{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2025,6,18]],"date-time":"2025-06-18T04:14:29Z","timestamp":1750220069664,"version":"3.41.0"},"reference-count":52,"publisher":"Association for Computing Machinery (ACM)","issue":"4","license":[{"start":{"date-parts":[[2023,7,24]],"date-time":"2023-07-24T00:00:00Z","timestamp":1690156800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.acm.org\/publications\/policies\/copyright_policy#Background"}],"funder":[{"name":"Programmi Operativi Nazionali"},{"name":"Ministero dell\u2019 Universit\u00e1e della Ricerca"},{"name":"eXact Lab S.R.L"},{"DOI":"10.13039\/501100005388","name":"Instituto Tecnol\u00f3gico de Costa Rica","doi-asserted-by":"crossref","id":[{"id":"10.13039\/501100005388","id-type":"DOI","asserted-by":"crossref"}]},{"name":"Master\u2019s scholarship programme from RidgeRun Embedded Solutions LLC"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Embed. Comput. Syst."],"published-print":{"date-parts":[[2023,7,31]]},"abstract":"<jats:p>\n            Low-power consumption and scarce computational resources limit the computation at the edge. Besides, the approximate computing paradigm reports promising techniques for designing accelerators to deal with inherent limitations of the edge, and high-level synthesis with C++ opens the opportunity to use meta-programming for specialisable generic design. This work proposes a framework for automatically generating synthesis-time configurable\n            <jats:bold>processing elements (PEs)<\/jats:bold>\n            for matrix multiplication-addition (GEMMA) and convolution. To evaluate our work, we perform a design exploration after varying data bit-width, operand sizes, and kernel sizes. Our analyses include resource consumption scaling, clocks-to-solution, design efficiency, and error distribution, presenting a comprehensive view of how the parameters affect the properties of our generic implementations. The GEMMA presented a trade-off between\n            <jats:italic>granularity vs efficiency<\/jats:italic>\n            , where large PEs with short data widths are favoured by the design efficiency, achieving, theoretically, up to 75 GMAC\/s on a Xilinx XC7Z020 @ 100 MHz with an efficiency of 27%. For design efficiency, we propose a figure of merit to evaluate operations per second and resource utilisation with respect to the maximum achievable by the FPGA. Regarding the convolution PEs, we implemented two algorithms: a window-based spatial convolution and Winograd. The former is the best in terms of performance with 150 GMAC\/s, reaching up to 47% of efficiency. Winograd also outperformed numerically using a 3\u00d7 3 kernel filter, presenting a mean error of 11.01% in 4-bits operands with a PSNR=16.28 dB, compared to the spatial convolution with 38.2% of mean error and PSNR=5.89 dB. Finally, we discuss how the error is mostly dependent on the PE\u2019s parameters. In the GEMMA, the error depends on the matrix size, causing limitations in the PE scaling but still applicable to accelerators. The PEs developed during this research will lead to further granular approximate accelerator research.\n          <\/jats:p>","DOI":"10.1145\/3594540","type":"journal-article","created":{"date-parts":[[2023,4,25]],"date-time":"2023-04-25T12:25:39Z","timestamp":1682425539000},"page":"1-27","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":6,"title":["Automatic Generation of Resource and Accuracy Configurable Processing Elements"],"prefix":"10.1145","volume":"22","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-3263-7853","authenticated-orcid":false,"given":"Luis G.","family":"Le\u00f3n-Vega","sequence":"first","affiliation":[{"name":"Universit\u00e1 degli Studi di Trieste, Italy and Instituto Tecnol\u00f3gico de Costa Rica, Costa Rica"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1715-4546","authenticated-orcid":false,"given":"Eduardo","family":"Salazar-Villalobos","sequence":"additional","affiliation":[{"name":"Instituto Tecnol\u00f3gico de Costa Rica, Costa Rica"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0007-6978-4826","authenticated-orcid":false,"given":"Alejandro","family":"Rodriguez-Figueroa","sequence":"additional","affiliation":[{"name":"Instituto