FPGA Acceleration of Probabilistic Sentential Decision Diagrams with High-level Synthesis

Probabilistic Sentential Decision Diagrams (PSDDs) provide efficient methods for modeling and reasoning with probability distributions in the presence of massive logical constraints. PSDDs can also be synthesized from graphical models such as Bayesian networks (BNs) therefore offering a new set of t...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:ACM transactions on reconfigurable technology and systems 2023-03, Vol.16 (2), p.1-22, Article 18
Hauptverfasser: Choi, Young-Kyu, Santillana, Carlos, Shen, Yujia, Darwiche, Adnan, Cong, Jason
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page 22
container_issue 2
container_start_page 1
container_title ACM transactions on reconfigurable technology and systems
container_volume 16
creator Choi, Young-Kyu
Santillana, Carlos
Shen, Yujia
Darwiche, Adnan
Cong, Jason
description Probabilistic Sentential Decision Diagrams (PSDDs) provide efficient methods for modeling and reasoning with probability distributions in the presence of massive logical constraints. PSDDs can also be synthesized from graphical models such as Bayesian networks (BNs) therefore offering a new set of tools for performing inference on these models (in time linear in the PSDD size). Despite these favorable characteristics of PSDDs, we have found multiple challenges in PSDD’s FPGA acceleration. Problems include limited parallelism, data dependency, and small pipeline iterations. In this article, we propose several optimization techniques to solve these issues with novel pipeline scheduling and parallelization schemes. We designed the PSDD kernel with a high-level synthesis (HLS) tool for ease of implementation and verified it on the Xilinx Alveo U250 board. Experimental results show that our methods improve the baseline FPGA HLS implementation performance by 2,200X and the multicore CPU implementation by 20X. The proposed design also outperforms state-of-the-art BN and Sum Product Network (SPN) accelerators that store the graph information in memory.
doi_str_mv 10.1145/3561514
format Article
fullrecord <record><control><sourceid>acm_cross</sourceid><recordid>TN_cdi_crossref_primary_10_1145_3561514</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>3561514</sourcerecordid><originalsourceid>FETCH-LOGICAL-a277t-2a1cff5cfdf59606977d4aaa9a0c4014d1cc4cb1dfef836973ee4280907c0a4a3</originalsourceid><addsrcrecordid>eNo9kM9LwzAcxYMoOKd495Sbp2rSpEl7LJvbhIGD6cFT-fbbZI30hyRF2X9vx-bgwXvwPrzDI-SesyfOZfIsEsUTLi_IhGdCRVpyeXnOTF2TmxC-GFNCpXJCPhebZU5zRNMYD4PrO9pbuvF9CaVrXBgc0q3phlEOGjo36MIBmjvYeWgD_XVDTVduV0eN-TEN3e67oTbBhVtyZaEJ5u7kU_KxeHmfraL12_J1lq8jiLUeohg4WpugrWySKaYyrSsJABkwlIzLiiNKLHlljU3FWAtjZJyyjGlkIEFMyeNxF30fgje2-PauBb8vOCsOjxSnR0by4UgCtmfov_wD_fJb8g</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>FPGA Acceleration of Probabilistic Sentential Decision Diagrams with High-level Synthesis</title><source>ACM Digital Library Complete</source><creator>Choi, Young-Kyu ; Santillana, Carlos ; Shen, Yujia ; Darwiche, Adnan ; Cong, Jason</creator><creatorcontrib>Choi, Young-Kyu ; Santillana, Carlos ; Shen, Yujia ; Darwiche, Adnan ; Cong, Jason</creatorcontrib><description>Probabilistic Sentential Decision Diagrams (PSDDs) provide efficient methods for modeling and reasoning with probability distributions in the presence of massive logical constraints. PSDDs can also be synthesized from graphical models such as Bayesian networks (BNs) therefore offering a new set of tools for performing inference on these models (in time linear in the PSDD size). Despite these favorable characteristics of PSDDs, we have found multiple challenges in PSDD’s FPGA acceleration. Problems include limited parallelism, data dependency, and small pipeline iterations. In this article, we propose several optimization techniques to solve these issues with novel pipeline scheduling and parallelization schemes. We designed the PSDD kernel with a high-level synthesis (HLS) tool for ease of implementation and verified it on the Xilinx Alveo U250 board. Experimental results show that our methods improve the baseline FPGA HLS implementation performance by 2,200X and the multicore CPU implementation by 20X. The proposed design also outperforms state-of-the-art BN and Sum Product Network (SPN) accelerators that store the graph information in memory.