Benoit Jacob

Thanks! The warning message could helpfully dump a list of accepted CPU values for the target architecture. At that point, passing the missing CPU flag would be so easy that...

I was going to advocate for least-surprisal as the overriding principle here, so I dug what compilers are doing, and unfortunately it's fragmented: - C/C++ compilers are of course defaulting...

Oh yes, just make the flag required, with diagnostic, and with helpful messages - so if people would like `host`, they can just copy and paste a flag from the...

@ScottTodd, @benvanik, @stellaraccident, here is what I have at the moment: https://github.com/iree-org/iree/pull/18682 . It folds some good ideas from this PR (the general idea, and defaulting CPU to `""` instead...

I have a MLIR-based compiler (IREE) that needs to perform various rewrites, including constant-evalutation / folding, ahead of converting to WGSL or SPIR-V or any other target that this compiler...

I guess both. But the former is a simpler case to think about. Suppose the source MLIR program contains ops like ```mlir %0 = arith.constant 0.0 : f32 %1 =...

> you can't really write INF or NaN in WGSL. That is exactly the issue that I meant to complain about, then. I am saying that a language that doesn't...

Thanks for sharing the meeting minutes. Not 100% sure I parse them correctly but it sounds like the point of divergence is where you suggest it's OK to "flush infinity...

@nirvedhmeshram , being on LLVM integration this week, I'm available (and interested) in testing any work-in-progress patch that you might have, and/or step in to help as needed.

Awesome progress @Yu-Zhewen ! FYI @sebvince .