Analysis of Data Transfer Bottlenecks in Commercial PIM Systems: A Study With UPMEM-PIM
Due to emerging workloads that require high memory bandwidth, Processing-in-Memory (PIM) has gained significant attention and led several industrial PIM products to be introduced which are integrated with conventional computing systems. This letter characterizes the data transfer overheads between c...
Gespeichert in:
| Veröffentlicht in: | IEEE computer architecture letters 2024-07, Vol.23 (2), p.179-182 |
|---|---|
| Hauptverfasser: | , , , |
| Format: | Artikel |
| Sprache: | eng |
| Schlagworte: | |
| Online-Zugang: | Volltext bestellen |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| container_end_page | 182 |
|---|---|
| container_issue | 2 |
| container_start_page | 179 |
| container_title | IEEE computer architecture letters |
| container_volume | 23 |
| creator | Lee, Dongjae Hyun, Bongjoon Kim, Taehun Rhu, Minsoo |
| description | Due to emerging workloads that require high memory bandwidth, Processing-in-Memory (PIM) has gained significant attention and led several industrial PIM products to be introduced which are integrated with conventional computing systems. This letter characterizes the data transfer overheads between conventional DRAM address space and PIM address space within a PIM-integrated system using the commercialized PIM device made by UPMEM. Our findings highlight the need for optimization in PIM-integrated systems to address these overheads, offering critical insights for future PIM technologies. |
| doi_str_mv | 10.1109/LCA.2024.3387472 |
| format | Article |
| fullrecord | <record><control><sourceid>crossref_RIE</sourceid><recordid>TN_cdi_ieee_primary_10496848</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><ieee_id>10496848</ieee_id><sourcerecordid>10_1109_LCA_2024_3387472</sourcerecordid><originalsourceid>FETCH-LOGICAL-c217t-f0a7e9f65ca3380421e14152c357cd0271de3fbd97ffd3bff83acc9243c1e8223</originalsourceid><addsrcrecordid>eNpNkE1Lw0AQhhdRsFbvHjzsH0jdr-wm3mKsWkix0JYew3Yzi9F8yO72kH9vSot4egfeeYbhQeiekhmlJH0s8mzGCBMzzhMlFLtAExrHMpJEisu_OZbX6Mb7L0KE5ImYoF3W6Wbwtce9xS86aLxxuvMWHH7uQ2igA_Ptcd3hvG9bcKbWDV4tlng9-ACtf8IZXodDNeBdHT7xdrWcL6Oxv0VXVjce7s45RdvX-SZ_j4qPt0WeFZFhVIXIEq0gtTI2evybCEaBChozw2NlKsIUrYDbfZUqayu-tzbh2piUCW4oJIzxKSKnu8b13juw5Y-rW-2GkpLyKKYcxZRHMeVZzIg8nJAaAP6ti1QmIuG_JfZeDw</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Analysis of Data Transfer Bottlenecks in Commercial PIM Systems: A Study With UPMEM-PIM</title><source>IEEE Xplore</source><creator>Lee, Dongjae ; Hyun, Bongjoon ; Kim, Taehun ; Rhu, Minsoo</creator><creatorcontrib>Lee, Dongjae ; Hyun, Bongjoon ; Kim, Taehun ; Rhu, Minsoo</creatorcontrib><description>Due to emerging workloads that require high memory bandwidth, Processing-in-Memory (PIM) has gained significant attention and led several industrial PIM products to be introduced which are integrated with conventional computing systems. This letter characterizes the data transfer overheads between conventional DRAM address space and PIM address space within a PIM-integrated system using the commercialized PIM device made by UPMEM. Our findings highlight the need for optimization in PIM-integrated systems to address these overheads, offering critical insights for future PIM technologies.