Scheduling Multi-Model AI Workloads On Heterogeneous MCM Accelerators (UC Irvine)


A technical paper titled “SCAR: Scheduling Multi-Model AI Workloads on Heterogeneous Multi-Chiplet Module Accelerators” was published by researchers at University of California Irvine. Abstract: "Emerging multi-model workloads with heavy models like recent large language models significantly increased the compute and memory demands on hardware. To address such increasing demands, designin... » read more

Heterogeneous Multi-Core HW Architectures With Fine-Grained Scheduling of Layer-Fused DNNs


A technical paper titled "Towards Heterogeneous Multi-core Accelerators Exploiting Fine-grained Scheduling of Layer-Fused Deep Neural Networks" was published by researchers at KU Leuven and TU Munich. Abstract "To keep up with the ever-growing performance demand of neural networks, specialized hardware (HW) accelerators are shifting towards multi-core and chiplet architectures. So far, thes... » read more

MIT & UC Berkeley: “Exo” Programming Language Writes High Performance Code For HW Accelerators


New research paper titled "Exocompilation for productive programming of hardware accelerators," from researchers at MIT and UC Berkeley. From their abstract: "To better support development of high-performance libraries for specialized hardware, we propose a new programming language, Exo, based on the principle of exocompilation: externalizing target-specific code generation support and op... » read more

New Approaches For Processor Architectures


Processor vendors are starting to emphasize microarchitectural improvements and data movement over process node scaling, setting the stage for much bigger performance gains in devices that narrowly target what end users are trying to accomplish. The changes are a recognition that domain specificity, and the ability to adjust or adapt designs to unique workloads, are now the best way to impro... » read more

NN-Baton: DNN Workload Orchestration & Chiplet Granularity Exploration for Multichip Accelerators


"Abstract—The revolution of machine learning poses an unprecedented demand for computation resources, urging more transistors on a single monolithic chip, which is not sustainable in the Post-Moore era. The multichip integration with small functional dies, called chiplets, can reduce the manufacturing cost, improve the fabrication yield, and achieve die-level reuse for different system scales... » read more

The Unexpected Impact Of Lots On Hold


One of the biggest bottlenecks in any Subcon is Lots on Hold. The problem occurs many times a week on most factory floors. It’s something you’ve grown to loathe or endure. But, is there something you can do to reduce the amount of time lots spend on hold? In this article, we will explain what Lots on Hold are and how you can make the process less painful for your team and help improve on-ti... » read more

Spreadsheets In Virtuoso


The looming tape-out deadline is the nightmare that keeps most design managers up at night. Managing schedules and tracking progress is always a black art that few, if any, can master. Various project management tools and methodologies have been developed that can help, if followed diligently. However, the learning curve of the tools, or the training and overhead of the process, often result in... » read more

What “Hamilton – An American Musical” Tickets And Emulation Have In Common


During a recent trip to New York, I managed to see “Hamilton, An American Musical”—despite the running joke about how hard it is to get tickets. The sale of “Hamilton” tickets teaches an interesting lesson about what I would call an “automatic feedback loop of value adjustment”. And believe it or not, it bears some resemblance to how verification users actually choose what engine ... » read more

Tear Down The Wall Between Front-End And Back-End Teams


As complexity of system-on-chip devices increases, it's becoming imperative for design teams and organizations to re-examine how they work with one another in order to improve productivity. One giant step in this direction is to bridge the divide between the front-end design process and the physical back-end design process. We often refer to this as a figurative “wall,” but there is real... » read more