{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,19]],"date-time":"2026-06-19T18:09:14Z","timestamp":1781892554220,"version":"3.54.5"},"reference-count":66,"publisher":"Association for Computing Machinery (ACM)","issue":"2","license":[{"start":{"date-parts":[[2021,12,6]],"date-time":"2021-12-06T00:00:00Z","timestamp":1638748800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.acm.org\/publications\/policies\/copyright_policy#Background"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Reconfigurable Technol. Syst."],"published-print":{"date-parts":[[2022,6,30]]},"abstract":"<jats:p>Customized compute acceleration in the datacenter is key to the wider roll-out of applications based on deep neural network (DNN) inference. In this article, we investigate how to maximize the performance and scalability of field-programmable gate array (FPGA)-based pipeline dataflow DNN inference accelerators (DFAs) automatically on computing infrastructures consisting of multi-die, network-connected FPGAs. We present Elastic-DF, a novel resource partitioning tool and associated FPGA runtime infrastructure that integrates with the DNN compiler FINN. Elastic-DF allocates FPGA resources to DNN layers and layers to individual FPGA dies to maximize the total performance of the multi-FPGA system. In the resulting Elastic-DF mapping, the accelerator may be instantiated multiple times, and each instance may be segmented across multiple FPGAs transparently, whereby the segments communicate peer-to-peer through 100\u00a0Gbps Ethernet FPGA infrastructure, without host involvement. When applied to ResNet-50, Elastic-DF provides a 44% latency decrease on Alveo U280. For MobileNetV1 on Alveo U200 and U280, Elastic-DF enables a 78% throughput increase, eliminating the performance difference between these cards and the larger Alveo U250. Elastic-DF also increases operating frequency in all our experiments, on average by over 20%. Elastic-DF therefore increases performance portability between different sizes of FPGA and increases the critical throughput per cost metric of datacenter inference.<\/jats:p>","DOI":"10.1145\/3470567","type":"journal-article","created":{"date-parts":[[2021,12,6]],"date-time":"2021-12-06T21:40:47Z","timestamp":1638826847000},"page":"1-34","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":32,"title":["Elastic-DF: Scaling Performance of DNN Inference in FPGA Clouds through Automatic Partitioning"],"prefix":"10.1145","volume":"15","author":[{"given":"Tobias","family":"Alonso","sequence":"first","affiliation":[{"name":"Universidad Aut\u00f3noma de Madrid, Madrid, Spain"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lucian","family":"Petrica","sequence":"additional","affiliation":[{"name":"Xilinx Research, Dublin, Ireland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mario","family":"Ruiz","sequence":"additional","affiliation":[{"name":"Xilinx University Program, Dublin, Ireland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jakoba","family":"Petri-Koenig","sequence":"additional","affiliation":[{"name":"Delft University of Technology, Delft, Netherlands"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yaman","family":"Umuroglu","sequence":"additional","affiliation":[{"name":"Xilinx