Speeding Up with Accelerate: Experience Optimizing C++ Code
summary
Rewriting cross-platform C++ audio-processing code for the Marusya voice assistant into Swift using Apple’s Accelerate framework, explaining why the port was done, the encountered challenges, and the results. The talk compares optimization techniques and benchmarked performance of pure Swift versus Accelerate, and concludes with pros and cons of Swift+Accelerate compared to the original C++ implementation for compute-heavy tasks like DSP and matrix math.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 11:49Performance problems
- 22:02Keyword code analysis
- 34:20Feature allocation optimization
- 49:41Optimization results
- 1:02:28Summary
related talks
19mDigital Signal Processing with Swift
25mAI-Assisted Swift in Gamifying an iOS App
18mScientific Swift
30mFrom Slow to Flow: Tips for Accelerating Your App Performance
41mSwift + Metal: The future of scientific programming?
01h 00mThe Mysterious Swift Performance
47mTime Traveling With Swift: Where We Started, Where We Are, and Where We’re Going
26mSwift beyond Apple: Making Software for the Physical World