Nathan V. Morrical

@csyonghe What's the correct approach to constructing vectors for the CUDA backend? bitfieldExtract is a component-wise intrinsic, so I need to support a function like this as a right hand...

For the time being, looks like a special case lookup table works for CUDA.

On adding these tests, I noticed that the current logic in glsl.meta.slang for bitfield extraction in the signed integer case was incorrect. For signed integers, the extracted bits are expected...

Looks like there are some newly failing tests on my end after adding these flags for the other targets. I'll see if I can figure out what's going on. For...

I'm a bit confused with the Vulkan spec... I think I might have found a typo?... According to the spec: > The Base operand of any OpBitCount, OpBitReverse, **OpBitFieldInsert**, **OpBitFieldSExtract**,...

It appears the two outstanding issues are: 1. When `SLANG_RUN_SPIRV_VALIDATION=1`, SPIR-V validation does not correctly consider how `SpvCapabilityInt64` extends the capabilities of `OpBitfieldSExtract`/`OpBitfieldUExtract`/`OpBitfieldInsert` intrinsics. This is causing the one windows...

> You can add -skip-spirv-validation in the //TEST line to disable validation for that specific test. You can search for it in the codebase to see how it is used...

> For 2, this seems like a bug with the current Slang Metal backend, and isn't directly tied to my changes as far as I can tell. > > @csyonghe...

@csyonghe looks like all relevant tests are passing. There’s one windows CI machine which is reporting that bash is missing, but I see the same error on ToT, so I’m...

Apologies on the delay here. Been a bit tied up with a move between states. I should have a bit more time to work on this in the upcoming week.