REDUCING POWER CONSUMPTION IN A NEURAL NETWORK PROCESSOR BY SKIPPING PROCESSING OPERATIONS

REDUCING POWER CONSUMPTION IN A NEURAL NETWORK PROCESSOR BY SKIPPING PROCESSING OPERATIONS

A deep neural network ("DNN") module compresses and decompresses neuron-generated activation data to reduce the utilization of memory bus bandwidth. The compression unit receives an uncompressed chunk of data generated by a neuron in the DNN module. The compression unit generates a mask po...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: AMBARDEKAR, Amol Ashok, CEDOLA, Kent D, MCBRIDE, Chad Balling, BOBROV, Boris, PETRE, George, WALL, Larry Marvin
Format: Patent
Sprache:eng ; fre ; ger
Schlagworte:
CALCULATING
COMPUTING
COUNTING
ELECTRIC DIGITAL DATA PROCESSING
PHYSICS
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 AMBARDEKAR, Amol Ashok
CEDOLA, Kent D
MCBRIDE, Chad Balling
BOBROV, Boris
PETRE, George
WALL, Larry Marvin
description A deep neural network ("DNN") module compresses and decompresses neuron-generated activation data to reduce the utilization of memory bus bandwidth. The compression unit receives an uncompressed chunk of data generated by a neuron in the DNN module. The compression unit generates a mask portion and a data portion of a compressed output chunk. The mask portion encodes the presence and location of the zero and non-zero bytes in the uncompressed chunk of data. The data portion stores truncated non-zero bytes from the uncompressed chunk of data. A decompression unit receives a compressed chunk of data from memory in the DNN processor or memory of an application host. The decompression unit decompresses the compressed chunk of data using the mask portion and the data portion.
format Patent
fullrecord <record><control><sourceid>epo_EVB</sourceid><recordid>TN_cdi_epo_espacenet_EP3612936B1</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>EP3612936B1</sourcerecordid><originalsourceid>FETCH-epo_espacenet_EP3612936B13</originalsourceid><addsrcrecordid>eNrjZIgKcnUJdfb0c1cI8A93DVJw9vcLDvUNCPH091Pw9FNwVPBzDQ1y9AFSIeH-Qd4KAUH-zq7Bwf5BCk6RCsHengEBYL0QURDTP8A1yBGkPZiHgTUtMac4lRdKczMouLmGOHvophbkx6cWFyQmp-allsS7BhibGRpZGps5GRoToQQAP-syDg</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>patent</recordtype></control><display><type>patent</type><title>REDUCING POWER CONSUMPTION IN A NEURAL NETWORK PROCESSOR BY SKIPPING PROCESSING OPERATIONS</title><source>esp@cenet</source><creator>AMBARDEKAR, Amol Ashok ; CEDOLA, Kent D ; MCBRIDE, Chad Balling ; BOBROV, Boris ; PETRE, George ; WALL, Larry Marvin</creator><creatorcontrib>AMBARDEKAR, Amol Ashok ; CEDOLA, Kent D ; MCBRIDE, Chad Balling ; BOBROV, Boris ; PETRE, George ; WALL, Larry Marvin</creatorcontrib><description>A deep neural network ("DNN") module compresses and decompresses neuron-generated activation data to reduce the utilization of memory bus bandwidth. The compression unit receives an uncompressed chunk of data generated by a neuron in the DNN module. The compression unit generates a mask portion and a data portion of a compressed output chunk. The mask portion encodes the presence and location of the zero and non-zero bytes in the uncompressed chunk of data. The data portion stores truncated non-zero bytes from the uncompressed chunk of data. A decompression unit receives a compressed chunk of data from memory in the DNN processor or memory of an application host. The decompression unit decompresses the compressed chunk of data using the mask portion and the data portion.</description><language>eng ; fre ; ger</language><subject>CALCULATING ; COMPUTING ; COUNTING ; ELECTRIC DIGITAL DATA PROCESSING ; 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=20231018&amp;DB=EPODOC&amp;CC=EP&amp;NR=3612936B1$$EHTML$$P50$$Gepo$$Hfree_for_read</linktohtml><link.rule.ids>230,308,776,881,25542,76290</link.rule.ids><linktorsrc>$$Uhttps://worldwide.espacenet.com/publicationDetails/biblio?FT=D&amp;date=20231018&amp;DB=EPODOC&amp;CC=EP&amp;NR=3612936B1$$EView_record_in_European_Patent_Office$$FView_record_in_$$GEuropean_Patent_Office$$Hfree_for_read</linktorsrc></links><search><creatorcontrib>AMBARDEKAR, Amol Ashok</creatorcontrib><creatorcontrib>CEDOLA, Kent D</creatorcontrib><creatorcontrib>MCBRIDE, Chad Balling</creatorcontrib><creatorcontrib>BOBROV, Boris</creatorcontrib><creatorcontrib>PETRE, George</creatorcontrib><creatorcontrib>WALL, Larry