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

[Mehdi Amini]

================
@@ -73,12 +85,16 @@ static std::optional<uint64_t> getKnownLaunchDim(Op op, LaunchDims type) {
       return value.getZExtValue();
   }
 
-  if (auto func = op->template getParentOfType<GPUFuncOp>()) {
+  if (auto func = op->template getParentOfType<FunctionOpInterface>()) {
     switch (type) {
     case LaunchDims::Block:
-      return llvm::transformOptional(func.getKnownBlockSize(dim), zext);
+      return llvm::transformOptional(
+          getKnownLaunchAttr(func, GPUFuncOp::getKnownBlockSizeAttrName(), dim),
+          zext);
     case LaunchDims::Grid:
-      return llvm::transformOptional(func.getKnownGridSize(dim), zext);
+      return llvm::transformOptional(
+          getKnownLaunchAttr(func, GPUFuncOp::getKnownGridSizeAttrName(), dim),
+          zext);
----------------
joker-eph wrote:

> I think that functions have, in practice, stronger semantics re discardability, similarly to how modules do.

That's a possible view, I'm on the side that we should define proper inherent attributes for what we intend to support instead (including DLTI for example).

Anyway: stepping back we still need to get back at the intended flow of GPU compilation: in particular whether GPU-specific features should be tied to specific containers or not. 

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