Occam: Optimal Data Reuse for Convolutional Neural Networks Article Swipe
Ashish Gondimalla
,
Jianqiao Liu
,
Mithuna Thottethodi
,
T. N. Vijaykumar
·
YOU?
·
· 2022
· Open Access
·
· DOI: https://doi.org/10.1145/3566052
YOU?
·
· 2022
· Open Access
·
· DOI: https://doi.org/10.1145/3566052
Convolutional neural networks (CNNs) are emerging as powerful tools for image processing in important commercial applications. We focus on the important problem of improving the latency of image recognition. While CNNs are highly amenable to prefetching and multithreading to avoid memory latency issues, CNNs’ large data – each layer’s input, filters, and output – poses a memory bandwidth problem. While previous work captures only some of the enormous data reuse, full reuse implies that the initial input image and filters are read once from off-chip and the final output is written once off-chip without spilling the intermediate layers’ data to off-chip. We propose Occam to capture full reuse via four contributions. First, we identify the necessary conditions for full reuse. Second, we identify the dependence closure as the sufficient condition to capture full reuse using the least on-chip memory. Third, because the dependence closure is often too large to fit in on-chip memory, we propose a dynamic programming algorithm that optimally partitions a given CNN to guarantee the least off-chip traffic at the partition boundaries for a given on-chip capacity. While tiling is well-known, our contribution determines the optimal cross-layer tiles. Occam’s partitions reside on different chips, forming a pipeline so that a partition’s filters and dependence closure remain on-chip as different images pass through (i.e., each partition incurs off-chip traffic only for its inputs and outputs). Finally, because the optimal partitions may result in an unbalanced pipeline, we propose staggered asynchronous pipelines (STAPs) that replicate bottleneck stages to improve throughput by staggering mini-batches across replicas. Importantly, STAPs achieve balanced pipelines without changing Occam’s optimal partitioning. Our simulations show that, on average, Occam cuts off-chip transfers by 21× and achieves 2.04× and 1.21× better performance, and 33% better energy than the base case, respectively. Using a field-programmable gate array (FPGA) implementation, Occam performs 6.1× and 1.5× better, on average, than the base case and Layer Fusion, respectively.
Related Topics
Concepts
Metadata
- Type
- article
- Language
- en
- Landing Page
- https://doi.org/10.1145/3566052
- https://dl.acm.org/doi/pdf/10.1145/3566052
- OA Status
- diamond
- Cited By
- 4
- References
- 73
- Related Works
- 10
- OpenAlex ID
- https://openalex.org/W3177229224
All OpenAlex metadata
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W3177229224Canonical identifier for this work in OpenAlex
- DOI
-
https://doi.org/10.1145/3566052Digital Object Identifier
- Title
-
Occam: Optimal Data Reuse for Convolutional Neural NetworksWork title
- Type
-
articleOpenAlex work type
- Language
-
enPrimary language
- Publication year
-
2022Year of publication
- Publication date
-
2022-12-16Full publication date if available
- Authors
-
Ashish Gondimalla, Jianqiao Liu, Mithuna Thottethodi, T. N. VijaykumarList of authors in order
- Landing page
-
https://doi.org/10.1145/3566052Publisher landing page
- PDF URL
-
https://dl.acm.org/doi/pdf/10.1145/3566052Direct link to full text PDF
- Open access
-
YesWhether a free full text is available
- OA status
-
diamondOpen access status per OpenAlex
- OA URL
-
https://dl.acm.org/doi/pdf/10.1145/3566052Direct OA link when available
- Concepts
-
Computer science, Parallel computing, Convolutional neural network, occam, Reuse, Chip, Pipeline (software), Multithreading, Speedup, Asynchronous communication, Computer engineering, Algorithm, Artificial intelligence, Thread (computing), Computer network, Programming language, Biology, Ecology, TelecommunicationsTop concepts (fields/topics) attached by OpenAlex
- Cited by
-
4Total citation count in OpenAlex
- Citations by year (recent)
-
2025: 1, 2024: 2, 2023: 1Per-year citation counts (last 5 years)
- References (count)
-
73Number of works referenced by this work
- Related works (count)
-
10Other works algorithmically related by OpenAlex
Full payload
| id | https://openalex.org/W3177229224 |