Tecnol\u00f3gico de Costa Rica, Costa Rica"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4808-4904","authenticated-orcid":false,"given":"Jorge","family":"Castro-God\u00ednez","sequence":"additional","affiliation":[{"name":"Instituto Tecnol\u00f3gico de Costa Rica, Costa Rica"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"320","published-online":{"date-parts":[[2023,7,24]]},"reference":[{"key":"e_1_3_2_2_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.heliyon.2018.e00938"},{"key":"e_1_3_2_3_2","doi-asserted-by":"publisher","DOI":"10.7551\/mitpress\/13811.003.0007"},{"key":"e_1_3_2_4_2","doi-asserted-by":"publisher","DOI":"10.1109\/iCCECE49321.2020.9231243"},{"key":"e_1_3_2_5_2","doi-asserted-by":"publisher","DOI":"10.1109\/MDAT.2019.2952350"},{"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\/TCSI.2020.3048260"},{"key":"e_1_3_2_8_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICMTMA54903.2022.00008"},{"key":"e_1_3_2_9_2","doi-asserted-by":"publisher","DOI":"10.1145\/3289602.3293915"},{"key":"e_1_3_2_10_2","volume-title":"TinyML Research Symposium","author":"Fahim Farah","year":"2021","unstructured":"Farah Fahim, Benjamin Hawks, Christian Herwig, James Hirschauer, Sergo Jindariani, Nhan Tran, Luca P. Carloni, Giuseppe Di Guglielmo, Philip Harris, Jeffrey Krupa, Dylan Rankin, Manuel Blanco, Valentin Josiah Hester, Yingyi Luo, John Mamish, Seda Orgrenci-Memik, Thea Aarestaad, Hamza Javed, Vladimir Loncar, Maurizio Pierini, Adrian Alan Pol, Sioni Summers, Javier Duarte, Scott Hauck, Shih-Chieh Hsu, Jennifer Ngadiuba, Mia Liu, Duc Hoang, Edward Kreinar, Hawkeye360 Herndon, Zhenbin Wu, Manuel Blanco Valentin, and Josiah Hester. 2021. hls4ml: An open-source codesign workflow to empower scientific low-power machine learning devices. In TinyML Research Symposium. arxiv:2103.05579v1"},{"key":"e_1_3_2_11_2","doi-asserted-by":"publisher","DOI":"10.1016\/0031-3203(82)90024-3"},{"key":"e_1_3_2_12_2","unstructured":"Volta GPU and Carmel CPU. 2019. DATA SHEET NVIDIA Jetson AGX Xavier series System-on-Module. (2019) 1\u201372. https:\/\/developer.nvidia.com\/embedded\/downloads#?search=data&tx=$product jetson_agx_xavier."},{"key":"e_1_3_2_13_2","doi-asserted-by":"publisher","DOI":"10.1145\/3289185"},{"key":"e_1_3_2_14_2","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1510.00149"},{"key":"e_1_3_2_15_2","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1503.02531"},{"key":"e_1_3_2_16_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICPR.2010.579"},{"key":"e_1_3_2_17_2","doi-asserted-by":"publisher","DOI":"10.23919\/FPL.2017.8056786"},{"key":"e_1_3_2_18_2","volume-title":"Computer Architecture - A Quantitative Approach 5th Edition","author":"Patterson John L. Hennessy and David A.","year":"2012","unstructured":"John L. Hennessy and David A. Patterson. 2012. Computer Architecture - A Quantitative Approach 5th Edition. Vol. 53."},{"key":"e_1_3_2_19_2","doi-asserted-by":"publisher","DOI":"10.1145\/3341105.3373852"},{"key":"e_1_3_2_20_2","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1703.07131"},{"key":"e_1_3_2_21_2","doi-asserted-by":"crossref","unstructured":"Andrew Lavin and Scott Gray. 2015. Fast Algorithms for Convolutional Neural Networks. arxiv:1509.09308 [cs.NE]","DOI":"10.1109\/CVPR.2016.435"},{"key":"e_1_3_2_22_2","doi-asserted-by":"publisher","DOI":"10.1109\/5.726791"},{"key":"e_1_3_2_23_2","doi-asserted-by":"publisher","DOI":"10.1109\/ISOCC50952.2020.9333012"},{"key":"e_1_3_2_24_2","doi-asserted-by":"publisher","DOI":"10.1109\/ISOCC53507.2021.9613977"},{"key":"e_1_3_2_25_2","doi-asserted-by":"publisher","DOI":"10.1109\/DCAS53974.2022.9845501"},{"key":"e_1_3_2_26_2","doi-asserted-by":"publisher","DOI":"10.1109\/TSUSC.2021.3060690"},{"key":"e_1_3_2_27_2","doi-asserted-by":"crossref","unstructured":"Tailin Liang John Glossner Lei Wang Shaobo Shi and Xiaotong Zhang. 