Jukka Larja (roskakori2006.delete@this.gmail.com) on November 1, 2020 10:11 am wrote:
> hobold (hobold.delete@this.vectorizer.org) on November 1, 2020 7:22 am wrote:
[...]
> > I was thinking of the contrast illustrated, for example, here between a plain
> > formula on the one hand, and a mess of nested intrinsics on the other hand:
> >
> > https://nullprogram.com/blog/2015/07/10/
> >
> > The GCC extension allows writing a plain formula, too, when targeting a SIMD backend.
>
> Mandelbrot example is such a toy that I'm not even sure what it should illustrate
> compared to our game code. I don't see such nice loops there.

My apologies. I didn't mean to load the SIMD burden on your shoulders. Any performance tuning work, SIMD or otherwise, is doomed when the underlying codebase is already unmaintainable or keeps shifting day to day.

I was just adding my two cents that even for something impractically tiny and simple like a Mandelbrot core loop, intrinsics are making for a bad language. Stilted, verbose, very redundant yet not error resilient, obscure ... the opposite of efficient transmission of information. Yet another unnecessary obstacle.

The GPGPU language dialects did not take that route. For good reason: you don't want to actively discourage programmers from using a new platform.