MMStencil: Optimizing High-order Stencils on Multicore CPU using Matrix Unit Article Swipe
YOU?
·
· 2025
· Open Access
·
· DOI: https://doi.org/10.48550/arxiv.2507.11067
Matrix-accelerated stencil computation is a hot research topic, yet its application to three-dimensional (3D) high-order stencils and HPC remains underexplored. With the emergence of matrix units on multicore CPUs, we analyze matrix-based acceleration strategies and tailor an optimal approach for 3D high-order stencils. We introduce algorithmic optimizations based on SIMD and matrix units to address strided memory accesses, alignment conflicts, and redundant accesses. We propose memory optimizations to boost on-package memory efficiency, and a novel multi-thread parallelism paradigm to overcome data-sharing challenges caused by the absence of shared data caches. MMStencil sustains consistently high hardware utilization across diverse stencil shapes and dimensions. Our DMA-based inter-NUMA communication further mitigates NUMA effects and MPI limitations in hybrid parallelism. Combining all the innovations, MMStencil outperforms state-of-the-art libraries on Nvidia A100 GPGPU by up to 2.1x. Moreover, the performance improvements translate directly to real-world HPC applications and enable RTM applications to yield 1.8x speedup versus a highly optimized industrial Nvidia A100 GPGPU version.
Related Topics
- Type
- preprint
- Language
- en
- Landing Page
- http://arxiv.org/abs/2507.11067
- https://arxiv.org/pdf/2507.11067
- OA Status
- green
- OpenAlex ID
- https://openalex.org/W4414944741
Raw OpenAlex JSON
- OpenAlex ID
-
https://openalex.org/W4414944741Canonical identifier for this work in OpenAlex
- DOI
-
https://doi.org/10.48550/arxiv.2507.11067Digital Object Identifier
- Title
-
MMStencil: Optimizing High-order Stencils on Multicore CPU using Matrix UnitWork title
- Type
-
preprintOpenAlex work type
- Language
-
enPrimary language
- Publication year
-
2025Year of publication
- Publication date
-
2025-07-15Full publication date if available
- Authors
-
Yinuo Wang, Mao Tian-qi, Lin Gan, Wubing Wan, Zeyu Song, Jiayu Fu, Ling He, Wenqiang Wang, Zhiya Yin, Wei Xue, Guangwen YangList of authors in order
- Landing page
-
https://arxiv.org/abs/2507.11067Publisher landing page
- PDF URL
-
https://arxiv.org/pdf/2507.11067Direct link to full text PDF
- Open access
-
YesWhether a free full text is available
- OA status
-
greenOpen access status per OpenAlex
- OA URL
-
https://arxiv.org/pdf/2507.11067Direct OA link when available
- Cited by
-
0Total citation count in OpenAlex
Full payload
| id | https://openalex.org/W4414944741 |
|---|---|
| doi | https://doi.org/10.48550/arxiv.2507.11067 |
| ids.doi | https://doi.org/10.48550/arxiv.2507.11067 |
| ids.openalex | https://openalex.org/W4414944741 |
| fwci | |
| type | preprint |
| title | MMStencil: Optimizing High-order Stencils on Multicore CPU using Matrix Unit |
| biblio.issue | |
| biblio.volume | |
| biblio.last_page | |
| biblio.first_page | |
| topics[0].id | https://openalex.org/T10739 |
| topics[0].field.id | https://openalex.org/fields/31 |
| topics[0].field.display_name | Physics and Astronomy |
| topics[0].score | 0.4440999925136566 |
| topics[0].domain.id | https://openalex.org/domains/3 |
| topics[0].domain.display_name | Physical Sciences |
| topics[0].subfield.id | https://openalex.org/subfields/3107 |
| topics[0].subfield.display_name | Atomic and Molecular Physics, and Optics |
| topics[0].display_name | Electromagnetic Scattering and Analysis |
| is_xpac | False |
| apc_list | |
| apc_paid | |
