krogovin

I freely admit that I am .. not really for pointers in a shading language. Here is an example which makes me sweat which pointers actually encourage: ``` struct flop...

so, the rules for pointer have that then pointer assignment is determined by compile time.. That makes implementation doable, but then I wonder what is the point of them really?...

hmm.... I think that for `pointer`, that is future proofing for something that should not ever happen since local variables should be registers. I can see `pointer` which is what...

just out of curiosity, what were the highlights? was it severe as rethink even having `pointer` ?

.. and virtual texture is not going to be fun because then you really want sparse texture. Personally, I very much want bindless as an optional feature, even as a...

I admit that that having a texture or buffer as a "pointer" is really useful.. and the only way to emulate that is with an unsized array. The use case...

I just want to second the point that a shader author needs to have an option of "I know what I am doing" in that one can declare that use...

There are a few bits about uniform control flow and it is worth spelling out fully. The main point was for SIMD architecture so that if a value depended on...

I'd advise for have a "strict/safe" and "liberal/unsafe" versions for those functions that have derivatives. On a related note, here is a real world example of using early out like...

> they're just going to start habitually replacing if(...) {} with if(uniform(...)) {} whenever they see one (or flipping the global escape hatch if one is provided) without even stopping...