Sean (Dec 18 2020 at 18:41): Sean (Dec 18 2020 at 18:41):/me attempts to revive adrt_slave
Our simulation needs are starting to exceed rt's current capabilities... need a working path tracer...
Sean (Dec 18 2020 at 18:42):oh and related, @starseeker any way to get facetize to not decimate the heck out of what it's working on when it's in a fallback mode?
starseeker (Dec 18 2020 at 18:44):--decimation-feature-size can be used, but I don't recall the semantics of how it behaves offhand. I think it's a scale factor compared to feature size...
starseeker (Dec 18 2020 at 18:45):let me do a quick test... stand by
Sean (Dec 18 2020 at 18:45):I'm currently running things like: facetize --feature-size .0003 --max-time 9999 newbot region
Sean (Dec 18 2020 at 18:46):I usually just get "Continuous Meshing: Fatal error, could not find crossing!"
CM: surface reconstruction failed: region
CM: continuation polygonization surface build failed.
Sean (Dec 18 2020 at 18:46):No feature size succeeds, but I go from 200k poly to like 5k poly and it's unusable
starseeker (Dec 18 2020 at 19:05):OK, --decimation-feature-size with smaller values should produce bigger meshes.
Sean (Dec 18 2020 at 19:05):different from --feature-size?
starseeker (Dec 18 2020 at 19:05):Yes. This was my test:
mged> facetize --CM --feature-size 0.06 --decimation-feature-size 2.0 ellipse.r ell_06.bot
CM: completed in 16 seconds with size 0.06
CM: completed in 23 seconds with size 0.054
CM: timed out after 30 seconds with size 0.0486
CM: successfully created 6960500 faces, feature size 0.054
CM: decimation succeeded, final BoT has 104 faces
mged> facetize --CM --feature-size 0.06 --decimation-feature-size .01 ellipse.r ell_06.bot
CM: completed in 16 seconds with size 0.06
CM: completed in 23 seconds with size 0.054
CM: timed out after 30 seconds with size 0.0486
CM: successfully created 6960500 faces, feature size 0.054
CM: decimation succeeded, final BoT has 247642 faces
Frankly I really don't like those settings - they're just direct exposures of numerical inputs in the underlying code, IIRC, and they're both sensitive and unintuitive.
starseeker (Dec 18 2020 at 19:08):IIRC "decimation-feature-size" is the one from the "new" decimation routine - I should probably go read the paper to try and get a better feel for what it's doing
starseeker (Dec 18 2020 at 19:13):In the long run, once we've got proper plugins-as-executables in place, I'd like to look at setting up some of the alternatives (CGAL, libigl) that might be slower but useful fallbacks - license considerations are a stopper for CGAL as-is, but with plugins as executables we might have more options.
Sean (Dec 18 2020 at 19:22):I'll give that a try, thank you
Erik (Dec 22 2020 at 02:07):I think if you're looking for the distributed aspect, it may be worth reconsidering mpi. Particularly if it's going to run on ib equipped clusters.
Sean (Dec 22 2020 at 21:09):we're just trying to get something that reliably doesn't result in error.
Sean (Dec 22 2020 at 21:10):I didn't mind it taking 24 hours the first pass, more impressed that it was initially a good result. Its this second pass that's destroying it via decimation.
Sean (Dec 22 2020 at 21:13):/me tried facetize --CM --feature-size .006 --decimation-feature-size .001 tip2.r tip3.bot ... actually fails very quickly now
Sean (Dec 22 2020 at 21:14):oh, promising, .0006/.0001 ...
Last updated: Oct 09 2024 at 00:44 UTC