Here is something quite interesting. I wonder if you, Ashley , already know about it? The interesting thing is that it does not improve on the gpu side but on the cpu side (which was about time!). 4x and more faster execution of HTML5 apps could also mean 4x and more faster C2
(I'm not as frequently active anymore, so if this has already been discussed before, I beg your pardon)
There was a discussion of SIMD in general a month or two ago. At the time ashley commented that it wouldn't really have many advantages for construct...though there wasn't a library available at the time, so maybe that was in the context of the effort required exceeding the potential benefits.
I really don't understand enough about this to comment more.
Ashley: would this library be at all useful?
No, I don't think there's anywhere in the C2 engine where SIMD would be useful or make a realistic performance difference. It can provide a big gain when working with large amounts of numerical data, like when encoding/decoding audio and video, doing software rendering (but we have GPU rendering). The article actually suggests gaming as a use case, but most games aren't really that data parallel, so I don't really understand why they suggest that. Also if you use SIMD wrong, it can actually degrade performance.
Develop games in your browser. Powerful, performant & highly capable.
"Dart's support for SIMD (single instruction, multiple data) processing allows for a big performance boost. Without SIMD, this animation simulation can show 34 monsters with acceptable performance, but with SIMD, it jumps to 124."
That's a 3D game, and the algorithms it uses are likely significantly different to anything our engine does.