David Neto

C99's spec for strtod doesn't go into detail on nuisance limits.

This boils down to a request to Microsoft: Does DirectX have a functional spec (effectively) for floating point accuracy? The starting point in WGSL is "what Vulkan says" but for...

The model for this is module-scope OpVariable in SPIR-V. That allows any OpConstant* or specialization constant. So yes, it should actually be override expression.

Another aspect is whether the value can be of structure type. With #3056 we decided to allow creation-time constants of array type (with creation-fixed footprint, of course). But we shied...

I mentioned an annoying case for rewriting an out param as a return value. It was the case where the function call is part of a larger expression. It gets...

In discussions I've heard the argument that, if I may paraphrase: * the pointer feature doesn't have enough functionality * future expansion of the feature would require the use of...

In a real sense Apple's feedback is that the word 'ptr' is too suggestive of features programmers are used to from other languages. It's a conceptualization problem primarily(?), and also...

One more thing: The atomic operations very strongly depend on *not* copying in and out. They must be able to name exactly what memory locations are going to be affected....

Memory has layout subsection which assumes you've seen other types. So it would be better to move type aliases up.' I think: Type checking Types for numeric expressions computed at...

I want to amplify @dj2's point that whatever we do has to scale up to future types like vectors of 16bit floats. It's best if that's done in a very...