|---|---|
| doi | https://doi.org/10.1145/3566052 |
| ids.doi | https://doi.org/10.1145/3566052 |
| ids.mag | 3177229224 |
| ids.openalex | https://openalex.org/W3177229224 |
| fwci | 0.4951672 |
| type | article |
| title | Occam: Optimal Data Reuse for Convolutional Neural Networks |
| biblio.issue | 1 |
| biblio.volume | 20 |
| biblio.last_page | 25 |
| biblio.first_page | 1 |
| topics[0].id | https://openalex.org/T10036 |
| topics[0].field.id | https://openalex.org/fields/17 |
| topics[0].field.display_name | Computer Science |
| topics[0].score | 0.9998000264167786 |
| topics[0].domain.id | https://openalex.org/domains/3 |
| topics[0].domain.display_name | Physical Sciences |
| topics[0].subfield.id | https://openalex.org/subfields/1707 |
| topics[0].subfield.display_name | Computer Vision and Pattern Recognition |
| topics[0].display_name | Advanced Neural Network Applications |
| topics[1].id | https://openalex.org/T11992 |
| topics[1].field.id | https://openalex.org/fields/22 |
| topics[1].field.display_name | Engineering |
| topics[1].score | 0.9998000264167786 |
| topics[1].domain.id | https://openalex.org/domains/3 |
| topics[1].domain.display_name | Physical Sciences |
| topics[1].subfield.id | https://openalex.org/subfields/2208 |
| topics[1].subfield.display_name | Electrical and Electronic Engineering |
| topics[1].display_name | CCD and CMOS Imaging Sensors |
| topics[2].id | https://openalex.org/T10502 |
| topics[2].field.id | https://openalex.org/fields/22 |
| topics[2].field.display_name | Engineering |
| topics[2].score | 0.9994999766349792 |
| topics[2].domain.id | https://openalex.org/domains/3 |
| topics[2].domain.display_name | Physical Sciences |
| topics[2].subfield.id | https://openalex.org/subfields/2208 |
| topics[2].subfield.display_name | Electrical and Electronic Engineering |
| topics[2].display_name | Advanced Memory and Neural Computing |
| is_xpac | False |
| apc_list | |
| apc_paid | |
| concepts[0].id | https://openalex.org/C41008148 |
| concepts[0].level | 0 |
| concepts[0].score | 0.8779021501541138 |
| concepts[0].wikidata | https://www.wikidata.org/wiki/Q21198 |
| concepts[0].display_name | Computer science |
| concepts[1].id | https://openalex.org/C173608175 |
| concepts[1].level | 1 |
| concepts[1].score | 0.6097621917724609 |
| concepts[1].wikidata | https://www.wikidata.org/wiki/Q232661 |
| concepts[1].display_name | Parallel computing |
| concepts[2].id | https://openalex.org/C81363708 |
| concepts[2].level | 2 |
| concepts[2].score | 0.5409556031227112 |
| concepts[2].wikidata | https://www.wikidata.org/wiki/Q17084460 |
| concepts[2].display_name | Convolutional neural network |
| concepts[3].id | https://openalex.org/C78469957 |
| concepts[3].level | 2 |
| concepts[3].score | 0.524736225605011 |
| concepts[3].wikidata | https://www.wikidata.org/wiki/Q838062 |
| concepts[3].display_name | occam |
| concepts[4].id | https://openalex.org/C206588197 |
| concepts[4].level | 2 |
| concepts[4].score | 0.5087361931800842 |
| concepts[4].wikidata | https://www.wikidata.org/wiki/Q846574 |
| concepts[4].display_name | Reuse |
| concepts[5].id | https://openalex.org/C165005293 |
| concepts[5].level | 2 |
| concepts[5].score | 0.5039469599723816 |
| concepts[5].wikidata | https://www.wikidata.org/wiki/Q1074500 |
| concepts[5].display_name | Chip |
| concepts[6].id | https://openalex.org/C43521106 |
| concepts[6].level | 2 |
| concepts[6].score | 0.49973559379577637 |
| concepts[6].wikidata | https://www.wikidata.org/wiki/Q2165493 |
| concepts[6].display_name | Pipeline (software) |
| concepts[7].id | https://openalex.org/C201410400 |
| concepts[7].level | 3 |
| concepts[7].score | 0.47680041193962097 |
| concepts[7].wikidata | https://www.wikidata.org/wiki/Q1064412 |
| concepts[7].display_name | Multithreading |
| concepts[8].id | https://openalex.org/C68339613 |
| concepts[8].level | 2 |
| concepts[8].score | 0.46002355217933655 |
| concepts[8].wikidata | https://www.wikidata.org/wiki/Q1549489 |
| concepts[8].display_name | Speedup |
| concepts[9].id | https://openalex.org/C151319957 |
| concepts[9].level | 2 |
| concepts[9].score | 0.43010854721069336 |
| concepts[9].wikidata | https://www.wikidata.org/wiki/Q752739 |
| concepts[9].display_name | Asynchronous communication |
| concepts[10].id | https://openalex.org/C113775141 |
