Arseny Kapoulkine

By the way, maybe I'm doing something wrong but while the transform itself works (at least in node), it doesn't seem to work in gltf.report in Chrome. Here's the code...

Something that came up in this topic in two different contexts is restrictions on the size of RS in D3D12 (64 DWORDs) and push constants in Vulkan (128 bytes). I'd...

Oh - also - Metal setBytes is a pretty different beast. Its implementation is up to the driver of course, but I believe it's somewhat more akin to vkCmdUpdateBuffer in...

@litherum Sure - details. The ring buffer fundamentally has to be synchronized with the command buffer wrt lifetime so you can think of it as being separate or reusing the...

I want to clarify the meaning and the implication of the non-determinism classes highlighted here as I couldn't find them defined anywhere. - "set" non-determinism says "when the spec provides...

Yeah this one is pretty fundamental for many workflows, e.g. in rendering domains it's common to store data as fixed-point integers for GPU consumption but to prepare this data you...

Should these instructions have `_sat` in the name? In the SIMD MVP _sat stands for saturating, but these instructions don't specify exact behavior for out of range inputs.

> I think it's beneficial to have 2048 or any reasonable static upper bound for vector sizes to be enforced by implementations rather than being limited by spec. This will...

> However, @zeux, what do you mean by unbounded length? What do you think about the size of SIMD registers on the platform being the upper bound? I mean that...

FWIW I've also found the name "blend" confusing even though my main experience lies within the SSE domain. "laneselect" seems much more clear.