[Mlir-commits] [mlir] [mlir] Let GPU ID bounds work on any FunctionOpInterfaces (PR #95166)

[Krzysztof Drewniak]

krzysz00 wrote:

During the conversion, the surrounding context could either be the source function (say, `gpu.func`) or the target function (say `llvm.func`) depending on if the target code gets the surrounding function or its body gets rewritten first, hence the original conversion code having cases for both GPU and LLVM's func ops, which this change generalizes

The known sizes also aren't inherent attributes of `gpu.func` either - see https://github.com/llvm/llvm-project/blob/main/mlir/include/mlir/Dialect/GPU/IR/GPUOps.td#L434 , which goes to a discardable tag

https://github.com/llvm/llvm-project/pull/95166