Ruby Supercomputing: Using The GPU For Massive Performance Speedup
summary
Uses the laptop GPU via OpenCL/CUDA and an NVIDIA toolkit-backed Ruby binding to run thousands of threads and convert a 16 million–ring synthetic workload into massively parallel kernel launches. Pure Ruby 1.9 single-thread takes ~7.6s, Ruby 1.9 threads ~7.7s, JRuby parallel drops to roughly ~0.5s, and native C + GPU kernels are orders of magnitude faster; memory transfers (about 60 MB copied back) and PCIe latency limit gains. Implementation uses the Barracuda binding to compile kernels and automatically copy results into Ruby arrays.
AI-written summary · about
chapters (6)
- 0:00Start
- 3:062nd working attempt.
- 10:53Common CPU Issues
- 13:23GPU Brief History
- 14:15Hardware Examples
- 18:22NVidia Tesla C1060
related talks
23mHigh Performance GPU Computing with Ruby
38m"Parallel and Thread-Safe Ruby at High-Speed with TruffleRuby"
41mHigh Performance Ruby
43mParallel and Thread-Safe Ruby at High-Speed with TruffleRuby
23mParallel Processing With Ruby
47mHigh Performance Ruby
18mPerformance Optimization in Ruby
28mConcurrency for Rubyists