</description><identifier>ISSN: 1936-7406</identifier><identifier>EISSN: 1936-7414</identifier><identifier>DOI: 10.1145/3561514</identifier><language>eng</language><publisher>New York, NY: ACM</publisher><subject>Computer systems organization ; High-level language architectures ; Reconfigurable computing</subject><ispartof>ACM transactions on reconfigurable technology and systems, 2023-03, Vol.16 (2), p.1-22, Article 18</ispartof><rights>Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee. Request permissions from</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-a277t-2a1cff5cfdf59606977d4aaa9a0c4014d1cc4cb1dfef836973ee4280907c0a4a3</citedby><cites>FETCH-LOGICAL-a277t-2a1cff5cfdf59606977d4aaa9a0c4014d1cc4cb1dfef836973ee4280907c0a4a3</cites><orcidid>0000-0001-5829-4425 ; 0000-0002-5512-6343 ; 0000-0003-3976-6735 ; 0000-0003-2887-6963 ; 0000-0002-4350-7720</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktopdf>$$Uhttps://dl.acm.org/doi/pdf/10.1145/3561514$$EPDF$$P50$$Gacm$$H</linktopdf><link.rule.ids>314,780,784,2282,27924,27925,40196,76228</link.rule.ids></links><search><creatorcontrib>Choi, Young-Kyu</creatorcontrib><creatorcontrib>Santillana, Carlos</creatorcontrib><creatorcontrib>Shen, Yujia</creatorcontrib><creatorcontrib>Darwiche, Adnan</creatorcontrib><creatorcontrib>Cong, Jason</creatorcontrib><title>FPGA Acceleration of Probabilistic Sentential Decision Diagrams with High-level Synthesis</title><title>ACM transactions on reconfigurable technology and systems</title><addtitle>ACM TRETS</addtitle><description>Probabilistic Sentential Decision Diagrams (PSDDs) provide efficient methods for modeling and reasoning with probability distributions in the presence of massive logical constraints. PSDDs can also be synthesized from graphical models such as Bayesian networks (BNs) therefore offering a new set of tools for performing inference on these models (in time linear in the PSDD size). Despite these favorable characteristics of PSDDs, we have found multiple challenges in PSDD’s FPGA acceleration. Problems include limited parallelism, data dependency, and small pipeline iterations. In this article, we propose several optimization techniques to solve these issues with novel pipeline scheduling and parallelization schemes. We designed the PSDD kernel with a high-level synthesis (HLS) tool for ease of implementation and verified it on the Xilinx Alveo U250 board. Experimental results show that our methods improve the baseline FPGA HLS implementation performance by 2,200X and the multicore CPU implementation by 20X. The proposed design also outperforms state-of-the-art BN and Sum Product Network (SPN) accelerators that store the graph information in memory.</description><subject>Computer systems organization</subject><subject>High-level language architectures</subject><subject>Reconfigurable computing</subject><issn>1936-7406</issn><issn>1936-7414</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><recordid>eNo9kM9LwzAcxYMoOKd495Sbp2rSpEl7LJvbhIGD6cFT-fbbZI30hyRF2X9vx-bgwXvwPrzDI-SesyfOZfIsEsUTLi_IhGdCRVpyeXnOTF2TmxC-GFNCpXJCPhebZU5zRNMYD4PrO9pbuvF9CaVrXBgc0q3phlEOGjo36MIBmjvYeWgD_XVDTVduV0eN-TEN3e67oTbBhVtyZaEJ5u7kU_KxeHmfraL12_J1lq8jiLUeohg4WpugrWySKaYyrSsJABkwlIzLiiNKLHlljU3FWAtjZJyyjGlkIEFMyeNxF30fgje2-PauBb8vOCsOjxSnR0by4UgCtmfov_wD_fJb8g</recordid><startdate>20230311</startdate><enddate>20230311</enddate><creator>Choi, Young-Kyu</creator><creator>Santillana, Carlos</creator><creator>Shen, Yujia</creator><creator>Darwiche, Adnan</creator><creator>Cong, Jason</creator><general>ACM</general><scope>AAYXX</scope><scope>CITATION</scope><orcidid>https://orcid.org/0000-0001-5829-4425</orcidid><orcidid>https://orcid.org/0000-0002-5512-6343</orcidid><orcidid>https://orcid.org/0000-0003-3976-6735</orcidid><orcidid>https://orcid.org/0000-0003-2887-6963</orcidid><orcidid>https://orcid.org/0000-0002-4350-7720</orcidid></search><sort><creationdate>20230311</creationdate><title>FPGA