<J23k87>
but as i said - it could take some time until it stops
<ndnihil>
yeah, definitely not an immediate thing
<ndnihil>
but might help in the middle of a 15minute snooze
<ndnihil>
Total rendering time: 0:00:06.246
J23k87 has quit [Quit: Client closed]
<ndnihil>
rewrote it though, so it's not too bad
J23k87 has joined #openscad
<J23k87>
today i found geometry that will render much faster with manifold than the preview .. so render was some seconds while the preview took a minute
<ndnihil>
hah, all that rendering and shit
<ndnihil>
and I just realized I was doing it wrong
<InPhase>
J23k87: Difference of intersections?
<ndnihil>
still might kinda work, but wrong
<J23k87>
InPhase it was generated by a hull of a lot of tetrahedrons
<J23k87>
just seeing it is not a lot .. it is just one but with strange faces
<J23k87>
maybe preview get faster if that would be not such a mess
<J23k87>
oh yes using multiple faces with 3 points instead of one face with 1500 is much better
guerd871 has joined #openscad
guerd87 has joined #openscad
guerd871 has quit [Client Quit]
<InPhase>
Yeah, that looks like a pretty weird usage of polyhedron's face input.
<J23k87>
probably you showed me that in a way Ü ( iirc it was from the guy doing fillets without minkowski )
<InPhase>
Well I don't remember it, but sometimes I do really horrible hackery for fun. ;)
J23k87 has quit [Quit: Client closed]
J23k87 has joined #openscad
<J23k87>
no it wasn't your code - you just pointed me to that guy and while digging through the code i found this for in isosphere
<J23k87>
I wonder if normal face generation is possible in that approach .. i thought they might not be ordered - but maybe he was just lazy and used this shortcut
<InPhase>
Ah.
<InPhase>
And yeah, those are some wildly different times. I forgot I even opened the file since it took so long, but I'm getting 12 minutes preview, 0.2 seconds render even after fushing the cache.
<InPhase>
I like the sphere it generates though. That has nice properties.
<InPhase>
Not quite unbiased in the coordinates with respect to the axes, but it's a mild bias.
<J23k87>
the 0.1 and 1.2 value change the twist
<J23k87>
at least some faces are just [n, n+5 , n +13] not sure if that is working for all q
<peeps[work]>
i wonder if preview-based hull (or scale) is still going through cgal rather than manifold. its possible that slightly different code path got overlooked
<peeps[work]>
definitely seems like something is going wrong there
<J23k87>
hm https://imgur.com/E6t8R6s iam not try to close that .. also the triangles change with q ..
<peeps[work]>
seems like that's it. change q down to 1000 for faster test iterations (about 3.5s preview) then disable manifold, and time seems the same between render and preview.
<J23k87>
didn't olivier also mentioned that with fast CSG that there are some preview operations different from render - at least i had different results when render with preview and without in the beginning