| concepts[10].level | 1 |
| concepts[10].score | 0.40509775280952454 |
| concepts[10].wikidata | https://www.wikidata.org/wiki/Q428691 |
| concepts[10].display_name | Computer engineering |
| concepts[11].id | https://openalex.org/C11413529 |
| concepts[11].level | 1 |
| concepts[11].score | 0.35612186789512634 |
| concepts[11].wikidata | https://www.wikidata.org/wiki/Q8366 |
| concepts[11].display_name | Algorithm |
| concepts[12].id | https://openalex.org/C154945302 |
| concepts[12].level | 1 |
| concepts[12].score | 0.17872759699821472 |
| concepts[12].wikidata | https://www.wikidata.org/wiki/Q11660 |
| concepts[12].display_name | Artificial intelligence |
| concepts[13].id | https://openalex.org/C138101251 |
| concepts[13].level | 2 |
| concepts[13].score | 0.15554490685462952 |
| concepts[13].wikidata | https://www.wikidata.org/wiki/Q213092 |
| concepts[13].display_name | Thread (computing) |
| concepts[14].id | https://openalex.org/C31258907 |
| concepts[14].level | 1 |
| concepts[14].score | 0.12033155560493469 |
| concepts[14].wikidata | https://www.wikidata.org/wiki/Q1301371 |
| concepts[14].display_name | Computer network |
| concepts[15].id | https://openalex.org/C199360897 |
| concepts[15].level | 1 |
| concepts[15].score | 0.10310021042823792 |
| concepts[15].wikidata | https://www.wikidata.org/wiki/Q9143 |
| concepts[15].display_name | Programming language |
| concepts[16].id | https://openalex.org/C86803240 |
| concepts[16].level | 0 |
| concepts[16].score | 0.0 |
| concepts[16].wikidata | https://www.wikidata.org/wiki/Q420 |
| concepts[16].display_name | Biology |
| concepts[17].id | https://openalex.org/C18903297 |
| concepts[17].level | 1 |
| concepts[17].score | 0.0 |
| concepts[17].wikidata | https://www.wikidata.org/wiki/Q7150 |
| concepts[17].display_name | Ecology |
| concepts[18].id | https://openalex.org/C76155785 |
| concepts[18].level | 1 |
| concepts[18].score | 0.0 |
| concepts[18].wikidata | https://www.wikidata.org/wiki/Q418 |
| concepts[18].display_name | Telecommunications |
| keywords[0].id | https://openalex.org/keywords/computer-science |
| keywords[0].score | 0.8779021501541138 |
| keywords[0].display_name | Computer science |
| keywords[1].id | https://openalex.org/keywords/parallel-computing |
| keywords[1].score | 0.6097621917724609 |
| keywords[1].display_name | Parallel computing |
| keywords[2].id | https://openalex.org/keywords/convolutional-neural-network |
| keywords[2].score | 0.5409556031227112 |
| keywords[2].display_name | Convolutional neural network |
| keywords[3].id | https://openalex.org/keywords/occam |
| keywords[3].score | 0.524736225605011 |
| keywords[3].display_name | occam |
| keywords[4].id | https://openalex.org/keywords/reuse |
| keywords[4].score | 0.5087361931800842 |
| keywords[4].display_name | Reuse |
| keywords[5].id | https://openalex.org/keywords/chip |
| keywords[5].score | 0.5039469599723816 |
| keywords[5].display_name | Chip |
| keywords[6].id | https://openalex.org/keywords/pipeline |
| keywords[6].score | 0.49973559379577637 |
| keywords[6].display_name | Pipeline (software) |
| keywords[7].id | https://openalex.org/keywords/multithreading |
| keywords[7].score | 0.47680041193962097 |
| keywords[7].display_name | Multithreading |
| keywords[8].id | https://openalex.org/keywords/speedup |
| keywords[8].score | 0.46002355217933655 |
| keywords[8].display_name | Speedup |
| keywords[9].id | https://openalex.org/keywords/asynchronous-communication |
| keywords[9].score | 0.43010854721069336 |
| keywords[9].display_name | Asynchronous communication |
| keywords[10].id | https://openalex.org/keywords/computer-engineering |
| keywords[10].score | 0.40509775280952454 |
| keywords[10].display_name | Computer engineering |
| keywords[11].id | https://openalex.org/keywords/algorithm |
| keywords[11].score | 0.35612186789512634 |
| keywords[11].display_name | Algorithm |
| keywords[12].id | https://openalex.org/keywords/artificial-intelligence |
| keywords[12].score | 0.17872759699821472 |
| keywords[12].display_name | Artificial intelligence |
| keywords[13].id | https://openalex.org/keywords/thread |
| keywords[13].score | 0.15554490685462952 |
| keywords[13].display_name | Thread (computing) |
| keywords[14].id | https://openalex.org/keywords/computer-network |
| keywords[14].score | 0.12033155560493469 |
