David Neto
David Neto
Maybe adding "distinct" would help, to make: > Execution of a WGSL function [must not](https://www.w3.org/TR/WGSL/#shader-creation-error) potentially access memory through distinct aliased root identifiers, where one access is a write and...
> However, I think the proposed verbiage is even more confusing. I'm not sure if distinct is referring to the memory, or the identifier itself. "distinct" is meant to modify...
More about wording. This is the sentence that defines the term 'root identifier' > Within a function, each [memory view](https://gpuweb.github.io/gpuweb/wgsl/#memory-view) has a particular **root identifier**, which names the variable or...
I think this issue is resolved. Closing. Please reopen if you think it's not resolved.
> We have a separate discussion about the scope of the "only used in a .... stage" being scoped to only be talking about shader input/output. https://crbug.com/350478877 I filed https://github.com/gpuweb/gpuweb/issues/4748...
i agree with the proposed choices. For the TBD items, I think they should all be arrayable.
ack. we should do better editorially.
what do you recommend?
I think these are good arguments: * You're right, the function makes sense with edge0 > edge1. It just traverses the curve backward. * It saves you a subtraction. I'm...
Hm. MSL, SPIR-V, and GLSL say results are undefined when edge0 >= edge 1. So giving the guarantee in WGSL implies polyfilling it everywhere, which means bloating and slowing down...