High Performance GPU Computing with Ruby
summary
Shows a Ruby wrapper for ArrayFire gem and a lower‑level RbCUDA gem that run BLAS/LAPACK on GPUs; benchmarks report determinant ~10x faster than NMatrix‑LAPACK, elementwise add ~10,000x faster and matmul up to 100000x (RbCUDA slightly faster than ArrayFire), demonstrates host→device memory copies, custom CUDA kernel compile/load workflow, works on CUDA and OpenCL/CPU backends, and exposes image‑processing and ML routines.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 1:27SciRuby and background
- 3:56Arrays and matrices
- 5:53GPU computing challenges
- 7:20ArrayFire gem
- 9:56ArrayFire examples
- 12:51Benchmarking ArrayFire
- 15:29Matrix multiplication benchmarks
- 16:52ArrayFire hardware benefits
- 17:21RbCUDA gem
- 18:27RbCUDA custom kernels
- 20:17RbCUDA benchmarks
- 21:24Future work
- 22:51Acknowledgements
related talks
36mRuby Supercomputing: Using The GPU For Massive Performance Speedup
28mData Science in Ruby? Is it possible? Is it Fast? Should we use it?
23mSpeeding up NMatrix by 100x
39mProgress of Ruby/Numo: Numerical Computing for Ruby
18mPerformance Optimization in Ruby
21mScientific Computing in Ruby
41mHigh Performance Ruby
38m"Parallel and Thread-Safe Ruby at High-Speed with TruffleRuby"