| keywords[14].display_name | Computer network |
| keywords[15].id | https://openalex.org/keywords/programming-language |
| keywords[15].score | 0.10310021042823792 |
| keywords[15].display_name | Programming language |
| language | en |
| locations[0].id | doi:10.1145/3566052 |
| locations[0].is_oa | True |
| locations[0].source.id | https://openalex.org/S26056741 |
| locations[0].source.issn | 1544-3566, 1544-3973 |
| locations[0].source.type | journal |
| locations[0].source.is_oa | True |
| locations[0].source.issn_l | 1544-3566 |
| locations[0].source.is_core | True |
| locations[0].source.is_in_doaj | False |
| locations[0].source.display_name | ACM Transactions on Architecture and Code Optimization |
| locations[0].source.host_organization | https://openalex.org/P4310319798 |
| locations[0].source.host_organization_name | Association for Computing Machinery |
| locations[0].source.host_organization_lineage | https://openalex.org/P4310319798 |
| locations[0].source.host_organization_lineage_names | Association for Computing Machinery |
| locations[0].license | |
| locations[0].pdf_url | https://dl.acm.org/doi/pdf/10.1145/3566052 |
| locations[0].version | publishedVersion |
| locations[0].raw_type | journal-article |
| locations[0].license_id | |
| locations[0].is_accepted | True |
| locations[0].is_published | True |
| locations[0].raw_source_name | ACM Transactions on Architecture and Code Optimization |
| locations[0].landing_page_url | https://doi.org/10.1145/3566052 |
| indexed_in | crossref |
| authorships[0].author.id | https://openalex.org/A5001458388 |
| authorships[0].author.orcid | https://orcid.org/0000-0002-3370-3576 |
| authorships[0].author.display_name | Ashish Gondimalla |
| authorships[0].countries | US |
| authorships[0].affiliations[0].institution_ids | https://openalex.org/I219193219 |
| authorships[0].affiliations[0].raw_affiliation_string | Purdue University, West Lafayette, Indiana, USA |
| authorships[0].institutions[0].id | https://openalex.org/I219193219 |
| authorships[0].institutions[0].ror | https://ror.org/02dqehb95 |
| authorships[0].institutions[0].type | education |
| authorships[0].institutions[0].lineage | https://openalex.org/I219193219 |
| authorships[0].institutions[0].country_code | US |
| authorships[0].institutions[0].display_name | Purdue University West Lafayette |
| authorships[0].author_position | first |
| authorships[0].raw_author_name | Ashish Gondimalla |
| authorships[0].is_corresponding | False |
| authorships[0].raw_affiliation_strings | Purdue University, West Lafayette, Indiana, USA |
| authorships[1].author.id | https://openalex.org/A5101876435 |
| authorships[1].author.orcid | https://orcid.org/0000-0003-4022-8897 |
| authorships[1].author.display_name | Jianqiao Liu |
| authorships[1].countries | US |
| authorships[1].affiliations[0].institution_ids | https://openalex.org/I1291425158 |
| authorships[1].affiliations[0].raw_affiliation_string | Google, USA, Mountain View, California, USA |
| authorships[1].institutions[0].id | https://openalex.org/I1291425158 |
| authorships[1].institutions[0].ror | https://ror.org/00njsd438 |
| authorships[1].institutions[0].type | company |
| authorships[1].institutions[0].lineage | https://openalex.org/I1291425158, https://openalex.org/I4210128969 |
| authorships[1].institutions[0].country_code | US |
| authorships[1].institutions[0].display_name | Google (United States) |
| authorships[1].author_position | middle |
| authorships[1].raw_author_name | Jianqiao Liu |
| authorships[1].is_corresponding | False |
| authorships[1].raw_affiliation_strings | Google, USA, Mountain View, California, USA |
| authorships[2].author.id | https://openalex.org/A5069139257 |
| authorships[2].author.orcid | https://orcid.org/0000-0003-4164-4542 |
| authorships[2].author.display_name | Mithuna Thottethodi |
| authorships[2].countries | US |
| authorships[2].affiliations[0].institution_ids | https://openalex.org/I219193219 |
| authorships[2].affiliations[0].raw_affiliation_string | Purdue University, West Lafayette, Indiana, USA |
| authorships[2].institutions[0].id | https://openalex.org/I219193219 |
| authorships[2].institutions[0].ror | https://ror.org/02dqehb95 |
| authorships[2].institutions[0].type | education |
| authorships[2].institutions[0].lineage | https://openalex.org/I219193219 |
| authorships[2].institutions[0].country_code | US |
