gstaichi 0.1.21.dev0__cp310-cp310-win_amd64.whl → 0.1.25.dev0__cp310-cp310-win_amd64.whl

taichi/_lib/core/taichi_python.pyi → gstaichi/_lib/core/gstaichi_python.pyi

@@ -1,10 +1,10 @@
 """
-taichi_python
+gstaichi_python
 """
 from __future__ import annotations
 import numpy
 import typing
-__all__ = ['ADJOINT', 'ADJOINT_CHECKBIT', 'AOS', 'ASTBuilder', 'AotModuleBuilder', 'Arch', 'Arg', 'ArgKind', 'ArgPack', 'AutodiffMode', 'Axis', 'Benchmark', 'BinaryOpType', 'BitStructType', 'BitStructTypeBuilder', 'BoundaryMode', 'CC', 'CE', 'CF', 'CGd', 'CGf', 'CLAMP', 'CUCG', 'CV', 'Cell', 'CompileConfig', 'CompiledGraph', 'CompiledKernelData', 'Config', 'ConvType', 'CuSparseMatrix', 'CuSparseSolver', 'DUAL', 'DataType', 'DataType_f16', 'DataType_f32', 'DataType_f64', 'DataType_gen', 'DataType_i16', 'DataType_i32', 'DataType_i64', 'DataType_i8', 'DataType_u1', 'DataType_u16', 'DataType_u32', 'DataType_u64', 'DataType_u8', 'DataType_unknown', 'DebugInfo', 'DeviceAllocation', 'DeviceCapabilityConfig', 'EC', 'EE', 'EF', 'EV', 'Edge', 'EigenSparseSolverfloat32LDLTAMD', 'EigenSparseSolverfloat32LDLTCOLAMD', 'EigenSparseSolverfloat32LLTAMD', 'EigenSparseSolverfloat32LLTCOLAMD', 'EigenSparseSolverfloat32LUAMD', 'EigenSparseSolverfloat32LUCOLAMD', 'EigenSparseSolverfloat64LDLTAMD', 'EigenSparseSolverfloat64LDLTCOLAMD', 'EigenSparseSolverfloat64LLTAMD', 'EigenSparseSolverfloat64LLTCOLAMD', 'EigenSparseSolverfloat64LUAMD', 'EigenSparseSolverfloat64LUCOLAMD', 'Expr', 'ExprGroup', 'Extension', 'FC', 'FE', 'FF', 'FORWARD', 'FV', 'Face', 'Format', 'Function', 'FunctionKey', 'GraphBuilder', 'HackedSignalRegister', 'InternalOp', 'Kernel', 'KernelLaunchContext', 'KernelProfileTracedRecord', 'KernelProfilerQueryResult', 'Layout', 'MATRIX', 'Mesh', 'MeshElementType', 'MeshPtr', 'MeshRelationType', 'MeshTopology', 'NDARRAY', 'NONE', 'NULL', 'Ndarray', 'Node', 'Operation', 'PRIMAL', 'Program', 'REVERSE', 'RWTEXTURE', 'SCALAR', 'SNode', 'SNodeAccessFlag', 'SNodeGradType', 'SNodeRegistry', 'SNodeTree', 'SNodeType', 'SOA', 'Sequential', 'SparseMatrix', 'SparseMatrixBuilder', 'SparseSolver', 'Stmt', 'TEXTURE', 'TaichiAssertionError', 'TaichiIndexError', 'TaichiRuntimeError', 'TaichiSyntaxError', 'TaichiTypeError', 'Task', 'Tetrahedron', 'Texture', 'TextureOpType', 'Triangle', 'Type', 'TypeFactory', 'UNSAFE', 'UnaryOpType', 'VALIDATION', 'VC', 'VE', 'VF', 'VV', 'Vector2d', 'Vector2f', 'Vector2i', 'Vector3d', 'Vector3f', 'Vector3i', 'Vector4d', 'Vector4f', 'Vector4i', 'Vertex', 'abs', 'acos', 'add', 'adstack', 'amdgpu', 'arch_from_name', 'arch_name', 'arch_uses_llvm', 'arm64', 'asin', 'assertion', 'atan2', 'bit_and', 'bit_not', 'bit_or', 'bit_sar', 'bit_shl', 'bit_shr', 'bit_struct', 'bit_xor', 'bitmasked', 'bits_cast', 'block_local', 'bls', 'cast_bits', 'cast_value', 'ceil', 'clean_offline_cache_files', 'clear_profile_info', 'clz', 'cmp_eq', 'cmp_ge', 'cmp_gt', 'cmp_le', 'cmp_lt', 'cmp_ne', 'cos', 'create_benchmark', 'create_initialized_benchmark', 'create_initialized_task', 'create_mesh', 'create_task', 'critical', 'cuda', 'cuda_version', 'dColMajor_EigenSparseMatrix', 'dRowMajor_EigenSparseMatrix', 'data64', 'data_type_name', 'data_type_size', 'debug', 'default_compile_config', 'dense', 'div', 'dx11', 'dx12', 'dynamic', 'element_order', 'element_type_name', 'error', 'exp', 'expr_abs', 'expr_acos', 'expr_add', 'expr_asin', 'expr_assume_in_range', 'expr_atan2', 'expr_atomic_add', 'expr_atomic_bit_and', 'expr_atomic_bit_or', 'expr_atomic_bit_xor', 'expr_atomic_max', 'expr_atomic_min', 'expr_atomic_mul', 'expr_atomic_sub', 'expr_bit_and', 'expr_bit_not', 'expr_bit_or', 'expr_bit_sar', 'expr_bit_shl', 'expr_bit_shr', 'expr_bit_xor', 'expr_ceil', 'expr_clz', 'expr_cmp_eq', 'expr_cmp_ge', 'expr_cmp_gt', 'expr_cmp_le', 'expr_cmp_lt', 'expr_cmp_ne', 'expr_cos', 'expr_div', 'expr_exp', 'expr_field', 'expr_floor', 'expr_floordiv', 'expr_frexp', 'expr_ifte', 'expr_inv', 'expr_log', 'expr_logic_not', 'expr_logical_and', 'expr_logical_or', 'expr_loop_unique', 'expr_matrix_field', 'expr_max', 'expr_min', 'expr_mod', 'expr_mul', 'expr_neg', 'expr_popcnt', 'expr_pow', 'expr_rcp', 'expr_round', 'expr_rsqrt', 'expr_select', 'expr_sin', 'expr_sqrt', 'expr_sub', 'expr_tan', 'expr_tanh', 'expr_truediv', 'extfunc', 'fColMajor_EigenSparseMatrix', 'fRowMajor_EigenSparseMatrix', 'finalize_snode_tree', 'floor', 'floordiv', 'flush_log', 'frexp', 'from_end_element_order', 'g2r', 'get_commit_hash', 'get_default_float_size', 'get_external_tensor_dim', 'get_external_tensor_element_dim', 'get_external_tensor_element_shape', 'get_external_tensor_element_type', 'get_external_tensor_needs_grad', 'get_external_tensor_real_func_args', 'get_external_tensor_shape_along_axis', 'get_llvm_target_support', 'get_max_num_args', 'get_max_num_indices', 'get_num_elements', 'get_python_package_dir', 'get_relation_access', 'get_relation_size', 'get_repo_dir', 'get_type_factory_instance', 'get_version_major', 'get_version_minor', 'get_version_patch', 'get_version_string', 'gles', 'hash', 'host_arch', 'info', 'insert_internal_func_call', 'inv', 'inverse_relation', 'is_extension_supported', 'is_integral', 'is_quant', 'is_real', 'is_signed', 'is_tensor', 'is_unsigned', 'js', 'kFetchTexel', 'kLoad', 'kSampleLod', 'kStore', 'kUndefined', 'l2g', 'l2r', 'libdevice_path', 'log', 'logging_effective', 'logic_not', 'make_arg_load_expr', 'make_binary_op_expr', 'make_const_expr_bool', 'make_const_expr_fp', 'make_const_expr_int', 'make_cucg_solver', 'make_cusparse_solver', 'make_double_cg_solver', 'make_external_tensor_expr', 'make_external_tensor_grad_expr', 'make_float_cg_solver', 'make_frontend_assign_stmt', 'make_get_element_expr', 'make_global_load_stmt', 'make_global_store_stmt', 'make_rand_expr', 'make_reference', 'make_rw_texture_ptr_expr', 'make_sparse_solver', 'make_texture_ptr_expr', 'make_unary_op_expr', 'max', 'mesh', 'mesh_local', 'metal', 'min', 'mod', 'mul', 'neg', 'opencl', 'opengl', 'place', 'pointer', 'pop_python_print_buffer', 'popcnt', 'pow', 'print_all_units', 'print_profile_info', 'promoted_type', 'quant', 'quant_array', 'quant_basic', 'query_int64', 'rcp', 'read_only', 'relation_by_orders', 'reset_default_compile_config', 'root', 'round', 'rsqrt', 'set_core_state_python_imported', 'set_core_trigger_gdb_when_crash', 'set_index_mapping', 'set_lib_dir', 'set_logging_level', 'set_logging_level_default', 'set_num_elements', 'set_num_patches', 'set_owned_offset', 'set_patch_max_element_num', 'set_python_package_dir', 'set_relation_dynamic', 'set_relation_fixed', 'set_tmp_dir', 'set_total_offset', 'set_vulkan_visible_device', 'sgn', 'sin', 'sparse', 'sqrt', 'start_memory_monitoring', 'sub', 'subscript_with_multiple_indices', 'tan', 'tanh', 'test_cpp_exception', 'test_logging', 'test_printf', 'test_raise_error', 'test_threading', 'test_throw', 'to_end_element_order', 'toggle_python_print_buffer', 'trace', 'trigger_crash', 'trigger_sig_fpe', 'truediv', 'undefined', 'value_cast', 'vulkan', 'wait_for_debugger', 'warn', 'with_amdgpu', 'with_cuda', 'with_dx11', 'with_dx12', 'with_metal', 'with_opengl', 'with_vulkan', 'x64']
