Geometry on Steroids: Practical Experience Porting Heavy Computation to GPU
summary
The talk covers using GPUs to accelerate computational geometry algorithms in CAD and engineering workflows, showing practical experience moving heavy geometric computations from CPU to GPU. It discusses technologies around GPU programming (graphics cards rather than just neural-network use) and applies them to real engineering tasks such as designing electrical networks for cities or industrial sites.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 3:24Engineering network design
- 8:43GPU architecture and OpenCL
- 16:36Running OpenCL program
- 16:56Data structures: buckets
- 19:04Point-in-polygon problem
- 21:33CPU vs GPU implementation
- 26:16Benchmarking and first optimizations
- 29:54Parallelism improvements
- 31:42Microarchitectural optimizations
- 34:38Summary and conclusions
- 35:41Q&A
related talks
01h 04mThe challenge of managing parallelism and data-movement for tensor computations on GPUs
45mHow to Become a GPU Engineer in an Hour
49mGraph Based Processing of Big Images
31mPainting With Light
14m3D Rendering, the Grand Illusion Explained
01h 27m“Patterns and Techniques Used in the Houdini 3D Graphics Application ”
01h 21mPanel+Closing
22mDreamSketch: Early Stage 3D Design Explorations with Sketching and Generative Design