Yeah, those deep-pocketed government lab and university researchers are clearly the dominant revenue source for MathWorks. Everybody in the corporate world is super nimble and moved on to Numpy as soon as they could.
The rules of HN state: "Essentially there are two rules here: don't post or upvote crap links, and don't be rude or dumb in comment threads."
I think you have unintentionally violated the last provision. Please don't post about things you have not taken the time to investigate at all, which is true by your own admission: "I've never touched [ScaLAPACK]".
Indeed, we didn't chose Elemental over ScaLAPACK; we used Elemental to replace LAPACK because it was impossible to refactor our code to use ScaLAPACK without absolutely ruining it forever. The ScaLAPACK interface is horrible and should only be used inside of a higher-level library interface that hides its ugly, ugly control structures.
From my perspective, Elemental is primarily about human performance. That it scales to thousands of nodes of Blue Gene and achieves performance on par with (and in some cases significantly better than) ScaLAPACK is really icing on the cake. Granted, I really like scalability and performance, but I will not ruin my code in the name of these things.
Given that a few dozen US-based students attend ACTS every year, I hardly see how it is a game-changer w.r.t. the widespread acceptance of Elemental. I think the internet is doing just fine so far.
CPU programming on the other hand, especially with OpenMP, is hard to get into at first, but once you have the right formula you can apply it pretty easily to most common tasks.
My guess is that you've spent a lot more time trying to understand CUDA and NVIDIA hardware. Or maybe all of your arrays are dimensions that are multiples of 16.
