This repository is used to see what implementation of the ScaleX algorithm is the fastest, and uses the least memory. This code will Scale a bitmap 2 or 3 times with the algorithm described here: https://www.scale2x.it/algorithm
This implementation is used by Greenshot, to scale the icons for HighDPI screens so they are not so small. It should only solve a temporary issue, until Greenshot can use vector graphics...
The reason to place it in this repository, is to see & discus what tricks can be used to optimize the code.
Usage: Change the solution to Release. The ScaleX.Benchmarks has the benchmarks, set this as startup project. Run with Ctrl+F5
The Scale2x_FastBitmap and Scale3x_FastBitmap tests are for the old and "current" Greenshot code, using an implementation to directly go to the bitmap raw data. It uses Parallel.ForEach, and thus blocks the CPU.
The new implementation uses Span and is using a single thread, which is faster and uses less CPU power. They also don't directly use System.Drawing.Bitmap, as an experiment.
The tests Scale2x_Unmanaged_Reference & Scale3x_Unmanaged_Reference use my initial code , this is used as a reference. The Scale2x_Unmanaged & Scale3x_Unmanaged is a copy of the reference implementation, used to make improvements and being able to compare them, a few changes are already in there, using refs, which does seem to improve a bit.
Besides being curious about what performance improvements are possible, one question does bother me: When running on .NET 472 I noticed a huge increase vs dotnet core 3.0 in the Scale3x_Unmanaged memory usage (from 56 B to 4096 B), I would love to understand why!
Running it on an old Windows 7 PC:
BenchmarkDotNet=v0.11.4, OS=Windows 7 SP1 (6.1.7601.0)
Intel Core i5-2400 CPU 3.10GHz (Sandy Bridge), 1 CPU, 4 logical and 4 physical cores
Frequency=3020634 Hz, Resolution=331.0563 ns, Timer=TSC
.NET Core SDK=3.0.100-preview4-010937
[Host] : .NET Core 3.0.0-preview4-27521-07 (CoreCLR 4.6.27521.73, CoreFX 4.7.19.16407), 64bit RyuJIT
DefaultJob : .NET Core 3.0.0-preview4-27521-07 (CoreCLR 4.6.27521.73, CoreFX 4.7.19.16407), 64bit RyuJIT| Method | Mean | Error | StdDev | Min | Max | Gen 0/1k Op | Gen 1/1k Op | Gen 2/1k Op | Allocated Memory/Op |
|---|---|---|---|---|---|---|---|---|---|
| Scale2x_FastBitmap | 7.013 ms | 0.1335 ms | 0.1249 ms | 6.843 ms | 7.305 ms | - | - | - | 2140 B |
| Scale2x_Unmanaged | 2.482 ms | 0.0027 ms | 0.0024 ms | 2.476 ms | 2.486 ms | - | - | - | 40 B |
| Scale2x_Unmanaged_Reference | 2.707 ms | 0.0040 ms | 0.0037 ms | 2.702 ms | 2.715 ms | - | - | - | 40 B |
| Scale3x_FastBitmap | 13.906 ms | 0.0724 ms | 0.0677 ms | 13.815 ms | 14.050 ms | - | - | - | 2115 B |
| Scale3x_Unmanaged | 5.616 ms | 0.0061 ms | 0.0054 ms | 5.605 ms | 5.625 ms | 500.0000 | 39.0625 | - | 40 B |
| Scale3x_Unmanaged_Reference | 5.890 ms | 0.0154 ms | 0.0144 ms | 5.857 ms | 5.911 ms | 500.0000 | 46.8750 | - | 40 B |
BenchmarkDotNet=v0.11.4, OS=Windows 7 SP1 (6.1.7601.0)
Intel Core i5-2400 CPU 3.10GHz (Sandy Bridge), 1 CPU, 4 logical and 4 physical cores
Frequency=3020634 Hz, Resolution=331.0563 ns, Timer=TSC
[Host] : .NET Framework 4.7.2 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3324.0
DefaultJob : .NET Framework 4.7.2 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3324.0
| Method | Mean | Error | StdDev | Median | Min | Max | Gen 0/1k Op | Gen 1/1k Op | Gen 2/1k Op | Allocated Memory/Op |
|---|---|---|---|---|---|---|---|---|---|---|
| Scale2x_FastBitmap | 7.017 ms | 0.1592 ms | 0.3179 ms | 6.851 ms | 6.780 ms | 7.874 ms | - | - | - | 1920 B |
| Scale2x_Unmanaged | 3.375 ms | 0.0039 ms | 0.0036 ms | 3.375 ms | 3.370 ms | 3.382 ms | - | - | - | 64 B |
| Scale2x_Unmanaged_Reference | 3.763 ms | 0.0050 ms | 0.0047 ms | 3.763 ms | 3.753 ms | 3.770 ms | - | - | - | 64 B |
| Scale3x_FastBitmap | 13.659 ms | 0.0269 ms | 0.0239 ms | 13.656 ms | 13.626 ms | 13.704 ms | - | - | - | 2048 B |
| Scale3x_Unmanaged | 7.196 ms | 0.0103 ms | 0.0092 ms | 7.194 ms | 7.183 ms | 7.212 ms | 500.0000 | 7.8125 | - | 4096 B |
| Scale3x_Unmanaged_Reference | 7.499 ms | 0.0078 ms | 0.0073 ms | 7.497 ms | 7.486 ms | 7.509 ms | 500.0000 | 7.8125 | - | 4096 B |