Slug algorithm for text rendering

[wh1t3lord]

@mikke89 hi, I will return to my PRs soon :) but it happened officially and got a lot of hype from senior devs

Author's tweet of slug: https://x.com/ericlengyel/status/2033973707547545643

Probable implementation but not SOTA (I didn’t research that repo but as reference implementation?): https://github.com/behdad/glyphy

Q: what do you think guys should we integrate this and what general benefits over freetype rasterization?

If it out performs or equally replacement to freetype then we can make freetype dependency optionally and it is kinda cool?

For rmlui it might get an additional promotion due to hype over slug (but as we know all hype will end at some point and as marketing thing it might not be effective in terms of promotion and gaining popularity).

And as was stated before we can provide native implementation of font rasterzation and live without freetype (as I understood the purpose of slug’s algorithm at first glance?).

So if there are guys who integrated or used slug could you tell us what benefits of slug over freetype please?

Answer from AI:

Slug is a proprietary, GPU-accelerated font rendering library that generally outperforms Freetype by providing perfectly antialiased text at any scale without precomputed textures. It uses a patented algorithm to render resolution-independent vector glyphs directly on the GPU, offering superior quality for complex diagrams and text at high speed

sounds cool what ai answered

[mikke89]

Hello again!

Yeah, it's a very interesting technique. Would be really cool to see how it would work in the library. Some thoughts:

• We should consider this against other techniques too, particularly SDF and multi-channel SDF. I believe those serve similar benefits.
• For this to replace freetype, it would need to handle fonts just as good as freetype does, including quirks. Maybe it could make sense to be able to use them together.
• We are moving towards better support for international scripts such as CJK, I wonder how it compares performance-wise when there are a lot of distinct glyphs.

It would be very interesting to hear from anyone with implementation experience. I'm not so concerned about promotion exactly, it needs to make sense from a combination of performance and visual quality primarily.

[wh1t3lord]

harfbuzz/harfbuzz#5878

[wh1t3lord]

@mikke89 https://x.com/harfbuzz/status/2039345746328375631?s=46

[wh1t3lord]

https://harfbuzz.github.io/hb-gpu-demo/

[mikke89]

Very cool! Looks quite good too.