| language | en |
| locations[0].id | pmh:oai:arXiv.org:2507.11067 |
| locations[0].is_oa | True |
| locations[0].source.id | https://openalex.org/S4306400194 |
| locations[0].source.issn | |
| locations[0].source.type | repository |
| locations[0].source.is_oa | True |
| locations[0].source.issn_l | |
| locations[0].source.is_core | False |
| locations[0].source.is_in_doaj | False |
| locations[0].source.display_name | arXiv (Cornell University) |
| locations[0].source.host_organization | https://openalex.org/I205783295 |
| locations[0].source.host_organization_name | Cornell University |
| locations[0].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[0].license | |
| locations[0].pdf_url | https://arxiv.org/pdf/2507.11067 |
| locations[0].version | submittedVersion |
| locations[0].raw_type | text |
| locations[0].license_id | |
| locations[0].is_accepted | False |
| locations[0].is_published | False |
| locations[0].raw_source_name | |
| locations[0].landing_page_url | http://arxiv.org/abs/2507.11067 |
| locations[1].id | doi:10.48550/arxiv.2507.11067 |
| locations[1].is_oa | True |
| locations[1].source.id | https://openalex.org/S4306400194 |
| locations[1].source.issn | |
| locations[1].source.type | repository |
| locations[1].source.is_oa | True |
| locations[1].source.issn_l | |
| locations[1].source.is_core | False |
| locations[1].source.is_in_doaj | False |
| locations[1].source.display_name | arXiv (Cornell University) |
| locations[1].source.host_organization | https://openalex.org/I205783295 |
| locations[1].source.host_organization_name | Cornell University |
| locations[1].source.host_organization_lineage | https://openalex.org/I205783295 |
| locations[1].license | |
| locations[1].pdf_url | |
| locations[1].version | |
| locations[1].raw_type | article |
| locations[1].license_id | |
| locations[1].is_accepted | False |
| locations[1].is_published | |
| locations[1].raw_source_name | |
| locations[1].landing_page_url | https://doi.org/10.48550/arxiv.2507.11067 |
| indexed_in | arxiv, datacite |
| authorships[0].author.id | https://openalex.org/A5029347245 |
| authorships[0].author.orcid | |
| authorships[0].author.display_name | Yinuo Wang |
| authorships[0].author_position | first |
| authorships[0].raw_author_name | Wang, Yinuo |
| authorships[0].is_corresponding | False |
| authorships[1].author.id | https://openalex.org/A5023448415 |
| authorships[1].author.orcid | |
| authorships[1].author.display_name | Mao Tian-qi |
| authorships[1].author_position | middle |
| authorships[1].raw_author_name | Mao, Tianqi |
| authorships[1].is_corresponding | False |
| authorships[2].author.id | https://openalex.org/A5038750675 |
| authorships[2].author.orcid | https://orcid.org/0000-0003-1297-4462 |
| authorships[2].author.display_name | Lin Gan |
| authorships[2].author_position | middle |
| authorships[2].raw_author_name | Gan, Lin |
| authorships[2].is_corresponding | False |
| authorships[3].author.id | https://openalex.org/A5033706115 |
| authorships[3].author.orcid | https://orcid.org/0009-0001-5174-1832 |
| authorships[3].author.display_name | Wubing Wan |
| authorships[3].author_position | middle |
| authorships[3].raw_author_name | Wan, Wubing |
| authorships[3].is_corresponding | False |
| authorships[4].author.id | https://openalex.org/A5101941615 |
| authorships[4].author.orcid | https://orcid.org/0000-0003-4992-4143 |
| authorships[4].author.display_name | Zeyu Song |
| authorships[4].author_position | middle |
| authorships[4].raw_author_name | Song, Zeyu |
| authorships[4].is_corresponding | False |
