Robi395

I completely agree with you on such matter, but I don't need any convincing. From the point of view of a guy who does not believe in climate change, saying that Dr. Murry Salby should no be trusted might actually have the opposite effect, and present this Dr. as a person fighting against all odds for the truth.

I quickly glanced over the post, it looks great. This kind of rebuttal is more convincing to a misinformed audience than saying, "the guy is bad". On the other hand, the frustration from people continuously spreading scientifically wrong ideas is understandable.

While this guy does have a shady past, It would be better to shown why his data or logic is false/inaccurate; and not base argument against him on his résumé, if only to better educate people watching the video.

"We believe that run-time performance can be improved significantly" is an understatement. Jacobi and Gauss Seidel have very poor convergence property on Poisson-like problem [0]. A state-of-the-art Poisson solver is likely to use a Krylov-space method [1], preconditioned with a multi-grid [2] or some sort of coarse preconditioner. This of course does not make such proof of principle less valuable.

[0] https://doi.ieeecomputersociety.org/cms/Computer.org/dl/mags... (On this graph, MG means MultiGrid)

[1] https://en.wikipedia.org/wiki/Krylov_subspace#Existing_metho...

[2] https://en.wikipedia.org/wiki/Multigrid_method

You are right, I thought vessenes meant: "FP64 seems like a very small use case for most of the parallelized workflows I can imagine" for any platform.

For scientific workloads, double precision is a must have. The 7 digits of FP32 is not enough. In my lab, we haven't updated our Kepler based GPU since 2013 for this reason.

Thanks, I'll take a look at those links. In the meantime, I have another question. How much do the Mill team plan to share with the general public before any product release, can we expect a emulator, it would be great to toy with.

Hi Will, A recurring question by the audience in the talks is that you will not get many benefits from such a wide architecture on general purpose code because of there is not enough "available parallelism". Do you have any concrete code examples where the Mill will find stuff to do where a OOO will not?