Research, Dublin, Ireland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ioannis","family":"Stamelos","sequence":"additional","affiliation":[{"name":"InAccel, US"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Elias","family":"Koromilas","sequence":"additional","affiliation":[{"name":"InAccel, US"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Michaela","family":"Blott","sequence":"additional","affiliation":[{"name":"Xilinx Research, Dublin, Ireland"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Kees","family":"Vissers","sequence":"additional","affiliation":[{"name":"Xilinx Research, San Jos\u00e9, US"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"320","published-online":{"date-parts":[[2021,12,6]]},"reference":[{"key":"e_1_3_2_2_2","doi-asserted-by":"publisher","DOI":"10.1109\/HOTI.2017.13"},{"key":"e_1_3_2_3_2","unstructured":"Amazon AWS. 2018. Retrieved from https:\/\/aws.amazon.com\/ec2\/instance-types\/f1\/."},{"key":"e_1_3_2_4_2","doi-asserted-by":"publisher","DOI":"10.1109\/IPDPSW.2018.00032"},{"key":"e_1_3_2_5_2","doi-asserted-by":"publisher","DOI":"10.1145\/3234200.3234237"},{"key":"e_1_3_2_6_2","doi-asserted-by":"publisher","DOI":"10.1145\/3242897"},{"key":"e_1_3_2_7_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICFPT51103.2020.00037"},{"key":"e_1_3_2_8_2","volume-title":"Proceedings of the Asia-Pacific Intertional Symposium on Lattice Field Theory","volume":"2020","author":"Calg\u0300 Salvatore","year":"2020","unstructured":"Salvatore Calg\u0300, Grzegorz Korcyl, and Piotr Korcyl. 2020. Using Xilinx Alveo accelerators for lattice QCD. In Proceedings of the Asia-Pacific Intertional Symposium on Lattice Field Theory, Vol. 2020."},{"key":"e_1_3_2_9_2","doi-asserted-by":"publisher","DOI":"10.5555\/3195638.3195647"},{"key":"e_1_3_2_10_2","doi-asserted-by":"publisher","DOI":"10.1109\/MM.2018.022071131"},{"key":"e_1_3_2_11_2","article-title":"ONNX: Open Neural Network Exchange","author":"Community ONNX","unstructured":"ONNX Community. [n.d.]. ONNX: Open Neural Network Exchange. Retrieved from https:\/\/github.com\/onnx.","journal-title":"Retrieved from https:\/\/github.com\/onnx"},{"key":"e_1_3_2_12_2","doi-asserted-by":"publisher","DOI":"10.5555\/500824"},{"key":"e_1_3_2_13_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2009.5206848"},{"key":"e_1_3_2_14_2","doi-asserted-by":"publisher","DOI":"10.1109\/FPT.2016.7929550"},{"key":"e_1_3_2_15_2","doi-asserted-by":"publisher","DOI":"10.5555\/795659.795888"},{"key":"e_1_3_2_16_2","doi-asserted-by":"publisher","DOI":"10.1109\/ISCA.2018.00012"},{"key":"e_1_3_2_17_2","doi-asserted-by":"publisher","DOI":"10.1109\/MM.2019.2910506"},{"key":"e_1_3_2_18_2","doi-asserted-by":"publisher","DOI":"10.1109\/FCCM.2018.00018"},{"key":"e_1_3_2_19_2","article-title":"Vitis with 100 Gbps TCP\/IP Network Stack","author":"Group ETH Zurich Systems","unstructured":"ETH Zurich Systems Group. [n.d.]. Vitis with 100 Gbps TCP\/IP Network Stack. Retrieved from https:\/\/github.com\/fpgasystems\/Vitis_with_100Gbps_TCP-IP.","journal-title":"Retrieved from https:\/\/github.com\/fpgasystems\/Vitis_with_100Gbps_TCP-IP"},{"key":"e_1_3_2_20_2","doi-asserted-by":"publisher","DOI":"10.1145\/2554688.2554776"},{"key":"e_1_3_2_21_2","doi-asserted-by":"crossref","unstructured":"Mathew Hall and Vaughn Betz. 2020. HPIPE: Heterogeneous layer-pipelined and sparse-aware CNN inference for FPGAs. Retrieved from https:\/\/arXiv:2007.10451.","DOI":"10.1145\/3373087.3375380"},{"key":"e_1_3_2_22_2","unstructured":"Kaiming He Xiangyu Zhang Shaoqing Ren and Jian Sun. 2015. Deep residual learning for image recognition. Retrieved from https:\/\/arXiv:1512.03385."