+__all__ = ['ADJOINT', 'ADJOINT_CHECKBIT', 'AOS', 'ASTBuilder', 'Arch', 'ArgPackCxx', 'AutodiffMode', 'Axis', 'Benchmark', 'BinaryOpType', 'BitStructType', 'BitStructTypeBuilder', 'BoundaryMode', 'CC', 'CE', 'CF', 'CGd', 'CGf', 'CLAMP', 'CUCG', 'CV', 'Cell', 'CompileConfig', 'CompiledKernelData', 'Config', 'ConvType', 'CuSparseMatrix', 'CuSparseSolver', 'DUAL', 'DataTypeCxx', 'DataType_f16', 'DataType_f32', 'DataType_f64', 'DataType_gen', 'DataType_i16', 'DataType_i32', 'DataType_i64', 'DataType_i8', 'DataType_u1', 'DataType_u16', 'DataType_u32', 'DataType_u64', 'DataType_u8', 'DataType_unknown', 'DebugInfo', 'DeviceAllocation', 'DeviceCapabilityConfig', 'EC', 'EE', 'EF', 'EV', 'Edge', 'EigenSparseSolverfloat32LDLTAMD', 'EigenSparseSolverfloat32LDLTCOLAMD', 'EigenSparseSolverfloat32LLTAMD', 'EigenSparseSolverfloat32LLTCOLAMD', 'EigenSparseSolverfloat32LUAMD', 'EigenSparseSolverfloat32LUCOLAMD', 'EigenSparseSolverfloat64LDLTAMD', 'EigenSparseSolverfloat64LDLTCOLAMD', 'EigenSparseSolverfloat64LLTAMD', 'EigenSparseSolverfloat64LLTCOLAMD', 'EigenSparseSolverfloat64LUAMD', 'EigenSparseSolverfloat64LUCOLAMD', 'ExprCxx', 'ExprGroup', 'Extension', 'FC', 'FE', 'FF', 'FORWARD', 'FV', 'Face', 'Format', 'Function', 'FunctionKey', 'GsTaichiAssertionError', 'GsTaichiIndexError', 'GsTaichiRuntimeError', 'GsTaichiSyntaxError', 'GsTaichiTypeError', 'HackedSignalRegister', 'InternalOp', 'KernelCxx', 'KernelLaunchContext', 'KernelProfileTracedRecord', 'KernelProfilerQueryResult', 'Layout', 'Mesh', 'MeshElementType', 'MeshPtr', 'MeshRelationType', 'MeshTopology', 'NONE', 'NULL', 'NdarrayCxx', 'Operation', 'PRIMAL', 'Program', 'REVERSE', 'SNodeAccessFlag', 'SNodeCxx', 'SNodeGradType', 'SNodeRegistry', 'SNodeTreeCxx', 'SNodeType', 'SOA', 'SparseMatrix', 'SparseMatrixBuilder', 'SparseSolver', 'Stmt', 'Task', 'Tetrahedron', 'TextureCxx', 'TextureOpType', 'Triangle', 'Type', 'TypeFactory', 'UNSAFE', 'UnaryOpType', 'VALIDATION', 'VC', 'VE', 'VF', 'VV', 'Vector2d', 'Vector2f', 'Vector2i', 'Vector3d', 'Vector3f', 'Vector3i', 'Vector4d', 'Vector4f', 'Vector4i', 'Vertex', 'abs', 'acos', 'add', 'adstack', 'amdgpu', 'arch_from_name', 'arch_name', 'arch_uses_llvm', 'arm64', 'asin', 'assertion', 'atan2', 'bit_and', 'bit_not', 'bit_or', 'bit_sar', 'bit_shl', 'bit_shr', 'bit_struct', 'bit_xor', 'bitmasked', 'bits_cast', 'block_local', 'bls', 'cast_bits', 'cast_value', 'ceil', 'clean_offline_cache_files', 'clear_profile_info', 'clz', 'cmp_eq', 'cmp_ge', 'cmp_gt', 'cmp_le', 'cmp_lt', 'cmp_ne', 'cos', 'create_benchmark', 'create_initialized_benchmark', 'create_initialized_task', 'create_mesh', 'create_task', 'critical', 'cuda', 'cuda_version', 'dColMajor_EigenSparseMatrix', 'dRowMajor_EigenSparseMatrix', 'data64', 'data_type_name', 'data_type_size', 'debug', 'default_compile_config', 'dense', 'div', 'dynamic', 'element_order', 'element_type_name', 'error', 'exp', 'expr_abs', 'expr_acos', 'expr_add', 'expr_asin', 'expr_assume_in_range', 'expr_atan2', 'expr_atomic_add', 'expr_atomic_bit_and', 'expr_atomic_bit_or', 'expr_atomic_bit_xor', 'expr_atomic_max', 'expr_atomic_min', 'expr_atomic_mul', 'expr_atomic_sub', 'expr_bit_and', 'expr_bit_not', 'expr_bit_or', 'expr_bit_sar', 'expr_bit_shl', 'expr_bit_shr', 'expr_bit_xor', 'expr_ceil', 'expr_clz', 'expr_cmp_eq', 'expr_cmp_ge', 'expr_cmp_gt', 'expr_cmp_le', 'expr_cmp_lt', 'expr_cmp_ne', 'expr_cos', 'expr_div', 'expr_exp', 'expr_field', 'expr_floor', 'expr_floordiv', 'expr_frexp', 'expr_ifte', 'expr_inv', 'expr_log', 'expr_logic_not', 'expr_logical_and', 'expr_logical_or', 'expr_loop_unique', 'expr_matrix_field', 'expr_max', 'expr_min', 'expr_mod', 'expr_mul', 'expr_neg', 'expr_popcnt', 'expr_pow', 'expr_rcp', 'expr_round', 'expr_rsqrt', 'expr_select', 'expr_sin', 'expr_sqrt', 'expr_sub', 'expr_tan', 'expr_tanh', 'expr_truediv', 'extfunc', 'fColMajor_EigenSparseMatrix', 'fRowMajor_EigenSparseMatrix', 'finalize_snode_tree', 'floor', 'floordiv', 'flush_log', 'frexp', 'from_end_element_order', 'g2r', 'get_commit_hash', 'get_default_float_size', 'get_external_tensor_dim', 'get_external_tensor_element_dim', 'get_external_tensor_element_shape', 'get_external_tensor_element_type', 'get_external_tensor_needs_grad', 'get_external_tensor_real_func_args', 'get_external_tensor_shape_along_axis', 'get_llvm_target_support', 'get_max_num_indices', 'get_num_elements', 'get_python_package_dir', 'get_relation_access', 'get_relation_size', 'get_repo_dir', 'get_type_factory_instance', 'get_version_major', 'get_version_minor', 'get_version_patch', 'get_version_string', 'hash', 'host_arch', 'info', 'insert_internal_func_call', 'inv', 'inverse_relation', 'is_extension_supported', 'is_integral', 'is_quant', 'is_real', 'is_signed', 'is_tensor', 'is_unsigned', 'js', 'kFetchTexel', 'kLoad', 'kSampleLod', 'kStore', 'kUndefined', 'l2g', 'l2r', 'libdevice_path', 'log', 'logging_effective', 'logic_not', 'make_arg_load_expr', 'make_binary_op_expr', 'make_const_expr_bool', 'make_const_expr_fp', 'make_const_expr_int', 'make_cucg_solver', 'make_cusparse_solver', 'make_double_cg_solver', 'make_external_tensor_expr', 'make_external_tensor_grad_expr', 'make_float_cg_solver', 'make_frontend_assign_stmt', 'make_get_element_expr', 'make_global_load_stmt', 'make_global_store_stmt', 'make_rand_expr', 'make_reference', 'make_rw_texture_ptr_expr', 'make_sparse_solver', 'make_texture_ptr_expr', 'make_unary_op_expr', 'max', 'mesh', 'mesh_local', 'metal', 'min', 'mod', 'mul', 'neg', 'opencl', 'place', 'pointer', 'pop_python_print_buffer', 'popcnt', 'pow', 'print_all_units', 'print_profile_info', 'promoted_type', 'quant', 'quant_array', 'quant_basic', 'query_int64', 'rcp', 'read_only', 'relation_by_orders', 'reset_default_compile_config', 'root', 'round', 'rsqrt', 'set_core_state_python_imported', 'set_core_trigger_gdb_when_crash', 'set_index_mapping', 'set_lib_dir', 'set_logging_level', 'set_logging_level_default', 'set_num_elements', 'set_num_patches', 'set_owned_offset', 'set_patch_max_element_num', 'set_python_package_dir', 'set_relation_dynamic', 'set_relation_fixed', 'set_tmp_dir', 'set_total_offset', 'sgn', 'sin', 'sparse', 'sqrt', 'start_memory_monitoring', 'sub', 'subscript_with_multiple_indices', 'tan', 'tanh', 'test_cpp_exception', 'test_logging', 'test_printf', 'test_raise_error', 'test_threading', 'test_throw', 'to_end_element_order', 'toggle_python_print_buffer', 'trace', 'trigger_crash', 'trigger_sig_fpe', 'truediv', 'undefined', 'value_cast', 'vulkan', 'wait_for_debugger', 'warn', 'with_amdgpu', 'with_cuda', 'with_metal', 'with_vulkan', 'x64']
 class ASTBuilder:
     def begin_frontend_if(self, arg0: ..., arg1: DebugInfo) -> None:
         ...
@@ -46,7 +46,7 @@ class ASTBuilder:
         ...
     def expr_alloca(self, arg0: DebugInfo) -> ...:
         ...
-    def expr_alloca_shared_array(self, arg0: list[int], arg1: DataType, arg2: DebugInfo) -> ...:
+    def expr_alloca_shared_array(self, arg0: tuple[int, ...], arg1: DataTypeCxx, arg2: DebugInfo) -> ...:
         ...
     def expr_assign(self, arg0: ..., arg1: ..., arg2: DebugInfo) -> None:
         ...
@@ -84,7 +84,7 @@ class ASTBuilder:
         ...
     def make_id_expr(self, arg0: str) -> ...:
         ...
-    def make_matrix_expr(self, arg0: list[int], arg1: DataType, arg2: list[...], arg3: DebugInfo) -> ...:
+    def make_matrix_expr(self, arg0: tuple[int, ...], arg1: DataTypeCxx, arg2: list[...], arg3: DebugInfo) -> ...:
         ...
     def make_texture_op_expr(self, arg0: ..., arg1: ..., arg2: ..., arg3: DebugInfo) -> ...:
         ...
@@ -104,17 +104,6 @@ class ASTBuilder:
         ...
     def strictly_serialize(self) -> None:
         ...
-class AotModuleBuilder:
-    def add(self, arg0: str, arg1: ...) -> None:
-        ...
-    def add_field(self, arg0: str, arg1: ..., arg2: bool, arg3: DataType, arg4: list[int], arg5: int, arg6: int) -> None:
-        ...
-    def add_graph(self, arg0: str, arg1: ...) -> None:
-        ...
-    def add_kernel_template(self, arg0: str, arg1: str, arg2: ...) -> None:
-        ...