| authorships[5].author.id | https://openalex.org/A5109486609 |
| authorships[5].author.orcid | https://orcid.org/0009-0000-8610-2945 |
| authorships[5].author.display_name | Jiayu Fu |
| authorships[5].author_position | middle |
| authorships[5].raw_author_name | Fu, Jiayu |
| authorships[5].is_corresponding | False |
| authorships[6].author.id | https://openalex.org/A5048198045 |
| authorships[6].author.orcid | https://orcid.org/0000-0003-3242-3641 |
| authorships[6].author.display_name | Ling He |
| authorships[6].author_position | middle |
| authorships[6].raw_author_name | He, Lanke |
| authorships[6].is_corresponding | False |
| authorships[7].author.id | https://openalex.org/A5100628303 |
| authorships[7].author.orcid | https://orcid.org/0000-0002-4425-6636 |
| authorships[7].author.display_name | Wenqiang Wang |
| authorships[7].author_position | middle |
| authorships[7].raw_author_name | Wang, Wenqiang |
| authorships[7].is_corresponding | False |
| authorships[8].author.id | https://openalex.org/A5112735786 |
| authorships[8].author.orcid | |
| authorships[8].author.display_name | Zhiya Yin |
| authorships[8].author_position | middle |
| authorships[8].raw_author_name | Yin, Zekun |
| authorships[8].is_corresponding | False |
| authorships[9].author.id | https://openalex.org/A5100408966 |
| authorships[9].author.orcid | https://orcid.org/0000-0003-0124-9433 |
| authorships[9].author.display_name | Wei Xue |
| authorships[9].author_position | middle |
| authorships[9].raw_author_name | Xue, Wei |
| authorships[9].is_corresponding | False |
| authorships[10].author.id | https://openalex.org/A5101678397 |
| authorships[10].author.orcid | https://orcid.org/0000-0002-8673-8254 |
| authorships[10].author.display_name | Guangwen Yang |
| authorships[10].author_position | last |
| authorships[10].raw_author_name | Yang, Guangwen |
| authorships[10].is_corresponding | False |
| has_content.pdf | False |
| has_content.grobid_xml | False |
| is_paratext | False |
| open_access.is_oa | True |
| open_access.oa_url | https://arxiv.org/pdf/2507.11067 |
| open_access.oa_status | green |
| open_access.any_repository_has_fulltext | False |
| created_date | 2025-10-08T00:00:00 |
| display_name | MMStencil: Optimizing High-order Stencils on Multicore CPU using Matrix Unit |
| has_fulltext | False |
| is_retracted | False |
| updated_date | 2025-11-06T06:51:31.235846 |
| primary_topic.id | https://openalex.org/T10739 |
| primary_topic.field.id | https://openalex.org/fields/31 |
| primary_topic.field.display_name | Physics and Astronomy |
| primary_topic.score | 0.4440999925136566 |
| primary_topic.domain.id | https://openalex.org/domains/3 |
| primary_topic.domain.display_name | Physical Sciences |
| primary_topic.subfield.id | https://openalex.org/subfields/3107 |
| primary_topic.subfield.display_name | Atomic and Molecular Physics, and Optics |
| primary_topic.display_name | Electromagnetic Scattering and Analysis |
| cited_by_count | 0 |
| locations_count | 2 |
| best_oa_location.id | pmh:oai:arXiv.org:2507.11067 |
| best_oa_location.is_oa | True |
| best_oa_location.source.id | https://openalex.org/S4306400194 |
| best_oa_location.source.issn | |
| best_oa_location.source.type | repository |
| best_oa_location.source.is_oa | True |
| best_oa_location.source.issn_l | |
| best_oa_location.source.is_core | False |
| best_oa_location.source.is_in_doaj | False |
| best_oa_location.source.display_name | arXiv (Cornell University) |
| best_oa_location.source.host_organization | https://openalex.org/I205783295 |