},{"key":"e_1_3_2_23_2","unstructured":"Andrew G. Howard Menglong Zhu Bo Chen Dmitry Kalenichenko Weijun Wang Tobias Weyand Marco Andreetto and Hartwig Adam. 2017. MobileNets: Efficient convolutional neural networks for mobile vision applications. Retrieved from http:\/\/arxiv.org\/abs\/1704.04861."},{"key":"e_1_3_2_24_2","unstructured":"Huawei. 2019. Retrieved from https:\/\/www.huaweicloud.com\/en-us\/product\/fcs.html."},{"key":"e_1_3_2_25_2","doi-asserted-by":"publisher","DOI":"10.1109\/WCSP.2011.6096913"},{"key":"e_1_3_2_26_2","unstructured":"Zhihao Jia Matei Zaharia and Alex Aiken. 2018. Beyond Data and Model Parallelism for Deep Neural Networks. Retrieved from https:\/\/arXiv:1807.05358."},{"key":"e_1_3_2_27_2","doi-asserted-by":"publisher","DOI":"10.1145\/3358192"},{"key":"e_1_3_2_28_2","doi-asserted-by":"publisher","DOI":"10.1145\/3140659.3080246"},{"key":"e_1_3_2_29_2","volume-title":"Performance evaluation of InAccel ML scalable suite","author":"Kachris Christoforos","year":"2018","unstructured":"Christoforos Kachris. 2018. Performance evaluation of InAccel ML scalable suite. Technical Report. InAccel. Retrieved from https:\/\/www.inaccel.com\/wp-content\/uploads\/inaccel_white_paper.pdf."},{"key":"e_1_3_2_30_2","doi-asserted-by":"publisher","DOI":"10.1145\/3373087.3375887"},{"key":"e_1_3_2_31_2","doi-asserted-by":"publisher","DOI":"10.1109\/FPL50879.2020.00053"},{"key":"e_1_3_2_32_2","doi-asserted-by":"publisher","DOI":"10.1145\/3377930.3389808"},{"key":"e_1_3_2_33_2","article-title":"FINN Dataflow Accelerator Examples","author":"Labs Xilinx Research","unstructured":"Xilinx Research Labs. [n.d.]. FINN Dataflow Accelerator Examples. Retrieved from https:\/\/github.com\/Xilinx\/finn-examples.","journal-title":"Retrieved from https:\/\/github.com\/Xilinx\/finn-examples"},{"key":"e_1_3_2_34_2","article-title":"FINN: Dataflow compiler for QNN inference on FPGAs","author":"Labs Xilinx Research","unstructured":"Xilinx Research Labs. [n.d.]. FINN: Dataflow compiler for QNN inference on FPGAs. Retrieved from https:\/\/github.com\/Xilinx\/finn.","journal-title":"Retrieved from https:\/\/github.com\/Xilinx\/finn"},{"key":"e_1_3_2_35_2","doi-asserted-by":"publisher","DOI":"10.1145\/2902961.2903025"},{"key":"e_1_3_2_36_2","doi-asserted-by":"publisher","DOI":"10.1109\/FPT.2015.7393136"},{"key":"e_1_3_2_37_2","article-title":"AIgean: An Open Framework for Machine Learning on a Heterogeneous Cluster","author":"Tarafdar Naif","unstructured":"Naif Tarafdar, Giuseppe Di Guglielmo, Philip C. Harris, Jeffrey D. Krupa, Vladimir Loncar, Dylan S. Rankin, Nhan Tran, Zhenbin Wu, Qianfeng Shen, and Paul Chow. [n.d.]. AIgean: An Open Framework for Machine Learning on a Heterogeneous Cluster. Retrieved from https:\/\/indico.cern.ch\/event\/924283\/contributions\/4105333\/attachments\/2154984\/3634529\/aigean_fastml.pdf.","journal-title":"Retrieved from https:\/\/indico.cern.ch\/event\/924283\/contributions\/4105333\/attachments\/2154984\/3634529\/aigean_fastml.pdf"},{"key":"e_1_3_2_38_2","doi-asserted-by":"publisher","DOI":"10.1109\/FCCM48280.2020.00010"},{"key":"e_1_3_2_39_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVLSI.2015.2478280"},{"key":"e_1_3_2_40_2","unstructured":"Nimbix. 2019. Retrieved from https:\/\/www.nimbix.net\/alveo."},{"key":"e_1_3_2_41_2","article-title":"Brevitas: Quantization-Aware Training in PyTorch","author":"Pappalardo Alessandro","unstructured":"Alessandro Pappalardo. [n.d.]