ARCHIDE: 3rd Workshop on Architecture Design Methodologies and Ecosystems for HPC and Scientific Edge Computing

The demand for specialized architectures and innovative approaches is growing to sustain performance and enhance energy efficiency in the post-Moore era of high-performance and scientific edge computing. In the realm of scientific instruments and HPC systems, the development of custom and specialized accelerators -- such as those based on ASICs, FPGAs, and other custom or reconfigurable architectures -- is crucial not only for advancing scientific discoveries but also for meeting efficiency goals. This workshop focuses on architectural designs for HPC and scientific instruments, along with related technologies, methodologies, and tools. We aim to explore topics and opportunities in innovative architectural approaches, hardware design languages, simulation, emulation, verification methodologies and tools, chip prototyping, emerging accelerators and architectures, compiler stacks, programming paradigms, and relevant application areas -- with particular emphasis on energy-efficient computing solutions. Participants will present their recent work in these fields. The workshop also offers opportunities to share knowledge, techniques, tools, and source code related to hardware specialization, encouraging an open and practical dialogue. Additionally, we aim to foster collaborative discussions aligned with DOE/MEXT collaborations.

The topics of interest include but are not limited to:

• Open-source hardware tool ecosystems

  • Chisel, Firesim, Verilator, cocotb, OpenRoad, Chipyard, MoSAIC, PDKs, etc.

• Custom hardware designs (ASIC, FPGA)

  • HPC/AI accelerator designs

  • Near-detector data processing hardware designs for X-ray science, high-energy physics, quantum computing, etc

    • Compression, deep neural network, spiking neural network

    • Surface codes for quantum computing error correction

  • Tight integration with AI accelerators such as Groq 

  • 5G acceleration designs

• AI-assisted chip development

  • Benefits, limitations, potential risks, and demonstrated advantages over conventional scripting

• Common research testbeds/resources and source-code repositories

  • Chameleon cloud, an experimental testbed for computer science, for example

• Discussions about collaborations

  • Future events/workshop

  • Potential roadblocks such as NDA, licenses, resources

Organizers: Kazutomo Yoshii (ANL), Kentaro Sano (RIKEN), John Shalf (LBNL), Jeff Vetter (ORNL), Rajesh Sankaran (ANL)