GPU Acceleration of Sparse Fully Homomorphic Encrypted DNNs Fully homomorphic encryption (FHE) has recently attracted significant attention as both a cryptographic primitive and a systems challenge. Given the latest advances in accelerated computing, FHE prese...

🚀 The preprint of our paper "𝑮𝑷𝑼 𝑨𝒄𝒄𝒆𝒍𝒆𝒓𝒂𝒕𝒊𝒐𝒏 𝒐𝒇 𝑺𝒑𝒂𝒓𝒔𝒆 𝑭𝒖𝒍𝒍𝒚 𝑯𝒐𝒎𝒐𝒎𝒐𝒓𝒑𝒉𝒊𝒄 𝑬𝒏𝒄𝒓𝒚𝒑𝒕𝒆𝒅 𝑫𝑵𝑵𝒔" is now available on arXiv, check it out! arxiv.org/abs/2604.11659

📅 We’ll be presenting this work on April 27 at the EuroMLSys workshop (part of EuroSys ’26) in Edinburgh.

Recurring theme of @hipeac.bsky.social this week has been the intersection of compilers and architectures to address challenges in #ai, especially llm. Really exciting when @jocare.bsky.social announced #HiPEAC 2027 will be held in Glasgow!

Thanks @nickbrown-hpc.bsky.social, see you in Glasgow next week!

I’m delighted to announce that #HiPEAC2027 will take place in Glasgow in January 2027, and I’m honored to serve as the General Chair! 🙂

We’re also excited to welcome Reetuparna Das as the Program Chair.

More details will follow — stay tuned!

www.hipeac.net

8th AccML Workshop at HiPEAC 2026 AccML workshop homepage

📢 The program of our 8th AccML Workshop at #HiPEAC26 is available!
We have 4 Invited talks (Cristina Silvano, Heiko Joerg Schick, Nicholas Fraser and Jacques Pienaar) and 9 paper presentations!
🗓️ 27 January | 📌 Kraków, Poland
👉 accml.dcs.gla.ac.uk

C4ML 7th C4ML workshop, at CGO 2026 Sunday, February 1, 2026 In person: Sydney, Australia

Happy to share that our paper "𝘿𝙚𝙛𝙚𝙖𝙩 𝙩𝙝𝙚 𝙃𝙚𝙖𝙥: 𝙕𝙚𝙧𝙤-𝘾𝙤𝙥𝙮 𝘿𝙖𝙩𝙖 𝙈𝙤𝙫𝙚𝙢𝙚𝙣𝙩 𝙞𝙣 𝘼𝙓𝙄4𝙈𝙇𝙄𝙍" has been accepted to the C4ML workshop at #CGO2026! 🙂 www.c4ml.org

Congrats to the main authors, Elam Cohavi and Nícolas Agostini, and to the co-authors Jude Haris, Antonino Tumeo, and David Kaeli!

⏳ Last 7 days to submit your work (𝗗𝗲𝗮𝗱𝗹𝗶𝗻𝗲 𝟱 𝗗𝗲𝗰 𝟮𝟬𝟮𝟱) to the 8th AccML Workshop at #HiPEAC26!

📢 Invited speakers: Cristina Silvano, Heiko Joerg Schick, Nicholas Fraser and Jacques Pienaar

🗓️ 27 January | 📌 Kraków, Poland
👉 accml.dcs.gla.ac.uk

⏳ DEADLINE EXTENDED (5 Dec 2025): 8th AccML Workshop at #HiPEAC26
📢 Confirmed invited speakers: Cristina Silvano, Heiko Joerg Schick, Nicholas Fraser and Jacques Pienaar
🗓️ 27 January | 📌 Kraków, Poland
👉 accml.dcs.gla.ac.uk

F-BFQ: Flexible Block Floating-Point Quantization Accelerator for LLMs Large Language Models (LLMs) have become increasingly prominent for daily tasks, from improving sound-totext translation to generating additional frames for the latest video games. With the help of LL...