-    def dump(self, arg0: str, arg1: str) -> None:
-        ...
 class Arch:
     """
     Members:
@@ -129,32 +118,20 @@ class Arch:
     
       metal
     
-      opengl
-    
-      dx11
-    
-      dx12
-    
       opencl
     
       amdgpu
     
       vulkan
-    
-      gles
     """
-    __members__: typing.ClassVar[dict[str, Arch]]  # value = {'x64': <Arch.x64: 0>, 'arm64': <Arch.arm64: 1>, 'js': <Arch.js: 2>, 'cuda': <Arch.cuda: 3>, 'metal': <Arch.metal: 4>, 'opengl': <Arch.opengl: 5>, 'dx11': <Arch.dx11: 6>, 'dx12': <Arch.dx12: 7>, 'opencl': <Arch.opencl: 8>, 'amdgpu': <Arch.amdgpu: 9>, 'vulkan': <Arch.vulkan: 10>, 'gles': <Arch.gles: 11>}
-    amdgpu: typing.ClassVar[Arch]  # value = <Arch.amdgpu: 9>
+    __members__: typing.ClassVar[dict[str, Arch]]  # value = {'x64': <Arch.x64: 0>, 'arm64': <Arch.arm64: 1>, 'js': <Arch.js: 2>, 'cuda': <Arch.cuda: 3>, 'metal': <Arch.metal: 4>, 'opencl': <Arch.opencl: 5>, 'amdgpu': <Arch.amdgpu: 6>, 'vulkan': <Arch.vulkan: 7>}
+    amdgpu: typing.ClassVar[Arch]  # value = <Arch.amdgpu: 6>
     arm64: typing.ClassVar[Arch]  # value = <Arch.arm64: 1>
     cuda: typing.ClassVar[Arch]  # value = <Arch.cuda: 3>
-    dx11: typing.ClassVar[Arch]  # value = <Arch.dx11: 6>
-    dx12: typing.ClassVar[Arch]  # value = <Arch.dx12: 7>
-    gles: typing.ClassVar[Arch]  # value = <Arch.gles: 11>
     js: typing.ClassVar[Arch]  # value = <Arch.js: 2>
     metal: typing.ClassVar[Arch]  # value = <Arch.metal: 4>
-    opencl: typing.ClassVar[Arch]  # value = <Arch.opencl: 8>
-    opengl: typing.ClassVar[Arch]  # value = <Arch.opengl: 5>
-    vulkan: typing.ClassVar[Arch]  # value = <Arch.vulkan: 10>
+    opencl: typing.ClassVar[Arch]  # value = <Arch.opencl: 5>
+    vulkan: typing.ClassVar[Arch]  # value = <Arch.vulkan: 7>
     x64: typing.ClassVar[Arch]  # value = <Arch.x64: 0>
     def __eq__(self, other: typing.Any) -> bool:
         ...
@@ -190,80 +167,8 @@ class Arch:
     @property
     def value(self) -> int:
         ...
-class Arg:
-    @typing.overload
-    def __init__(self, tag: ArgKind, name: str, dtype: DataType, field_dim: int, element_shape: list[int]) -> None:
-        ...
-    @typing.overload
-    def __init__(self, tag: ArgKind, name: str, channel_format: DataType, num_channels: int, shape: list[int]) -> None:
-        ...
-    def channel_format(self) -> DataType:
-        ...
-    def dtype(self) -> DataType:
-        ...
-    @property
-    def element_shape(self) -> list[int]:
-        ...
-    @property
-    def field_dim(self) -> int:
-        ...
-    @property
-    def name(self) -> str:
-        ...
-    @property
-    def num_channels(self) -> int:
-        ...
-    @property
-    def texture_shape(self) -> list[int]:
-        ...
-class ArgKind:
-    """
-    Members:
-    
-      SCALAR
-    
-      NDARRAY
-    
-      MATRIX
-    
-      TEXTURE
-    
-      RWTEXTURE
-    """
-    MATRIX: typing.ClassVar[ArgKind]  # value = <ArgKind.MATRIX: 1>
-    NDARRAY: typing.ClassVar[ArgKind]  # value = <ArgKind.NDARRAY: 2>
-    RWTEXTURE: typing.ClassVar[ArgKind]  # value = <ArgKind.RWTEXTURE: 4>
-    SCALAR: typing.ClassVar[ArgKind]  # value = <ArgKind.SCALAR: 0>
-    TEXTURE: typing.ClassVar[ArgKind]  # value = <ArgKind.TEXTURE: 3>
-    __members__: typing.ClassVar[dict[str, ArgKind]]  # value = {'SCALAR': <ArgKind.SCALAR: 0>, 'NDARRAY': <ArgKind.NDARRAY: 2>, 'MATRIX': <ArgKind.MATRIX: 1>, 'TEXTURE': <ArgKind.TEXTURE: 3>, 'RWTEXTURE': <ArgKind.RWTEXTURE: 4>}
-    def __eq__(self, other: typing.Any) -> bool:
-        ...
-    def __getstate__(self) -> int:
-        ...
-    def __hash__(self) -> int:
-        ...
-    def __index__(self) -> int:
-        ...
-    def __init__(self, value: int) -> None:
-        ...
-    def __int__(self) -> int:
-        ...
-    def __ne__(self, other: typing.Any) -> bool:
-        ...
-    def __repr__(self) -> str:
-        ...
-    def __setstate__(self, state: int) -> None:
-        ...
-    def __str__(self) -> str:
-        ...
-    @property
-    def name(self) -> str:
-        ...
-    @property
-    def value(self) -> int:
-        ...
-class ArgPack:
-    def data_type(self) -> DataType:
+class ArgPackCxx:
+    def data_type(self) -> DataTypeCxx:
         ...
     def device_allocation(self) -> DeviceAllocation:
         ...
@@ -271,16 +176,16 @@ class ArgPack:
         ...
     def nelement(self) -> int:
         ...
-    def set_arg_float(self, arg0: list[int], arg1: float) -> None:
+    def set_arg_float(self, arg0: tuple[int, ...], arg1: float) -> None:
         ...
-    def set_arg_int(self, arg0: list[int], arg1: int) -> None:
+    def set_arg_int(self, arg0: tuple[int, ...], arg1: int) -> None:
         ...
-    def set_arg_nested_argpack(self, arg0: int, arg1: ArgPack) -> None:
+    def set_arg_nested_argpack(self, arg0: int, arg1: ArgPackCxx) -> None:
         ...
-    def set_arg_uint(self, arg0: list[int], arg1: int) -> None:
+    def set_arg_uint(self, arg0: tuple[int, ...], arg1: int) -> None:
         ...
     @property
-    def dtype(self) -> DataType:
+    def dtype(self) -> DataTypeCxx:
         ...
 class AutodiffMode:
     """
@@ -521,11 +426,11 @@ class CGd:
         ...
     def set_b(self, arg0: numpy.ndarray[numpy.float64[m, 1]]) -> None:
         ...
-    def set_b_ndarray(self, arg0: Program, arg1: Ndarray) -> None:
+    def set_b_ndarray(self, arg0: Program, arg1: NdarrayCxx) -> None:
         ...
     def set_x(self, arg0: numpy.ndarray[numpy.float64[m, 1]]) -> None:
         ...
-    def set_x_ndarray(self, arg0: Program, arg1: Ndarray) -> None:
+    def set_x_ndarray(self, arg0: Program, arg1: NdarrayCxx) -> None:
         ...
     def solve(self) -> None:
         ...
@@ -538,21 +443,20 @@ class CGf:
         ...
     def set_b(self, arg0: numpy.ndarray[numpy.float32[m, 1]]) -> None:
         ...
-    def set_b_ndarray(self, arg0: Program, arg1: Ndarray) -> None:
+    def set_b_ndarray(self, arg0: Program, arg1: NdarrayCxx) -> None:
         ...
     def set_x(self, arg0: numpy.ndarray[numpy.float32[m, 1]]) -> None:
         ...
-    def set_x_ndarray(self, arg0: Program, arg1: Ndarray) -> None:
+    def set_x_ndarray(self, arg0: Program, arg1: NdarrayCxx) -> None:
         ...
     def solve(self) -> None:
         ...
 class CUCG:
-    def solve(self, arg0: Program, arg1: Ndarray, arg2: Ndarray) -> None:
+    def solve(self, arg0: Program, arg1: NdarrayCxx, arg2: NdarrayCxx) -> None:
         ...
 class CompileConfig:
     ad_stack_size: int
     advanced_optimization: bool
-    allow_nv_shader_extension: bool
     arch: Arch
     auto_mesh_local_default_occupacy: int
     cache_loop_invariant_global_vars: bool
@@ -563,10 +467,10 @@ class CompileConfig:
     cuda_stack_limit: int
     debug: bool
     default_cpu_block_dim: int
-    default_fp: DataType
+    default_fp: DataTypeCxx
     default_gpu_block_dim: int
-    default_ip: DataType
-    default_up: DataType
+    default_ip: DataTypeCxx
+    default_up: DataTypeCxx
     demote_dense_struct_fors: bool
     demote_no_access_mesh_fors: bool
     detect_read_only: bool
@@ -620,9 +524,6 @@ class CompileConfig:
     vk_api_version: str
     def __init__(self) -> None:
         ...
-class CompiledGraph:
-    def jit_run(self, arg0: CompileConfig, arg1: dict) -> None:
-        ...
 class CompiledKernelData:
     pass
 class Config:
@@ -679,7 +580,7 @@ class CuSparseMatrix(SparseMatrix):
     def __add__(self, arg0: CuSparseMatrix) -> SparseMatrix:
         ...
     @typing.overload
-    def __init__(self, arg0: int, arg1: int, arg2: DataType) -> None:
+    def __init__(self, arg0: int, arg1: int, arg2: DataTypeCxx) -> None:
         ...
     @typing.overload
     def __init__(self, arg0: CuSparseMatrix) -> None:
@@ -694,7 +595,7 @@ class CuSparseMatrix(SparseMatrix):
         ...
     def matmul(self, arg0: CuSparseMatrix) -> SparseMatrix:
         ...
-    def spmv(self, arg0: Program, arg1: Ndarray, arg2: Ndarray) -> None:
+    def spmv(self, arg0: Program, arg1: NdarrayCxx, arg2: NdarrayCxx) -> None:
         ...
     def to_string(self) -> str:
         ...
@@ -709,12 +610,12 @@ class CuSparseSolver(SparseSolver):
         ...
     def info(self) -> bool:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
-class DataType:
+class DataTypeCxx:
     def __call__(self, *args, **kwargs) -> None:
         ...
-    def __eq__(self, arg0: DataType) -> bool:
+    def __eq__(self, arg0: DataTypeCxx) -> bool:
         ...
     def __getstate__(self) -> tuple:
         ...
@@ -726,13 +627,13 @@ class DataType:
         ...
     def __str__(self) -> str:
         ...
-    def element_type(self) -> DataType:
+    def element_type(self) -> DataTypeCxx:
         ...
     def get_ptr(self) -> ...:
         ...
-    def ptr_removed(self) -> DataType:
+    def ptr_removed(self) -> DataTypeCxx:
         ...
-    def shape(self) -> list[int]:
+    def shape(self) -> tuple[int, ...]:
         ...
     def to_string(self) -> str:
         ...
