[clang] [CIR] Address Space support for GlobalOps (PR #179082)

[Andy Kaylor via cfe-commits]

================
@@ -0,0 +1,30 @@
+// RUN: %clang_cc1 -triple nvptx64-nvidia-cuda -x cuda \
+// RUN:   -fcuda-is-device -emit-cir %s -o %t.cir
+// RUN: FileCheck --input-file=%t.cir %s
+
+// Verifies CIR emits correct address spaces for CUDA globals.
+
+#include "Inputs/cuda.h"
+
+// CHECK: cir.global external  lang_address_space(offload_global) @i = #cir.int<0> : !s32i
----------------
andykaylor wrote:

I'd like to see this lowered to LLVM IR.

https://github.com/llvm/llvm-project/pull/179082