Actually, I'll take that back about the emphasis bit. Boswell pretty well nails it right on the head. Now I'm looking through some of his other articles, and they're excellent.

Well, ok. Though there's not much more that I could have written in that short of a space that can teach the subject.

I linked the Calgary Herald / Postmedia News article because it's an astonishingly well-written bit of science journalism that lays it all out superbly – kudos to Randy Boswell. He didn't put *exactly* the same emphasis on exactly the same things that Proemse (the principal author) would have, but it's minor. That's the "public" piece, and it's full of tons of great information.

I also linked the official research article. Unfortunately it's behind a paywall. However, if that's the kind of thing that really turns your crank you probably already have access to it one way or another (in the worst case: via a physical trip to your local university). If you can't, well, correspondence with an author is a time-honored method for obtaining your own copy.

Nah, he's not wrong. But neither are you.

Seismic processing is about as embarrassingly parallel as it comes. Just about every processing step can be split up into e.g. single shot record steps, taking advantage of assumed linearity in wave equations. Furthermore, most production industrial imaging codes weren't actually using my original example of a full finite difference solution until quite recently, and instead they were using algorithms that have been developed for decades under the limitations of very old computers. Sure, some of the big shops have full blown "proper" HPC, shared-ram setups, etc. However, it's common to see much more simplistic parallelization with very ad-hoc clusters being used.

In short, there are loads of processing shops that run off-the-shelf servers on gigabit ethernet, and they do a good business with it. Heck, there are loads of processing shops out there that do a good business running relatively crude time migrations.

Seismic imaging. Imagine solving a wave equation (acoustic, elastic, or worse) over a 3D grid many kilometers on a side with grid spacing on the order of meters. Imagine you're doing it with a strong high-order finite-difference code. Calculate for tens of thousands of timesteps. Now repeat that entire thing thousands of times for a given full survey.

No matter how much computer you have, it's never nearly enough for seismic imaging.

No, the public won't see these results as a rule, at least not right away (while it's still commercially valuable to protect as a secret), though strictly speaking it depends on the countries involved. Nor would it matter for property value, as the "land owners" usually don't own the mineral rights in most places. Furthermore, this setup will be used probably mostly for marine/offshore seismic imaging, ie not much land involved.

Thanks!

It's sort of amazing that this site has consistently been my browser home page for sooooo many years.