I’ve seen on your UMG tutorial video the usage of small “separated” textures to display item icons.
On the other hand I built a Material Function that allows Texture Atlases usage before start any UMG work and built almost “everything” using Dynamic Material Instances and widgets with a common material and variables to access UV sectors.
Imagining on a game with “90+” kinds of items and where the player could access ~24 simultaneously I “thought” that could be better get just one texture copy in memory and get the desired areas, but on watch one from your pro guys doing that way I’m not sure anymore about what I did.
I’ve built them this way because:
- I’ve read a bit about Memory Fragmentation, thought that this could also occurs on video memory (where I suppose U4 loads textures).
- This method allows Texture Streaming to non-power-of-two sized icons, I can have lots of 52x44px icons inside a 1024 power of two texture.
- While the benefits could not be greatly noticed unless on mobile (remembering a teacher’s tip about battery usage on create/load objects) I thought “less loads, better”*.
Please notice that I’m just interested on don’t make the things “wrong”, I understand that both approaches “works”, my only fear is on try to sophisticate my implementation be in really messing everything, so if someone could give me some feedback about if I “can keep” or “better change” I’ll thank you a lot.