Hi everyone, happy Friday! Notes for Release 597
56 Likes
Whoa. :o
Were these edge cases or very overkill meshparts? What’s the lower bound for these improvements?
31 Likes
MeshParts with detailed collisions load up to 50% faster and use up to 30% less memory.
Yay now my computer won’t take 5 minutes to load a Mesh from blender 
13 Likes
Everyone is tryna speedrun to get the first replies lol
Adds TextTruncate support for rich text.
Cool, I thought it was already a feature, I guess not?
6 Likes
They just mentioned that just now, they must have done that earilier
5 Likes
This is a fix for Luau typechecking does not carry type guards after return statements , Intellisense should treat the "continue" keyword as a guard clause and Loop continue/break on conditional with check does not adjust type
18 Likes
Is that so? Then why is my map with over 6,000 MeshParts only giving me 30 FPS? (This is a joke, and also not a joke at the same time. My newest Backrooms map DOES have over 6,000 MeshParts.)
6 Likes
come on dont leave this guy hanging
9 Likes
AWESOME!! I thought I was going crazy when this wasn’t working. Thanks!
5 Likes
What was this new scroll wheel behavior in the first place?
5 Likes
I think it was the dynamic camera speed thing that let you change your camera speed between 0.1 and 20. A new slider could be dialed to change it, or by holding down the right mouse button and scrolling.
7 Likes
Reinstall roblox!
6 Likes
I’m still annoyed that all my Avatar > Head & Body > Head won’t load all my bought items, and the one I need is not shown even at the bottom. (Website)
At least make a pagination feature for it if you guys don’t want the long height of the webpage.
8 Likes
Whoops, the release notes left out that it improves unions too.
We wrote a new mesh decoder that runs about twice as fast. This applies to every TriangleMesh,so MeshParts and unions. The savings are multiplicative, so this only impacts you if you have a ton of high complexity meshes that aren’t box fidelity. This also means somewhat faster CSG operations.
The new decoder is a bit more careful with memory; it allocates less and doesn’t overallocate - This usually means 0-30% savings in kdtree memory and 0-50% in mesh convex storage memory depending on the mesh and how many vertices/convexes you have.
To get a good idea if this will impact you, you can check the memory readout under ctrl+f6 → mode → counters. physics/geometry/kdtree and physics/geometry/mesh are optimized, so if these are using a lot of memory in your place then this should be a small memory improvement for you.
We’ll probably turn it on next week when it makes it to most platforms.
35 Likes