</description><identifier>ISSN: 1556-6056</identifier><identifier>EISSN: 1556-6064</identifier><identifier>DOI: 10.1109/LCA.2024.3387472</identifier><identifier>CODEN: ICALC3</identifier><language>eng</language><publisher>IEEE</publisher><subject>Bandwidth ; Data transfer ; Memory management ; near-memory processing ; parallel architecture ; Processing-in-memory (PIM) ; Random access memory ; Runtime library ; Software ; Throughput</subject><ispartof>IEEE computer architecture letters, 2024-07, Vol.23 (2), p.179-182</ispartof><lds50>peer_reviewed</lds50><woscitedreferencessubscribed>false</woscitedreferencessubscribed><cites>FETCH-LOGICAL-c217t-f0a7e9f65ca3380421e14152c357cd0271de3fbd97ffd3bff83acc9243c1e8223</cites><orcidid>0000-0002-2529-7264 ; 0000-0001-9678-6228 ; 0000-0003-3303-8681 ; 0000-0001-8104-8150</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://ieeexplore.ieee.org/document/10496848$$EHTML$$P50$$Gieee$$H</linktohtml><link.rule.ids>314,776,780,792,27901,27902,54733</link.rule.ids><linktorsrc>$$Uhttps://ieeexplore.ieee.org/document/10496848$$EView_record_in_IEEE$$FView_record_in_$$GIEEE</linktorsrc></links><search><creatorcontrib>Lee, Dongjae</creatorcontrib><creatorcontrib>Hyun, Bongjoon</creatorcontrib><creatorcontrib>Kim, Taehun</creatorcontrib><creatorcontrib>Rhu, Minsoo</creatorcontrib><title>Analysis of Data Transfer Bottlenecks in Commercial PIM Systems: A Study With UPMEM-PIM</title><title>IEEE computer architecture letters</title><addtitle>LCA</addtitle><description>Due to emerging workloads that require high memory bandwidth, Processing-in-Memory (PIM) has gained significant attention and led several industrial PIM products to be introduced which are integrated with conventional computing systems. This letter characterizes the data transfer overheads between conventional DRAM address space and PIM address space within a PIM-integrated system using the commercialized PIM device made by UPMEM. Our findings highlight the need for optimization in PIM-integrated systems to address these overheads, offering critical insights for future PIM technologies.</description><subject>Bandwidth</subject><subject>Data transfer</subject><subject>Memory management</subject><subject>near-memory processing</subject><subject>parallel architecture</subject><subject>Processing-in-memory (PIM)</subject><subject>Random access memory</subject><subject>Runtime library</subject><subject>Software</subject><subject>Throughput</subject><issn>1556-6056</issn><issn>1556-6064</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>RIE</sourceid><recordid>eNpNkE1Lw0AQhhdRsFbvHjzsH0jdr-wm3mKsWkix0JYew3Yzi9F8yO72kH9vSot4egfeeYbhQeiekhmlJH0s8mzGCBMzzhMlFLtAExrHMpJEisu_OZbX6Mb7L0KE5ImYoF3W6Wbwtce9xS86aLxxuvMWHH7uQ2igA_Ptcd3hvG9bcKbWDV4tlng9-ACtf8IZXodDNeBdHT7xdrWcL6Oxv0VXVjce7s45RdvX-SZ_j4qPt0WeFZFhVIXIEq0gtTI2evybCEaBChozw2NlKsIUrYDbfZUqayu-tzbh2piUCW4oJIzxKSKnu8b13juw5Y-rW-2GkpLyKKYcxZRHMeVZzIg8nJAaAP6ti1QmIuG_JfZeDw</recordid><startdate>20240701</startdate><enddate>20240701</enddate><creator>Lee, Dongjae</creator><creator>Hyun, Bongjoon</creator><creator>Kim, Taehun</creator><creator>Rhu, Minsoo</creator><general>IEEE</general><scope>97E</scope><scope>RIA</scope><scope>RIE</scope><scope>AAYXX</scope><scope>CITATION</scope><orcidid>https://orcid.org/0000-0002-2529-7264</orcidid><orcidid>https://orcid.org/0000-0001-9678-6228</orcidid><orcidid>https://orcid.org/0000-0003-3303-8681</orcidid><orcidid>https://orcid.org/0000-0001-8104-8150</orcidid></search><sort><creationdate>20240701</creationdate><title>Analysis of Data Transfer Bottlenecks in Commercial PIM Systems: A Study With UPMEM-PIM</title><author>Lee, Dongjae ; Hyun, Bongjoon ; Kim, Taehun ; Rhu, Minsoo</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c217t-f0a7e9f65ca3380421e14152c357cd0271de3fbd97ffd3bff83acc9243c1e8223</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Bandwidth</topic><topic>Data