Abstract:The advent of chiplet technology introduces cutting-edge opportunities for constructing highly heterogeneous platforms with specialized accelerators. However, the HPC community currently lacks expertise in hardware development, a gap that must be bridged to leverage these advancements. Additionally, technologies like chiplet is cutting-edge with limited educational resource available. This paper addresses potential hardware specialization direction in HPC and how to cultivate these skills among students and staff, emphasizing the importance of understanding and developing custom hardware (e.g., rapid prototyping and resource estimation). We have been mentoring graduate-level students and new staff in hardware designs in a hands-on manner, encouraging them to utilize modern open-source hardware tools for their designs, which facilitates the sharing of research ideas. Additionally, we provide a summary of theses tools as part of our approach to prototyping and mentoring.

What problem does this paper attempt to address?

The main problems that this paper attempts to solve are the current lack of expertise in hardware development in the high - performance computing (HPC) community, especially in chiplet technology and dedicated accelerator design. Specifically: 1. **Lack of hardware development skills**: With the rise of chiplet technology, it has become possible to build highly heterogeneous platforms and dedicated accelerators, but the HPC community currently lacks the necessary hardware development skills to fully utilize these technologies. 2. **Scarcity of educational resources**: Frontier technologies such as chiplet have limited resources, and educational materials and training opportunities are scarce, making it difficult to cultivate a new generation of hardware design talents. 3. **Integration challenges of dedicated accelerators**: Although the heterogeneous configuration of CPU and GPU is widely adopted in HPC platforms, dedicated accelerators have not been significantly integrated in HPC yet. This is mainly because adding new functions will increase the complexity of the entire chip development process, especially in the verification stage. 4. **Cultivation of innovative talents**: To meet the above challenges, the paper emphasizes the importance of cultivating the hardware design, verification, and simulation skills of existing employees and students, and in particular, promoting hardware design knowledge among those with a background in computer science or field science. To solve these problems, the paper proposes the following ways to cultivate hardware expertise: - Utilize modern open - source hardware tools for rapid prototyping and resource sharing. - Provide practical guidance and support, especially for postgraduate students and new employees. - Promote the use of open - source hardware tools such as Chisel, Cocotb, and OpenRoad to lower the barriers to learning and use and promote the exchange of research ideas. Through these measures, the paper aims to provide a systematic path for the HPC community to cultivate and develop hardware specialization capabilities, so as to better utilize the opportunities brought by chiplet technology.

Educating for Hardware Specialization in the Chiplet Era: A Path for the HPC Community

Modular High-Performance Computing Using Chiplets

EdUCAS: An In-house CI/CD Platform with Cloud FPGAs for Agilely Conducting Computer Systems Course Projects

A Hardware Co-design Workflow for Scientific Instruments at the Edge

An accelerator-aware microarchitecture simulator for design space exploration

Lessons Learned from a Decade of Providing Interactive, On-Demand High Performance Computing to Scientists and Engineers

AHA: An Agile Approach to the Design of Coarse-Grained Reconfigurable Accelerators and Compilers

The evolution of heterogeneous integration and packaging for the age of chiplets

Programming Bare-Metal Accelerators with Heterogeneous Threading Models: A Case Study of Matrix-3000

Challenges and Opportunities to Enable Large-Scale Computing via Heterogeneous Chiplets

Advanced Packaging: Enabling Moore’s Law’s Next Frontier Through Heterogeneous Integration

Early experiences evaluating the HPE/Cray ecosystem for AMD GPUs

RapidChiplet: A Toolchain for Rapid Design Space Exploration of Chiplet Architectures

Centrifuge: Evaluating full-system HLS-generated heterogenous-accelerator SoCs using FPGA-Acceleration

Design Space Exploration of HW Accelerators and Network Infrastructure for FPGA-Based MPSoC

Reinventing High Performance Computing: Challenges and Opportunities

Monad: Towards Cost-effective Specialization for Chiplet-based Spatial Accelerators

Customizable Computing—From Single Chip to Datacenters

Design Space Exploration for Device and Architectural Heterogeneity in Chip-Multiprocessors

Chiplets and the Codelet Model

Programming Heterogeneous Systems from an Image Processing DSL