Tempted to switch to https://github.com/shader-slang/slang for compute shader language. It has some very nice features but I'm a bit worried that I might hit some limits with no workaround (for instance, it doesn't look like they support pointers into groupshared memory)

Anyone has experience with it? Thoughts?

GitHub - shader-slang/slang: Making it easier to work with shaders

Making it easier to work with shaders. Contribute to shader-slang/slang development by creating an account on GitHub.

GitHub

@xoofx I don't have experience, but positive signs are that it is in active development (https://github.com/shader-slang/slang/releases), and is now the "official future language" of Khronos et al (compared to GLSL which is effectively frozen)

[upd: it is not “official language”, just one of Khronos projects]

Releases · shader-slang/slang

Making it easier to work with shaders. Contribute to shader-slang/slang development by creating an account on GitHub.

GitHub

@aras @xoofx

Sweet Jesus no it is not.

Slang is NOT the official future language of Khronos.

I told them again and again people would assume that.

None of this is your fault, dear friends and readers.

Hey @thekhronosgroup take note.

@aras @xoofx @thekhronosgroup

Slang is under active development, yes. NVIDIA has been actively working on it for years and gotten great mileage out of that. Great!
The compiler has been open source for a while.

What has changed is that NVIDIA wanted to open up governance of the language. Instead of making a new corporate body (like a Rust Foundation setup), they took it to Khronos. It's about a having a corporate body to absorb IP and other legal risks when corporations collaborate.

@aras @xoofx @thekhronosgroup

In no way is this to be interpreted as Slang being a blessed or preferred language for Vulkan or any other Khronos related property.

Excuse me if I'm shouting but I suffered months of meetings making this point and shaping the charter of the Slang Working Group at Khronos.

Neither Google or I participate in the working group. I wish them well but it's not where I can spend energy.

@dneto @aras @xoofx @thekhronosgroup can confirm. I was also in those months of meetings.
@beanz @dneto @xoofx @thekhronosgroup ohh. well. as you can see, simpletons like me indeed took that announcement as “this is official language now”. communication is hard!
@aras @beanz @dneto @xoofx @thekhronosgroup It's an official language, just not _the designated successor to GLSL_ or similar. One more language in a pool of languages.
@anteru @aras @beanz @dneto @thekhronosgroup thanks folks, no worries, didn't want to trigger bad feelings about this! 😅
I'm going to rely on SPIR-V for my experiment. Seems to be the "perfect" intermediate vehicle these days to target GPU workloads, and will prototype higher level constructs on top of C# instead.
@xoofx @aras @beanz @dneto @thekhronosgroup SPIR-V is fine, D3D will also target it natively, and it's definitely one of the nicer IRs to work with (especially if you want to instrument things and what not.) At some point I wrote a C# disassembler for SPIR-V and it's mostly auto-generated and still works, which shows how well designed the IR is.
@anteru @xoofx @aras @dneto @thekhronosgroup SPIRV is by far the friendliest compilation target for GPUs today. I love me some LLVM, but All the LLVM-based GPU interchange formats are a lot more difficult to work with.