I'm recently learning the stuffs related to TBN Matrix. I already got that how this TBN Matrix is calculated and in which way we're going to use it to convert normals from tangent space into object/world space(depends on the basis vector of TBN) so that they're able to be computed by lights in the same space.
But I also noticed that almost in every tutorial teaching this subject, there is always a section telling how to calculate the inverse of TBN so that you can convert lights into tangent space as an option.
Well really, this is just an option, but is there any reason we do as such? Is there any difference between converting normals from tangent space to object/world space AND converting lights from object/world space to tangent space? Can we gain some performance or can we save a lot of memory or calculation cost? I just can't figure it out, since I think they have the same quantity of calculations, due to each vertex having a unique TBN Matrix on its own.
Thanks for your help~