. Brevitas: Quantization-Aware Training in PyTorch. Retrieved from https:\/\/github.com\/Xilinx\/brevitas.","journal-title":"Retrieved from https:\/\/github.com\/Xilinx\/brevitas"},{"key":"e_1_3_2_42_2","article-title":"Elastic-DF partitioner and FINN integration","author":"Petrica Lucian","unstructured":"Lucian Petrica. [n.d.]. Elastic-DF partitioner and FINN integration. Retrieved from https:\/\/github.com\/Xilinx\/finn-experimental\/blob\/main\/src\/finn\/util\/platforms.py.","journal-title":"Retrieved from https:\/\/github.com\/Xilinx\/finn-experimental\/blob\/main\/src\/finn\/util\/platforms.py"},{"key":"e_1_3_2_43_2","article-title":"Quantized ResNet-50 Dataflow Acceleration on Alveo","author":"Petrica Lucian","unstructured":"Lucian Petrica. [n.d.]. Quantized ResNet-50 Dataflow Acceleration on Alveo. Retrieved from https:\/\/github.com\/Xilinx\/ResNet50-PYNQ.","journal-title":"Retrieved from https:\/\/github.com\/Xilinx\/ResNet50-PYNQ"},{"key":"e_1_3_2_44_2","doi-asserted-by":"crossref","unstructured":"Lucian Petrica Tobias Alonso Mairin Kroes Nicholas Fraser Sorin Cotofana and Michaela Blott. 2020. Memory-Efficient Dataflow Inference for Deep CNNs on FPGA. Retrieved from https:\/\/arXiv:2011.07317.","DOI":"10.1109\/ICFPT51103.2020.00016"},{"key":"e_1_3_2_45_2","article-title":"Xilinx Adaptive Compute Cluster (XACC) Program","author":"Program Xilinx University","unstructured":"Xilinx University Program. [n.d.]. Xilinx Adaptive Compute Cluster (XACC) Program. Retrieved from https:\/\/www.xilinx.com\/support\/university\/XUP-XACC.html.","journal-title":"Retrieved from https:\/\/www.xilinx.com\/support\/university\/XUP-XACC.html"},{"key":"e_1_3_2_46_2","article-title":"XUP Vitis Network Example (VNx)","author":"Ruiz Mario","unstructured":"Mario Ruiz. [n.d.]. XUP Vitis Network Example (VNx). Retrieved from https:\/\/github.com\/Xilinx\/xup_vitis_network_example.","journal-title":"Retrieved from https:\/\/github.com\/Xilinx\/xup_vitis_network_example"},{"key":"e_1_3_2_47_2","doi-asserted-by":"publisher","DOI":"10.1109\/FPL.2019.00053"},{"key":"e_1_3_2_48_2","unstructured":"Kirk Saban. 2012. Xilinx Stacked Silicon Interconnect Technology Delivers Breakthrough FPGA Capacity Bandwidth and Power Efficiency. https:\/\/www.xilinx.com\/support\/documentation\/white_papers\/wp380_Stacked_Silicon_Interconnect_Technology.pdf."},{"key":"e_1_3_2_49_2","doi-asserted-by":"publisher","DOI":"10.1109\/TCAD.2020.2994256"},{"key":"e_1_3_2_50_2","doi-asserted-by":"publisher","DOI":"10.1109\/FCCM.2015.12"},{"key":"e_1_3_2_51_2","doi-asserted-by":"publisher","DOI":"10.1109\/FPL.2016.7577319"},{"key":"e_1_3_2_52_2","article-title":"IEEE standard for ethernet-amendment 2: Physical layer specifications and management parameters for 100 Gb\/s operation over backplanes and copper cables","author":"Society IEEE Computer","year":"2014","unstructured":"IEEE Computer Society. 2014. IEEE standard for ethernet-amendment 2: Physical layer specifications and management parameters for 100 Gb\/s operation over backplanes and copper cables. IEEE Std 802.3 bj-2014 (Amendment to IEEE Std 802.3-2014). Retrieved from https:\/\/standards.ieee.org\/standard\/802-3bj-2014.html.","journal-title":"IEEE Std 802.3 bj-2014 (Amendment to IEEE Std 802.3-2014)"},{"key":"e_1_3_2_53_2","article-title":"Inter-Kernel Links for Direct Inter-FPGA Communication","author":"Tetreault Roberto Dicecco Susanne M. Balle, Mark","year":"2020","unstructured":"Roberto Dicecco Susanne M. Balle, Mark Tetreault. 