2021. Pruning and Quantization for Deep Neural Network Acceleration: A Survey. arxiv:2101.09671 [cs.CV]","DOI":"10.1016\/j.neucom.2021.07.045"},{"key":"e_1_3_2_28_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCD46524.2019.00076"},{"key":"e_1_3_2_29_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCSS55260.2022.9802420"},{"key":"e_1_3_2_30_2","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA.2016.7446050"},{"key":"e_1_3_2_31_2","unstructured":"Daisuke Miyashita Edward H. Lee and Boris Murmann. 2016. Convolutional neural networks using logarithmic data representation. (2016). arxiv:1603.01025http:\/\/arxiv.org\/abs\/1603.01025"},{"key":"e_1_3_2_32_2","unstructured":"Ehab Mohsen. 2018. Performance and bandwidth in a cost-optimized. 423 (2018) 1\u201314."},{"key":"e_1_3_2_33_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-662-00551-4_4"},{"key":"e_1_3_2_34_2","unstructured":"NVIDIA. 2014. DATA SHEET NVIDIA Jetson Nano System-on-Module. https:\/\/developer.nvidia.com\/embedded\/downloads."},{"key":"e_1_3_2_35_2","doi-asserted-by":"publisher","DOI":"10.5281\/zenodo.6413243"},{"key":"e_1_3_2_36_2","doi-asserted-by":"publisher","DOI":"10.1109\/CIRSYSSIM.2019.8935599"},{"key":"e_1_3_2_37_2","doi-asserted-by":"publisher","DOI":"10.5281\/zenodo.6413238"},{"key":"e_1_3_2_38_2","doi-asserted-by":"publisher","DOI":"10.1109\/MCAS.2020.3005467"},{"key":"e_1_3_2_39_2","doi-asserted-by":"publisher","DOI":"10.1109\/RIVF.2019.8713613"},{"key":"e_1_3_2_40_2","doi-asserted-by":"publisher","DOI":"10.1145\/3020078.3021744"},{"key":"e_1_3_2_41_2","doi-asserted-by":"publisher","DOI":"10.1109\/TVLSI.2009.2032289"},{"key":"e_1_3_2_42_2","doi-asserted-by":"publisher","DOI":"10.1109\/TIP.2003.819861"},{"key":"e_1_3_2_43_2","doi-asserted-by":"publisher","DOI":"10.1145\/1498765.1498785"},{"key":"e_1_3_2_44_2","doi-asserted-by":"publisher","DOI":"10.1137\/1.9781611970364"},{"key":"e_1_3_2_45_2","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA.2019.00048"},{"key":"e_1_3_2_46_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICFPT47387.2019.00084"},{"key":"e_1_3_2_47_2","unstructured":"Xilinx. 2021. Kria K26 SOM: The Ideal Platform for Vision AI at the Edge 18 pages. https:\/\/www.xilinx.com\/content\/dam\/xilinx\/support\/documentation\/white_papers\/wp529-som-benchmarks.pdf."},{"key":"e_1_3_2_48_2","unstructured":"Xilinx. u.d. Zynq-7000 SoC 18 pages. https:\/\/www.xilinx.com\/products\/silicon-devices\/soc\/zynq-7000.html."},{"key":"e_1_3_2_49_2","article-title":"Zynq-7000 SoC data sheet: Overview","volume":"190","author":"Inc. Xilinx","year":"2018","unstructured":"Xilinx Inc.2018. Zynq-7000 SoC data sheet: Overview. Xilinx DS190 (2018). https:\/\/www.xilinx.com\/support\/documentation\/data_sheets\/ds190-Zynq-7000-Overview.pdf.","journal-title":"Xilinx"},{"key":"e_1_3_2_50_2","unstructured":"Xilinx Inc.2021. DPUCAHX8L for Convolutional Neural Networks 27 pages. https:\/\/www.xilinx.com\/content\/dam\/xilinx\/support\/documentation\/ip_documentation\/dpucahx8l\/v1_0\/pg366-dpucahx8l.pdf."},{"key":"e_1_3_2_51_2","doi-asserted-by":"publisher","DOI":"10.1145\/3394885.3431632"},{"key":"e_1_3_2_52_2","doi-asserted-by":"publisher","DOI":"10.1109\/TCAD.2017.2785257"},{"key":"e_1_3_2_53_2","doi-asserted-by":"publisher","DOI":"10.1145\/3400302.3415609"}],"container-title":["ACM Transactions on Embedded Computing Systems"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3594540","content-type":"unspecified","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3594540","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,17]],"date-time":"2025-06-17T18:09:07Z","timestamp":1750183747000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3594540"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2023,7,24]]},"references-count":52,"journal-issue":{"issue":"4","published-print":{"date-parts":[[2023,7,31]]}},"alternative-id":["10.1145\/3594540"],"URL":"https:\/\/doi.org\/10.1145\/3594540","relation":{},"ISSN":["1539-9087","1558-3465"],"issn-type":[{"type":"print","value":"1539-9087"},{"type":"electronic","value":"1558-3465"}],"subject":[],"published":{"date-parts":[[2023,7,24]]},"assertion":[{"value":"2022-04-05","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2023-04-16","order":1,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2023-07-24","order":2,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}