[clang] [llvm] [mlir] [AMDGPU] Add the support for 45-bit buffer resource (PR #159702)

[Shilei Tian via llvm-commits]

================
@@ -1431,7 +1431,7 @@ def int_amdgcn_make_buffer_rsrc : DefaultAttrsIntrinsic <
   [llvm_anyptr_ty],
   [llvm_anyptr_ty, // base
    llvm_i16_ty,    // stride (and swizzle control)
-   llvm_i32_ty,    // NumRecords / extent
+   llvm_i64_ty,    // NumRecords / extent
----------------
shiltian wrote:

There's no real benefit to making it variadic, because:

1. From a backward compatibility view, even with the variadic version, the user code still has to be updated to use the new intrinsic signature.
2. From a semantics view, we don't have front-end like sema check to enforce types for the targets. So we could still end up in the same situation.

https://github.com/llvm/llvm-project/pull/159702