robtfm
robtfm
> I'm curious how this improves compile times. It looks like just as much functions need to be codegened, if not even more for the type erased variants. yes it's...
> This PR doesn't compile in release on my mac fixed. will address the other feedback tomorrow if there's positive disposition.
maybe related: https://fasterthanli.me/articles/when-rustc-explodes#what-now ~sensitivy check: is `BlackBox` acceptable for a wrapper that contains a boxed type-erased value?~
you sent me down a rabbit hole of newtypes not being zero-cost before i read the edit :) ... apparently newtypes are only zero cost as long as they have...
I thought it was needed so that we can take it in try_unwrap, and not subsequently decrement it on drop. Maybe there is a better way, I’m not really sure...
> I would be happy to help develop and maintain the `naga_oil` crate. thanks, i appreciate it - worth noting though that even if it's part of bevy it can...
added overrides, so functions can be replaced dynamically, and added an example `shader_material_override` to demonstrate. todo still: provide a mechanism for general overrides (so you can override the default lighting...
and added global/default overrides, with `pbr_global_override` example
> 2. naga_oil covers cases that are "core" enough to the bevy shader experience that I would want that to be under the bevyengine org umbrella and go through our...
> structs can be used in entry-point IO, but only one level deep i've moved @bulitin(position) into the MeshVertexOutput structs so we can use them directly as the vertex stage...