Assertion `nowEffectiveSizeRatio <= (1.+1.e-8)' failed

[dmcdougall]

Ticket I got at TACC:

While I've gotten my test code to run correctly on Stampede, I'm running into a new issue as I move to the full calculations. When I increase the size of my all-atom reference set of dihedral angle samples, which QUESO is trying to match with coarse-grained parameters, I get the following error which appears to be from the QUESO library:

Assertion `nowEffectiveSizeRatio <= (1.+1.e-8)' failed.
nowEffectiveSizeRatio = -nan
(1.+1.e-8) = 1
effective sample size ratio cannot be > 1
stats/src/MLSampling.C, line 2794, compiled Oct 27 2017 at 16:59:05

This happens even though I can successfully run my test system using a larger number of all-atom samples, so I'm not clear on how this is happening.