SIMD in .NET: Image Processing with Intrinsics
summary
Covers using .NET SIMD intrinsics (hardware-specific CPU instructions exposed as intrinsics since .NET Core 3) to accelerate image-processing algorithms. Describes several real image-processing algorithms, the performance bottlenecks encountered, and the optimization stages and techniques applied to achieve best throughput.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 1:32Why SIMD in project
- 4:23What is SIMD and implementation
- 7:17SIMD intrinsics in .NET
- 11:20Hello World with SIMD
- 16:21Image processing and pixel format
- 17:06Erosion algorithm optimization
- 23:28Erosion horizontal pass optimization
- 27:23Loop unrolling for performance
- 28:31Blur algorithm optimization
- 32:46Box Blur and ring buffers
- 36:02Division via multiplication optimization
- 41:06Comparison with compiler optimizations
- 42:52AVX2 transition and frequency limits
- 43:37Memory alignment for SIMD
- 47:15Combining filters and results
- 48:20Conclusion and resources
- 49:52Q&A
related talks
58m.NET Core Internals and Optimizations
56mSpeedrun Into Massive Data: Using SIMD in C#
01h 01mBeating CoreCLR's own C++ code with CoreCLR 3.0 intrinsics
47mUnlocking Performance Improvements in .NET
01h 00mWindows 10 internals for .NET developers
56mState of the .NET Performance
44m.NET 11 in depth: Runtime, libraries, and SDK for the AI era
01h 03mF# for Performance-Critical Code