| best_oa_location.source.host_organization_name | Cornell University |
| best_oa_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| best_oa_location.license | |
| best_oa_location.pdf_url | https://arxiv.org/pdf/2507.11067 |
| best_oa_location.version | submittedVersion |
| best_oa_location.raw_type | text |
| best_oa_location.license_id | |
| best_oa_location.is_accepted | False |
| best_oa_location.is_published | False |
| best_oa_location.raw_source_name | |
| best_oa_location.landing_page_url | http://arxiv.org/abs/2507.11067 |
| primary_location.id | pmh:oai:arXiv.org:2507.11067 |
| primary_location.is_oa | True |
| primary_location.source.id | https://openalex.org/S4306400194 |
| primary_location.source.issn | |
| primary_location.source.type | repository |
| primary_location.source.is_oa | True |
| primary_location.source.issn_l | |
| primary_location.source.is_core | False |
| primary_location.source.is_in_doaj | False |
| primary_location.source.display_name | arXiv (Cornell University) |
| primary_location.source.host_organization | https://openalex.org/I205783295 |
| primary_location.source.host_organization_name | Cornell University |
| primary_location.source.host_organization_lineage | https://openalex.org/I205783295 |
| primary_location.license | |
| primary_location.pdf_url | https://arxiv.org/pdf/2507.11067 |
| primary_location.version | submittedVersion |
| primary_location.raw_type | text |
| primary_location.license_id | |
| primary_location.is_accepted | False |
| primary_location.is_published | False |
| primary_location.raw_source_name | |
| primary_location.landing_page_url | http://arxiv.org/abs/2507.11067 |
| publication_date | 2025-07-15 |
| publication_year | 2025 |
| referenced_works_count | 0 |
| abstract_inverted_index.a | 4, 73, 151 |
| abstract_inverted_index.3D | 40 |
| abstract_inverted_index.We | 43, 63 |
| abstract_inverted_index.an | 36 |
| abstract_inverted_index.by | 83, 128 |
| abstract_inverted_index.in | 113 |
| abstract_inverted_index.is | 3 |
| abstract_inverted_index.of | 23, 86 |
| abstract_inverted_index.on | 26, 48, 124 |
| abstract_inverted_index.to | 11, 53, 67, 78, 130, 138, 146 |
| abstract_inverted_index.up | 129 |
| abstract_inverted_index.we | 29 |
| abstract_inverted_index.HPC | 17, 140 |
| abstract_inverted_index.MPI | 111 |
| abstract_inverted_index.Our | 102 |
| abstract_inverted_index.RTM | 144 |
| abstract_inverted_index.all | 117 |
| abstract_inverted_index.and | 16, 34, 50, 60, 72, 100, 110, 142 |
| abstract_inverted_index.for | 39 |
| abstract_inverted_index.hot | 5 |
| abstract_inverted_index.its | 9 |
| abstract_inverted_index.the | 21, 84, 118, 133 |
| abstract_inverted_index.yet | 8 |
| abstract_inverted_index.(3D) | 13 |
| abstract_inverted_index.1.8x | 148 |
| abstract_inverted_index.A100 | 126, 156 |
| abstract_inverted_index.NUMA | 108 |
| abstract_inverted_index.SIMD | 49 |
| abstract_inverted_index.With | 20 |
| abstract_inverted_index.data | 88 |
| abstract_inverted_index.high | 93 |
| abstract_inverted_index.2.1x. | 131 |
| abstract_inverted_index.CPUs, | 28 |
| abstract_inverted_index.GPGPU | 127, 157 |
| abstract_inverted_index.based | 47 |
| abstract_inverted_index.boost | 68 |
| abstract_inverted_index.novel | 74 |
| abstract_inverted_index.units | 25, 52 |
| abstract_inverted_index.yield | 147 |
| abstract_inverted_index.Nvidia | 125, 155 |
| abstract_inverted_index.across | 96 |
| abstract_inverted_index.caused | 82 |
| abstract_inverted_index.enable | 143 |
| abstract_inverted_index.highly | 152 |
| abstract_inverted_index.hybrid | 114 |
