Revert auto tiling merge and more to_gpu tests

tests/transformations/gpu_transform_test.py

@@ -119,9 +119,60 @@ def write_subset_dynamic(A: dace.int32[20, 20], x: dace.int32[20], y: dace.int32
     assert np.array_equal(ref, val)
 
 
+@pytest.mark.parametrize("transient", [False, True])
+def test_free_tasklet_and_array(transient):
+    sdfg = dace.SDFG("assign")
+
+    state = sdfg.add_state("main")
+    arr_name, arr = sdfg.add_array("A", (4,), dace.float32, transient=transient)
+    an = state.add_access(arr_name)
+
+    t = state.add_tasklet("assign", {}, {"_out"}, "_out = 2.0")
+    state.add_edge(t, "_out", an, None, dace.memlet.Memlet("A[0]"))
+
+    sdfg.validate()
+
+    sdfg.apply_gpu_transformations(
+        validate = True,
+        validate_all = True,
+        permissive = True,
+        sequential_innermaps=True,
+        register_transients=False,
+        simplify=False
+    )
+
+    sdfg.validate()
+
+@pytest.mark.parametrize("transient", [False, True])
+def test_free_tasklet_and_scalar(transient):
+    sdfg = dace.SDFG("assign")
+
+    state = sdfg.add_state("main")
+    arr_name, arr = sdfg.add_scalar("A", dace.float32, transient=transient)
+    an = state.add_access(arr_name)
+
+    t = state.add_tasklet("assign", {}, {"_out"}, "_out = 2.0")
+    state.add_edge(t, "_out", an, None, dace.memlet.Memlet("A"))
+
+    sdfg.validate()
+
+    sdfg.apply_gpu_transformations(
+        validate = True,
+        validate_all = True,
+        permissive = True,
+        sequential_innermaps=True,
+        register_transients=False,
+        simplify=False
+    )
+
+    sdfg.validate()
+
 if __name__ == '__main__':
     test_toplevel_transient_lifetime()
     test_scalar_to_symbol_in_nested_sdfg()
     test_write_subset()
     test_write_full()
     test_write_subset_dynamic()
+    for transient in [False, True]:
+        test_free_tasklet_and_array(transient)
+        test_free_tasklet_and_array(transient)