The developer behind the Lossless Scaling Frame Generation on Linux (lsfg-vk) project has made an [announcement](https://lsfg-vk.dev/blog/important-changes-to-lsfg-vk/) about a bunch of major changes.
For starters - they've moved away from GitHub due to it "getting increasingly worse over the past few years" and it constantly shoving "AI and Copilot in your face". Can't really argue with that, Microsoft loves AI. They're now [self-hosting the repository](https://git.lsfg-vk.dev/) along with having a [Codeberg](https://codeberg.org/PancakeTAS/lsfg-vk) mirror.
Another big change is the license. It was originally MIT licensed, then they moved to the GPLv3 for for next major version but now they've changed it again to CC BY-NC-ND 4.0. This is quite a major change and technically means it is no longer open source but source available due to the restrictions placed on it.
Why such a major licensing change?
"Especially with GitHub shoving AI and Copilot into every corner, it is no surprise that over 87% of all forks (not including lsfg-vk v1 or plain mirrors) are AI-generated, most of which abandoned after a few commits that never worked in the first place.
There have also been way too many cases of people making an AI slop fork with the sole intention of either misleading users or making money off of it."
As a personal passion project - they said they don't want to have to deal with any of it. And people can reach out to the developer directly if they want to work on it, as they're willing to discuss forks and such but this is their personal way to try and stop all the AI slop.
They also released the first testing build of v2 with some major changes:
-
An entirely new bindless pipeline has been written. This pipeline brings a bunch of new features with it.
-
FP16 acceleration: On systems with a 2:1 FP16 ratio, such as most AMD laptops and handhelds, performance will be 2x-3x as fast as before, without losing any quality at all.
-
Memory reuse: The new bindless pipeline reuses memory wherever possible, using only 30% as much memory as before. This beats even the Windows version by a landslide!
-
Dropping Vulkan to 1.2: This might seem like a downgrade at first, but dropping the required Vulkan version from 1.3 to 1.2 actually has no effect on lsfg-vk itself. Instead, it enables support for lsfg-vk to (as far as I know) every single Vulkan-capable GPU ever!
-
Processes can be identified via Linux binary, Windows executable and Steam App ID, meaning it’s easier than ever to set up a profile for your game.
-
A new tool
lsfg-vk-cli
can be used to benchmark lsfg-vk and validate the configuration file. - The user interface is now Qt6-based, no longer GTK+Rust. This brings the filesize down from 70 MB to just a few kilobytes. It will also integrate neatly into the KDE desktop.
- Prebuilt lsfg-vk now also contains a layer for 32-bit games, which works seamlessly.
- Various bug fixes and minor quality improvements.
External Sources:
[git.lsfg-vk.dev](https://git.lsfg-vk.dev/),
[codeberg.org](https://codeberg.org/PancakeTAS/lsfg-vk),
[lsfg-vk.dev](https://lsfg-vk.dev/blog/important-changes-to-lsfg-vk/),
[creativecommons.org](https://creativecommons.org/licenses/by-nc-nd/4.0/deed.en)
Some you may have missed, popular articles from the last month:
All posts need to
follow our rules. Please hit the Report Flag icon on any post that breaks the rules or contains illegal / harmful content. Readers can alsoemail usfor any issues or concerns.