| abstract_inverted_index.matrix | 24, 51 |
| abstract_inverted_index.memory | 56, 65, 70 |
| abstract_inverted_index.shapes | 99 |
| abstract_inverted_index.shared | 87 |
| abstract_inverted_index.tailor | 35 |
| abstract_inverted_index.topic, | 7 |
| abstract_inverted_index.versus | 150 |
| abstract_inverted_index.absence | 85 |
| abstract_inverted_index.address | 54 |
| abstract_inverted_index.analyze | 30 |
| abstract_inverted_index.caches. | 89 |
| abstract_inverted_index.diverse | 97 |
| abstract_inverted_index.effects | 109 |
| abstract_inverted_index.further | 106 |
| abstract_inverted_index.optimal | 37 |
| abstract_inverted_index.propose | 64 |
| abstract_inverted_index.remains | 18 |
| abstract_inverted_index.speedup | 149 |
| abstract_inverted_index.stencil | 1, 98 |
| abstract_inverted_index.strided | 55 |
| abstract_inverted_index.approach | 38 |
| abstract_inverted_index.directly | 137 |
| abstract_inverted_index.hardware | 94 |
| abstract_inverted_index.overcome | 79 |
| abstract_inverted_index.paradigm | 77 |
| abstract_inverted_index.research | 6 |
| abstract_inverted_index.stencils | 15 |
| abstract_inverted_index.sustains | 91 |
| abstract_inverted_index.version. | 158 |
| abstract_inverted_index.Combining | 116 |
| abstract_inverted_index.DMA-based | 103 |
| abstract_inverted_index.MMStencil | 90, 120 |
| abstract_inverted_index.Moreover, | 132 |
| abstract_inverted_index.accesses, | 57 |
| abstract_inverted_index.accesses. | 62 |
| abstract_inverted_index.alignment | 58 |
| abstract_inverted_index.emergence | 22 |
| abstract_inverted_index.introduce | 44 |
| abstract_inverted_index.libraries | 123 |
| abstract_inverted_index.mitigates | 107 |
| abstract_inverted_index.multicore | 27 |
| abstract_inverted_index.optimized | 153 |
| abstract_inverted_index.redundant | 61 |
| abstract_inverted_index.stencils. | 42 |
| abstract_inverted_index.translate | 136 |
| abstract_inverted_index.challenges | 81 |
| abstract_inverted_index.conflicts, | 59 |
| abstract_inverted_index.high-order | 14, 41 |
| abstract_inverted_index.industrial | 154 |
| abstract_inverted_index.inter-NUMA | 104 |
| abstract_inverted_index.on-package | 69 |
| abstract_inverted_index.real-world | 139 |
| abstract_inverted_index.strategies | 33 |
| abstract_inverted_index.algorithmic | 45 |
| abstract_inverted_index.application | 10 |
| abstract_inverted_index.computation | 2 |
| abstract_inverted_index.dimensions. | 101 |
| abstract_inverted_index.efficiency, | 71 |
| abstract_inverted_index.limitations | 112 |
| abstract_inverted_index.outperforms | 121 |
| abstract_inverted_index.parallelism | 76 |
| abstract_inverted_index.performance | 134 |
| abstract_inverted_index.utilization | 95 |
| abstract_inverted_index.acceleration | 32 |
| abstract_inverted_index.applications | 141, 145 |
| abstract_inverted_index.consistently | 92 |
| abstract_inverted_index.data-sharing | 80 |
| abstract_inverted_index.improvements | 135 |
| abstract_inverted_index.innovations, | 119 |
| abstract_inverted_index.matrix-based | 31 |
| abstract_inverted_index.multi-thread | 75 |
| abstract_inverted_index.parallelism. | 115 |
| abstract_inverted_index.communication | 105 |
| abstract_inverted_index.optimizations | 46, 66 |
| abstract_inverted_index.underexplored. | 19 |
| abstract_inverted_index.state-of-the-art | 122 |
| abstract_inverted_index.three-dimensional | 12 |
| abstract_inverted_index.Matrix-accelerated | 0 |
| cited_by_percentile_year | |
| countries_distinct_count | 0 |
| institutions_distinct_count | 11 |
| citation_normalized_percentile |