Generalized acceleration of matrix multiply accumulate operations

A method, computer readable medium, and processor are disclosed for performing matrix multiply and accumulate (MMA) operations. The processor includes a datapath configured to execute the MMA operation to generate a plurality of elements of a result matrix at an output of the datapath. Each element...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Oberman, Stuart, Siu, Ming Y, Choquette, Jack H, Boswell, Brent Ralph, Alben, Jonah M
Format: Patent
Sprache: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 Oberman, Stuart
Siu, Ming Y
Choquette, Jack H
Boswell, Brent Ralph
Alben, Jonah M
description A method, computer readable medium, and processor are disclosed for performing matrix multiply and accumulate (MMA) operations. The processor includes a datapath configured to execute the MMA operation to generate a plurality of elements of a result matrix at an output of the datapath. Each element of the result matrix is generated by calculating at least one dot product of corresponding pairs of vectors associated with matrix operands specified in an instruction for the MMA operation. A dot product operation includes the steps of: generating a plurality of partial products by multiplying each element of a first vector with a corresponding element of a second vector; aligning the plurality of partial products based on the exponents associated with each element of the first vector and each element of the second vector; and accumulating the plurality of aligned partial products into a result queue utilizing at least one adder.
format Patent
fullrecord <record><control><sourceid>epo_EVB</sourceid><recordid>TN_cdi_epo_espacenet_US11797301B2</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>US11797301B2</sourcerecordid><originalsourceid>FETCH-epo_espacenet_US11797301B23</originalsourceid><addsrcrecordid>eNrjZHB0T81LLUrMyaxKTVFITE5OzQHySjLz8xTy0xRyE0uKMisUcktzSjILcipB8qVATmJJqkJ-AVRdMQ8Da1piTnEqL5TmZlB0cw1x9tBNLciPTy0uSEwG2lASHxpsaGhuaW5sYOhkZEyMGgBtSTK8</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>patent</recordtype></control><display><type>patent</type><title>Generalized acceleration of matrix multiply accumulate operations</title><source>esp@cenet</source><creator>Oberman, Stuart ; Siu, Ming Y ; Choquette, Jack H ; Boswell, Brent Ralph ; Alben, Jonah M</creator><creatorcontrib>Oberman, Stuart ; Siu, Ming Y ; Choquette, Jack H ; Boswell, Brent Ralph ; Alben, Jonah M</creatorcontrib><description>A method, computer readable medium, and processor are disclosed for performing matrix multiply and accumulate (MMA) operations. The processor includes a datapath configured to execute the MMA operation to generate a plurality of elements of a result matrix at an output of the datapath. Each element of the result matrix is generated by calculating at least one dot product of corresponding pairs of vectors associated with matrix operands specified in an instruction for the MMA operation. A dot product operation includes the steps of: generating a plurality of partial products by multiplying each element of a first vector with a corresponding element of a second vector; aligning the plurality of partial products based on the exponents associated with each element of the first vector and each element of the second vector; and accumulating the plurality of aligned partial products into a result queue utilizing at least one adder.</description><language>eng</language><subject>CALCULATING ; COMPUTING ; COUNTING ; ELECTRIC DIGITAL DATA PROCESSING ; IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ; PHYSICS</subject><creationdate>2023</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&amp;date=20231024&amp;DB=EPODOC&amp;CC=US&amp;NR=11797301B2$$EHTML$$P50$$Gepo$$Hfree_for_read</linktohtml><link.rule.ids>230,308,776,881,25542,76289</link.rule.ids><linktorsrc>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&amp;date=20231024&amp;DB=EPODOC&amp;CC=US&amp;NR=11797301B2$$EView_record_in_European_Patent_Office$$FView_record_in_$$GEuropean_Patent_Office$$Hfree_for_read</linktorsrc></links><search><creatorcontrib>Oberman, Stuart</creatorcontrib><creatorcontrib>Siu, Ming Y</creatorcontrib><creatorcontrib>Choquette, Jack H</creatorcontrib><creatorcontrib>Boswell, Brent Ralph</creatorcontrib><creatorcontrib>Alben, Jonah M</creatorcontrib><title>Generalized acceleration of matrix multiply accumulate operations</title><description>A method, computer readable medium, and processor are disclosed for performing