Abstracts:To meet the increasingly demanding compute requirements of modern workloads, systems on chip (SoCs) must provide an accelerator-rich hardware architecture and software programming interface. However, scalability remains a first-order concern, as introducing additional unmanageable complexity to either physical design or software integration may prohibit the deployment of new accelerators. To address these challenges, this work presents AuRORA, an accelerator integration methodology that provides a scalable physical accelerator interface while preserving software semantics with minimal overhead for accelerator access. AuRORA provides a new accelerator integration methodology that preserves the software and hardware interface of a tightly CPU-coupled accelerator while physically disaggregating the accelerators away from a host CPU. To address software scalability, AuRORA also includes a lightweight software runtime for an SoC with heterogeneous accelerators, providing low-overhead access to these accelerators for multitenant applications.

Abstracts:Although online reinforcement learning (RL) has shown promise for microarchitecture decision making, processor vendors are still reluctant to adopt it. There are two main reasons that make RL-based solutions unattractive. First, they have high complexity and storage overhead. Second, many RL agents are engineered for a specific problem and are not reusable. In this work, we propose a way to tackle these shortcomings. We find that, in diverse microarchitecture problems, only a few actions are useful in a given time window. Motivated by this property, we design Micro-Armed Bandit (or Bandit for short), an RL agent that is based on the low-complexity Multi-Armed Bandit algorithms. We show that Bandit can match or exceed the performance of more complex RL and non-RL alternatives in two different problems: data prefetching and instruction fetch thread selection in simultaneous multithreaded processors. We believe that Bandit’s simplicity, reusability, and small storage overhead make online RL more practical for microarchitecture.

Abstracts:The lack of publicly available cloud services has been a recurring problem in architecture and systems. Although open source benchmarks exist, they do not capture the complexity of cloud services. Application cloning is a promising approach, however, prior work is limited to CPU-/cache-centric, single-node services. We present Ditto, a framework for cloning end-to-end cloud applications and monolithic and microservices that captures input–output and network activity as well as kernel operations, in addition to application logic. Ditto takes a hierarchical approach to application cloning, capturing the dependency graph across services, recreating each tier’s control/dataflow, and generating system calls and assembly that mimics individual applications. Ditto does not reveal the logic of the original application, facilitating publicly sharing clones of production services. We show that across a diverse set of applications, Ditto accurately captures their resource characteristics as well as their performance metrics, is portable across platforms, and facilitates a wide range of studies.

Abstracts:It is our pleasure to introduce the IEEE Micro Special Issue on Top Picks from the 2023 Computer Architecture Conferences. This special issue includes 12 articles chosen by a selection committee (SC) as being the most significant research articles in computer architecture in 2023 in terms of their potential for long-term impact. The committee also selected an additional 12 articles to be recognized with an honorable mention (see “Honorable Mentions”).

Abstracts:This IEEE Micro Special Issue on Top Picks includes 12 outstanding papers selected from those published in 2023 computer architecture conferences. When we were prepping and putting together all the articles in this special issue, another significant event in the computer industry was unfolding in Taipei, Taiwan, in June: the 2024 COMPUTEX Taipei. The event hosted more than 1500 exhibitors from 136 countries and regions, showcasing the latest innovations and sleek products in information and computing technologies. Unparalleled to previous equivalent events, the 2024 COMPUTEX organizers invited an elite group of influential leaders from today’s technology juggernauts to deliver keynotes. The program features AMD CEO Lisa Su, Delta Electronics General Director Tzi-cker Chiueh, Intel CEO Pat Gelsinger, MediaTek CEO Rick Tsai, NXP CTO Lars Reger, Qualcomm CEO Cristiano Amon, and Supermicro CEO Charles Liang. They were joined by a couple of additional speakers, ARM CEO Rene Haas and Nvidia CEO Jensen Huang, who delivered their keynotes outside the COMPUTEX exhibition hall. These Top Picks speakers represent the ecosystem companies of the computing infrastructure supply chain that help shape, bolster, and accelerate the Fourth Industrial Revolution, driven by the prevailing and unstoppable artificial intelligence (AI) tsunami. The unveiled systems push the frontiers of technology, reaching the pinnacle of computer architecture design, system integration, advanced packaging techniques, and leading-edge process technologies.