Repository navigation
Use .NET Core Hardware Intrinsics to optimize the code? #56
Description
It's great to see a C# machine learning framework!
It seems some linear algebra algorithms are implemented by calling into native cpp code. I assume this is due to the rich SIMD instructions in cpp. Since .NET Core 2.1 has the preview feature of Hardware Intrinsics, using hardware intrinsics is another option to use SIMD instructions.
Reacted by Fei Peng, Dan Moseley, M. Zeeshan Siddiqui, Justin Ormont, Tom Finley, Shahid Khan, Gani Nazirov, Adam Jones, Sorrien and jp2masa
Activity
helloguo commentedon May 8, 2018
cc\ @asthana86
[-]Use Hardware Intrinsics to optimize the code?[/-][+]Use .NET Core Hardware Intrinsics to optimize the code?[/+]helloguo commentedon May 9, 2018
Relate to #97
danmoseley commentedon Sep 6, 2018
This is done for x86/x64 with some remaining issues here #823
Perhaps you would like to tackle any of those, @helloguo ?
When ARM intrinsics are offered by .NET, we could add implementations for those.
I believe this issue is done, otherwise. Thanks for suggesting it to us.
danmoseley commentedon Sep 6, 2018
And please try it!
misc fixes & test additions, towards 0.1 release (dotnet#56)
code migration from machinelearning-automl to machinelearning repo fe…
misc fixes & test additions, towards 0.1 release (dotnet#56)
Added onnx export functionality for LpNormNormalizingTransformer (#4161)
Added Onnx export functionality to PCATransformer (#4188)