{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,5]],"date-time":"2026-03-05T15:47:51Z","timestamp":1772725671070,"version":"3.50.1"},"reference-count":23,"publisher":"IEEE","license":[{"start":{"date-parts":[[2025,7,6]],"date-time":"2025-07-06T00:00:00Z","timestamp":1751760000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2025,7,6]],"date-time":"2025-07-06T00:00:00Z","timestamp":1751760000000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2025,7,6]]},"DOI":"10.1109\/isvlsi65124.2025.11130221","type":"proceedings-article","created":{"date-parts":[[2025,8,27]],"date-time":"2025-08-27T18:20:15Z","timestamp":1756318815000},"page":"1-6","source":"Crossref","is-referenced-by-count":2,"title":["FlatAttention: Dataflow and Fabric Collectives Co-Optimization for Efficient Multi-Head Attention on Tile-Based Many-PE Accelerators"],"prefix":"10.1109","author":[{"given":"Chi","family":"Zhang","sequence":"first","affiliation":[{"name":"Integrated Systems Laboratory (IIS), ETH Zurich,Zurich,Switzerland"}]},{"given":"Luca","family":"Colagrande","sequence":"additional","affiliation":[{"name":"Integrated Systems Laboratory (IIS), ETH Zurich,Zurich,Switzerland"}]},{"given":"Renzo","family":"Andri","sequence":"additional","affiliation":[{"name":"Computing Systems Lab, Huawei Zurich Research Center,Zurich,Switzerland"}]},{"given":"Thomas","family":"Benz","sequence":"additional","affiliation":[{"name":"Integrated Systems Laboratory (IIS), ETH Zurich,Zurich,Switzerland"}]},{"given":"Gamze","family":"Islamoglu","sequence":"additional","affiliation":[{"name":"Integrated Systems Laboratory (IIS), ETH Zurich,Zurich,Switzerland"}]},{"given":"Alessandro","family":"Nadalini","sequence":"additional","affiliation":[{"name":"University of Bologna,Department of Electrical, Electronic, and Information Engineering (DEI),Bologna,Italy"}]},{"given":"Francesco","family":"Conti","sequence":"additional","affiliation":[{"name":"University of Bologna,Department of Electrical, Electronic, and Information Engineering (DEI),Bologna,Italy"}]},{"given":"Yawei","family":"Li","sequence":"additional","affiliation":[{"name":"Integrated Systems Laboratory (IIS), ETH Zurich,Zurich,Switzerland"}]},{"given":"Luca","family":"Benini","sequence":"additional","affiliation":[{"name":"University of Bologna,Department of Electrical, Electronic, and Information Engineering (DEI),Bologna,Italy"}]}],"member":"263","reference":[{"key":"ref1","article-title":"On the computational complexity of self-attention","author":"Keles","year":"2023","journal-title":"ALT"},{"key":"ref2","article-title":"LLM inference unveiled: Survey and roofline model insights","author":"Yuan","year":"2024","journal-title":"xarXiv preprint arXiv:2402.16363"},{"key":"ref3","article-title":"Data movement is all you need: A case study on optimizing transformers","author":"Ivanov","year":"2021","journal-title":"MLSys"},{"key":"ref4","article-title":"FlashAttention: Fast and memory-efficient exact attention with IO-awareness","author":"Dao","year":"2022","journal-title":"NeurIPS"},{"key":"ref5","article-title":"FlashAttention-2: Faster attention with better parallelism and work partitioning","author":"Dao","year":"2024","journal-title":"ICLR"},{"key":"ref6","article-title":"Flashattention-3: Fast and accurate attention with asynchrony and low-precision","author":"Shah","year":"2024","journal-title":"arXiv preprint arXiv:2407.08608"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO61859.2024.00100"},{"key":"ref8","doi-asserted-by":"publisher","DOI":"10.1109\/HCS61935.2024.10664673"},{"key":"ref9","doi-asserted-by":"publisher","DOI":"10.1109\/HCS61935.2024.10664810"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1109\/ISSCC49661.2025.10904727"},{"key":"ref11","article-title":"Attention in SRAM on tenstorrent grayskull","author":"Th\u00fcning","year":"2024","journal-title":"arXiv preprint arXiv:2407.13885"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.1145\/3575693.3575747"},{"key":"ref13","doi-asserted-by":"publisher","DOI":"10.1109\/MICRO61859.2024.00107"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1109\/HPCA57654.2024.00022"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1007\/978-0-387-09766-4_28"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.1145\/2155620.2155630"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/ICCD53106.2021.00071"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.1109\/TCAD.2025.3528349"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.1109\/TC.2023.3329930"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.1016\/j.future.2023.07.002"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.1109\/TVLSI.2025.3527225"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.2197\/ipsjtsldm.8.63"},{"key":"ref23","article-title":"MI300X vs H100 vs H200 Benchmark Part 1: Training","author":"Schor","year":"2024"}],"event":{"name":"2025 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)","location":"Kalamata, Greece","start":{"date-parts":[[2025,7,6]]},"end":{"date-parts":[[2025,7,9]]}},"container-title":["2025 IEEE Computer Society Annual Symposium on VLSI (ISVLSI)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11129697\/11130193\/11130221.pdf?arnumber=11130221","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,8,28]],"date-time":"2025-08-28T04:31:15Z","timestamp":1756355475000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11130221\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,7,6]]},"references-count":23,"URL":"https:\/\/doi.org\/10.1109\/isvlsi65124.2025.11130221","relation":{},"subject":[],"published":{"date-parts":[[2025,7,6]]}}}