[Mlir-commits] [mlir] [mlir][gpu] Support Cluster of Thread Blocks in `gpu.launch_func` (PR #72871)

[Guray Ozen]

================
@@ -550,12 +601,23 @@ def GPU_LaunchFuncOp :GPU_Op<"launch_func", [
     /// The name of the kernel.
     StringAttr getKernelName();
 
+    /// Returns true if cluster size is specified.
+    bool hasClusterSize() {
+      if (getClusterSizeX() && getClusterSizeY() && getClusterSizeZ())
----------------
grypp wrote:

All dimensions must be explicitly defined just like block or grid dimensions.
(We could implicitly the non-used dimensions set 1)

https://github.com/llvm/llvm-project/pull/72871