Add SIMD acceleration for Matrix4x4.Invert #34394 by eanova · Pull Request #36323 · dotnet/runtime (original) (raw)
BenchmarkDotNet=v0.12.1, OS=Windows 10.0.18363.836 (1909/November2018Update/19H2) Intel Core i9-9900K CPU 3.60GHz (Coffee Lake), 1 CPU, 16 logical and 8 physical cores .NET Core SDK=5.0.100-preview.3.20216.6 [Host] : .NET Core 5.0.0 (CoreCLR 5.0.20.21406, CoreFX 5.0.20.21406), X64 RyuJIT DefaultJob : .NET Core 5.0.0 (CoreCLR 5.0.20.21406, CoreFX 5.0.20.21406), X64 RyuJIT
| Method | Mean | Error | StdDev |
|---|---|---|---|
| OriginalDotNetInverse | 41.18 ns | 0.288 ns | 0.269 ns |
| FinalSIMDInverseWithSSEStore | 31.30 ns | 0.023 ns | 0.020 ns |
These are the results corresponding to the current commit, which uses the CompilerServices Unknown class to store the final output using SSE Store intrinsics. I followed the the benchmarking guidelines in the docs to generate the tests.