[clang] [llvm] [mlir] [OpenMP] Reserve the main thread's warp for generic mode kernels (PR #218790)

[Joseph Huber via cfe-commits]

https://github.com/jhuber6 commented:

Okay, I get it now. We do the runtime widening but the issue was the codegen writing a value to the maximum threads attributes that prevented us from widening it. Seems like this has a really simple solution.

https://github.com/llvm/llvm-project/pull/218790