The popular creative app paint.net now natively supports ARM64. Previously, the app would run in emulation on devices like the Surface Pro X and HP Elite Folio. The move to ARM64 is only available for alpha testers at the moment but should bring improved performance on Windows 10 on ARM PCs.
Its recent update brings paint.net alpha to build 7894. It includes a long list of features and improvements, including faster Magic Wand and Paint Bucket tools. The app’s author, Rick Brewster, says that he’s “spent some time optimizing the snot out of the Magic Wand and Paint Bucket tools.”
This update also moves to self-contained deployment, which means people don’t have to install .NET anymore. All needed components are now bundled with the app. This shouldn’t result in any changes in functionality, but it simplifies some things for app installation.
Brewster is also testing a web installer. This has a few benefits, the most notable of which is that it ensures people with Windows 10 on ARM PCs get the ARM64 version of the app.
Here are all of the changes compared to the previous alpha build of paint.net:
- Improved performance of the Magic Wand tool by a substantial amount overall
- Improved performance of the Paint Bucket tool by a substantial amount overall
- Improved performance of the Paint Bucket when changing the color or certain other toolbar settings by not recalculating the stencil buffer for the filled area. Changes should now be reflected immediately.
- Improved: Plugins are now loaded into isolated AssemblyLoadContexts, which improves stability and resiliency to various problems with versioning, loading dependencies, or incorrect installation.
- Fixed a rendering bug in the Gradient Tool, first reported here: https://forums.getpaint.net/topic/118486-paintnet-43-alpha-build-7881/?do=findComment&comment=585139 (regression from 4.2.16)
- Fixed a crash when trying to draw a fixed-size rectangle selection that was larger than the image
- Fixed a crash with the Shapes dropdown, which was due to a bug in WinForms
- Fixed a bug with the Tolerance slider where both 58% and 59% would be displayed as 58%, due to a rounding issue. The correct value was actually being used, it just wasn’t being displayed correctly.
- Added a /disablePlugins command-line argument to aid in troubleshooting various issues with app startup and incorrect plugin installation
- Changed: The requirement to have SSSE3 on x86/x64 has been reverted to just SSE2. However, CPUs without SSSE3 are no longer being optimized for and will run slower as a result.
- Updated the bundled AvidFileType plugin to v126.96.36.199, courtesy of @null54, which includes support for multi-layer AVIF files. See the project’s GitHub Releases page for more information.
If you’re updating from version 4.2.16, you’ll see the following changes:
- New: The app has been migrated to .NET 5, and performance has been greatly improved as a result.
- New: ARM64 is now natively supported
- Installation is much faster now due to no longer needing the “Optimizing performance for your system” stage (NGEN has been replaced with ReadyToRun via crossgen).
- Portable ZIPs are now officially available, utilizing self-contained deployment, and do not require the installation of .NET 5. You must manually update to newer versions of the app, however.
- Improved performance of most effects and other compute-intensive tasks by about 20% on average, due to .NET 5’s improved code generation
- Improved: Many areas of the app have been optimized to use SIMD (SSE2/3/4.1, AVX2) on x86/x64
- Drawing on an image with a large number of layers is faster
- Drawing with the Gradient Tool’s alpha mode is faster
- Drawing with the Recolor tool is faster
- Drawing with the Erase or Clone Stamp tool is faster when using a color whose alpha value is less than 255
- Drawing with the Overwrite blend mode is faster
- Drawing with the Shapes tool is faster, due to improved compositing performance
- Drawing with a brush tool without antialiasing is faster
- Tile compression has lower overhead due to being optimized for SSE2 and AVX2
- Working with complex selections is faster
- Many other optimizations all throughout the application and rendering engine
- Improved: Selection antialiasing quality is now better by way of using a 4×4 super sampling filter instead of 3×3. The code has also been optimized to use SSSE3 on x86/x64.
- Improved: Performance should be significantly better on systems without HyperThreading (by reserving 1 core for the UI), and also a bit better on systems with it (by utilizing more of the logical cores)
- Improved: Effect and File Type plugins can now be organized into folders — the plugin loader will recursively search for DLLs up to 1 folder deep
- Fixed a bug in Polar Inversion that was causing hangs
- Improved performance of Polar Inversion by 3x, due to the aforementioned bug fix (and a little bit is due to .NET 5)
- Fixed: There were some issues with zooming in/out using the keyboard, where the anchoring (centering) point was not calculated correctly, causing the canvas to drift in the wrong direction.
- Fixed a bug when saving with “Auto Detect” bit-depth that was causing fidelity loss on some images that already had 256 colors or less. This was reported here: https://forums.getpaint.net/topic/118401-images-already-with-256-or-fewer-colors-being-dithered-when-saved-with-palette/
- Changed: A processor that supports SSSE3 (yes 3 S’s) is now required for x86/x64 systems (previously only SSE2 was required). All CPUs released since about 2006 (e.g. Core 2 Duo) support this.
- Known Issue: Due to crashing, GPU accelerated effects (Gaussian Blur, Motion Blur, Radial Blur) on ARM64 will actually use the CPU for rendering.
A blog post from Brewster explains how to get the latest alpha build. At this time, you cannot get alpha or beta releases of the app through the Microsoft Store.