Matches in SemOpenAlex for { <https://semopenalex.org/work/W3216953604> ?p ?o ?g. }
Showing items 1 to 100 of
100
with 100 items per page.
- W3216953604 abstract "This work focuses on an efficient Agile design methodology for domain-specific accelerators. We employ feature-by-feature enhancement of a vertical development stack and apply it to the TVM/VTA inference accelerator. We have enhanced the VTA design space and enabled end-to-end support for additional workloads. This has been accomplished by augmenting the VTA micro-architecture and instruction set architecture (ISA), as well as by enhancing the TVM compilation stack to support a wide range of VTA configs. The VTA tsim implementation (CHISEL-based) has been enhanced with fully pipelined versions of the ALU/GEMM execution units. In tsim, memory width can now range between 8-64 bytes. Field widths have been made more flexible to support larger scratchpads. New instructions have been added: element-wise 8-bit multiplication to support depthwise convolution, and load with a choice of pad values to support max pooling. Support for more layers and better double buffering has also been added. Fully pipelining ALU/GEMM helps significantly: 4.9x fewer cycles with minimal area change to run ResNet-18 under the default config. Configs featuring a further 11.5x decrease in cycle count at a cost of 12x greater area can be instantiated. Many points on the area-performance pareto curve are shown, showcasing the balance of execution unit sizing, memory interface width, and scratchpad sizing. Finally, VTA is now able to run Mobilenet 1.0 and all layers for ResNets, including the previously disabled pooling and fully connected layers. The TVM/VTA architecture has always featured end-to-end workload evaluation on RTL in minutes. With our modifications, it now offers a much greater number of feasible configurations with a wide range of cost vs. performance. All capabilities mentioned are available in opensource forks while a subset of these capabilities have already been upstreamed." @default.
- W3216953604 created "2021-12-06" @default.
- W3216953604 creator A5002419350 @default.
- W3216953604 creator A5027376294 @default.
- W3216953604 creator A5031030336 @default.
- W3216953604 creator A5052037169 @default.
- W3216953604 creator A5053009380 @default.
- W3216953604 creator A5057395052 @default.
- W3216953604 creator A5061222925 @default.
- W3216953604 creator A5081904415 @default.
- W3216953604 creator A5083178603 @default.
- W3216953604 date "2021-11-29" @default.
- W3216953604 modified "2023-09-26" @default.
- W3216953604 title "A Highly Configurable Hardware/Software Stack for DNN Inference Acceleration" @default.
- W3216953604 cites W1983394510 @default.
- W3216953604 cites W2002555321 @default.
- W3216953604 cites W2003313945 @default.
- W3216953604 cites W2006312753 @default.
- W3216953604 cites W2055312318 @default.
- W3216953604 cites W2100218206 @default.
- W3216953604 cites W2154790323 @default.
- W3216953604 cites W2186615578 @default.
- W3216953604 cites W2402144811 @default.
- W3216953604 cites W2471164860 @default.
- W3216953604 cites W2553303224 @default.
- W3216953604 cites W2562773490 @default.
- W3216953604 cites W2612445135 @default.
- W3216953604 cites W2786320458 @default.
- W3216953604 cites W2804032941 @default.
- W3216953604 cites W2804500013 @default.
- W3216953604 cites W2810610794 @default.
- W3216953604 cites W2868091835 @default.
- W3216953604 cites W2905135312 @default.
- W3216953604 cites W2906737788 @default.
- W3216953604 cites W2912012512 @default.
- W3216953604 cites W2961619211 @default.
- W3216953604 cites W2963114857 @default.
- W3216953604 cites W2963947383 @default.
- W3216953604 cites W2964259004 @default.
- W3216953604 cites W2970971581 @default.
- W3216953604 cites W3007772124 @default.
- W3216953604 cites W3008788679 @default.
- W3216953604 cites W3012249773 @default.
- W3216953604 cites W3031264475 @default.
- W3216953604 cites W3037712104 @default.
- W3216953604 cites W3088415669 @default.
- W3216953604 cites W3092300766 @default.
- W3216953604 cites W3096395190 @default.
- W3216953604 doi "https://doi.org/10.48550/arxiv.2111.15024" @default.
- W3216953604 hasPublicationYear "2021" @default.
- W3216953604 type Work @default.
- W3216953604 sameAs 3216953604 @default.
- W3216953604 citedByCount "0" @default.
- W3216953604 crossrefType "posted-content" @default.
- W3216953604 hasAuthorship W3216953604A5002419350 @default.
- W3216953604 hasAuthorship W3216953604A5027376294 @default.
- W3216953604 hasAuthorship W3216953604A5031030336 @default.
- W3216953604 hasAuthorship W3216953604A5052037169 @default.
- W3216953604 hasAuthorship W3216953604A5053009380 @default.
- W3216953604 hasAuthorship W3216953604A5057395052 @default.
- W3216953604 hasAuthorship W3216953604A5061222925 @default.
- W3216953604 hasAuthorship W3216953604A5081904415 @default.
- W3216953604 hasAuthorship W3216953604A5083178603 @default.
- W3216953604 hasBestOaLocation W32169536041 @default.
- W3216953604 hasConcept C118524514 @default.
- W3216953604 hasConcept C13280743 @default.
- W3216953604 hasConcept C149635348 @default.
- W3216953604 hasConcept C173608175 @default.
- W3216953604 hasConcept C185798385 @default.
- W3216953604 hasConcept C205649164 @default.
- W3216953604 hasConcept C41008148 @default.
- W3216953604 hasConcept C42935608 @default.
- W3216953604 hasConcept C68339613 @default.
- W3216953604 hasConceptScore W3216953604C118524514 @default.
- W3216953604 hasConceptScore W3216953604C13280743 @default.
- W3216953604 hasConceptScore W3216953604C149635348 @default.
- W3216953604 hasConceptScore W3216953604C173608175 @default.
- W3216953604 hasConceptScore W3216953604C185798385 @default.
- W3216953604 hasConceptScore W3216953604C205649164 @default.
- W3216953604 hasConceptScore W3216953604C41008148 @default.
- W3216953604 hasConceptScore W3216953604C42935608 @default.
- W3216953604 hasConceptScore W3216953604C68339613 @default.
- W3216953604 hasLocation W32169536041 @default.
- W3216953604 hasLocation W32169536042 @default.
- W3216953604 hasOpenAccess W3216953604 @default.
- W3216953604 hasPrimaryLocation W32169536041 @default.
- W3216953604 hasRelatedWork W1509211761 @default.
- W3216953604 hasRelatedWork W1800827217 @default.
- W3216953604 hasRelatedWork W1869243490 @default.
- W3216953604 hasRelatedWork W2116172876 @default.
- W3216953604 hasRelatedWork W2143904576 @default.
- W3216953604 hasRelatedWork W2470589840 @default.
- W3216953604 hasRelatedWork W2777249922 @default.
- W3216953604 hasRelatedWork W27867058 @default.
- W3216953604 hasRelatedWork W4294734199 @default.
- W3216953604 hasRelatedWork W4306156203 @default.
- W3216953604 isParatext "false" @default.
- W3216953604 isRetracted "false" @default.
- W3216953604 magId "3216953604" @default.
- W3216953604 workType "article" @default.