| authorships[2].institutions[0].display_name | Purdue University West Lafayette |
| authorships[2].author_position | middle |
| authorships[2].raw_author_name | Mithuna Thottethodi |
| authorships[2].is_corresponding | False |
| authorships[2].raw_affiliation_strings | Purdue University, West Lafayette, Indiana, USA |
| authorships[3].author.id | https://openalex.org/A5103145581 |
| authorships[3].author.orcid | https://orcid.org/0000-0001-6624-4372 |
| authorships[3].author.display_name | T. N. Vijaykumar |
| authorships[3].countries | US |
| authorships[3].affiliations[0].institution_ids | https://openalex.org/I219193219 |
| authorships[3].affiliations[0].raw_affiliation_string | Purdue University, West Lafayette, Indiana, USA |
| authorships[3].institutions[0].id | https://openalex.org/I219193219 |
| authorships[3].institutions[0].ror | https://ror.org/02dqehb95 |
| authorships[3].institutions[0].type | education |
| authorships[3].institutions[0].lineage | https://openalex.org/I219193219 |
| authorships[3].institutions[0].country_code | US |
| authorships[3].institutions[0].display_name | Purdue University West Lafayette |
| authorships[3].author_position | last |
| authorships[3].raw_author_name | T. N. Vijaykumar |
| authorships[3].is_corresponding | False |
| authorships[3].raw_affiliation_strings | Purdue University, West Lafayette, Indiana, USA |
| has_content.pdf | True |
| has_content.grobid_xml | True |
| is_paratext | False |
| open_access.is_oa | True |
| open_access.oa_url | https://dl.acm.org/doi/pdf/10.1145/3566052 |
| open_access.oa_status | diamond |
| open_access.any_repository_has_fulltext | False |
| created_date | 2025-10-10T00:00:00 |
| display_name | Occam: Optimal Data Reuse for Convolutional Neural Networks |
| has_fulltext | True |
| is_retracted | False |
| updated_date | 2025-11-06T03:46:38.306776 |
| primary_topic.id | https://openalex.org/T10036 |
| primary_topic.field.id | https://openalex.org/fields/17 |
| primary_topic.field.display_name | Computer Science |
| primary_topic.score | 0.9998000264167786 |
| primary_topic.domain.id | https://openalex.org/domains/3 |
| primary_topic.domain.display_name | Physical Sciences |
| primary_topic.subfield.id | https://openalex.org/subfields/1707 |
| primary_topic.subfield.display_name | Computer Vision and Pattern Recognition |
| primary_topic.display_name | Advanced Neural Network Applications |
| related_works | https://openalex.org/W1993191611, https://openalex.org/W3089117407, https://openalex.org/W2030271374, https://openalex.org/W1850053445, https://openalex.org/W2010970156, https://openalex.org/W2023373638, https://openalex.org/W2003336045, https://openalex.org/W1509211761, https://openalex.org/W606126729, https://openalex.org/W2133693067 |
| cited_by_count | 4 |
| counts_by_year[0].year | 2025 |
| counts_by_year[0].cited_by_count | 1 |
| counts_by_year[1].year | 2024 |
| counts_by_year[1].cited_by_count | 2 |
| counts_by_year[2].year | 2023 |
| counts_by_year[2].cited_by_count | 1 |
| locations_count | 1 |
| best_oa_location.id | doi:10.1145/3566052 |
| best_oa_location.is_oa | True |
| best_oa_location.source.id | https://openalex.org/S26056741 |
| best_oa_location.source.issn | 1544-3566, 1544-3973 |
| best_oa_location.source.type | journal |
| best_oa_location.source.is_oa | True |
| best_oa_location.source.issn_l | 1544-3566 |
| best_oa_location.source.is_core | True |
| best_oa_location.source.is_in_doaj | False |
| best_oa_location.source.display_name | ACM Transactions on Architecture and Code Optimization |
| best_oa_location.source.host_organization | https://openalex.org/P4310319798 |
| best_oa_location.source.host_organization_name | Association for Computing Machinery |
| best_oa_location.source.host_organization_lineage | https://openalex.org/P4310319798 |
| best_oa_location.source.host_organization_lineage_names | Association for Computing Machinery |
| best_oa_location.license | |
| best_oa_location.pdf_url | https://dl.acm.org/doi/pdf/10.1145/3566052 |
| best_oa_location.version | publishedVersion |
| best_oa_location.raw_type | journal-article |
| best_oa_location.license_id | |
| best_oa_location.is_accepted | True |
| best_oa_location.is_published | True |
| best_oa_location.raw_source_name | ACM Transactions on Architecture and Code Optimization |