@@ -770,7 +671,7 @@ class EigenSparseSolverfloat32LDLTAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float32[m, 1]]) -> numpy.ndarray[numpy.float32[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat32LDLTCOLAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -783,7 +684,7 @@ class EigenSparseSolverfloat32LDLTCOLAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float32[m, 1]]) -> numpy.ndarray[numpy.float32[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat32LLTAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -796,7 +697,7 @@ class EigenSparseSolverfloat32LLTAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float32[m, 1]]) -> numpy.ndarray[numpy.float32[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat32LLTCOLAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -809,7 +710,7 @@ class EigenSparseSolverfloat32LLTCOLAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float32[m, 1]]) -> numpy.ndarray[numpy.float32[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat32LUAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -822,7 +723,7 @@ class EigenSparseSolverfloat32LUAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float32[m, 1]]) -> numpy.ndarray[numpy.float32[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat32LUCOLAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -835,7 +736,7 @@ class EigenSparseSolverfloat32LUCOLAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float32[m, 1]]) -> numpy.ndarray[numpy.float32[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat64LDLTAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -848,7 +749,7 @@ class EigenSparseSolverfloat64LDLTAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float64[m, 1]]) -> numpy.ndarray[numpy.float64[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat64LDLTCOLAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -861,7 +762,7 @@ class EigenSparseSolverfloat64LDLTCOLAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float64[m, 1]]) -> numpy.ndarray[numpy.float64[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat64LLTAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -874,7 +775,7 @@ class EigenSparseSolverfloat64LLTAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float64[m, 1]]) -> numpy.ndarray[numpy.float64[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat64LLTCOLAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -887,7 +788,7 @@ class EigenSparseSolverfloat64LLTCOLAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float64[m, 1]]) -> numpy.ndarray[numpy.float64[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat64LUAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -900,7 +801,7 @@ class EigenSparseSolverfloat64LUAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float64[m, 1]]) -> numpy.ndarray[numpy.float64[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
 class EigenSparseSolverfloat64LUCOLAMD(SparseSolver):
     def analyze_pattern(self, arg0: SparseMatrix) -> None:
@@ -913,9 +814,9 @@ class EigenSparseSolverfloat64LUCOLAMD(SparseSolver):
         ...
     def solve(self, arg0: numpy.ndarray[numpy.float64[m, 1]]) -> numpy.ndarray[numpy.float64[m, 1]]:
         ...
-    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: Ndarray, arg3: Ndarray) -> None:
+    def solve_rf(self, arg0: Program, arg1: SparseMatrix, arg2: NdarrayCxx, arg3: NdarrayCxx) -> None:
         ...
-class Expr:
+class ExprCxx:
     def get_dbg_info(self) -> DebugInfo:
         ...
     def get_dt(self) -> ...:
@@ -928,11 +829,11 @@ class Expr:
         ...
     def get_raw_address(self) -> int:
         ...
-    def get_ret_type(self) -> DataType:
+    def get_ret_type(self) -> DataTypeCxx:
         ...
-    def get_rvalue_type(self) -> DataType:
+    def get_rvalue_type(self) -> DataTypeCxx:
         ...
-    def get_shape(self) -> list[int] | None:
+    def get_shape(self) -> tuple[int, ...] | None:
         ...
     def get_underlying_ptr_address(self) -> int:
         ...
@@ -948,13 +849,13 @@ class Expr:
         ...
     def is_tensor(self) -> bool:
         ...
-    def set_adjoint(self, arg0: Expr) -> None:
+    def set_adjoint(self, arg0: ExprCxx) -> None:
         ...
-    def set_adjoint_checkbit(self, arg0: Expr) -> None:
+    def set_adjoint_checkbit(self, arg0: ExprCxx) -> None:
         ...
     def set_dbg_info(self, arg0: DebugInfo) -> None:
         ...
-    def set_dual(self, arg0: Expr) -> None:
+    def set_dual(self, arg0: ExprCxx) -> None:
         ...
     def set_dynamic_index_stride(self, arg0: int) -> None:
         ...
@@ -962,14 +863,14 @@ class Expr:
         ...
     def set_name(self, arg0: str) -> None:
         ...
-    def snode(self) -> SNode:
+    def snode(self) -> SNodeCxx:
         ...
     def type_check(self, arg0: CompileConfig) -> None:
         ...
 class ExprGroup:
     def __init__(self) -> None:
         ...
-    def push_back(self, arg0: Expr) -> None:
+    def push_back(self, arg0: ExprCxx) -> None:
         ...
     def size(self) -> int:
         ...
@@ -1209,19 +1110,19 @@ class Function:
         ...
     def finalize_rets(self) -> None:
         ...
-    def insert_arr_param(self, arg0: DataType, arg1: int, arg2: list[int], arg3: str) -> list[int]:
+    def insert_arr_param(self, arg0: DataTypeCxx, arg1: int, arg2: tuple[int, ...], arg3: str) -> tuple[int, ...]:
         ...
-    def insert_ndarray_param(self, arg0: DataType, arg1: int, arg2: str, arg3: bool) -> list[int]:
+    def insert_ndarray_param(self, arg0: DataTypeCxx, arg1: int, arg2: str, arg3: bool) -> tuple[int, ...]:
         ...
-    def insert_pointer_param(self, arg0: DataType, arg1: str) -> list[int]:
+    def insert_pointer_param(self, arg0: DataTypeCxx, arg1: str) -> tuple[int, ...]:
         ...
-    def insert_ret(self, arg0: DataType) -> int:
+    def insert_ret(self, arg0: DataTypeCxx) -> int:
         ...
-    def insert_rw_texture_param(self, arg0: int, arg1: Format, arg2: str) -> list[int]:
+    def insert_rw_texture_param(self, arg0: int, arg1: Format, arg2: str) -> tuple[int, ...]:
         ...
-    def insert_scalar_param(self, arg0: DataType, arg1: str) -> list[int]:
+    def insert_scalar_param(self, arg0: DataTypeCxx, arg1: str) -> tuple[int, ...]:
         ...
-    def insert_texture_param(self, arg0: int, arg1: str) -> list[int]:
+    def insert_texture_param(self, arg0: int, arg1: str) -> tuple[int, ...]:
         ...
     def set_function_body(self, arg0: typing.Callable[[], None]) -> None:
         ...
@@ -1231,140 +1132,139 @@ class FunctionKey:
     @property
     def instance_id(self) -> int:
         ...
-class GraphBuilder:
-    def __init__(self) -> None:
-        ...
-    def compile(self) -> ...:
-        ...
-    def create_sequential(self) -> Sequential:
-        ...
-    def dispatch(self, arg0: ..., arg1: list[Arg]) -> None:
-        ...
-    def seq(self) -> Sequential:
-        ...
+class GsTaichiAssertionError(AssertionError):
+    pass
+class GsTaichiIndexError(IndexError):
+    pass
+class GsTaichiRuntimeError(RuntimeError):
+    pass
+class GsTaichiSyntaxError(SyntaxError):
+    pass
+class GsTaichiTypeError(TypeError):
+    pass
 class HackedSignalRegister:
     def __init__(self) -> None:
         ...
 class InternalOp:
-    block_barrier: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    block_barrier_and_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    block_barrier_count_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    block_barrier_or_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    composite_extract_0: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    composite_extract_1: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    composite_extract_2: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    composite_extract_3: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_active_mask: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_all_sync_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_any_sync_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_ballot_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_match_all_sync_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_match_any_sync_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_shfl_down_sync_f32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_shfl_down_sync_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_shfl_sync_f32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_shfl_sync_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_shfl_up_sync_f32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_shfl_up_sync_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_shfl_xor_sync_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    cuda_uni_sync_i32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    do_nothing: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    grid_memfence: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    insert_triplet_f32: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    insert_triplet_f64: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    linear_thread_idx: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    localInvocationId: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    refresh_counter: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupAdd: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupAnd: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupBarrier: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupBroadcast: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupElect: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupInclusiveAdd: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupInclusiveAnd: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupInclusiveMax: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupInclusiveMin: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupInclusiveMul: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupInclusiveOr: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupInclusiveXor: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupInvocationId: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupMax: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupMemoryBarrier: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupMin: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupMul: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupOr: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupSize: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    subgroupXor: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    test_active_mask: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    test_internal_func_args: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    test_list_manager: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    test_node_allocator: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    test_node_allocator_gc_cpu: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    test_shfl: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    test_stack: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    vkGlobalThreadIdx: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    warp_barrier: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    workgroupBarrier: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-    workgroupMemoryBarrier: typing.ClassVar[Operation]  # value = <taichi._lib.core.taichi_python.Operation object>
-class Kernel:
+    block_barrier: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    block_barrier_and_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    block_barrier_count_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    block_barrier_or_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    composite_extract_0: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    composite_extract_1: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    composite_extract_2: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    composite_extract_3: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_active_mask: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_all_sync_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_any_sync_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_ballot_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_match_all_sync_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_match_any_sync_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_shfl_down_sync_f32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_shfl_down_sync_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_shfl_sync_f32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_shfl_sync_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_shfl_up_sync_f32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_shfl_up_sync_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_shfl_xor_sync_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    cuda_uni_sync_i32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    do_nothing: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    grid_memfence: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    insert_triplet_f32: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    insert_triplet_f64: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    linear_thread_idx: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    localInvocationId: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    refresh_counter: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupAdd: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupAnd: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupBarrier: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupBroadcast: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupElect: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupInclusiveAdd: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupInclusiveAnd: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupInclusiveMax: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupInclusiveMin: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupInclusiveMul: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupInclusiveOr: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupInclusiveXor: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupInvocationId: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupMax: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupMemoryBarrier: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupMin: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupMul: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupOr: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupSize: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    subgroupXor: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    test_active_mask: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    test_internal_func_args: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    test_list_manager: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    test_node_allocator: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    test_node_allocator_gc_cpu: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    test_shfl: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    test_stack: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    vkGlobalThreadIdx: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    warp_barrier: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    workgroupBarrier: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+    workgroupMemoryBarrier: typing.ClassVar[Operation]  # value = <gstaichi._lib.core.gstaichi_python.Operation object>
+class KernelCxx:
     def ast_builder(self) -> ASTBuilder:
         ...
     def finalize_params(self) -> None:
         ...
     def finalize_rets(self) -> None:
         ...
-    def insert_argpack_param_and_push(self, arg0: str) -> list[int]:
+    def insert_argpack_param_and_push(self, arg0: str) -> tuple[int, ...]:
         ...
-    def insert_arr_param(self, arg0: DataType, arg1: int, arg2: list[int], arg3: str) -> list[int]:
+    def insert_arr_param(self, arg0: DataTypeCxx, arg1: int, arg2: tuple[int, ...], arg3: str) -> tuple[int, ...]:
         ...
-    def insert_ndarray_param(self, arg0: DataType, arg1: int, arg2: str, arg3: bool) -> list[int]:
+    def insert_ndarray_param(self, arg0: DataTypeCxx, arg1: int, arg2: str, arg3: bool) -> tuple[int, ...]:
         ...
-    def insert_pointer_param(self, arg0: DataType, arg1: str) -> list[int]:
+    def insert_pointer_param(self, arg0: DataTypeCxx, arg1: str) -> tuple[int, ...]:
         ...
