[llvm] [AMDGPU] Add intrinsics and MIs for ds_bvh_stack_* (PR #130007)

[Jay Foad via llvm-commits]

================
@@ -2801,6 +2803,32 @@ def int_amdgcn_wmma_bf16_16x16x16_bf16 : AMDGPUWmmaIntrinsicOPSEL<llvm_anyint_ty
 // GFX12 Intrinsics
 //===----------------------------------------------------------------------===//
 
+def int_amdgcn_ds_bvh_stack_push4_pop1_rtn : IntDSBVHStackRtn;
+
+def int_amdgcn_ds_bvh_stack_push8_pop1_rtn :
+  Intrinsic<
+    [llvm_i32_ty, llvm_i32_ty], // %vdst, %addr
+    [
+      llvm_i32_ty,   // %addr
+      llvm_i32_ty,   // %data0
+      llvm_v8i32_ty, // %data1
+      llvm_i32_ty,   // %offset
+    ],
+    [ImmArg<ArgIndex<3>>, IntrWillReturn]
----------------
jayfoad wrote:

Should at least have `IntrNoCallback, IntrNoFree` to match `IntDSBVHStackRtn`.

https://github.com/llvm/llvm-project/pull/130007