Jeffrey Regier

LD_LBFGS. I've never encountered a "roundoff-limited" exception, only "nlopt failure".

I don't, unfortunately. I realize that makes it pretty difficult to diagnose the problem without one. The problem only appears for fairly complicated optimization problems. If I can recreate it...

Are there any workarounds in the meantime?

I need threading, unfortunately. On Sun, Mar 6, 2016 at 10:21 AM Stefan Karpinski [email protected] wrote: > If you stay on the stable release version of Julia, this shouldn't be...

Correct, we want to keep the parameters contiguous in memory. @keno has a branch of Celeste that does away this `ids`, I think. I believe he used function calls to...

The catalog is stored in ``` /global/cscratch1/sd/jregier/celeste_sdds_catalog_jld ``` That directory contains ~180,000 files in the `jld` format. Please see PR #496 for how to convert them to `fits` files.

Yes, that's fine if a few are zero sized. Glad to hear the conversion went smoothly.

@rcthomas Is this still something you're working on? It would help us to publicize the GB results. We'll also have "model" images once #663 is done.

Setting `max_iters=200` seems to help with the `simple_star` galsim test case, but not with `validate_on_stripe82.jl --fft`. Not that I'm working on this issue, just figured I'd mention it.

What's happening here? This object is very bright -- 16 star magnitude is about right. It's a galaxy but it has a small radius. Therefore `star_mag_r` and `gal_mag_r` should be...