-    def insert_ret(self, arg0: DataType) -> int:
+    def insert_ret(self, arg0: DataTypeCxx) -> int:
         ...
-    def insert_rw_texture_param(self, arg0: int, arg1: Format, arg2: str) -> list[int]:
+    def insert_rw_texture_param(self, arg0: int, arg1: Format, arg2: str) -> tuple[int, ...]:
         ...
-    def insert_scalar_param(self, arg0: DataType, arg1: str) -> list[int]:
+    def insert_scalar_param(self, arg0: DataTypeCxx, arg1: str) -> tuple[int, ...]:
         ...
-    def insert_texture_param(self, arg0: int, arg1: str) -> list[int]:
+    def insert_texture_param(self, arg0: int, arg1: str) -> tuple[int, ...]:
         ...
-    def make_launch_context(self) -> ...:
+    def make_launch_context(self) -> KernelLaunchContext:
         ...
-    def no_activate(self, arg0: SNode) -> None:
+    def no_activate(self, arg0: SNodeCxx) -> None:
         ...
     def pop_argpack_stack(self) -> None:
         ...
 class KernelLaunchContext:
-    def get_struct_ret_float(self, arg0: list[int]) -> float:
+    def get_struct_ret_float(self, arg0: tuple[int, ...]) -> float:
         ...
-    def get_struct_ret_int(self, arg0: list[int]) -> int:
+    def get_struct_ret_int(self, arg0: tuple[int, ...]) -> int:
         ...
-    def get_struct_ret_uint(self, arg0: list[int]) -> int:
+    def get_struct_ret_uint(self, arg0: tuple[int, ...]) -> int:
         ...
-    def set_arg_argpack(self, arg0: list[int], arg1: ArgPack) -> None:
+    def set_arg_argpack(self, arg0: tuple[int, ...], arg1: ArgPackCxx) -> None:
         ...
-    def set_arg_external_array_with_shape(self, arg0: list[int], arg1: int, arg2: int, arg3: list[int], arg4: int) -> None:
+    def set_arg_external_array_with_shape(self, arg0: tuple[int, ...], arg1: int, arg2: int, arg3: tuple[int, ...], arg4: int) -> None:
         ...
-    def set_arg_float(self, arg0: list[int], arg1: float) -> None:
+    def set_arg_float(self, arg0: tuple[int, ...], arg1: float) -> None:
         ...
-    def set_arg_int(self, arg0: list[int], arg1: int) -> None:
+    def set_arg_int(self, arg0: tuple[int, ...], arg1: int) -> None:
         ...
-    def set_arg_ndarray(self, arg0: list[int], arg1: Ndarray) -> None:
+    def set_arg_ndarray(self, arg0: tuple[int, ...], arg1: Ndarray) -> None:
         ...
-    def set_arg_ndarray_with_grad(self, arg0: list[int], arg1: Ndarray, arg2: Ndarray) -> None:
+    def set_arg_ndarray_with_grad(self, arg0: tuple[int, ...], arg1: NdarrayCxx, arg2: NdarrayCxx) -> None:
         ...
-    def set_arg_rw_texture(self, arg0: list[int], arg1: Texture) -> None:
+    def set_arg_rw_texture(self, arg0: tuple[int, ...], arg1: TextureCxx) -> None:
         ...
-    def set_arg_texture(self, arg0: list[int], arg1: Texture) -> None:
+    def set_arg_texture(self, arg0: tuple[int, ...], arg1: TextureCxx) -> None:
         ...
-    def set_arg_uint(self, arg0: list[int], arg1: int) -> None:
+    def set_arg_uint(self, arg0: tuple[int, ...], arg1: int) -> None:
         ...
-    def set_struct_arg_float(self, arg0: list[int], arg1: float) -> None:
+    def set_struct_arg_float(self, arg0: tuple[int, ...], arg1: float) -> None:
         ...
-    def set_struct_arg_int(self, arg0: list[int], arg1: int) -> None:
+    def set_struct_arg_int(self, arg0: tuple[int, ...], arg1: int) -> None:
         ...
-    def set_struct_arg_uint(self, arg0: list[int], arg1: int) -> None:
+    def set_struct_arg_uint(self, arg0: tuple[int, ...], arg1: int) -> None:
         ...
 class KernelProfileTracedRecord:
     active_blocks_per_multiprocessor: int
@@ -1616,39 +1516,37 @@ class MeshTopology:
     @property
     def value(self) -> int:
         ...
-class Ndarray:
+class NdarrayCxx:
     def device_allocation(self) -> DeviceAllocation:
         ...
     def device_allocation_ptr(self) -> int:
         ...
-    def element_data_type(self) -> DataType:
+    def element_data_type(self) -> DataTypeCxx:
         ...
-    def element_shape(self) -> list[int]:
+    def element_shape(self) -> tuple[int, ...]:
         ...
     def element_size(self) -> int:
         ...
     def nelement(self) -> int:
         ...
-    def read_float(self, arg0: list[int]) -> float:
+    def read_float(self, arg0: tuple[int, ...]) -> float:
         ...
-    def read_int(self, arg0: list[int]) -> int:
+    def read_int(self, arg0: tuple[int, ...]) -> int:
         ...
-    def read_uint(self, arg0: list[int]) -> int:
+    def read_uint(self, arg0: tuple[int, ...]) -> int:
         ...
-    def total_shape(self) -> list[int]:
+    def total_shape(self) -> tuple[int, ...]:
         ...
-    def write_float(self, arg0: list[int], arg1: float) -> None:
+    def write_float(self, arg0: tuple[int, ...], arg1: float) -> None:
         ...
-    def write_int(self, arg0: list[int], arg1: int) -> None:
+    def write_int(self, arg0: tuple[int, ...], arg1: int) -> None:
         ...
     @property
-    def dtype(self) -> DataType:
+    def dtype(self) -> DataTypeCxx:
         ...
     @property
-    def shape(self) -> list[int]:
+    def shape(self) -> tuple[int, ...]:
         ...
-class Node:
-    pass
 class Operation:
     pass
 class Program:
@@ -1660,17 +1558,17 @@ class Program:
         ...
     def config(self) -> CompileConfig:
         ...
-    def create_argpack(self, dt: DataType) -> ...:
+    def create_argpack(self, dt: DataTypeCxx) -> ...:
         ...
     def create_function(self, arg0: ...) -> Function:
         ...
-    def create_kernel(self, arg0: typing.Callable[[...], None], arg1: str, arg2: AutodiffMode) -> ...:
+    def create_kernel(self, arg0: typing.Callable[[...], None], arg1: str, arg2: AutodiffMode) -> KernelCxx:
         ...
-    def create_ndarray(self, dt: DataType, shape: list[int], layout: Layout = ..., zero_fill: bool = False, dbg_info: DebugInfo = ...) -> ...:
+    def create_ndarray(self, dt: DataTypeCxx, shape: tuple[int, ...], layout: Layout = ..., zero_fill: bool = False, dbg_info: DebugInfo = ...) -> ...:
         ...
-    def create_sparse_matrix(self, arg0: int, arg1: int, arg2: DataType, arg3: str) -> ...:
+    def create_sparse_matrix(self, arg0: int, arg1: int, arg2: DataTypeCxx, arg3: str) -> ...:
         ...
-    def create_texture(self, fmt: ..., shape: list[int] = ()) -> ...:
+    def create_texture(self, fmt: ..., shape: tuple[int, ...] = ()) -> ...:
         ...
     def delete_argpack(self, arg0: ...) -> None:
         ...
@@ -1706,8 +1604,6 @@ class Program:
         ...
     def launch_kernel(self, arg0: CompiledKernelData, arg1: ...) -> None:
         ...
-    def make_aot_module_builder(self, arg0: Arch, arg1: list[str]) -> ...:
-        ...
     def make_id_expr(self, arg0: str) -> ...:
         ...
     def make_sparse_matrix_from_ndarray(self, arg0: ..., arg1: ...) -> None:
@@ -1732,29 +1628,77 @@ class Program:
         ...
     def update_kernel_profiler(self) -> None:
         ...
-class SNode:
-    parent: SNode
+class SNodeAccessFlag:
+    """
+    Members:
+    
+      block_local
+    
+      read_only
+    
+      mesh_local
+    """
+    __members__: typing.ClassVar[dict[str, SNodeAccessFlag]]  # value = {'block_local': <SNodeAccessFlag.block_local: 0>, 'read_only': <SNodeAccessFlag.read_only: 1>, 'mesh_local': <SNodeAccessFlag.mesh_local: 2>}
+    block_local: typing.ClassVar[SNodeAccessFlag]  # value = <SNodeAccessFlag.block_local: 0>
+    mesh_local: typing.ClassVar[SNodeAccessFlag]  # value = <SNodeAccessFlag.mesh_local: 2>
+    read_only: typing.ClassVar[SNodeAccessFlag]  # value = <SNodeAccessFlag.read_only: 1>
+    def __eq__(self, other: typing.Any) -> bool:
+        ...
+    def __ge__(self, other: typing.Any) -> bool:
+        ...
+    def __getstate__(self) -> int:
+        ...
+    def __gt__(self, other: typing.Any) -> bool:
+        ...
+    def __hash__(self) -> int:
+        ...
+    def __index__(self) -> int:
+        ...
+    def __init__(self, value: int) -> None:
+        ...
+    def __int__(self) -> int:
+        ...
+    def __le__(self, other: typing.Any) -> bool:
+        ...
+    def __lt__(self, other: typing.Any) -> bool:
+        ...
+    def __ne__(self, other: typing.Any) -> bool:
+        ...
+    def __repr__(self) -> str:
+        ...
+    def __setstate__(self, state: int) -> None:
+        ...
+    def __str__(self) -> str:
+        ...
+    @property
+    def name(self) -> str:
+        ...
+    @property
+    def value(self) -> int:
+        ...
+class SNodeCxx:
+    parent: SNodeCxx
     def __init__(self) -> None:
         ...
     def allocate_adjoint_checkbit(self) -> None:
         ...
-    def bit_struct(self, arg0: ..., arg1: DebugInfo) -> SNode:
+    def bit_struct(self, arg0: ..., arg1: DebugInfo) -> SNodeCxx:
         ...
-    def bitmasked(self, arg0: list[Axis], arg1: list[int], arg2: DebugInfo) -> SNode:
+    def bitmasked(self, arg0: list[Axis], arg1: tuple[int, ...], arg2: DebugInfo) -> SNodeCxx:
         ...
-    def data_type(self) -> DataType:
+    def data_type(self) -> DataTypeCxx:
         ...
-    def dense(self, arg0: list[Axis], arg1: list[int], arg2: DebugInfo) -> SNode:
+    def dense(self, arg0: list[Axis], arg1: tuple[int, ...], arg2: DebugInfo) -> SNodeCxx:
         ...
