Gauss Centre for Supercomputing’s Post

Turing defined a mathematical model of a sequential machine that can compute any function that a real computer can compute. For a concurrent machine there is no single universal model of computation, because the computational power of the distributed system depends on various parameters, such as the reliability of its components and the available means of communication. But even for the most basic situation where asynchronous processes that may crash communicate by read/write shared registers, it was not known if a universal model requires a shared memory of unbounded size. In the paper presented this week at ACM PODC (Principles of Distributed Computing) this question has finally been resolved (almost) https://lnkd.in/g7i6c7kf

Are you a researcher, data scientist or computer scientist working in industry or a tech start-up in Ontario? Does your company use significant compute power, do data-intensive computing or work with AI? We want to hear from you. CO is conducting a survey to understand the current use of computing technologies and resources (such as GPUs for AI) in Ontario businesses. We want to identify challenges, limitations, and future needs. Your responses will help shape strategies to support businesses like yours in accessing and utilizing DRI effectively. The survey is open until December 6. Participate here: https://lnkd.in/gRsfe-Am

We value your input if you're an #Ontario business that would benefit from access to digital research infrastructure to up your game in AI or other compute intensive workloads. "To out-compete, you must out-compute" the US council on competitiveness. #AI #DRI #HPC #competitiveness

Heading to #ISCHPC2024 in Hamburg this weekend. There is a fantastic line-up of tutorials and talks. Some that may be of particular interest: --Tutorial on Performance Engineering for Linear Solvers. Hartwig Anzt, Christie Alappat, Jonas Thies, and Georg Hager are presenting on code analysis, performance modeling, and optimization for linear solvers on CPU and GPU nodes. Sunday May 2, 2-6pm in Hall Y6. --Kathy Yelick is giving a keynote on Monday entitled 'Beyond Exascale Computing'. Interested in what the challenges and opportunities the research community is facing? Her talk is a must attend. -- Jeff Hammond will be presenting on the MPI ABI at the High Performance Container workshop. Jeff is always a knowledgable and entertaining presenter, be sure to stop by to learn more about the need for a standard MPI ABI. --Andrew Richards is giving a talk titled "Unlocking the Next 35 Years of Software for HPC and AI". I love that he's giving the Randles Lab work with HARVEY a shout out, but more than that it should be an insightful look at next steps for HPC and AI. --And of course, I'm excited to be presenting in the Jack Dongarra Early Career Award Series. I'll be talking about our latest efforts to build digital twins for healthcare and fellow winner Edgar Solomonik will be talking about high performance tensor network contraction and decomposition algorithms. Join us at 10:45am on Monday the 13th in Hall Z. Looking forward to catching up with friends in Germany! If you're attending, let me know! I'm only there briefly but would love to catch up. I'm also looking to find better storage solutions so looking forward to meeting with folks from VAST Data. Storage is one of our big headaches now, so even if you're not attending ISC and you just have storage solution suggestions, advice is welcome as well. #hpc #computationalmodeling

Congratulations to University of Washington, Department of Electrical & Computer Engineering professor Sajjad Moazeni for receiving a three-year grant from the National Science Foundation (NSF) to develop the next generation of AI and machine learning! 🎉 As computer power, capacity and speed increase, so does the need for new data center infrastructure. Data centers (also known as “the cloud”) are only as fast and powerful as their optical interconnects that transfer data between devices and servers. The grant will fund Moazeni’s team as they develop a new kind of optical interconnect – one that’s more compact, energy-efficient and faster. “[The new optical interconnect] is not just computer simulations or abstract papers. This research is going to result in actual chips in the lab being tested, verified and characterized,” says Moazeni. Learn more: https://bit.ly/4f7udGR

In a paper published in Nature, Google introduces AlphaQubit, an AI-based decoder that identifies quantum computing errors with state-of-the-art accuracy. This collaborative work brought together Google DeepMind’s machine learning knowledge and Google Quantum AI’s error correction expertise to accelerate progress on building a reliable quantum computer. Accurately identifying errors is a critical step towards making quantum computers capable of performing long computations at scale, opening the doors to scientific breakthroughs and many new areas of discovery. #GoogleCloud, #Google, #Alphabet, #Quantum, #QuantumComputing https://google.smh.re/4Zae

#LearnaboutaerOS: The aerOS journal paper entitled "Federated Deep Reinforcement Learning for Prediction-Based Network Slice Mobility in 6G Mobile Networks" has been accepted at IEEE Transactions on Mobile Computing. Stay tuned for more information. #metaos #journalpaper #research #6G #HorizonEurope #IEEE EUCloudEdgeIoT HiPEAC

On the 7th Day of HPC, Red Oak showed to me... Seven Systems Soaring 🚀 Like Santa's reindeer hitting their stride on Christmas Eve, these seven performance peaks show what happens when HPC systems really take flight! 𝟏. 𝐁𝐫𝐞𝐚𝐤𝐢𝐧𝐠 𝐭𝐡𝐞 𝐄𝐱𝐚𝐅𝐋𝐎𝐏 𝐁𝐚𝐫𝐫𝐢𝐞𝐫 • Faster than Santa's round-the-world trip • A quadrillion calculations per second • Making the impossible possible 𝟐. 𝐐𝐮𝐚𝐧𝐭𝐮𝐦 𝐋𝐞𝐚𝐩𝐬 • Next-gen computing taking off • Solving previously impossible problems • Future-ready architecture 𝟑. 𝐒𝐮𝐬𝐭𝐚𝐢𝐧𝐚𝐛𝐥𝐞 𝐒𝐩𝐞𝐞𝐝 • Green computing at full power • Energy efficiency that sparkles • Performance per watt that glows 𝟒. 𝐀𝐈 𝐀𝐜𝐜𝐞𝐥𝐞𝐫𝐚𝐭𝐢𝐨𝐧 • Machine learning at light speed • Neural networks that know the way • Training times tumbling down 𝟓. 𝐒𝐭𝐨𝐫𝐚𝐠𝐞 𝐒𝐮𝐩𝐫𝐞𝐦𝐚𝐜𝐲 • I/O faster than elves wrapping presents • Data access in microseconds • Cache hits sweeter than candy canes 𝟔. 𝐌𝐞𝐦𝐨𝐫𝐲 𝐌𝐚𝐠𝐢𝐜 • RAM faster than reindeer • Cache coherency cleaner than snow • Bandwidth broader than the North Pole 𝟕. 𝐂𝐥𝐨𝐮𝐝 𝐂𝐚𝐩𝐚𝐛𝐢𝐥𝐢𝐭𝐢𝐞𝐬 • Scaling higher than Santa's sleigh • Elasticity that stretches like stockings • Resources as abundant as Christmas wishes Need help making your systems soar? 𝐑𝐞𝐝 𝐎𝐚𝐤 𝐂𝐨𝐧𝐬𝐮𝐥𝐭𝐢𝐧𝐠'𝐬 𝐩𝐞𝐫𝐟𝐨𝐫𝐦𝐚𝐧𝐜𝐞 𝐞𝐱𝐩𝐞𝐫𝐭𝐬 can help your computing take flight! #HPC #12DaysOfHPC #TechChristmas #HPCPerformance #TechExcellence

For this piece, I really tried dissecting what computing means. What a mind-boggling journey. Give my blog a read. And wonder at the wonder known as computing. https://lnkd.in/dHCR587U