Troels Henriksen

I think the CI got stuck on this one. Could you give it a kick?

I pushed a commit that enables some more browser crud. With that change, it works on my home machine. (Probably not on my laptop, which I think has a GPU...

Yes, sure. We just need to figure out how to make the runners listen to this repository as well. Fortunately, we have hired a mercenary to do things like that.

Why is that useful? It's almost the same program as [the one we have](https://github.com/diku-dk/futhark-benchmarks/tree/master/accelerate/mandelbrot), only with a slightly different color function (but that isn't where the time is spent). If...

It is more precise (uses double precision), but on most GPUs it will be significantly slower.

Is there much to gain by exploiting the symmetry, when compared to just doing a normal matrix multiplication? It seems like it would only reduce the work by a factor...

No, Futhark is pure. For testing and benchmarking programs, the compiler automatically inserts IO for reading input and writing results ([see here](https://futhark-book.readthedocs.io/en/latest/practical-matters.html#testing-and-debugging)), but for actual production use, Futhark code must...

Are there interesting newer papers that use it? I'm a bit reluctant to incorporate more unmaintained benchmark suites, unless the code quality is much better than is usually the case.

I guess the PBBS link for this one is dead.

Futhark does not fare particularly well against ClBlast (at least I'd not expect that it does). This is based on only a little data: for the [PPoPP'19 experiments](https://github.com/diku-dk/futhark-ppopp19) we implemented...