-    def dynamic(self, arg0: Axis, arg1: int, arg2: int, arg3: DebugInfo) -> SNode:
+    def dynamic(self, arg0: Axis, arg1: int, arg2: int, arg3: DebugInfo) -> SNodeCxx:
         ...
-    def get_ch(self, arg0: int) -> SNode:
+    def get_ch(self, arg0: int) -> SNodeCxx:
         ...
     def get_expr(self) -> ...:
         ...
     def get_num_ch(self) -> int:
         ...
-    def get_physical_index_position(self) -> list[int]:
+    def get_physical_index_position(self) -> tuple[int, ...]:
         ...
     def get_shape_along_axis(self, arg0: int) -> int:
         ...
@@ -1768,7 +1712,7 @@ class SNode:
         ...
     def has_dual(self) -> bool:
         ...
-    def hash(self, arg0: list[Axis], arg1: list[int], arg2: DebugInfo) -> SNode:
+    def hash(self, arg0: list[Axis], arg1: tuple[int, ...], arg2: DebugInfo) -> SNodeCxx:
         ...
     def is_place(self) -> bool:
         ...
@@ -1782,23 +1726,23 @@ class SNode:
         ...
     def num_active_indices(self) -> int:
         ...
-    def place(self, arg0: ..., arg1: list[int], arg2: int) -> None:
+    def place(self, arg0: ..., arg1: tuple[int, ...], arg2: int) -> None:
         ...
-    def pointer(self, arg0: list[Axis], arg1: list[int], arg2: DebugInfo) -> SNode:
+    def pointer(self, arg0: list[Axis], arg1: tuple[int, ...], arg2: DebugInfo) -> SNodeCxx:
         ...
-    def quant_array(self, arg0: list[Axis], arg1: list[int], arg2: int, arg3: DebugInfo) -> SNode:
+    def quant_array(self, arg0: list[Axis], arg1: tuple[int, ...], arg2: int, arg3: DebugInfo) -> SNodeCxx:
         ...
-    def read_float(self, arg0: list[int]) -> float:
+    def read_float(self, arg0: tuple[int, ...]) -> float:
         ...
-    def read_int(self, arg0: list[int]) -> int:
+    def read_int(self, arg0: tuple[int, ...]) -> int:
         ...
-    def read_uint(self, arg0: list[int]) -> int:
+    def read_uint(self, arg0: tuple[int, ...]) -> int:
         ...
-    def write_float(self, arg0: list[int], arg1: float) -> None:
+    def write_float(self, arg0: tuple[int, ...], arg1: float) -> None:
         ...
-    def write_int(self, arg0: list[int], arg1: int) -> None:
+    def write_int(self, arg0: tuple[int, ...], arg1: int) -> None:
         ...
-    def write_uint(self, arg0: list[int], arg1: int) -> None:
+    def write_uint(self, arg0: tuple[int, ...], arg1: int) -> None:
         ...
     @property
     def cell_size_bytes(self) -> int:
@@ -1807,7 +1751,7 @@ class SNode:
     def id(self) -> int:
         ...
     @property
-    def offset(self) -> list[int]:
+    def offset(self) -> tuple[int, ...]:
         ...
     @property
     def offset_bytes_in_parent_cell(self) -> int:
@@ -1815,54 +1759,6 @@ class SNode:
     @property
     def type(self) -> SNodeType:
         ...
-class SNodeAccessFlag:
-    """
-    Members:
-    
-      block_local
-    
-      read_only
-    
-      mesh_local
-    """
-    __members__: typing.ClassVar[dict[str, SNodeAccessFlag]]  # value = {'block_local': <SNodeAccessFlag.block_local: 0>, 'read_only': <SNodeAccessFlag.read_only: 1>, 'mesh_local': <SNodeAccessFlag.mesh_local: 2>}
-    block_local: typing.ClassVar[SNodeAccessFlag]  # value = <SNodeAccessFlag.block_local: 0>
-    mesh_local: typing.ClassVar[SNodeAccessFlag]  # value = <SNodeAccessFlag.mesh_local: 2>
-    read_only: typing.ClassVar[SNodeAccessFlag]  # value = <SNodeAccessFlag.read_only: 1>
-    def __eq__(self, other: typing.Any) -> bool:
-        ...
-    def __ge__(self, other: typing.Any) -> bool:
-        ...
-    def __getstate__(self) -> int:
-        ...
-    def __gt__(self, other: typing.Any) -> bool:
-        ...
-    def __hash__(self) -> int:
-        ...
-    def __index__(self) -> int:
-        ...
-    def __init__(self, value: int) -> None:
-        ...
-    def __int__(self) -> int:
-        ...
-    def __le__(self, other: typing.Any) -> bool:
-        ...
-    def __lt__(self, other: typing.Any) -> bool:
-        ...
-    def __ne__(self, other: typing.Any) -> bool:
-        ...
-    def __repr__(self) -> str:
-        ...
-    def __setstate__(self, state: int) -> None:
-        ...
-    def __str__(self) -> str:
-        ...
-    @property
-    def name(self) -> str:
-        ...
-    @property
-    def value(self) -> int:
-        ...
 class SNodeGradType:
     """
     Members:
@@ -1917,9 +1813,9 @@ class SNodeGradType:
 class SNodeRegistry:
     def __init__(self) -> None:
         ...
-    def create_root(self, arg0: Program) -> SNode:
+    def create_root(self, arg0: Program) -> SNodeCxx:
         ...
-class SNodeTree:
+class SNodeTreeCxx:
     def destroy_snode_tree(self, arg0: Program) -> None:
         ...
     def id(self) -> int:
@@ -1993,24 +1889,17 @@ class SNodeType:
     @property
     def value(self) -> int:
         ...
-class Sequential(Node):
-    def __init__(self, arg0: ...) -> None:
-        ...
-    def append(self, arg0: Node) -> None:
-        ...
-    def dispatch(self, arg0: ..., arg1: list[Arg]) -> None:
-        ...
 class SparseMatrix:
     @typing.overload
     def __init__(self) -> None:
         ...
     @typing.overload
-    def __init__(self, rows: int, cols: int, dt: DataType = ...) -> None:
+    def __init__(self, rows: int, cols: int, dt: DataTypeCxx = ...) -> None:
         ...
     @typing.overload
     def __init__(self, arg0: SparseMatrix) -> None:
         ...
-    def get_data_type(self) -> DataType:
+    def get_data_type(self) -> DataTypeCxx:
         ...
     def get_element(self, arg0: int, arg1: int) -> float:
         ...
@@ -2025,7 +1914,7 @@ class SparseMatrix:
     def to_string(self) -> str:
         ...
 class SparseMatrixBuilder:
-    def __init__(self, rows: int, cols: int, max_num_triplets: int, dt: DataType = ..., storage_format: str = 'col_major') -> None:
+    def __init__(self, rows: int, cols: int, max_num_triplets: int, dt: DataTypeCxx = ..., storage_format: str = 'col_major') -> None:
         ...
     def build(self) -> ...:
         ...
@@ -2054,27 +1943,17 @@ class SparseSolver:
         ...
 class Stmt:
     pass
-class TaichiAssertionError(AssertionError):
-    pass
-class TaichiIndexError(IndexError):
-    pass
-class TaichiRuntimeError(RuntimeError):
-    pass
-class TaichiSyntaxError(SyntaxError):
-    pass
-class TaichiTypeError(TypeError):
-    pass
 class Task:
     def initialize(self, arg0: Config) -> None:
         ...
     def run(self, arg0: list[str]) -> str:
         ...
-class Texture:
+class TextureCxx:
     def device_allocation_ptr(self) -> int:
         ...
-    def from_ndarray(self, arg0: Ndarray) -> None:
+    def from_ndarray(self, arg0: NdarrayCxx) -> None:
         ...
-    def from_snode(self, arg0: SNode) -> None:
+    def from_snode(self, arg0: SNodeCxx) -> None:
         ...
 class TextureOpType:
     """
@@ -2134,9 +2013,9 @@ class Type:
     def to_string(self) -> str:
         ...
 class TypeFactory:
-    def get_argpack_type(self, arg0: list[tuple[DataType, str]]) -> DataType:
+    def get_argpack_type(self, arg0: list[tuple[DataTypeCxx, str]]) -> DataTypeCxx:
         ...
-    def get_ndarray_struct_type(self, dt: DataType, ndim: int, needs_grad: bool) -> Type:
+    def get_ndarray_struct_type(self, dt: DataTypeCxx, ndim: int, needs_grad: bool) -> Type:
         ...
     def get_quant_fixed_type(self, digits_type: Type, compute_type: Type, scale: float) -> Type:
         ...
@@ -2146,11 +2025,11 @@ class TypeFactory:
         ...
     def get_rwtexture_struct_type(self) -> Type:
         ...
-    def get_struct_type(self, arg0: list[tuple[DataType, str]]) -> DataType:
+    def get_struct_type(self, arg0: list[tuple[DataTypeCxx, str]]) -> DataTypeCxx:
         ...
-    def get_struct_type_for_argpack_ptr(self, dt: DataType, layout: str = 'none') -> Type:
+    def get_struct_type_for_argpack_ptr(self, dt: DataTypeCxx, layout: str = 'none') -> Type:
         ...
-    def get_tensor_type(self, shape: tuple[int, ...], element_type: DataType) -> DataType:
+    def get_tensor_type(self, shape: tuple[int, ...], element_type: DataType) -> DataTypeCxx:
         ...
 class UnaryOpType:
     """
@@ -2403,7 +2282,7 @@ class dColMajor_EigenSparseMatrix(SparseMatrix):
     def __imul__(self, arg0: float) -> dColMajor_EigenSparseMatrix:
         ...
     @typing.overload
-    def __init__(self, arg0: int, arg1: int, arg2: DataType) -> None:
+    def __init__(self, arg0: int, arg1: int, arg2: DataTypeCxx) -> None:
         ...
     @typing.overload
     def __init__(self, arg0: dColMajor_EigenSparseMatrix) -> None:
@@ -2431,7 +2310,7 @@ class dColMajor_EigenSparseMatrix(SparseMatrix):
         ...
     def set_element(self, arg0: int, arg1: int, arg2: float) -> None:
         ...
-    def spmv(self, arg0: Program, arg1: Ndarray, arg2: Ndarray) -> None:
+    def spmv(self, arg0: Program, arg1: NdarrayCxx, arg2: NdarrayCxx) -> None:
         ...
     def transpose(self) -> dColMajor_EigenSparseMatrix:
         ...
@@ -2443,7 +2322,7 @@ class dRowMajor_EigenSparseMatrix(SparseMatrix):
     def __imul__(self, arg0: float) -> dRowMajor_EigenSparseMatrix:
         ...
     @typing.overload
-    def __init__(self, arg0: int, arg1: int, arg2: DataType) -> None:
+    def __init__(self, arg0: int, arg1: int, arg2: DataTypeCxx) -> None:
         ...
