TUM.ai Applied Accelerated Computing - Session 2: Deep Dive

Name: TUM.ai Applied Accelerated Computing - Session 2: Deep Dive
Start: 2025-06-07T10:00:00.000+02:00
End: 2025-06-07T13:00:00.000+02:00
Location: TUM.ai

TUM.ai AAC

TUM.ai

München, Bayern

Past Event

Welcome! To join the event, please register below.

You will be asked to verify token ownership with your wallet.

About Event

Hey TUM.ai crew!

Your Applied Accelerated Computing Task Force is back for Round 2! Last time, we got a taste of CPU to GPU speedups. This week, we're diving deep into what truly makes different accelerators tick, revealing the unique architectural superpowers of NVIDIA CUDA, Apple Metal, AMD ROCm, and Intel's Xe/NPU. Discover why there's no "one size fits all" in high-performance computing, like how Intel NPUs can be incredibly power-efficient for tasks such as face recognition, making them ideal for continuous, on-device AI. We'll also kick off our journey into OpenCL, the universal language for heterogeneous computing, showing you how to write code that runs across a huge range of devices.

Workshop Agenda

Beyond CUDA: A deep dive into the unique architectures of NVIDIA (CUDA), Apple (Metal), AMD (ROCm), and Intel (Xe/NPU). Discover what makes each platform uniquely powerful for specific workloads.
OpenCL: The Universal Language: How this standard emerged from OpenGL shaders and why it's your go-to for cross-platform parallel programming.
Hands-on with OpenCL: We'll guide you through writing and running your first OpenCL kernel, getting your multi-vendor hardware to do some serious number-crunching!

Important Details

TLDW Session (Optional): Missed last week? No worries! Join us at 9:30 AM for a quick 30-minute recap of the previous session's core concepts.
Main Workshop: Saturday, June 7 (It's a deep dive, but we keep it engaging!)
Place: TUM.ai Homebase

What to Bring

Your laptop, if you can. We'll be doing some hands-on PyOpenCL coding, and you'll get to see your own machine accelerate!
Curiosity and questions – this is an open discussion, and your insights are always welcome.
Please install pyopencl beforehand. We'll provide basic troubleshooting, but having it ready will give you a significant head start for the hands-on session! Feel free to text me if you have issues at this. You should essentially have a pyopencl setup that detects all the HW your PC has (CPUs, GPUs, NPUs). ChatGPT and Gemini are your best friends at this! :)