Marvin</creatorcontrib><title>REDUCING POWER CONSUMPTION IN A NEURAL NETWORK PROCESSOR BY SKIPPING PROCESSING OPERATIONS</title><description>A deep neural network ("DNN") module compresses and decompresses neuron-generated activation data to reduce the utilization of memory bus bandwidth. The compression unit receives an uncompressed chunk of data generated by a neuron in the DNN module. The compression unit generates a mask portion and a data portion of a compressed output chunk. The mask portion encodes the presence and location of the zero and non-zero bytes in the uncompressed chunk of data. The data portion stores truncated non-zero bytes from the uncompressed chunk of data. A decompression unit receives a compressed chunk of data from memory in the DNN processor or memory of an application host. The decompression unit decompresses the compressed chunk of data using the mask portion and the data portion.</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>2023</creationdate><recordtype>patent</recordtype><sourceid>EVB</sourceid><recordid>eNrjZIgKcnUJdfb0c1cI8A93DVJw9vcLDvUNCPH091Pw9FNwVPBzDQ1y9AFSIeH-Qd4KAUH-zq7Bwf5BCk6RCsHengEBYL0QURDTP8A1yBGkPZiHgTUtMac4lRdKczMouLmGOHvophbkx6cWFyQmp-allsS7BhibGRpZGps5GRoToQQAP-syDg</recordid><startdate>20231018</startdate><enddate>20231018</enddate><creator>AMBARDEKAR, Amol Ashok</creator><creator>CEDOLA, Kent D</creator><creator>MCBRIDE, Chad Balling</creator><creator>BOBROV, Boris</creator><creator>PETRE, George</creator><creator>WALL, Larry Marvin</creator><scope>EVB</scope></search><sort><creationdate>20231018</creationdate><title>REDUCING POWER CONSUMPTION IN A NEURAL NETWORK PROCESSOR BY SKIPPING PROCESSING OPERATIONS</title><author>AMBARDEKAR, Amol Ashok ; CEDOLA, Kent D ; MCBRIDE, Chad Balling ; BOBROV, Boris ; PETRE, George ; WALL, Larry Marvin</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-epo_espacenet_EP3612936B13</frbrgroupid><rsrctype>patents</rsrctype><prefilter>patents</prefilter><language>eng ; fre ; ger</language><creationdate>2023</creationdate><topic>CALCULATING</topic><topic>COMPUTING</topic><topic>COUNTING</topic><topic>ELECTRIC DIGITAL DATA PROCESSING</topic><topic>PHYSICS</topic><toplevel>online_resources</toplevel><creatorcontrib>AMBARDEKAR, Amol Ashok</creatorcontrib><creatorcontrib>CEDOLA, Kent D</creatorcontrib><creatorcontrib>MCBRIDE, Chad Balling</creatorcontrib><creatorcontrib>BOBROV, Boris</creatorcontrib><creatorcontrib>PETRE, George</creatorcontrib><creatorcontrib>WALL, Larry Marvin</creatorcontrib><collection>esp@cenet</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>AMBARDEKAR, Amol Ashok</au><au>CEDOLA, Kent D</au><au>MCBRIDE, Chad Balling</au><au>BOBROV, Boris</au><au>PETRE, George</au><au>WALL, Larry Marvin</au><format>patent</format><genre>patent</genre><ristype>GEN</ristype><title>REDUCING POWER CONSUMPTION IN A NEURAL NETWORK PROCESSOR BY SKIPPING PROCESSING OPERATIONS</title><date>2023-10-18</date><risdate>2023</risdate><abstract>A deep neural network ("DNN") module compresses and decompresses neuron-generated activation data to reduce the utilization of memory bus bandwidth. The compression unit receives an uncompressed chunk of data generated by a neuron in the DNN module. The compression unit generates a mask portion and a data portion of a compressed output chunk. The mask portion encodes the presence and location of the zero and non-zero bytes in the uncompressed chunk of data. The data portion stores truncated non-zero bytes from the uncompressed chunk of data. A decompression unit receives a compressed chunk of data from memory in the DNN processor or memory of an application host. The decompression unit decompresses the compressed chunk of data using the mask portion and the data portion.</abstract><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier
ispartof
issn
language eng ; fre ; ger
recordid cdi_epo_espacenet_EP3612936B1
source esp@cenet
subjects CALCULATING
COMPUTING
COUNTING
ELECTRIC DIGITAL DATA PROCESSING
PHYSICS
title REDUCING POWER CONSUMPTION IN A NEURAL NETWORK PROCESSOR BY SKIPPING PROCESSING OPERATIONS
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-28T18%3A11%3A28IST&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=AMBARDEKAR,%20Amol%20Ashok&rft.date=2023-10-18&rft_id=info:doi/&rft_dat=%3Cepo_EVB%3EEP3612936B1%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