FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W3155444316> ?p ?o ?g. }

• W3155444316 abstract "The success of Deep Artificial Neural Networks (DNNs) in many domains created a rich body of research concerned with hardware accelerators for compute-intensive DNN operators. However, implementing such operators efficiently with complex instructions such as matrix multiply is a task not yet automated gracefully. Solving this task often requires complex program and memory layout transformations. First solutions to this problem have been proposed, such as TVM or ISAMIR, which work on a loop-level representation of operators and rewrite the program before an instruction embedding into the operator is performed. This top-down approach creates a tension between exploration range and search space complexity. In this work, we propose a new approach to this problem. We have created a bottom-up method that allows the direct generation of implementations based on an accelerator's instruction set. By formulating the embedding as a constraint satisfaction problem over the scalar dataflow, every possible embedding solution is contained in the search space. By adding additional constraints, a solver can produce the subset of preferable solutions. A detailed evaluation using the VTA hardware accelerator with the Baidu DeepBench inference benchmark suite shows that our approach can automatically generate code competitive to reference implementations, and furthermore that memory layout flexibilty can be beneficial for overall performance. While the reference implementation achieves very low hardware utilization due to its fixed embedding strategy, we achieve a geomean speedup of up to x2.49, while individual operators can improve as much as x238." @default.
• W3155444316 created "2021-04-26" @default.
• W3155444316 creator A5014251134 @default.
• W3155444316 creator A5053327260 @default.
• W3155444316 creator A5069480038 @default.
• W3155444316 date "2021-04-10" @default.
• W3155444316 modified "2023-09-27" @default.
• W3155444316 title "The Programming of Deep Learning Accelerators as a Constraint Satisfaction Problem." @default.
• W3155444316 cites W1530262073 @default.
• W3155444316 cites W1547410739 @default.
• W3155444316 cites W1667652561 @default.
• W3155444316 cites W1735683409 @default.
• W3155444316 cites W2046968552 @default.
• W3155444316 cites W2152839228 @default.
• W3155444316 cites W2167058518 @default.
• W3155444316 cites W2402144811 @default.
• W3155444316 cites W2442974303 @default.
• W3155444316 cites W2522802439 @default.
• W3155444316 cites W2585460399 @default.
• W3155444316 cites W2594730095 @default.
• W3155444316 cites W2773664881 @default.
• W3155444316 cites W2780077279 @default.
• W3155444316 cites W2786320458 @default.
• W3155444316 cites W2804032941 @default.
• W3155444316 cites W2914775474 @default.
• W3155444316 cites W2940862705 @default.
• W3155444316 cites W2944745903 @default.
• W3155444316 cites W2962993073 @default.
• W3155444316 cites W2963114857 @default.
• W3155444316 cites W2963947383 @default.
• W3155444316 cites W2963960923 @default.
• W3155444316 cites W2970971581 @default.
• W3155444316 cites W2995723163 @default.
• W3155444316 cites W3007772124 @default.
• W3155444316 cites W3035582633 @default.
• W3155444316 cites W3120332656 @default.
• W3155444316 hasPublicationYear "2021" @default.
• W3155444316 type Work @default.
• W3155444316 sameAs 3155444316 @default.
• W3155444316 citedByCount "0" @default.
• W3155444316 crossrefType "posted-content" @default.
• W3155444316 hasAuthorship W3155444316A5014251134 @default.
• W3155444316 hasAuthorship W3155444316A5053327260 @default.
• W3155444316 hasAuthorship W3155444316A5069480038 @default.
• W3155444316 hasConcept C13164978 @default.
• W3155444316 hasConcept C13280743 @default.
• W3155444316 hasConcept C154945302 @default.
• W3155444316 hasConcept C166957645 @default.
• W3155444316 hasConcept C173608175 @default.
• W3155444316 hasConcept C185798385 @default.
• W3155444316 hasConcept C199360897 @default.
• W3155444316 hasConcept C199622910 @default.
• W3155444316 hasConcept C205649164 @default.
• W3155444316 hasConcept C26713055 @default.
• W3155444316 hasConcept C2777904410 @default.
• W3155444316 hasConcept C2778770139 @default.
• W3155444316 hasConcept C41008148 @default.
• W3155444316 hasConcept C41608201 @default.
• W3155444316 hasConcept C49937458 @default.
• W3155444316 hasConcept C68339613 @default.
• W3155444316 hasConcept C79581498 @default.
• W3155444316 hasConcept C80444323 @default.
• W3155444316 hasConcept C95457728 @default.
• W3155444316 hasConcept C96324660 @default.
• W3155444316 hasConceptScore W3155444316C13164978 @default.
• W3155444316 hasConceptScore W3155444316C13280743 @default.
• W3155444316 hasConceptScore W3155444316C154945302 @default.
• W3155444316 hasConceptScore W3155444316C166957645 @default.
• W3155444316 hasConceptScore W3155444316C173608175 @default.
• W3155444316 hasConceptScore W3155444316C185798385 @default.
• W3155444316 hasConceptScore W3155444316C199360897 @default.
• W3155444316 hasConceptScore W3155444316C199622910 @default.
• W3155444316 hasConceptScore W3155444316C205649164 @default.
• W3155444316 hasConceptScore W3155444316C26713055 @default.
• W3155444316 hasConceptScore W3155444316C2777904410 @default.
• W3155444316 hasConceptScore W3155444316C2778770139 @default.
• W3155444316 hasConceptScore W3155444316C41008148 @default.
• W3155444316 hasConceptScore W3155444316C41608201 @default.
• W3155444316 hasConceptScore W3155444316C49937458 @default.
• W3155444316 hasConceptScore W3155444316C68339613 @default.
• W3155444316 hasConceptScore W3155444316C79581498 @default.
• W3155444316 hasConceptScore W3155444316C80444323 @default.
• W3155444316 hasConceptScore W3155444316C95457728 @default.
• W3155444316 hasConceptScore W3155444316C96324660 @default.
• W3155444316 hasLocation W31554443161 @default.
• W3155444316 hasOpenAccess W3155444316 @default.
• W3155444316 hasPrimaryLocation W31554443161 @default.
• W3155444316 hasRelatedWork W1569519819 @default.
• W3155444316 hasRelatedWork W2000059748 @default.
• W3155444316 hasRelatedWork W2128563800 @default.
• W3155444316 hasRelatedWork W2164314331 @default.
• W3155444316 hasRelatedWork W2248142725 @default.
• W3155444316 hasRelatedWork W2585460399 @default.
• W3155444316 hasRelatedWork W2768571798 @default.
• W3155444316 hasRelatedWork W2770593200 @default.
• W3155444316 hasRelatedWork W2808828039 @default.
• W3155444316 hasRelatedWork W2954698599 @default.
• W3155444316 hasRelatedWork W2966915742 @default.
• W3155444316 hasRelatedWork W3034789126 @default.
• W3155444316 hasRelatedWork W3035032999 @default.

• next