Null space problem: Compute resources

I’m working on a story for the Advanced Scientific Computing (ASC) program. I’d like to talk about the compute resources required to do the global nullspace problem vs the U-Spline enabled local nullspace problems. Does Coreform have any internal documentation related to this?

Would someone at Coreform be able to guesstimate the compute resources required to find a P2-C1 smooth spline on this example mesh?

image

Essentially, the story I’d like to say would be something like:

Computing the C^1 smoothness constraint on this problem, globally, requires either:
a) A quantum computer
or
b) An exascale supercomputer running for millions of years

But using the U-Spline algorithm allows this problem to be solved in less than a second on a single CPU core.

Is there a story here?

Can you send us the mesh?

Not sure if quantum computers will do anything for this problem. It’s hard to say how long the brute force approach would take.

@derek,

I’ve sent you the files via email, as the formats aren’t able to be uploaded to this forum.

As a matter of record, the geometry is publicly available here:

https://cubit.sandia.gov/public/15.1/help_manual/WebHelp/step_by_step_tutorials/power_tools/step_01.htm

Note that the mesh files I sent you are slightly different than the one in the above picture.