Accelerator execution method and electronic equipment
A method performed by an accelerator and an electronic device are described herein. The method comprises the following steps: receiving a first tensor multiplication instruction for a first thread of an accelerator; the first thread set broadcasts a second set of factors in the second tensor to the...
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Patent |
| Sprache: | chi ; eng |
| Schlagworte: | |
| Online-Zugang: | Volltext bestellen |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| container_end_page | |
|---|---|
| container_issue | |
| container_start_page | |
| container_title | |
| container_volume | |
| creator | SANG YONGQI XIE GANGFENG LI JIA GE JIANMING QIU XIAOGANG YAO FEI YANG JINGWEI |
| description | A method performed by an accelerator and an electronic device are described herein. The method comprises the following steps: receiving a first tensor multiplication instruction for a first thread of an accelerator; the first thread set broadcasts a second set of factors in the second tensor to the second thread set based on a memory logical address for the second tensor; a first thread in a second set of threads dot-product operates a first set of factors and the second set of factors based on the first factor register representation to generate a first dot-product set in a first row of a third tensor. According to the method, the matrix is decomposed, and the threads are distributed according to the rows, so that the multiple threads can process multiple rows of the matrix tensor in parallel, and the processing efficiency of matrix multiplication is improved. Besides, programming personnel know the row and column structure of the matrix tensor and the thread condition in the accelerator during programming, |
| format | Patent |
| fullrecord | <record><control><sourceid>epo_EVB</sourceid><recordid>TN_cdi_epo_espacenet_CN114579929A</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>CN114579929A</sourcerecordid><originalsourceid>FETCH-epo_espacenet_CN114579929A3</originalsourceid><addsrcrecordid>eNrjZDB1TE5OzUktSizJL1JIrUhNLi3JzM9TyE0tychPUUjMS1EAyiaXFOXnZSYrpBaWZhbkpuaV8DCwpiXmFKfyQmluBkU31xBnD93Ugvz41OKCxOTUvNSSeGc_Q0MTU3NLSyNLR2Ni1AAAn5Qtxw</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>patent</recordtype></control><display><type>patent</type><title>Accelerator execution method and electronic equipment</title><source>esp@cenet</source><creator>SANG YONGQI ; XIE GANGFENG ; LI JIA ; GE JIANMING ; QIU XIAOGANG ; YAO FEI ; YANG JINGWEI</creator><creatorcontrib>SANG YONGQI ; XIE GANGFENG ; LI JIA ; GE JIANMING ; QIU XIAOGANG ; YAO FEI ; YANG JINGWEI</creatorcontrib><description>A method performed by an accelerator and an electronic device are described herein. The method comprises the following steps: receiving a first tensor multiplication instruction for a first thread of an accelerator; the first thread set broadcasts a second set of factors in the second tensor to the second thread set based on a memory logical address for the second tensor; a first thread in a second set of threads dot-product operates a first set of factors and the second set of factors based on the first factor register representation to generate a first dot-product set in a first row of a third tensor. According to the method, the matrix is decomposed, and the threads are distributed according to the rows, so that the multiple threads can process multiple rows of the matrix tensor in parallel, and the processing efficiency of matrix multiplication is improved. Besides, programming personnel know the row and column structure of the matrix tensor and the thread condition in the accelerator during programming,</description><language>chi ; eng</language><subject>CALCULATING ; COMPUTING ; COUNTING ; ELECTRIC DIGITAL DATA PROCESSING ; PHYSICS</subject><creationdate>2022</creationdate><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20220603&DB=EPODOC&CC=CN&NR=114579929A$$EHTML$$P50$$Gepo$$Hfree_for_read</linktohtml><link.rule.ids>230,308,780,885,25564,76547</link.rule.ids><linktorsrc>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&date=20220603&DB=EPODOC&CC=CN&NR=114579929A$$EView_record_in_European_Patent_Office$$FView_record_in_$$GEuropean_Patent_Office$$Hfree_for_read</linktorsrc></links><search><creatorcontrib>SANG YONGQI</creatorcontrib><creatorcontrib>XIE GANGFENG</creatorcontrib><creatorcontrib>LI JIA</creatorcontrib><creatorcontrib>GE JIANMING</creatorcontrib><creatorcontrib>QIU XIAOGANG</creatorcontrib><creatorcontrib>YAO FEI</creatorcontrib><creatorcontrib>YANG JINGWEI</creatorcontrib><title>Accelerator execution method and electronic equipment</title><description>A method performed by an accelerator and an electronic device are