David Neto
David Neto
If the only problem is the size of the converted shaders, or the time to generate it, then we can make Tint's converter smarter. Currently the GLSL emission path generates...
Ooof, I was too optimistic. > Currently the GLSL emission path generates individual uint padding members. I took a quick look and think it can be taught to emit sized...
Like @kangz said i++ and i-- are statements right now. Allowing `--i` as an expression actually breaks backward compatibility, because `a--i` is parsed as `(a)-(-(i))`. If we add prefix decrement...
Nice. Seems like a bug, and the fix should be treated like a bugfix.
For Vulkan: both dynamic and non-uniform indexing are required in the "Roadmap 2022" feature set. See https://docs.vulkan.org/spec/latest/appendices/roadmap.html > The Roadmap 2022 milestone is intended to be supported by newer mid-to-high-end...
The last 8 of those feature bits were promoted into Vulkan 1.2 (where they are queryable but not necessarily required); prior to that you have to look for VK_EXT_descriptor_indexing: https://docs.vulkan.org/spec/latest/appendices/extensions.html#VK_EXT_descriptor_indexing
For tooling that emits WGSL, we can fix the shadowing problem with a reserved `wgsl` namespace, as in #4308. That's in Milestone 2. General namespaces is in M3+. I recommend...
Fixed by https://github.com/gpuweb/gpuweb/pull/5394
Fixed by https://github.com/gpuweb/gpuweb/pull/5394
Fixed by https://github.com/gpuweb/gpuweb/pull/5394