Matches in SemOpenAlex for { <https://semopenalex.org/work/W3188766293> ?p ?o ?g. }
Showing items 1 to 100 of
100
with 100 items per page.
- W3188766293 abstract "Deep neural networks are widely used in personalized recommendation systems. Such models involve two major components: the memory-bound embedding layer and the computation-bound fully-connected layers. Existing solutions are either slow on both stages or only optimize one of them. To implement recommendation inference efficiently in the context of a real deployment, we design and implement an FPGA cluster optimizing the performance of both stages. To remove the memory bottleneck, we take advantage of the High-Bandwidth Memory (HBM) available on the latest FPGAs for highly concurrent embedding table lookups. To match the required DNN computation throughput, we partition the workload across multiple FPGAs interconnected via a 100 Gbps TCP/IP network. Compared to an optimized CPU baseline (16 vCPU, AVX2-enabled) and a one-node FPGA implementation, our system (four-node version) achieves 28.95× and 7.68× speedup in terms of throughput respectively. The proposed system also guarantees a latency of tens of microseconds per single inference, significantly better than CPU and GPU-based systems which take at least milliseconds." @default.
- W3188766293 created "2021-08-16" @default.
- W3188766293 creator A5018202004 @default.
- W3188766293 creator A5028753892 @default.
- W3188766293 creator A5032503782 @default.
- W3188766293 creator A5035131147 @default.
- W3188766293 creator A5057864403 @default.
- W3188766293 creator A5069651430 @default.
- W3188766293 date "2021-08-01" @default.
- W3188766293 modified "2023-10-03" @default.
- W3188766293 title "Distributed Recommendation Inference on FPGA Clusters" @default.
- W3188766293 cites W1563299586 @default.
- W3188766293 cites W2028863828 @default.
- W3188766293 cites W2048266589 @default.
- W3188766293 cites W2092736553 @default.
- W3188766293 cites W2154815337 @default.
- W3188766293 cites W2475334473 @default.
- W3188766293 cites W2475840367 @default.
- W3188766293 cites W2512971201 @default.
- W3188766293 cites W2605350416 @default.
- W3188766293 cites W2615256897 @default.
- W3188766293 cites W2625954420 @default.
- W3188766293 cites W2761070740 @default.
- W3188766293 cites W2798956872 @default.
- W3188766293 cites W2883929540 @default.
- W3188766293 cites W2901622660 @default.
- W3188766293 cites W2962745591 @default.
- W3188766293 cites W2962953210 @default.
- W3188766293 cites W2973172293 @default.
- W3188766293 cites W2979719709 @default.
- W3188766293 cites W2987319672 @default.
- W3188766293 cites W3016842236 @default.
- W3188766293 cites W3022949655 @default.
- W3188766293 cites W3042495273 @default.
- W3188766293 cites W3043023836 @default.
- W3188766293 cites W3043433718 @default.
- W3188766293 cites W3102169921 @default.
- W3188766293 cites W4239385313 @default.
- W3188766293 cites W4256629673 @default.
- W3188766293 doi "https://doi.org/10.1109/fpl53798.2021.00057" @default.
- W3188766293 hasPublicationYear "2021" @default.
- W3188766293 type Work @default.
- W3188766293 sameAs 3188766293 @default.
- W3188766293 citedByCount "4" @default.
- W3188766293 countsByYear W31887662932022 @default.
- W3188766293 countsByYear W31887662932023 @default.
- W3188766293 crossrefType "proceedings-article" @default.
- W3188766293 hasAuthorship W3188766293A5018202004 @default.
- W3188766293 hasAuthorship W3188766293A5028753892 @default.
- W3188766293 hasAuthorship W3188766293A5032503782 @default.
- W3188766293 hasAuthorship W3188766293A5035131147 @default.
- W3188766293 hasAuthorship W3188766293A5057864403 @default.
- W3188766293 hasAuthorship W3188766293A5069651430 @default.
- W3188766293 hasBestOaLocation W31887662932 @default.
- W3188766293 hasConcept C111919701 @default.
- W3188766293 hasConcept C120314980 @default.
- W3188766293 hasConcept C149635348 @default.
- W3188766293 hasConcept C154945302 @default.
- W3188766293 hasConcept C157764524 @default.
- W3188766293 hasConcept C173608175 @default.
- W3188766293 hasConcept C2776214188 @default.
- W3188766293 hasConcept C2780513914 @default.
- W3188766293 hasConcept C41008148 @default.
- W3188766293 hasConcept C42935608 @default.
- W3188766293 hasConcept C555944384 @default.
- W3188766293 hasConcept C68339613 @default.
- W3188766293 hasConcept C76155785 @default.
- W3188766293 hasConcept C82876162 @default.
- W3188766293 hasConceptScore W3188766293C111919701 @default.
- W3188766293 hasConceptScore W3188766293C120314980 @default.
- W3188766293 hasConceptScore W3188766293C149635348 @default.
- W3188766293 hasConceptScore W3188766293C154945302 @default.
- W3188766293 hasConceptScore W3188766293C157764524 @default.
- W3188766293 hasConceptScore W3188766293C173608175 @default.
- W3188766293 hasConceptScore W3188766293C2776214188 @default.
- W3188766293 hasConceptScore W3188766293C2780513914 @default.
- W3188766293 hasConceptScore W3188766293C41008148 @default.
- W3188766293 hasConceptScore W3188766293C42935608 @default.
- W3188766293 hasConceptScore W3188766293C555944384 @default.
- W3188766293 hasConceptScore W3188766293C68339613 @default.
- W3188766293 hasConceptScore W3188766293C76155785 @default.
- W3188766293 hasConceptScore W3188766293C82876162 @default.
- W3188766293 hasLocation W31887662931 @default.
- W3188766293 hasLocation W31887662932 @default.
- W3188766293 hasOpenAccess W3188766293 @default.
- W3188766293 hasPrimaryLocation W31887662931 @default.
- W3188766293 hasRelatedWork W1509211761 @default.
- W3188766293 hasRelatedWork W1531488649 @default.
- W3188766293 hasRelatedWork W1585350690 @default.
- W3188766293 hasRelatedWork W2133693067 @default.
- W3188766293 hasRelatedWork W2366027386 @default.
- W3188766293 hasRelatedWork W2391299576 @default.
- W3188766293 hasRelatedWork W2540591044 @default.
- W3188766293 hasRelatedWork W2582456645 @default.
- W3188766293 hasRelatedWork W3037767301 @default.
- W3188766293 hasRelatedWork W4318464993 @default.
- W3188766293 isParatext "false" @default.
- W3188766293 isRetracted "false" @default.
- W3188766293 magId "3188766293" @default.
- W3188766293 workType "article" @default.