[Inductor] Fallback scatter when src dtype is bf16 (#113204)

basic_gnn_gcn, basic_gnn_gin, basic_gnn_sage now pass

Pull Request resolved: #113204
Approved by: https://github.com/eellison

test/inductor/test_torchinductor.py

@@ -667,6 +667,20 @@ def fn(a):
 
         self.common(fn, [torch.linspace(-10, 10, 41)])
 
+    def test_scatter_bf16(self):
+        def fn(inp, src, index):
+            return inp.scatter_add(0, index, src)
+
+        for dtype in [torch.int64, torch.bool, torch.bfloat16]:
+            self.common(
+                fn,
+                [
+                    torch.zeros(3, 5, dtype=dtype),
+                    torch.ones((2, 5), dtype=dtype),
+                    torch.tensor([[0, 1, 2, 0, 0]]),
+                ],
+            )
+
     def test_randn_generator(self):
         def fn(a, generator):
             torch.randn([20, 20], generator=generator, device=a.device)

torch/_inductor/lowering.py

@@ -2927,6 +2927,11 @@ def _unsafe_index_put_(self, indices, values, accumulate=False):
     return index_put_impl_(self, indices, values, accumulate, check=False)
 
 
+def needs_fallback_due_to_atomic_add_limitations(dtype):
+    # tl.atomic_add does NOT support the following types
+    return dtype in {torch.int64, torch.bool, torch.bfloat16}
+
+
 def index_put_impl_(self, indices, values, accumulate, check):
     # Dispatch to masked fill for single boolean index with single value
     if (
@@ -2951,8 +2956,7 @@ def index_put_impl_(self, indices, values, accumulate, check):
     x_size = self.get_size()
     x_ndim = len(x_size)
 
-    # fallback to aten.index_put_, as tl.atomic_add does NOT support int64 or bool
-    if self.get_dtype() in {torch.int64, torch.bool}:
+    if needs_fallback_due_to_atomic_add_limitations(self.get_dtype()):
         # self is an scalar Tensor
         if x_ndim == 0:
             self = view(self, [1])
@@ -3080,6 +3084,10 @@ def scatter_fallback(
     reduce_ty = "add" if fn == "aten.scatter_" else "sum"
     if (
         reduce not in {None, reduce_ty}
+        or (
+            isinstance(src, TensorBox)
+            and needs_fallback_due_to_atomic_add_limitations(src.get_dtype())
+        )
         or (
             fn == "aten.scatter_reduce_"
             and reduce == "sum"