     @typing.overload
     def __init__(self, arg0: dRowMajor_EigenSparseMatrix) -> None:
@@ -2471,7 +2350,7 @@ class dRowMajor_EigenSparseMatrix(SparseMatrix):
         ...
     def set_element(self, arg0: int, arg1: int, arg2: float) -> None:
         ...
-    def spmv(self, arg0: Program, arg1: Ndarray, arg2: Ndarray) -> None:
+    def spmv(self, arg0: Program, arg1: NdarrayCxx, arg2: NdarrayCxx) -> None:
         ...
     def transpose(self) -> dRowMajor_EigenSparseMatrix:
         ...
@@ -2483,7 +2362,7 @@ class fColMajor_EigenSparseMatrix(SparseMatrix):
     def __imul__(self, arg0: float) -> fColMajor_EigenSparseMatrix:
         ...
     @typing.overload
-    def __init__(self, arg0: int, arg1: int, arg2: DataType) -> None:
+    def __init__(self, arg0: int, arg1: int, arg2: DataTypeCxx) -> None:
         ...
     @typing.overload
     def __init__(self, arg0: fColMajor_EigenSparseMatrix) -> None:
@@ -2511,7 +2390,7 @@ class fColMajor_EigenSparseMatrix(SparseMatrix):
         ...
     def set_element(self, arg0: int, arg1: int, arg2: float) -> None:
         ...
-    def spmv(self, arg0: Program, arg1: Ndarray, arg2: Ndarray) -> None:
+    def spmv(self, arg0: Program, arg1: NdarrayCxx, arg2: NdarrayCxx) -> None:
         ...
     def transpose(self) -> fColMajor_EigenSparseMatrix:
         ...
@@ -2523,7 +2402,7 @@ class fRowMajor_EigenSparseMatrix(SparseMatrix):
     def __imul__(self, arg0: float) -> fRowMajor_EigenSparseMatrix:
         ...
     @typing.overload
-    def __init__(self, arg0: int, arg1: int, arg2: DataType) -> None:
+    def __init__(self, arg0: int, arg1: int, arg2: DataTypeCxx) -> None:
         ...
     @typing.overload
     def __init__(self, arg0: fRowMajor_EigenSparseMatrix) -> None:
@@ -2551,7 +2430,7 @@ class fRowMajor_EigenSparseMatrix(SparseMatrix):
         ...
     def set_element(self, arg0: int, arg1: int, arg2: float) -> None:
         ...
-    def spmv(self, arg0: Program, arg1: Ndarray, arg2: Ndarray) -> None:
+    def spmv(self, arg0: Program, arg1: NdarrayCxx, arg2: NdarrayCxx) -> None:
         ...
     def transpose(self) -> fRowMajor_EigenSparseMatrix:
         ...
@@ -2561,7 +2440,7 @@ def arch_name(arg0: Arch) -> str:
     ...
 def arch_uses_llvm(arg0: Arch) -> bool:
     ...
-def bits_cast(arg0: Expr, arg1: DataType) -> Expr:
+def bits_cast(arg0: ExprCxx, arg1: DataTypeCxx) -> ExprCxx:
     ...
 def clean_offline_cache_files(arg0: str) -> int:
     ...
@@ -2581,9 +2460,9 @@ def critical(arg0: str) -> None:
     ...
 def cuda_version() -> str:
     ...
-def data_type_name(arg0: DataType) -> str:
+def data_type_name(arg0: DataTypeCxx) -> str:
     ...
-def data_type_size(arg0: DataType) -> int:
+def data_type_size(arg0: DataTypeCxx) -> int:
     ...
 def debug(arg0: str) -> None:
     ...
@@ -2595,129 +2474,129 @@ def element_type_name(arg0: MeshElementType) -> str:
     ...
 def error(arg0: str) -> None:
     ...
-def expr_abs(arg0: Expr) -> Expr:
+def expr_abs(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_acos(arg0: Expr) -> Expr:
+def expr_acos(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_add(arg0: Expr, arg1: Expr) -> Expr:
+def expr_add(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_asin(arg0: Expr) -> Expr:
+def expr_asin(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_assume_in_range(arg0: Expr, arg1: Expr, arg2: int, arg3: int, arg4: DebugInfo) -> Expr:
+def expr_assume_in_range(arg0: ExprCxx, arg1: ExprCxx, arg2: int, arg3: int, arg4: DebugInfo) -> ExprCxx:
     ...
-def expr_atan2(arg0: Expr, arg1: Expr) -> Expr:
+def expr_atan2(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_atomic_add(arg0: Expr, arg1: Expr) -> Expr:
+def expr_atomic_add(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_atomic_bit_and(arg0: Expr, arg1: Expr) -> Expr:
+def expr_atomic_bit_and(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_atomic_bit_or(arg0: Expr, arg1: Expr) -> Expr:
+def expr_atomic_bit_or(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_atomic_bit_xor(arg0: Expr, arg1: Expr) -> Expr:
+def expr_atomic_bit_xor(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_atomic_max(arg0: Expr, arg1: Expr) -> Expr:
+def expr_atomic_max(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_atomic_min(arg0: Expr, arg1: Expr) -> Expr:
+def expr_atomic_min(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_atomic_mul(arg0: Expr, arg1: Expr) -> Expr:
+def expr_atomic_mul(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_atomic_sub(arg0: Expr, arg1: Expr) -> Expr:
+def expr_atomic_sub(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_bit_and(arg0: Expr, arg1: Expr) -> Expr:
+def expr_bit_and(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_bit_not(arg0: Expr) -> Expr:
+def expr_bit_not(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_bit_or(arg0: Expr, arg1: Expr) -> Expr:
+def expr_bit_or(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_bit_sar(arg0: Expr, arg1: Expr) -> Expr:
+def expr_bit_sar(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_bit_shl(arg0: Expr, arg1: Expr) -> Expr:
+def expr_bit_shl(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_bit_shr(arg0: Expr, arg1: Expr) -> Expr:
+def expr_bit_shr(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_bit_xor(arg0: Expr, arg1: Expr) -> Expr:
+def expr_bit_xor(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_ceil(arg0: Expr) -> Expr:
+def expr_ceil(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_clz(arg0: Expr) -> Expr:
+def expr_clz(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_cmp_eq(arg0: Expr, arg1: Expr) -> Expr:
+def expr_cmp_eq(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_cmp_ge(arg0: Expr, arg1: Expr) -> Expr:
+def expr_cmp_ge(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_cmp_gt(arg0: Expr, arg1: Expr) -> Expr:
+def expr_cmp_gt(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_cmp_le(arg0: Expr, arg1: Expr) -> Expr:
+def expr_cmp_le(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_cmp_lt(arg0: Expr, arg1: Expr) -> Expr:
+def expr_cmp_lt(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_cmp_ne(arg0: Expr, arg1: Expr) -> Expr:
+def expr_cmp_ne(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_cos(arg0: Expr) -> Expr:
+def expr_cos(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_div(arg0: Expr, arg1: Expr) -> Expr:
+def expr_div(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_exp(arg0: Expr) -> Expr:
+def expr_exp(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_field(arg0: Expr, arg1: DataType) -> Expr:
+def expr_field(arg0: ExprCxx, arg1: DataTypeCxx) -> ExprCxx:
     ...
-def expr_floor(arg0: Expr) -> Expr:
+def expr_floor(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_floordiv(arg0: Expr, arg1: Expr) -> Expr:
+def expr_floordiv(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_frexp(arg0: Expr) -> Expr:
+def expr_frexp(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_ifte(arg0: Expr, arg1: Expr, arg2: Expr) -> Expr:
+def expr_ifte(arg0: ExprCxx, arg1: ExprCxx, arg2: ExprCxx) -> ExprCxx:
     ...
-def expr_inv(arg0: Expr) -> Expr:
+def expr_inv(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_log(arg0: Expr) -> Expr:
+def expr_log(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_logic_not(arg0: Expr) -> Expr:
+def expr_logic_not(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_logical_and(arg0: Expr, arg1: Expr) -> Expr:
+def expr_logical_and(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_logical_or(arg0: Expr, arg1: Expr) -> Expr:
+def expr_logical_or(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_loop_unique(arg0: Expr, arg1: list[SNode], arg2: DebugInfo) -> Expr:
+def expr_loop_unique(arg0: ExprCxx, arg1: list[SNodeCxx], arg2: DebugInfo) -> ExprCxx:
     ...
-def expr_matrix_field(arg0: list[Expr], arg1: list[int]) -> Expr:
+def expr_matrix_field(arg0: list[ExprCxx], arg1: tuple[int, ...]) -> ExprCxx:
     ...
-def expr_max(arg0: Expr, arg1: Expr) -> Expr:
+def expr_max(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_min(arg0: Expr, arg1: Expr) -> Expr:
+def expr_min(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_mod(arg0: Expr, arg1: Expr) -> Expr:
+def expr_mod(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_mul(arg0: Expr, arg1: Expr) -> Expr:
+def expr_mul(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_neg(arg0: Expr) -> Expr:
+def expr_neg(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_popcnt(arg0: Expr) -> Expr:
+def expr_popcnt(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_pow(arg0: Expr, arg1: Expr) -> Expr:
+def expr_pow(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_rcp(arg0: Expr) -> Expr:
+def expr_rcp(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_round(arg0: Expr) -> Expr:
+def expr_round(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_rsqrt(arg0: Expr) -> Expr:
+def expr_rsqrt(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_select(arg0: Expr, arg1: Expr, arg2: Expr) -> Expr:
+def expr_select(arg0: ExprCxx, arg1: ExprCxx, arg2: ExprCxx) -> ExprCxx:
     ...
-def expr_sin(arg0: Expr) -> Expr:
+def expr_sin(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_sqrt(arg0: Expr) -> Expr:
+def expr_sqrt(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_sub(arg0: Expr, arg1: Expr) -> Expr:
+def expr_sub(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def expr_tan(arg0: Expr) -> Expr:
+def expr_tan(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_tanh(arg0: Expr) -> Expr:
+def expr_tanh(arg0: ExprCxx) -> ExprCxx:
     ...
-def expr_truediv(arg0: Expr, arg1: Expr) -> Expr:
+def expr_truediv(arg0: ExprCxx, arg1: ExprCxx) -> ExprCxx:
     ...
-def finalize_snode_tree(arg0: SNodeRegistry, arg1: SNode, arg2: Program, arg3: bool) -> SNodeTree:
+def finalize_snode_tree(arg0: SNodeRegistry, arg1: SNodeCxx, arg2: Program, arg3: bool) -> SNodeTreeCxx:
     ...
 def flush_log() -> None:
     ...
@@ -2727,33 +2606,31 @@ def get_commit_hash() -> str:
     ...
 def get_default_float_size() -> int:
     ...
-def get_external_tensor_dim(arg0: Expr) -> int:
+def get_external_tensor_dim(arg0: ExprCxx) -> int:
     ...
-def get_external_tensor_element_dim(arg0: Expr) -> int:
+def get_external_tensor_element_dim(arg0: ExprCxx) -> int:
     ...
