diff --git a/taichi/aot/graph_data.cpp b/taichi/aot/graph_data.cpp
index 1ebefde10cbde..2cb9eafcd2acc 100644
--- a/taichi/aot/graph_data.cpp
+++ b/taichi/aot/graph_data.cpp
@@ -23,7 +23,8 @@ void CompiledGraph::run(
       TI_ERROR_IF(found == args.end(), "Missing runtime value for {}",
                   symbolic_arg.name);
       const aot::IValue &ival = found->second;
-      if (ival.tag == aot::ArgKind::kNdarray) {
+      if (symbolic_arg.tag == aot::ArgKind::kNdarray) {
+        TI_ASSERT(ival.tag == aot::ArgKind::kNdarray);
         Ndarray *arr = reinterpret_cast<Ndarray *>(ival.val);
 
         TI_ERROR_IF(arr->get_element_shape() != symbolic_arg.element_shape,
@@ -65,12 +66,17 @@ void CompiledGraph::run(
                     arr_primitive_dtype.to_string());
         ctx.set_arg_ndarray(i, arr->get_device_allocation_ptr_as_int(),
                             arr->shape);
-      } else if (ival.tag == aot::ArgKind::kScalar) {
+      } else if (symbolic_arg.tag == aot::ArgKind::kScalar ||
+                 symbolic_arg.tag == aot::ArgKind::kMatrix) {
+        TI_ASSERT(ival.tag == aot::ArgKind::kScalar);
+        // Matrix args are flattened so they're same as scalars.
         ctx.set_arg(i, ival.val);
-      } else if (ival.tag == aot::ArgKind::kTexture) {
+      } else if (symbolic_arg.tag == aot::ArgKind::kTexture) {
+        TI_ASSERT(ival.tag == aot::ArgKind::kTexture);
         Texture *tex = reinterpret_cast<Texture *>(ival.val);
         ctx.set_arg_texture(i, tex->get_device_allocation_ptr_as_int());
-      } else if (ival.tag == aot::ArgKind::kRWTexture) {
+      } else if (symbolic_arg.tag == aot::ArgKind::kRWTexture) {
+        TI_ASSERT(ival.tag == aot::ArgKind::kTexture);
         Texture *tex = reinterpret_cast<Texture *>(ival.val);
         ctx.set_arg_rw_texture(i, tex->get_device_allocation_ptr_as_int(),
                                tex->get_size());
diff --git a/tests/python/test_graph.py b/tests/python/test_graph.py
index dad64359702a1..1cd653e4a7133 100644
--- a/tests/python/test_graph.py
+++ b/tests/python/test_graph.py
@@ -448,3 +448,41 @@ def test_kernel(arr: ti.types.ndarray()):
     g = g_builder.compile()
     g.run({'arr': arr})
     assert arr.to_numpy() == 1
+
+
+@test_utils.test(arch=[ti.vulkan])
+def test_texture_struct_for():
+    res = (128, 128)
+    tex = ti.Texture(ti.Format.r32f, res)
+    arr = ti.ndarray(ti.f32, res)
+
+    @ti.kernel
+    def write(tex: ti.types.rw_texture(num_dimensions=2,
+                                       num_channels=1,
+                                       channel_format=ti.f32,
+                                       lod=0)):
+        for i, j in tex:
+            tex.store(ti.Vector([i, j]), ti.Vector([1.0, 0.0, 0.0, 0.0]))
+
+    @ti.kernel
+    def read(tex: ti.types.texture(num_dimensions=2), arr: ti.types.ndarray()):
+        for i, j in arr:
+            arr[i, j] = tex.fetch(ti.Vector([i, j]), 0).x
+
+    sym_tex = ti.graph.Arg(ti.graph.ArgKind.RWTEXTURE,
+                           'tex',
+                           channel_format=ti.f32,
+                           shape=res,
+                           num_channels=1)
+    sym_arr = ti.graph.Arg(ti.graph.ArgKind.NDARRAY,
+                           'arr',
+                           ti.f32,
+                           field_dim=2)
+
+    gb = ti.graph.GraphBuilder()
+    gb.dispatch(write, sym_tex)
+    gb.dispatch(read, sym_tex, sym_arr)
+    graph = gb.compile()
+
+    graph.run({'tex': tex, 'arr': arr})
+    assert arr.to_numpy().sum() == 128 * 128