Check out our paper "F-BFQ: Flexible Block Floating-Point Quantization Accelerator for LLMs", presented at the LG-ARC Workshop at #ISCA2025! Kudos to Jude Haris for the great work! 🙂 arxiv.org/abs/2510.13401

📢 Call for papers
8th Workshop on Accelerated Machine Learning (AccML) at #HiPEAC26
⏳Deadline: 21 November, 2025
👉 accml.dcs.gla.ac.uk

On my way to the University of Essex for the IDEAL project meeting! 🙂

ideal-epsrc.github.io

HeteroPar'2025 23rd International Workshop on Algorithms, Models and Tools for Parallel Computing on Heterogeneous Platforms

On my way to Euro-Par 2025 in Dresden, Germany!

Tomorrow I'm organizing the HeteroPar workshop, and we have a great program! 🙂

heteropar2025.dcs.gla.ac.uk

Wow, that's cool!

Ready to speak at the LG-ARC workshop at #ISCA2025! 🙂

llm-gnn.org

I'm delighted to share that our paper '𝐀𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐢𝐧𝐠 𝐓𝐫𝐚𝐧𝐬𝐩𝐨𝐬𝐞𝐝 𝐂𝐨𝐧𝐯𝐨𝐥𝐮𝐭𝐢𝐨𝐧𝐬 𝐨𝐧 𝐅𝐏𝐆𝐀-𝐛𝐚𝐬𝐞𝐝 𝐄𝐝𝐠𝐞 𝐃𝐞𝐯𝐢𝐜𝐞𝐬' has been accepted at FPL 2025! It has been a long journey for this work. Kudos to Jude Haris! 🙂

2025.fpl.org

@giclab.bsky.social‬, @uofgcompsci.bsky.social, daiedge.eu

Happy to share that our Abstract '𝐒𝐄𝐂𝐃𝐀 𝐃𝐞𝐬𝐢𝐠𝐧 𝐒𝐮𝐢𝐭𝐞: 𝐄𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐭 𝐇𝐚𝐫𝐝𝐰𝐚𝐫𝐞-𝐒𝐨𝐟𝐭𝐰𝐚𝐫𝐞 𝐂𝐨-𝐃𝐞𝐬𝐢𝐠𝐧 𝐨𝐟 𝐃𝐍𝐍 𝐀𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐨𝐫𝐬' has been accepted at the OSCAR workshop @ #ISCA2025! Kudos to Jude Haris 🙂

@giclab.bsky.social‬, @uofgcompsci.bsky.social

HeteroPar'2025 23rd International Workshop on Algorithms, Models and Tools for Parallel Computing on Heterogeneous Platforms

📢Final extension: 23rd International Workshop on Algorithms, Models and Tools for Parallel Computing on Heterogeneous Platforms (𝐇𝐞𝐭𝐞𝐫𝐨𝐏𝐚𝐫) at #EuroPar2025
⏳Deadline: 25 May, 2025 (hard deadline)
👉 heteropar2025.dcs.gla.ac.uk

Thanks Jan, we'll share the preprint soon!

Happy to share that our paper '𝐅-𝐁𝐅𝐐: 𝐅𝐥𝐞𝐱𝐢𝐛𝐥𝐞 𝐁𝐥𝐨𝐜𝐤-𝐅𝐥𝐨𝐚𝐭𝐢𝐧𝐠 𝐏𝐨𝐢𝐧𝐭 𝐐𝐮𝐚𝐧𝐭𝐢𝐳𝐚𝐭𝐢𝐨𝐧 𝐀𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐨𝐫 𝐟𝐨𝐫 𝐋𝐋𝐌𝐬' has been accepted at the ARC-LG workshop @ #ISCA2025! Kudos to Jude Haris 🙂

ICE-Pruning: An Iterative Cost-Efficient Pruning Pipeline for Deep Neural Networks Pruning is a widely used method for compressing Deep Neural Networks (DNNs), where less relevant parameters are removed from a DNN model to reduce its size. However, removing parameters reduces model ...

