Between this and the Xi editor, I am impressed how Raph has shown Rust’s strengths for system software. I hope these projects continue gaining traction moving beyond a useable proof of concept.
“Parsing: No allocation”: Nice to see laziness in action.
Allocation is the enemy of performance.
Impressive and interesting, but from the “What next?” section:
Possibly font-rs becomes production-ready.
I’m always skeptical of speed claims for non-production software. A lot of bloat and slow down can come from productizing software and making the interface actually useful and usable. Especially on something like font rendering, where being the absolute fastest possible may be less important than having some other features that may slow things down.
The author mentioned on Reddit that it’s mostly around the error handing and some features he’s not personally invested in; that doesn’t significantly change the core rendering loop.
I agree with your point in general, but this is also someone who’s an expert; he knows what he’s talking about.
Raph has been developing and shipping font and rendering software since the 90s. He knows this space as well as anyone.
[Comment removed by author]
The TLDR of this seems to be SIMD.
Really? The first graph shows that even without SIMD, there’s a pretty wide performance gap.
I was obviously reading this wrong.