[Mlir-commits] [mlir] [MLIR][GPU] Support grid constant, byval, byref on gpu.func (PR #172037)

[Asher Mancinelli]

================
@@ -369,7 +370,10 @@ GPUFuncOpLowering::matchAndRewrite(gpu::GPUFuncOp gpuFuncOp, OpAdaptor adaptor,
     }
 
     if (lowersToPointer) {
+      copyPointerAttribute(mlir::NVVM::NVVMDialect::getGridConstantAttrName());
----------------
ashermancinelli wrote:

Would it be acceptable to copy over all the argument attrs on llvm.ptrs? In that case, we wouldn't be looking for attributes from a specific dialect (which I incorrectly did in this PR), and all the attrs set by the user would still be passed along. That should be good enough for my current needs. In this case, would only attributes valid in the LLVM/NVVM dialects would be valid?

https://github.com/llvm/llvm-project/pull/172037