Acceleration of Probabilistic Sentential Decision Diagrams with High-level Synthesis</title><author>Choi, Young-Kyu ; Santillana, Carlos ; Shen, Yujia ; Darwiche, Adnan ; Cong, Jason</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a277t-2a1cff5cfdf59606977d4aaa9a0c4014d1cc4cb1dfef836973ee4280907c0a4a3</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Computer systems organization</topic><topic>High-level language architectures</topic><topic>Reconfigurable computing</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Choi, Young-Kyu</creatorcontrib><creatorcontrib>Santillana, Carlos</creatorcontrib><creatorcontrib>Shen, Yujia</creatorcontrib><creatorcontrib>Darwiche, Adnan</creatorcontrib><creatorcontrib>Cong, Jason</creatorcontrib><collection>CrossRef</collection><jtitle>ACM transactions on reconfigurable technology and systems</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Choi, Young-Kyu</au><au>Santillana, Carlos</au><au>Shen, Yujia</au><au>Darwiche, Adnan</au><au>Cong, Jason</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>FPGA Acceleration of Probabilistic Sentential Decision Diagrams with High-level Synthesis</atitle><jtitle>ACM transactions on reconfigurable technology and systems</jtitle><stitle>ACM TRETS</stitle><date>2023-03-11</date><risdate>2023</risdate><volume>16</volume><issue>2</issue><spage>1</spage><epage>22</epage><pages>1-22</pages><artnum>18</artnum><issn>1936-7406</issn><eissn>1936-7414</eissn><abstract>Probabilistic Sentential Decision Diagrams (PSDDs) provide efficient methods for modeling and reasoning with probability distributions in the presence of massive logical constraints. PSDDs can also be synthesized from graphical models such as Bayesian networks (BNs) therefore offering a new set of tools for performing inference on these models (in time linear in the PSDD size). Despite these favorable characteristics of PSDDs, we have found multiple challenges in PSDD’s FPGA acceleration. Problems include limited parallelism, data dependency, and small pipeline iterations. In this article, we propose several optimization techniques to solve these issues with novel pipeline scheduling and parallelization schemes. We designed the PSDD kernel with a high-level synthesis (HLS) tool for ease of implementation and verified it on the Xilinx Alveo U250 board. Experimental results show that our methods improve the baseline FPGA HLS implementation performance by 2,200X and the multicore CPU implementation by 20X. The proposed design also outperforms state-of-the-art BN and Sum Product Network (SPN) accelerators that store the graph information in memory.</abstract><cop>New York, NY</cop><pub>ACM</pub><doi>10.1145/3561514</doi><tpages>22</tpages><orcidid>https://orcid.org/0000-0001-5829-4425</orcidid><orcidid>https://orcid.org/0000-0002-5512-6343</orcidid><orcidid>https://orcid.org/0000-0003-3976-6735</orcidid><orcidid>https://orcid.org/0000-0003-2887-6963</orcidid><orcidid>https://orcid.org/0000-0002-4350-7720</orcidid><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier ISSN: 1936-7406
ispartof ACM transactions on reconfigurable technology and systems, 2023-03, Vol.16 (2), p.1-22, Article 18
issn 1936-7406
1936-7414
language eng
recordid cdi_crossref_primary_10_1145_3561514
source ACM Digital Library Complete
subjects Computer systems organization
High-level language architectures
Reconfigurable computing
title FPGA Acceleration of Probabilistic Sentential Decision Diagrams with High-level Synthesis
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-24T23%3A15%3A12IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-acm_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=FPGA%20Acceleration%20of%20Probabilistic%20Sentential%20Decision%20Diagrams%20with%20High-level%20Synthesis&rft.jtitle=ACM%20transactions%20on%20reconfigurable%20technology%20and%20systems&rft.au=Choi,%20Young-Kyu&rft.date=2023-03-11&rft.volume=16&rft.issue=2&rft.spage=1&rft.epage=22&rft.pages=1-22&rft.artnum=18&rft.issn=1936-7406&rft.eissn=1936-7414&rft_id=info:doi/10.1145/3561514&rft_dat=%3Cacm_cross%3E3561514%3C/acm_cross%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true