[clang] [llvm] [mlir] [NVPTX] Add support for Shared Cluster Memory address space. (PR #135444)

[Durgadoss R via llvm-commits]

================
@@ -43,7 +43,10 @@ enum NVVMMemorySpace {
   /// Tensor memory space identifier.
   /// Tensor memory is available only in arch-accelerated
   /// variants from sm100 onwards.
-  kTensorMemorySpace = 6
+  kTensorMemorySpace = 6,
+  /// Distributed shared memory space identifier.
+  /// Distributed shared memory is available only in sm80+.
----------------
durga4github wrote:

We have sm_86, sm_87 etc. that do not support dsmem.
So, let us say "sm 90 onwards" here.

https://docs.nvidia.com/cuda/parallel-thread-execution/index.html#memory-hierarchy


https://github.com/llvm/llvm-project/pull/135444