Had a really successful #NVIDIAGTC2025 , but soooo happy to be heading home. So many people, so many DLIs, so many meetings, so many sessions!
My voice is tired, my body is tired, but I'm happy to work for a company that is changing the world and exciting people.
To everyone who flagged me down by the DGX systems on the show floor, who came to one of my talks, or who participated in a DLI - THANK YOU!
Watch #NVIDIA #GPU Technology Conference (GTC) Keynote #NvidiaGTC2025 #NvidiaGTC #AI
https://www.elevenforum.com/t/watch-nvidia-gtc-keynote.34419/
Die Nvidia GTC Konferenz 2025 verspricht, richtungsweisende Entwicklungen zu präsentieren. Experten erwarten, dass diese Veranstaltung die Gewinner und Verlierer im sich rasant entwickelnden KI-Netzwerk-Markt offenbaren wird.
Attending NVIDIA GTC? Stop by Engin’s poster (Monday, 5-7 PM) to see how Chapel’s GPU support can be used for radiological imaging research to deliver more than 160x speedup over C with much higher-level code.
![Screenshot of the linked page. Poster title: GPU-Based Monte Carlo Simulation of Light Transport in Tissue: A Chapel Implementation [P73583]. Authors: Hui Wang and Engin Kayraklioglu.
This poster demonstrates how NVIDIA GPUs can be used to simulate photon propagation in tissue using the Chapel programming language. The Chapel implementation is concise and intuitive, while CUDA's device-side random number generators provide the backbone of the implementation, delivering more than 160X improvement over the sequential baseline.
Key Takeaways:
* Photon propagation in tissue is an embarrasingly parallel application, whose performance relies on efficient random number generation
* CUDA's curand_kernel device-side interface is up to the task, and delivers 160X improvement on GH200 over sequential version
* The Chapel parallel programming language's parallel-first design enables an intuitive interface for parallel architectures, like GPUs](https://files.mastodon.social/media_attachments/files/114/162/262/021/445/698/original/d3dea219b01e359a.png)