2020. Inter-Kernel Links for Direct Inter-FPGA Communication. Retrieved from https:\/\/www.intel.com\/content\/dam\/www\/programmable\/us\/en\/others\/literature\/wp\/wp-01305-inter-kernel-links-for-direct-inter-fpga-communication.pdf.","journal-title":"Retrieved from https:\/\/www.intel.com\/content\/dam\/www\/programmable\/us\/en\/others\/literature\/wp\/wp-01305-inter-kernel-links-for-direct-inter-fpga-communication.pdf"},{"key":"e_1_3_2_54_2","doi-asserted-by":"publisher","DOI":"10.1109\/JPROC.2017.2761740"},{"key":"e_1_3_2_55_2","doi-asserted-by":"publisher","DOI":"10.1109\/FCCM48280.2020.00072"},{"key":"e_1_3_2_56_2","doi-asserted-by":"publisher","DOI":"10.1109\/FCCM48280.2020.00072"},{"key":"e_1_3_2_57_2","doi-asserted-by":"publisher","DOI":"10.1109\/MM.2018.2877290"},{"key":"e_1_3_2_58_2","article-title":"The Python MIP Package","author":"Toffolo Haroldo G. Santos and Tulio A. M.","unstructured":"Haroldo G. Santos and Tulio A. M. Toffolo. [n.d.]. The Python MIP Package. Retrieved from https:\/\/www.python-mip.com\/.","journal-title":"Retrieved from https:\/\/www.python-mip.com\/"},{"key":"e_1_3_2_59_2","doi-asserted-by":"publisher","DOI":"10.1145\/3020078.3021744"},{"key":"e_1_3_2_60_2","unstructured":"Yaman Umuroglu and Magnus Jahre. 2017. Streamlined deployment for quantized neural networks. Retrieved from http:\/\/arxiv.org\/abs\/1709.04060."},{"key":"e_1_3_2_61_2","doi-asserted-by":"publisher","DOI":"10.1109\/FCCM.2016.22"},{"key":"e_1_3_2_62_2","doi-asserted-by":"publisher","DOI":"10.1109\/FCCM.2019.00021"},{"key":"e_1_3_2_63_2","doi-asserted-by":"publisher","DOI":"10.1109\/TC.2008.29"},{"key":"e_1_3_2_64_2","article-title":"Vitis AI Model Zoo","unstructured":"Xilinx. [n.d.]. Vitis AI Model Zoo. Retrieved from https:\/\/github.com\/Xilinx\/Vitis-AI\/tree\/master\/models\/AI-Model-Zoo.","journal-title":"https:\/\/github.com\/Xilinx\/Vitis-AI\/tree\/master\/models\/AI-Model-Zoo"},{"key":"e_1_3_2_65_2","doi-asserted-by":"publisher","DOI":"10.1109\/FPL.2019.00032"},{"key":"e_1_3_2_66_2","doi-asserted-by":"publisher","DOI":"10.1145\/2934583.2934644"},{"key":"e_1_3_2_67_2","doi-asserted-by":"publisher","DOI":"10.23919\/DATE.2019.8715174"}],"container-title":["ACM Transactions on Reconfigurable Technology and Systems"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3470567","content-type":"unspecified","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3470567","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,17]],"date-time":"2025-06-17T20:18:54Z","timestamp":1750191534000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3470567"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2021,12,6]]},"references-count":66,"journal-issue":{"issue":"2","published-print":{"date-parts":[[2022,6,30]]}},"alternative-id":["10.1145\/3470567"],"URL":"https:\/\/doi.org\/10.1145\/3470567","relation":{},"ISSN":["1936-7406","1936-7414"],"issn-type":[{"value":"1936-7406","type":"print"},{"value":"1936-7414","type":"electronic"}],"subject":[],"published":{"date-parts":[[2021,12,6]]},"assertion":[{"value":"2021-01-01","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2021-06-01","order":1,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2021-12-06","order":2,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}