Mostra el registre d'ítem simple
Fast energy-optimal multi-kernel DNN-like application allocation on multi-FPGA platforms
dc.contributor.author | Shan, Junnan |
dc.contributor.author | Lazarescu, Mihai T. |
dc.contributor.author | Cortadella, Jordi |
dc.contributor.author | Lavagno, Luciano |
dc.contributor.author | Casu, Mario R. |
dc.contributor.other | Universitat Politècnica de Catalunya. Departament de Ciències de la Computació |
dc.date.accessioned | 2021-05-31T10:04:25Z |
dc.date.available | 2021-05-31T10:04:25Z |
dc.date.issued | 2022-04 |
dc.identifier.citation | Shan, J. [et al.]. Fast energy-optimal multi-kernel DNN-like application allocation on multi-FPGA platforms. "IEEE transactions on computer-aided design of integrated circuits and systems", Abril 2022, vol. 41, núm. 4, p. 1186-1190. |
dc.identifier.issn | 0278-0070 |
dc.identifier.uri | http://hdl.handle.net/2117/346425 |
dc.description.abstract | Platforms with multiple Field Programmable Gate Arrays (FPGAs), such as Amazon Web Services (AWS) F1 instances, can efficiently accelerate multi-kernel pipelined applications, e.g., Convolutional Neural Networks for machine vision tasks or transformer networks for Natural Language Processing tasks. To reduce energy consumption when the FPGAs are underutilized, we propose a model to (1) find off-line the minimum-power solution for given throughput constraints, and (2) dynamically reprogram the FPGA at runtime (which is complementary to dynamic voltage and frequency scaling) to match best the workloads when they change. The off-line optimization model can be solved using a Mixed-Integer Non-Linear Programming (MINLP) solver, but it can be very slow. Hence, we provide two heuristic optimization methods that improve result quality within a bounded time. We use several very large designs to demonstrate that both heuristics obtain comparable results to MINLP, when it can find the best solution, and they obtain much better results than MINLP, when it cannot find the optimum within a bounded amount of time. The heuristic methods can also be thousands of times faster than the MINLP solver. |
dc.format.extent | 5 p. |
dc.language.iso | eng |
dc.subject | Àrees temàtiques de la UPC::Informàtica::Informàtica teòrica |
dc.subject.lcsh | Field programmable gate arrays |
dc.subject.lcsh | Energy consumption |
dc.subject.other | CNN |
dc.subject.other | NLP |
dc.subject.other | Transformer |
dc.subject.other | Multi-FPGA |
dc.subject.other | Allocation |
dc.subject.other | Optimization |
dc.subject.other | Heuristic |
dc.subject.other | AWS |
dc.title | Fast energy-optimal multi-kernel DNN-like application allocation on multi-FPGA platforms |
dc.type | Article |
dc.subject.lemac | Matrius de portes programables per l'usuari |
dc.subject.lemac | Energia -- Consum |
dc.contributor.group | Universitat Politècnica de Catalunya. ALBCOM - Algorismia, Bioinformàtica, Complexitat i Mètodes Formals |
dc.identifier.doi | 10.1109/TCAD.2021.3076958 |
dc.description.peerreviewed | Peer Reviewed |
dc.relation.publisherversion | https://ieeexplore.ieee.org/document/9419915 |
dc.rights.access | Open Access |
local.identifier.drac | 31307389 |
dc.description.version | Postprint (author's final draft) |
local.citation.author | Shan, J.; Lazarescu, M.; Cortadella, J.; Lavagno, L.; Casu, M. |
local.citation.publicationName | IEEE transactions on computer-aided design of integrated circuits and systems |
local.citation.volume | 41 |
local.citation.number | 4 |
local.citation.startingPage | 1186 |
local.citation.endingPage | 1190 |
Fitxers d'aquest items
Aquest ítem apareix a les col·leccions següents
-
Articles de revista [1.049]
-
Articles de revista [274]