-
-
Notifications
You must be signed in to change notification settings - Fork 888
Closed
Labels
Description
Prerequsite: #1354
- Top gain: reduced memory footprint
- There will be a noticeable LoadResizeSave speed regression on 2.1 and Framework (no SIMD for shuffles), but I hope the ROTR / ROTL path will keep it in an acceptable range.
- On 3.1 and .NET 5 the LoadResizeSave speed regression should be negligible thanks to the SIMD path in 3 <==> 4 Channel Shuffling with Hardware Intrinsics #1409
- Introduce PixelTypeInfo.HasAlpha #1396 combined with omitting premultipication for
Rgb24will cancel out these regressions, bringing noticeable speedup on 3.1 and .NET 5. We may want to consider milestoning that issue to 1.1.0
JimBobSquarePants