Now that VK_EXT_shader_object is out, I've got some thoughts...

Before I dig in, a disclaimer: I am NOT speaking on behalf of Khronos or the Vulkan working group. I also no longer work for a hardware vendor. I'm just very well-connected within the 3D graphics community. The way I see ESO is as a sort of speculative or presumptuous directional extension. In an ideal world, it's where we'd like to go. Will we be able to get there? We've done enough engineering work to know it's implementable on the three desktop vendors. Will it get good perf?🤷‍♀️

For those of you that don't know, today is #TransDayOfVisibility. I'm not going to do a whole thread because that's not my style. What I will say is that we exist, that we're fairly normal people (apart from getting assigned the wrong gender) and that there's a surprising number of us writing your graphics drivers. 🧐

As I continue playing with all the different 3D hardware out there, I've found a fascinating (but obvious) correlation: How the different IHVs are doing in terms of market dominance exactly correlates with how good their HW design is. Exactly. At this point, I've had in-depth conversations with driver engineers and also personally touched the drivers for basically all the 3D hardware in the industry so I'm pretty confident in this assertion.

I kind-of hate Vulkan pipeline objects.

Ok, yes, I realize that may be a bit of an inflammatory statement in some circles but let me explain... 🧵 Back in 2015 when we were designing Vulkan, much of the graphics industry was under a shared delusion that we could make desktop/mobile more like consoles. In particular, that we could get to shader pre-compilation nirvana, or at least pretty close.