transfer</topic><topic>Memory management</topic><topic>near-memory processing</topic><topic>parallel architecture</topic><topic>Processing-in-memory (PIM)</topic><topic>Random access memory</topic><topic>Runtime library</topic><topic>Software</topic><topic>Throughput</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Lee, Dongjae</creatorcontrib><creatorcontrib>Hyun, Bongjoon</creatorcontrib><creatorcontrib>Kim, Taehun</creatorcontrib><creatorcontrib>Rhu, Minsoo</creatorcontrib><collection>IEEE All-Society Periodicals Package (ASPP) 2005-present</collection><collection>IEEE All-Society Periodicals Package (ASPP) 1998–Present</collection><collection>IEEE Xplore</collection><collection>CrossRef</collection><jtitle>IEEE computer architecture letters</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Lee, Dongjae</au><au>Hyun, Bongjoon</au><au>Kim, Taehun</au><au>Rhu, Minsoo</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Analysis of Data Transfer Bottlenecks in Commercial PIM Systems: A Study With UPMEM-PIM</atitle><jtitle>IEEE computer architecture letters</jtitle><stitle>LCA</stitle><date>2024-07-01</date><risdate>2024</risdate><volume>23</volume><issue>2</issue><spage>179</spage><epage>182</epage><pages>179-182</pages><issn>1556-6056</issn><eissn>1556-6064</eissn><coden>ICALC3</coden><abstract>Due to emerging workloads that require high memory bandwidth, Processing-in-Memory (PIM) has gained significant attention and led several industrial PIM products to be introduced which are integrated with conventional computing systems. This letter characterizes the data transfer overheads between conventional DRAM address space and PIM address space within a PIM-integrated system using the commercialized PIM device made by UPMEM. Our findings highlight the need for optimization in PIM-integrated systems to address these overheads, offering critical insights for future PIM technologies.</abstract><pub>IEEE</pub><doi>10.1109/LCA.2024.3387472</doi><tpages>4</tpages><orcidid>https://orcid.org/0000-0002-2529-7264</orcidid><orcidid>https://orcid.org/0000-0001-9678-6228</orcidid><orcidid>https://orcid.org/0000-0003-3303-8681</orcidid><orcidid>https://orcid.org/0000-0001-8104-8150</orcidid></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | ISSN: 1556-6056 |
| ispartof | IEEE computer architecture letters, 2024-07, Vol.23 (2), p.179-182 |
| issn | 1556-6056 1556-6064 |
| language | eng |
| recordid | cdi_ieee_primary_10496848 |
| source | IEEE Xplore |
| subjects | Bandwidth Data transfer Memory management near-memory processing parallel architecture Processing-in-memory (PIM) Random access memory Runtime library Software Throughput |
| title | Analysis of Data Transfer Bottlenecks in Commercial PIM Systems: A Study With UPMEM-PIM |
| url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-12T06%3A03%3A25IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-crossref_RIE&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Analysis%20of%20Data%20Transfer%20Bottlenecks%20in%20Commercial%20PIM%20Systems:%20A%20Study%20With%20UPMEM-PIM&rft.jtitle=IEEE%20computer%20architecture%20letters&rft.au=Lee,%20Dongjae&rft.date=2024-07-01&rft.volume=23&rft.issue=2&rft.spage=179&rft.epage=182&rft.pages=179-182&rft.issn=1556-6056&rft.eissn=1556-6064&rft.coden=ICALC3&rft_id=info:doi/10.1109/LCA.2024.3387472&rft_dat=%3Ccrossref_RIE%3E10_1109_LCA_2024_3387472%3C/crossref_RIE%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rft_ieee_id=10496848&rfr_iscdi=true |