described herein. The method comprises the following steps: receiving a first tensor multiplication instruction for a first thread of an accelerator; the first thread set broadcasts a second set of factors in the second tensor to the second thread set based on a memory logical address for the second tensor; a first thread in a second set of threads dot-product operates a first set of factors and the second set of factors based on the first factor register representation to generate a first dot-product set in a first row of a third tensor. According to the method, the matrix is decomposed, and the threads are distributed according to the rows, so that the multiple threads can process multiple rows of the matrix tensor in parallel, and the processing efficiency of matrix multiplication is improved. Besides, programming personnel know the row and column structure of the matrix tensor and the thread condition in the accelerator during programming,</description><subject>CALCULATING</subject><subject>COMPUTING</subject><subject>COUNTING</subject><subject>ELECTRIC DIGITAL DATA PROCESSING</subject><subject>PHYSICS</subject><fulltext>true</fulltext><rsrctype>patent</rsrctype><creationdate>2022</creationdate><recordtype>patent</recordtype><sourceid>EVB</sourceid><recordid>eNrjZDB1TE5OzUktSizJL1JIrUhNLi3JzM9TyE0tychPUUjMS1EAyiaXFOXnZSYrpBaWZhbkpuaV8DCwpiXmFKfyQmluBkU31xBnD93Ugvz41OKCxOTUvNSSeGc_Q0MTU3NLSyNLR2Ni1AAAn5Qtxw</recordid><startdate>20220603</startdate><enddate>20220603</enddate><creator>SANG YONGQI</creator><creator>XIE GANGFENG</creator><creator>LI JIA</creator><creator>GE JIANMING</creator><creator>QIU XIAOGANG</creator><creator>YAO FEI</creator><creator>YANG JINGWEI</creator><scope>EVB</scope></search><sort><creationdate>20220603</creationdate><title>Accelerator execution method and electronic equipment</title><author>SANG YONGQI ; XIE GANGFENG ; LI JIA ; GE JIANMING ; QIU XIAOGANG ; YAO FEI ; YANG JINGWEI</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-epo_espacenet_CN114579929A3</frbrgroupid><rsrctype>patents</rsrctype><prefilter>patents</prefilter><language>chi ; eng</language><creationdate>2022</creationdate><topic>CALCULATING</topic><topic>COMPUTING</topic><topic>COUNTING</topic><topic>ELECTRIC DIGITAL DATA PROCESSING</topic><topic>PHYSICS</topic><toplevel>online_resources</toplevel><creatorcontrib>SANG YONGQI</creatorcontrib><creatorcontrib>XIE GANGFENG</creatorcontrib><creatorcontrib>LI JIA</creatorcontrib><creatorcontrib>GE JIANMING</creatorcontrib><creatorcontrib>QIU XIAOGANG</creatorcontrib><creatorcontrib>YAO FEI</creatorcontrib><creatorcontrib>YANG JINGWEI</creatorcontrib><collection>esp@cenet</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>SANG YONGQI</au><au>XIE GANGFENG</au><au>LI JIA</au><au>GE JIANMING</au><au>QIU XIAOGANG</au><au>YAO FEI</au><au>YANG JINGWEI</au><format>patent</format><genre>patent</genre><ristype>GEN</ristype><title>Accelerator execution method and electronic equipment</title><date>2022-06-03</date><risdate>2022</risdate><abstract>A method performed by an accelerator and an electronic device are described herein. The method comprises the following steps: receiving a first tensor multiplication instruction for a first thread of an accelerator; the first thread set broadcasts a second set of factors in the second tensor to the second thread set based on a memory logical address for the second tensor; a first thread in a second set of threads dot-product operates a first set of factors and the second set of factors based on the first factor register representation to generate a first dot-product set in a first row of a third tensor. According to the method, the matrix is decomposed, and the threads are distributed according to the rows, so that the multiple threads can process multiple rows of the matrix tensor in parallel, and the processing efficiency of matrix multiplication is improved. Besides, programming personnel know the row and column structure of the matrix tensor and the thread condition in the accelerator during programming,</abstract><oa>free_for_read</oa></addata></record> |
| fulltext | fulltext_linktorsrc |
| identifier | |
| ispartof | |
| issn | |
| language | chi ; eng |
| recordid | cdi_epo_espacenet_CN114579929A |
| source | esp@cenet |
| subjects | CALCULATING COMPUTING COUNTING ELECTRIC DIGITAL DATA PROCESSING PHYSICS |
| title | Accelerator execution method and electronic equipment |
| url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-04T22%3A39%3A59IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-epo_EVB&rft_val_fmt=info:ofi/fmt:kev:mtx:patent&rft.genre=patent&rft.au=SANG%20YONGQI&rft.date=2022-06-03&rft_id=info:doi/&rft_dat=%3Cepo_EVB%3ECN114579929A%3C/epo_EVB%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 |