[clang] [llvm] [mlir] [AMDGPU] Add a new amdgcn.load.to.lds intrinsic (PR #137425)

[Krzysztof Drewniak via cfe-commits]

================
@@ -564,6 +564,11 @@ Value *CodeGenFunction::EmitAMDGPUBuiltinExpr(unsigned BuiltinID,
     llvm::Function *F = CGM.getIntrinsic(IID, {LoadTy});
     return Builder.CreateCall(F, {Addr});
   }
+  case AMDGPU::BI__builtin_amdgcn_load_to_lds: {
+    // Should this have asan instrumentation?
----------------
krzysz00 wrote:

It might be a backend problem

However, https://github.com/llvm/llvm-project/blob/a9ce60ee885e25cba54c8900605563ff66e067b7/clang/lib/CodeGen/CGBuiltin.cpp#L4142 is the handling for memcpy(), and doesn't seem to rely on the backend handling it. I don't know if this same handling should be happening for the intrinsic

https://github.com/llvm/llvm-project/pull/137425