Matches in Ghent University Academic Bibliography for { <https://biblio.ugent.be/publication/01HQ0WT764375CFNS64KFDW3H3> ?p ?o. }
Showing items 1 to 27 of
27
with 100 items per page.
- 01HQ0WT764375CFNS64KFDW3H3 classification P1.
- 01HQ0WT764375CFNS64KFDW3H3 date "2023".
- 01HQ0WT764375CFNS64KFDW3H3 language "eng".
- 01HQ0WT764375CFNS64KFDW3H3 type conference.
- 01HQ0WT764375CFNS64KFDW3H3 hasPart 01HYG5P9V8KK30BVV44PNTEDFC.pdf.
- 01HQ0WT764375CFNS64KFDW3H3 hasPart urn:uuid:0d45f492-da72-4ea9-82e0-f5fc2892940e.
- 01HQ0WT764375CFNS64KFDW3H3 subject "Technology and Engineering".
- 01HQ0WT764375CFNS64KFDW3H3 doi "10.1145/3575693.3575745".
- 01HQ0WT764375CFNS64KFDW3H3 isbn "9781450399166".
- 01HQ0WT764375CFNS64KFDW3H3 presentedAt urn:uuid:b9983389-81da-473a-a7b9-2ae952f2dac0.
- 01HQ0WT764375CFNS64KFDW3H3 abstract "The parallel execution model of GPUs enables scaling to hundreds of thousands of threads, which is a key capability that many modern high-performance applications exploit. GPU vendors are hence increasing the compute and memory resources with every GPU generation D resulting in the need to efficiently stitch together a plethora of Symmetric Multiprocessors (SMs), Last-Level Cache (LLC) slices and memory controllers while maximizing bandwidth and keeping power consumption and design complexity in check. Conventional GPUs are Uniform Bandwidth Architectures (UBAs) as they provide equal bandwidth between all SMs and all LLC slices. UBA GPUs require a uniform high-bandwidth Network-onChip (NoC), and our key observation is that provisioning a NoC to match the LLC slice bandwidth incurs a hefty power and complexity overhead. We propose the Non-Uniform Bandwidth Architecture (NUBA), a GPU system architecture aimed at fully utilizing LLC slice bandwidth. A NUBA GPU consists of partitions that each feature a few SMs and LLC slices as well as a memory controller D hence exposing the complete LLC bandwidth to the SMs within a partition since they can be connected with point-to-point links D and a NoC between partitions D to enable access to remote data. Exploiting the potential of NUBA GPUs however requires carefully co-designing system software, the compiler and architectural policies. The critical system software component is our Local-And-Balanced (LAB) page placement policy which enables the GPU driver to place data in local partitions while avoiding load imbalance. Moreover, we propose Model-Driven Replication (MDR) which identifies readonly shared data with data-flow analysis at compile time. At run time, MDR leverages an architectural mechanism that replicates read-only shared data across LLC slices when this can be done without pressuring cache capacity. With LAB and MDR, our NUBA GPU improves average performance by 23.1% and 22.2% (and up to 183.9% and 182.4%) compared to iso-resource memory-side and SMside UBA GPUs, respectively. When the NUBA concept is leveraged to reduce overhead while maintaining similar performance, NUBA reduces NoC power consumption by 12.1x and 9.4x, respectively.".
- 01HQ0WT764375CFNS64KFDW3H3 author F57D866C-F0ED-11E1-A9DE-61C894A0A6B4.
- 01HQ0WT764375CFNS64KFDW3H3 author urn:uuid:64c93028-180d-4926-8518-ebebcc4f835a.
- 01HQ0WT764375CFNS64KFDW3H3 author urn:uuid:7528a574-181b-4b47-a70b-1e564f212751.
- 01HQ0WT764375CFNS64KFDW3H3 author urn:uuid:7cb38b82-03f9-460e-83b8-8c0717ac9291.
- 01HQ0WT764375CFNS64KFDW3H3 author urn:uuid:d72fcea2-ad70-4026-998c-b6ad132f0b59.
- 01HQ0WT764375CFNS64KFDW3H3 dateCreated "2024-02-19T14:37:41Z".
- 01HQ0WT764375CFNS64KFDW3H3 dateModified "2024-07-09T15:44:24Z".
- 01HQ0WT764375CFNS64KFDW3H3 editor urn:uuid:57c8a805-9785-43e7-90c9-287107ff5f93.
- 01HQ0WT764375CFNS64KFDW3H3 editor urn:uuid:5f81e99e-a947-47fb-9eb4-cb434cc340aa.
- 01HQ0WT764375CFNS64KFDW3H3 editor urn:uuid:81ff33cd-badf-451d-9a3b-a0493152cdba.
- 01HQ0WT764375CFNS64KFDW3H3 name "NUBA : non-uniform bandwidth GPUs".
- 01HQ0WT764375CFNS64KFDW3H3 pagination urn:uuid:73470388-d5ab-4b00-b65d-5d77220c314f.
- 01HQ0WT764375CFNS64KFDW3H3 publisher urn:uuid:51faf920-d221-4f90-910b-6048454a19a4.
- 01HQ0WT764375CFNS64KFDW3H3 sameAs LU-01HQ0WT764375CFNS64KFDW3H3.
- 01HQ0WT764375CFNS64KFDW3H3 sourceOrganization urn:uuid:8f4920f8-3e27-490c-96fa-7a90b3de756b.
- 01HQ0WT764375CFNS64KFDW3H3 type P1.