| best_oa_location.landing_page_url | https://doi.org/10.1145/3566052 |
| primary_location.id | doi:10.1145/3566052 |
| primary_location.is_oa | True |
| primary_location.source.id | https://openalex.org/S26056741 |
| primary_location.source.issn | 1544-3566, 1544-3973 |
| primary_location.source.type | journal |
| primary_location.source.is_oa | True |
| primary_location.source.issn_l | 1544-3566 |
| primary_location.source.is_core | True |
| primary_location.source.is_in_doaj | False |
| primary_location.source.display_name | ACM Transactions on Architecture and Code Optimization |
| primary_location.source.host_organization | https://openalex.org/P4310319798 |
| primary_location.source.host_organization_name | Association for Computing Machinery |
| primary_location.source.host_organization_lineage | https://openalex.org/P4310319798 |
| primary_location.source.host_organization_lineage_names | Association for Computing Machinery |
| primary_location.license | |
| primary_location.pdf_url | https://dl.acm.org/doi/pdf/10.1145/3566052 |
| primary_location.version | publishedVersion |
| primary_location.raw_type | journal-article |
| primary_location.license_id | |
| primary_location.is_accepted | True |
| primary_location.is_published | True |
| primary_location.raw_source_name | ACM Transactions on Architecture and Code Optimization |
| primary_location.landing_page_url | https://doi.org/10.1145/3566052 |
| publication_date | 2022-12-16 |
| publication_year | 2022 |
| referenced_works | https://openalex.org/W2541839172, https://openalex.org/W4251575795, https://openalex.org/W1979527452, https://openalex.org/W2028067303, https://openalex.org/W2528222261, https://openalex.org/W3174885745, https://openalex.org/W2152839228, https://openalex.org/W2048266589, https://openalex.org/W3024621361, https://openalex.org/W2158158791, https://openalex.org/W2037915256, https://openalex.org/W2108598243, https://openalex.org/W2067523571, https://openalex.org/W2883929540, https://openalex.org/W2921126577, https://openalex.org/W2759398875, https://openalex.org/W2950656546, https://openalex.org/W2807147113, https://openalex.org/W2969388332, https://openalex.org/W6687483927, https://openalex.org/W2901299405, https://openalex.org/W3115432371, https://openalex.org/W2296218291, https://openalex.org/W2786066748, https://openalex.org/W2606722458, https://openalex.org/W1598866093, https://openalex.org/W2952294252, https://openalex.org/W2098220211, https://openalex.org/W2112796928, https://openalex.org/W2093043622, https://openalex.org/W3201891881, https://openalex.org/W2000967104, https://openalex.org/W3167739300, https://openalex.org/W2163687928, https://openalex.org/W2084917734, https://openalex.org/W3169997198, https://openalex.org/W2940862705, https://openalex.org/W2625457103, https://openalex.org/W2055312318, https://openalex.org/W2117539524, https://openalex.org/W2954996726, https://openalex.org/W2911148381, https://openalex.org/W3047599609, https://openalex.org/W3081486497, https://openalex.org/W2119609467, https://openalex.org/W2562564656, https://openalex.org/W2094756095, https://openalex.org/W2963840672, https://openalex.org/W2953212265, https://openalex.org/W2604272474, https://openalex.org/W2603836393, https://openalex.org/W2952020226, https://openalex.org/W2963959650, https://openalex.org/W2285660444, https://openalex.org/W1503700136, https://openalex.org/W4300954432, https://openalex.org/W2949650786, https://openalex.org/W2194775991, https://openalex.org/W4299619069, https://openalex.org/W2163605009, https://openalex.org/W2286365479, https://openalex.org/W2593390416, https://openalex.org/W1849277567, https://openalex.org/W4293775467, https://openalex.org/W2565305208, https://openalex.org/W2626018039, https://openalex.org/W2289252105, https://openalex.org/W2991040477, https://openalex.org/W2516141709, https://openalex.org/W2963358710, https://openalex.org/W2900228909, https://openalex.org/W1507039213, https://openalex.org/W4230315356 |
| referenced_works_count | 73 |
| abstract_inverted_index.a | 55, 155, 162, 176, 198, 202, 295 |
| abstract_inverted_index.We | 16, 101 |
| abstract_inverted_index.an | 235 |
| abstract_inverted_index.as | 6, 126, 210 |
| abstract_inverted_index.at | 171 |
| abstract_inverted_index.by | 251, 276 |
| abstract_inverted_index.in | 12, 150, 234 |
| abstract_inverted_index.is | 89, 144, 182 |