The preprint of our IJCNN 2025 paper '𝐈𝐂𝐄-𝐏𝐫𝐮𝐧𝐢𝐧𝐠: 𝐀𝐧 𝐈𝐭𝐞𝐫𝐚𝐭𝐢𝐯𝐞 𝐂𝐨𝐬𝐭-𝐄𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐭 𝐏𝐫𝐮𝐧𝐢𝐧𝐠 𝐏𝐢𝐩𝐞𝐥𝐢𝐧𝐞 𝐟𝐨𝐫 𝐃𝐞𝐞𝐩 𝐍𝐞𝐮𝐫𝐚𝐥 𝐍𝐞𝐭𝐰𝐨𝐫𝐤𝐬'' is available, take a look! arxiv.org/abs/2505.07411

I just registered for #ISCA2025 in Tokyo, so really looking forward to it! 🙂

iscaconf.org/isca2025/

HeteroPar'2025 23rd International Workshop on Algorithms, Models and Tools for Parallel Computing on Heterogeneous Platforms

📢 Call for papers: 23rd HeteroPar Workshop at #EuroPar2025
⏳Deadline extended: 19 May, 2025 (AoE)
👉 heteropar2025.dcs.gla.ac.uk

@rosambadia.bsky.social, Anne Benoit, Alexey Lastovetsky, Leonel Sousa, Denis Trystram, Roman Wyrzykowski

Our 2 posters at the EuroMLSys workshop at EuroSys 2025! 🙂

1) "𝐄𝐱𝐩𝐥𝐨𝐢𝐭𝐢𝐧𝐠 𝐔𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝 𝐒𝐩𝐚𝐫𝐬𝐢𝐭𝐲 𝐢𝐧 𝐅𝐮𝐥𝐥𝐲 𝐇𝐨𝐦𝐨𝐦𝐨𝐫𝐩𝐡𝐢𝐜 𝐄𝐧𝐜𝐫𝐲𝐩𝐭𝐞𝐝 𝐃𝐍𝐍𝐬"
2) "𝐃𝐞𝐜𝐨𝐮𝐩𝐥𝐢𝐧𝐠 𝐒𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐚𝐥 𝐚𝐧𝐝 𝐐𝐮𝐚𝐧𝐭𝐢𝐭𝐚𝐭𝐢𝐯𝐞 𝐊𝐧𝐨𝐰𝐥𝐞𝐝𝐠𝐞 𝐢𝐧 𝐑𝐞𝐋𝐔-𝐛𝐚𝐬𝐞𝐝 𝐃𝐞𝐞𝐩

Kick-off of MLBench workshop at ASPLOS 2025! 🙂

I gave an invited talk on "Accelerating AI at the Edge" at Durham University today, nice campus! 🙂

Exploiting Unstructured Sparsity in Fully Homomorphic Encrypted DNNs The deployment of deep neural networks (DNNs) in privacy-sensitive environments is constrained by computational overheads in fully homomorphic encryption (FHE). This paper explores unstructured sparsi...

The preprint of our EuroMLSys 2025 paper '𝐄𝐱𝐩𝐥𝐨𝐢𝐭𝐢𝐧𝐠 𝐔𝐧𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝 𝐒𝐩𝐚𝐫𝐬𝐢𝐭𝐲 𝐢𝐧 𝐅𝐮𝐥𝐥𝐲 𝐇𝐨𝐦𝐨𝐦𝐨𝐫𝐩𝐡𝐢𝐜 𝐄𝐧𝐜𝐫𝐲𝐩𝐭𝐞𝐝 𝐃𝐍𝐍𝐬'' is available, take a look! arxiv.org/abs/2503.09184

I was there this morning too...

The 5th Workshop on Machine Learning and Systems (EuroMLSys) The 5th Workshop on Machine Learning and Systems (EuroMLSys)

We have 2 papers accepted at EuroMLSys 2025! 😃

1) "Exploiting Unstructured Sparsity in Fully Homomorphic Encrypted DNNs".

2) "Decoupling Structural and Quantitative Knowledge in ReLU-based Deep Neural Networks".

euromlsys.eu

PhD examination at my home university @upv.es earlier today! 🙂

Thanks Nathalie!