Sant’Abbondio Cloister (Como) & CNR Territorial Research Area (Bologna)
We are pleased to announce a three-module course on High-Performance and Parallel Computing for Astrophysics. This course provides comprehensive, hands-on expertise in high-performance computing and parallel programming, ranging from the inner workings of modern computing architectures to the development, optimization, debugging, and profiling of parallel code on multi-core, multi-node, and GPU-accelerated systems.
MODULE A — Modern Architectures, Serial Code, and Shared Memory (40h)
Learning Objectives:
Understand how coding choices translate into performance on microarchitecture; learn how to transition a serial kernel to a vectorized and cache-friendly version; master single-node parallelization with OpenMP while correctly managing shared memory; and learn to identify correctness bugs and serial bottlenecks using standard tools.
Course Syllabus:
Modern Architectures (CPUs)
Serial Code Optimization
Node Architecture and NUMA
Multithreading with OpenMP
Debugging & Profiling I (Serial)
MODULE B — Distributed Memory (MPI) and Computing Cluster Architectures (40h)
Learning Objectives:
Understand the distributed memory model and learn to write correct, efficient intermediate-level MPI applications; master advanced MPI techniques for complex communication patterns and parallel I/O; and perform debugging and profiling on multi-process parallel applications.
Course Syllabus:
Cluster Architectures
MPI — Basics
MPI — Intermediate
MPI — Advanced
Debugging & Profiling II (Parallel)
Containerization for HPC
MODULE C — GPUs, Algorithms, and Deployment (40h)
Learning Objectives:
Understand GPU architecture and learn to write and optimize basic kernels by leveraging the memory model; manage multi-GPU and multi-node scenarios using dedicated communication libraries; and profile production-scale applications and deploy them reproducibly.
Course Syllabus:
GPU I
GPU II (Advanced and Distributed)
Large-Scale Profiling
Registration Fee is €100 per person for each module. The three modules are designed to be taken independently.
The registration fee covers lunch and coffee breaks on all lecture days. Travel and accommodation expenses are not included and remain the sole responsibility of the participants.
There are a total of 40 seats available, and admission is granted on a first-come, first-served basis. If registration exceeds capacity, priority will be given to PhD students and early-career researchers, while still respecting the chronological order of registration.