| abstract_inverted_index.of | 22, 26, 65 |
| abstract_inverted_index.on | 18, 194, 270, 307 |
| abstract_inverted_index.so | 200 |
| abstract_inverted_index.to | 34, 38, 99, 104, 130, 148, 165, 248 |
| abstract_inverted_index.we | 112, 121, 153, 238 |
| abstract_inverted_index.33% | 286 |
| abstract_inverted_index.CNN | 164 |
| abstract_inverted_index.Our | 266 |
| abstract_inverted_index.and | 36, 51, 78, 85, 205, 225, 278, 281, 285, 304, 313 |
| abstract_inverted_index.are | 4, 31, 80 |
| abstract_inverted_index.fit | 149 |
| abstract_inverted_index.for | 9, 117, 175, 222 |
| abstract_inverted_index.its | 223 |
| abstract_inverted_index.may | 232 |
| abstract_inverted_index.our | 184 |
| abstract_inverted_index.the | 19, 24, 66, 74, 86, 95, 114, 123, 127, 135, 141, 167, 172, 187, 229, 290, 310 |
| abstract_inverted_index.too | 146 |
| abstract_inverted_index.via | 108 |
| abstract_inverted_index.– | 46, 53 |
| abstract_inverted_index.21× | 277 |
| abstract_inverted_index.CNNs | 30 |
| abstract_inverted_index.base | 291, 311 |
| abstract_inverted_index.case | 312 |
| abstract_inverted_index.cuts | 273 |
| abstract_inverted_index.data | 45, 68, 98 |
| abstract_inverted_index.each | 47, 216 |
| abstract_inverted_index.four | 109 |
| abstract_inverted_index.from | 83 |
| abstract_inverted_index.full | 70, 106, 118, 132 |
| abstract_inverted_index.gate | 297 |
| abstract_inverted_index.once | 82, 91 |
| abstract_inverted_index.only | 63, 221 |
| abstract_inverted_index.pass | 213 |
| abstract_inverted_index.read | 81 |
| abstract_inverted_index.show | 268 |
| abstract_inverted_index.some | 64 |
| abstract_inverted_index.than | 289, 309 |
| abstract_inverted_index.that | 73, 159, 201, 244 |
| abstract_inverted_index.work | 61 |
| abstract_inverted_index.1.5× | 305 |
| abstract_inverted_index.6.1× | 303 |
| abstract_inverted_index.Layer | 314 |
| abstract_inverted_index.Occam | 103, 272, 301 |
| abstract_inverted_index.STAPs | 257 |
| abstract_inverted_index.Using | 294 |
| abstract_inverted_index.While | 29, 59, 180 |
| abstract_inverted_index.array | 298 |
| abstract_inverted_index.avoid | 39 |
| abstract_inverted_index.case, | 292 |
| abstract_inverted_index.final | 87 |
| abstract_inverted_index.focus | 17 |
| abstract_inverted_index.given | 163, 177 |
| abstract_inverted_index.image | 10, 27, 77 |
| abstract_inverted_index.input | 76 |
| abstract_inverted_index.large | 44, 147 |
| abstract_inverted_index.least | 136, 168 |
| abstract_inverted_index.often | 145 |
| abstract_inverted_index.poses | 54 |
| abstract_inverted_index.reuse | 71, 107, 133 |
| abstract_inverted_index.that, | 269 |
| abstract_inverted_index.tools | 8 |
| abstract_inverted_index.using | 134 |
| abstract_inverted_index.(CNNs) | 3 |
| abstract_inverted_index.(FPGA) | 299 |
| abstract_inverted_index.(i.e., | 215 |
| abstract_inverted_index.1.21× | 282 |
| abstract_inverted_index.2.04× | 280 |
| abstract_inverted_index.First, | 111 |
| abstract_inverted_index.Third, | 139 |
| abstract_inverted_index.across | 254 |
| abstract_inverted_index.better | 283, 287 |
| abstract_inverted_index.chips, | 196 |
| abstract_inverted_index.energy | 288 |
| abstract_inverted_index.highly | 32 |
| abstract_inverted_index.images | 212 |
| abstract_inverted_index.incurs | 218 |
| abstract_inverted_index.input, | 49 |
| abstract_inverted_index.inputs | 224 |
| abstract_inverted_index.memory | 40, 56 |
| abstract_inverted_index.neural | 1 |
| abstract_inverted_index.output | 52, 88 |
| abstract_inverted_index.remain | 208 |
| abstract_inverted_index.reside | 193 |
| abstract_inverted_index.result | 233 |
| abstract_inverted_index.reuse, | 69 |
| abstract_inverted_index.reuse. | 119 |
| abstract_inverted_index.stages | 247 |
| abstract_inverted_index.tiles. | 190 |
| abstract_inverted_index.tiling | 181 |
| abstract_inverted_index.(STAPs) | 243 |
| abstract_inverted_index.CNNs’ | 43 |
| abstract_inverted_index.Fusion, | 315 |
| abstract_inverted_index.Second, | 120 |
| abstract_inverted_index.achieve | 258 |
| abstract_inverted_index.because | 140, 228 |
| abstract_inverted_index.better, | 306 |
| abstract_inverted_index.capture | 105, 131 |
| abstract_inverted_index.closure | 125, 143, 207 |
| abstract_inverted_index.dynamic | 156 |
| abstract_inverted_index.filters | 79, 204 |
| abstract_inverted_index.forming | 197 |