matrix multiply and accumulate (MMA) operations. The processor includes a datapath configured to execute the MMA operation to generate a plurality of elements of a result matrix at an output of the datapath. Each element of the result matrix is generated by calculating at least one dot product of corresponding pairs of vectors associated with matrix operands specified in an instruction for the MMA operation. A dot product operation includes the steps of: generating a plurality of partial products by multiplying each element of a first vector with a corresponding element of a second vector; aligning the plurality of partial products based on the exponents associated with each element of the first vector and each element of the second vector; and accumulating the plurality of aligned partial products into a result queue utilizing at least one adder.</description><subject>CALCULATING</subject><subject>COMPUTING</subject><subject>COUNTING</subject><subject>ELECTRIC DIGITAL DATA PROCESSING</subject><subject>IMAGE DATA PROCESSING OR GENERATION, IN GENERAL</subject><subject>PHYSICS</subject><fulltext>true</fulltext><rsrctype>patent</rsrctype><creationdate>2023</creationdate><recordtype>patent</recordtype><sourceid>EVB</sourceid><recordid>eNrjZHB0T81LLUrMyaxKTVFITE5OzQHySjLz8xTy0xRyE0uKMisUcktzSjILcipB8qVATmJJqkJ-AVRdMQ8Da1piTnEqL5TmZlB0cw1x9tBNLciPTy0uSEwG2lASHxpsaGhuaW5sYOhkZEyMGgBtSTK8</recordid><startdate>20231024</startdate><enddate>20231024</enddate><creator>Oberman, Stuart</creator><creator>Siu, Ming Y</creator><creator>Choquette, Jack H</creator><creator>Boswell, Brent Ralph</creator><creator>Alben, Jonah M</creator><scope>EVB</scope></search><sort><creationdate>20231024</creationdate><title>Generalized acceleration of matrix multiply accumulate operations</title><author>Oberman, Stuart ; Siu, Ming Y ; Choquette, Jack H ; Boswell, Brent Ralph ; Alben, Jonah M</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-epo_espacenet_US11797301B23</frbrgroupid><rsrctype>patents</rsrctype><prefilter>patents</prefilter><language>eng</language><creationdate>2023</creationdate><topic>CALCULATING</topic><topic>COMPUTING</topic><topic>COUNTING</topic><topic>ELECTRIC DIGITAL DATA PROCESSING</topic><topic>IMAGE DATA PROCESSING OR GENERATION, IN GENERAL</topic><topic>PHYSICS</topic><toplevel>online_resources</toplevel><creatorcontrib>Oberman, Stuart</creatorcontrib><creatorcontrib>Siu, Ming Y</creatorcontrib><creatorcontrib>Choquette, Jack H</creatorcontrib><creatorcontrib>Boswell, Brent Ralph</creatorcontrib><creatorcontrib>Alben, Jonah M</creatorcontrib><collection>esp@cenet</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Oberman, Stuart</au><au>Siu, Ming Y</au><au>Choquette, Jack H</au><au>Boswell, Brent Ralph</au><au>Alben, Jonah M</au><format>patent</format><genre>patent</genre><ristype>GEN</ristype><title>Generalized acceleration of matrix multiply accumulate operations</title><date>2023-10-24</date><risdate>2023</risdate><abstract>A method, computer readable medium, and processor are disclosed for performing matrix multiply and accumulate (MMA) operations. The processor includes a datapath configured to execute the MMA operation to generate a plurality of elements of a result matrix at an output of the datapath. Each element of the result matrix is generated by calculating at least one dot product of corresponding pairs of vectors associated with matrix operands specified in an instruction for the MMA operation. A dot product operation includes the steps of: generating a plurality of partial products by multiplying each element of a first vector with a corresponding element of a second vector; aligning the plurality of partial products based on the exponents associated with each element of the first vector and each element of the second vector; and accumulating the plurality of aligned partial products into a result queue utilizing at least one adder.</abstract><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier
ispartof
issn
language eng
recordid cdi_epo_espacenet_US11797301B2
source esp@cenet
subjects CALCULATING
COMPUTING
COUNTING
ELECTRIC DIGITAL DATA PROCESSING
IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
PHYSICS
title Generalized acceleration of matrix multiply accumulate operations
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-02-06T23%3A52%3A26IST&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=Oberman,%20Stuart&rft.date=2023-10-24&rft_id=info:doi/&rft_dat=%3Cepo_EVB%3EUS11797301B2%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