1 Fusionopolis Way,
Connexis South, #17-01,
Singapore 138632
Site Map
© National Supercomputing Centre (NSCC) Singapore. Updated as of 14 May 2024.
NSCC Singapore strives for products and services with the highest levels of security and reliability.
In November, NSCC Singapore continued to support and grow the high performance computing (HPC) community by conducting two hands-on training workshops for users of the ASPIRE 2A and ASPIRE 2A+ systems. These national supercomputers are widely used by researchers for complex simulations, data analytics, and AI workloads. The workshops were designed to help users better understand how to run their work more efficiently and effectively on these powerful systems.
ASPIRE 2A Advanced Workshop on Optimisation Techniques
This workshop provided an introduction to key concepts that can help speed up compute-intensive applications. Such workload optimisation matters as improving how code runs can save time and cost, and make better use of the national HPC resources. The key concepts covered include:
- Parallelism — breaking work into smaller parts so they can run at the same time
- CPU and memory hierarchy — understanding the inner workings of the hardware so users can write optimised code
- General-Purpose GPU Computing (GPGPU) — using GPUs to speed up tasks that can run in parallel
- Compilers and performance libraries — compilers translate code into machine instructions, while the choice of performance libraries can significantly affect performance
- MPI tuning with Cray MPICH — adjusting settings to improve communication between nodes
- Roofline Model — a visual way to understand what limits the performance of an application (e.g., memory bandwidth or compute power)
Missed this run of the workshop? Check out the upcoming workshop schedule here.
ASPIRE 2A+ Advanced Workshop: Parallel Programming Models
This session focused on the ASPIRE 2A+ system, which is designed for large-scale AI and machine learning workloads. Participants were introduced to the ASPIRE 2A+ architecture before gaining hands-on experience with deploying Enroot container-based workloads and using PyTorch’s parallelisation frameworks. The workshop also provided a short introduction to large language models (LLMs), showcasing examples of running inference and benchmarking. The participants learned how to apply a combination of OpenMP, MPI, and CUDA—three common parallel programming models—to accelerate performance, along with strategies to improve the efficiency of their submitted jobs on the system.
Couldn’t join us this time? Explore our upcoming workshop dates and secure your spot here.
Through these workshops, which provide hands-on learning and direct technical engagement, NSCC Singapore aims to lower barriers to adopting HPC in research, improve user experience, and enable more efficient and impactful use of national HPC resources. As research needs continue to evolve, we look forward to delivering more training opportunities that help users fully harness the capabilities of ASPIRE 2A and ASPIRE 2A+.