| abstract_inverted_index.implies | 72 |
| abstract_inverted_index.improve | 249 |
| abstract_inverted_index.initial | 75 |
| abstract_inverted_index.issues, | 42 |
| abstract_inverted_index.latency | 25, 41 |
| abstract_inverted_index.memory, | 152 |
| abstract_inverted_index.memory. | 138 |
| abstract_inverted_index.on-chip | 137, 151, 178, 209 |
| abstract_inverted_index.optimal | 188, 230, 264 |
| abstract_inverted_index.problem | 21 |
| abstract_inverted_index.propose | 102, 154, 239 |
| abstract_inverted_index.through | 214 |
| abstract_inverted_index.traffic | 170, 220 |
| abstract_inverted_index.without | 93, 261 |
| abstract_inverted_index.written | 90 |
| abstract_inverted_index.Finally, | 227 |
| abstract_inverted_index.achieves | 279 |
| abstract_inverted_index.amenable | 33 |
| abstract_inverted_index.average, | 271, 308 |
| abstract_inverted_index.balanced | 259 |
| abstract_inverted_index.captures | 62 |
| abstract_inverted_index.changing | 262 |
| abstract_inverted_index.emerging | 5 |
| abstract_inverted_index.enormous | 67 |
| abstract_inverted_index.filters, | 50 |
| abstract_inverted_index.identify | 113, 122 |
| abstract_inverted_index.networks | 2 |
| abstract_inverted_index.off-chip | 84, 92, 169, 219, 274 |
| abstract_inverted_index.performs | 302 |
| abstract_inverted_index.pipeline | 199 |
| abstract_inverted_index.powerful | 7 |
| abstract_inverted_index.previous | 60 |
| abstract_inverted_index.problem. | 58 |
| abstract_inverted_index.spilling | 94 |
| abstract_inverted_index.Occam’s | 191, 263 |
| abstract_inverted_index.algorithm | 158 |
| abstract_inverted_index.bandwidth | 57 |
| abstract_inverted_index.capacity. | 179 |
| abstract_inverted_index.condition | 129 |
| abstract_inverted_index.different | 195, 211 |
| abstract_inverted_index.guarantee | 166 |
| abstract_inverted_index.important | 13, 20 |
| abstract_inverted_index.improving | 23 |
| abstract_inverted_index.layers’ | 97 |
| abstract_inverted_index.layer’s | 48 |
| abstract_inverted_index.necessary | 115 |
| abstract_inverted_index.off-chip. | 100 |
| abstract_inverted_index.optimally | 160 |
| abstract_inverted_index.outputs). | 226 |
| abstract_inverted_index.partition | 173, 217 |
| abstract_inverted_index.pipeline, | 237 |
| abstract_inverted_index.pipelines | 242, 260 |
| abstract_inverted_index.replicas. | 255 |
| abstract_inverted_index.replicate | 245 |
| abstract_inverted_index.staggered | 240 |
| abstract_inverted_index.transfers | 275 |
| abstract_inverted_index.bottleneck | 246 |
| abstract_inverted_index.boundaries | 174 |
| abstract_inverted_index.commercial | 14 |
| abstract_inverted_index.conditions | 116 |
| abstract_inverted_index.dependence | 124, 142, 206 |
| abstract_inverted_index.determines | 186 |
| abstract_inverted_index.partitions | 161, 192, 231 |
| abstract_inverted_index.processing | 11 |
| abstract_inverted_index.staggering | 252 |
| abstract_inverted_index.sufficient | 128 |
| abstract_inverted_index.throughput | 250 |
| abstract_inverted_index.unbalanced | 236 |
| abstract_inverted_index.cross-layer | 189 |
| abstract_inverted_index.prefetching | 35 |
| abstract_inverted_index.programming | 157 |
| abstract_inverted_index.simulations | 267 |
| abstract_inverted_index.well-known, | 183 |
| abstract_inverted_index.Importantly, | 256 |
| abstract_inverted_index.asynchronous | 241 |
| abstract_inverted_index.contribution | 185 |
| abstract_inverted_index.intermediate | 96 |
| abstract_inverted_index.mini-batches | 253 |
| abstract_inverted_index.performance, | 284 |
| abstract_inverted_index.recognition. | 28 |
| abstract_inverted_index.Convolutional | 0 |
| abstract_inverted_index.applications. | 15 |
| abstract_inverted_index.partitioning. | 265 |
| abstract_inverted_index.partition’s | 203 |
| abstract_inverted_index.respectively. | 293, 316 |
| abstract_inverted_index.contributions. | 110 |
| abstract_inverted_index.multithreading | 37 |
| abstract_inverted_index.implementation, | 300 |
| abstract_inverted_index.field-programmable | 296 |
| cited_by_percentile_year.max | 96 |
| cited_by_percentile_year.min | 89 |
| countries_distinct_count | 1 |
| institutions_distinct_count | 4 |
| citation_normalized_percentile.value | 0.59594845 |
| citation_normalized_percentile.is_in_top_1_percent | False |
| citation_normalized_percentile.is_in_top_10_percent | False |