-def get_external_tensor_element_shape(arg0: Expr) -> list[int]:
+def get_external_tensor_element_shape(arg0: ExprCxx) -> tuple[int, ...]:
     ...
-def get_external_tensor_element_type(arg0: Expr) -> DataType:
+def get_external_tensor_element_type(arg0: ExprCxx) -> DataTypeCxx:
     ...
-def get_external_tensor_needs_grad(arg0: Expr) -> bool:
+def get_external_tensor_needs_grad(arg0: ExprCxx) -> bool:
     ...
-def get_external_tensor_real_func_args(arg0: Expr, arg1: DebugInfo) -> list[Expr]:
+def get_external_tensor_real_func_args(arg0: ExprCxx, arg1: DebugInfo) -> list[ExprCxx]:
     ...
-def get_external_tensor_shape_along_axis(arg0: Expr, arg1: int, arg2: DebugInfo) -> Expr:
+def get_external_tensor_shape_along_axis(arg0: ExprCxx, arg1: int, arg2: DebugInfo) -> ExprCxx:
     ...
 def get_llvm_target_support() -> str:
     ...
-def get_max_num_args() -> int:
-    ...
 def get_max_num_indices() -> int:
     ...
 def get_num_elements(arg0: MeshPtr, arg1: MeshElementType) -> int:
     ...
 def get_python_package_dir() -> str:
     ...
-def get_relation_access(arg0: ..., arg1: Expr, arg2: ..., arg3: Expr, arg4: DebugInfo) -> Expr:
+def get_relation_access(arg0: ..., arg1: ExprCxx, arg2: ..., arg3: ExprCxx, arg4: DebugInfo) -> ExprCxx:
     ...
-def get_relation_size(arg0: ..., arg1: Expr, arg2: ..., arg3: DebugInfo) -> Expr:
+def get_relation_size(arg0: ..., arg1: ExprCxx, arg2: ..., arg3: DebugInfo) -> ExprCxx:
     ...
 def get_repo_dir() -> str:
     ...
@@ -2771,69 +2648,69 @@ def host_arch() -> Arch:
     ...
 def info(arg0: str) -> None:
     ...
-def insert_internal_func_call(arg0: ..., arg1: ExprGroup) -> Expr:
+def insert_internal_func_call(arg0: ..., arg1: ExprGroup) -> ExprCxx:
     ...
 def inverse_relation(arg0: MeshRelationType) -> MeshRelationType:
     ...
 def is_extension_supported(arg0: Arch, arg1: Extension) -> bool:
     ...
-def is_integral(arg0: DataType) -> bool:
+def is_integral(arg0: DataTypeCxx) -> bool:
     ...
-def is_quant(arg0: DataType) -> bool:
+def is_quant(arg0: DataTypeCxx) -> bool:
     ...
-def is_real(arg0: DataType) -> bool:
+def is_real(arg0: DataTypeCxx) -> bool:
     ...
-def is_signed(arg0: DataType) -> bool:
+def is_signed(arg0: DataTypeCxx) -> bool:
     ...
-def is_tensor(arg0: DataType) -> bool:
+def is_tensor(arg0: DataTypeCxx) -> bool:
     ...
-def is_unsigned(arg0: DataType) -> bool:
+def is_unsigned(arg0: DataTypeCxx) -> bool:
     ...
 def libdevice_path() -> str:
     ...
 def logging_effective(arg0: str) -> bool:
     ...
-def make_arg_load_expr(arg_id: list[int], dt: DataType, is_ptr: bool = False, create_load: bool = True, arg_depth: int = 0, dbg_info: DebugInfo = ...) -> Expr:
+def make_arg_load_expr(arg_id: tuple[int, ...], dt: DataTypeCxx, is_ptr: bool = False, create_load: bool = True, arg_depth: int = 0, dbg_info: DebugInfo = ...) -> ExprCxx:
     ...
-def make_binary_op_expr(arg0: BinaryOpType, arg1: Expr, arg2: Expr) -> Expr:
+def make_binary_op_expr(arg0: BinaryOpType, arg1: ExprCxx, arg2: ExprCxx) -> ExprCxx:
     ...
-def make_const_expr_bool(arg0: DataType, arg1: bool | np.bool_) -> Expr:
+def make_const_expr_bool(arg0: DataType, arg1: bool | np.bool_) -> ExprCxx:
     ...
-def make_const_expr_fp(arg0: DataType, arg1: float | np.floating) -> Expr:
+def make_const_expr_fp(arg0: DataType, arg1: float | np.floating) -> ExprCxx:
     ...
-def make_const_expr_int(arg0: DataType, arg1: int | np.int32 | np.int64) -> Expr:
+def make_const_expr_int(arg0: DataType, arg1: int | np.int32 | np.int64) -> ExprCxx:
     ...
 def make_cucg_solver(arg0: SparseMatrix, arg1: int, arg2: float, arg3: bool) -> CUCG:
     ...
-def make_cusparse_solver(arg0: DataType, arg1: str, arg2: str) -> SparseSolver:
+def make_cusparse_solver(arg0: DataTypeCxx, arg1: str, arg2: str) -> SparseSolver:
     ...
 def make_double_cg_solver(arg0: SparseMatrix, arg1: int, arg2: float, arg3: bool) -> CGd:
     ...
-def make_external_tensor_expr(arg0: DataType, arg1: int, arg2: list[int], arg3: bool, arg4: int, arg5: BoundaryMode) -> Expr:
+def make_external_tensor_expr(arg0: DataTypeCxx, arg1: int, arg2: tuple[int, ...], arg3: bool, arg4: int, arg5: BoundaryMode) -> ExprCxx:
     ...
-def make_external_tensor_grad_expr(arg0: Expr) -> Expr:
+def make_external_tensor_grad_expr(arg0: ExprCxx) -> ExprCxx:
     ...
 def make_float_cg_solver(arg0: SparseMatrix, arg1: int, arg2: float, arg3: bool) -> CGf:
     ...
-def make_frontend_assign_stmt(arg0: Expr, arg1: Expr, arg2: DebugInfo) -> Stmt:
+def make_frontend_assign_stmt(arg0: ExprCxx, arg1: ExprCxx, arg2: DebugInfo) -> Stmt:
     ...
-def make_get_element_expr(arg0: Expr, arg1: tuple[int, ...], arg2: DebugInfo) -> Expr:
+def make_get_element_expr(arg0: Expr, arg1: tuple[int, ...], arg2: DebugInfo) -> ExprCxx:
     ...
 def make_global_load_stmt(arg0: Stmt) -> Stmt:
     ...
 def make_global_store_stmt(arg0: Stmt, arg1: Stmt) -> Stmt:
     ...
-def make_rand_expr(arg0: DataType, arg1: DebugInfo) -> Expr:
+def make_rand_expr(arg0: DataTypeCxx, arg1: DebugInfo) -> ExprCxx:
     ...
-def make_reference(arg0: Expr, arg1: DebugInfo) -> Expr:
+def make_reference(arg0: ExprCxx, arg1: DebugInfo) -> ExprCxx:
     ...
-def make_rw_texture_ptr_expr(arg0: list[int], arg1: int, arg2: int, arg3: Format, arg4: int, arg5: DebugInfo) -> Expr:
+def make_rw_texture_ptr_expr(arg0: tuple[int, ...], arg1: int, arg2: int, arg3: Format, arg4: int, arg5: DebugInfo) -> ExprCxx:
     ...
-def make_sparse_solver(arg0: DataType, arg1: str, arg2: str) -> SparseSolver:
+def make_sparse_solver(arg0: DataTypeCxx, arg1: str, arg2: str) -> SparseSolver:
     ...
-def make_texture_ptr_expr(arg0: list[int], arg1: int, arg2: int, arg3: DebugInfo) -> Expr:
+def make_texture_ptr_expr(arg0: tuple[int, ...], arg1: int, arg2: int, arg3: DebugInfo) -> ExprCxx:
     ...
-def make_unary_op_expr(arg0: UnaryOpType, arg1: Expr) -> Expr:
+def make_unary_op_expr(arg0: UnaryOpType, arg1: ExprCxx) -> ExprCxx:
     ...
 def pop_python_print_buffer() -> str:
     ...
@@ -2841,7 +2718,7 @@ def print_all_units() -> None:
     ...
 def print_profile_info() -> None:
     ...
-def promoted_type(arg0: DataType, arg1: DataType) -> DataType:
+def promoted_type(arg0: DataTypeCxx, arg1: DataTypeCxx) -> DataTypeCxx:
     ...
 def query_int64(arg0: str) -> int:
     ...
@@ -2853,7 +2730,7 @@ def set_core_state_python_imported(arg0: bool) -> None:
     ...
 def set_core_trigger_gdb_when_crash(arg0: bool) -> None:
     ...
-def set_index_mapping(arg0: MeshPtr, arg1: MeshElementType, arg2: ConvType, arg3: SNode) -> None:
+def set_index_mapping(arg0: MeshPtr, arg1: MeshElementType, arg2: ConvType, arg3: SNodeCxx) -> None:
     ...
 def set_lib_dir(arg0: str) -> None:
     ...
@@ -2865,25 +2742,23 @@ def set_num_elements(arg0: MeshPtr, arg1: MeshElementType, arg2: int) -> None:
     ...
 def set_num_patches(arg0: MeshPtr, arg1: int) -> None:
     ...
-def set_owned_offset(arg0: MeshPtr, arg1: MeshElementType, arg2: SNode) -> None:
+def set_owned_offset(arg0: MeshPtr, arg1: MeshElementType, arg2: SNodeCxx) -> None:
     ...
 def set_patch_max_element_num(arg0: MeshPtr, arg1: MeshElementType, arg2: int) -> None:
     ...
 def set_python_package_dir(arg0: str) -> None:
     ...
-def set_relation_dynamic(arg0: MeshPtr, arg1: MeshRelationType, arg2: SNode, arg3: SNode, arg4: SNode) -> None:
+def set_relation_dynamic(arg0: MeshPtr, arg1: MeshRelationType, arg2: SNodeCxx, arg3: SNodeCxx, arg4: SNodeCxx) -> None:
     ...
-def set_relation_fixed(arg0: MeshPtr, arg1: MeshRelationType, arg2: SNode) -> None:
+def set_relation_fixed(arg0: MeshPtr, arg1: MeshRelationType, arg2: SNodeCxx) -> None:
     ...
 def set_tmp_dir(arg0: str) -> None:
     ...
-def set_total_offset(arg0: MeshPtr, arg1: MeshElementType, arg2: SNode) -> None:
-    ...
-def set_vulkan_visible_device(arg0: str) -> None:
+def set_total_offset(arg0: MeshPtr, arg1: MeshElementType, arg2: SNodeCxx) -> None:
     ...
 def start_memory_monitoring(arg0: str, arg1: int, arg2: float) -> None:
     ...
-def subscript_with_multiple_indices(arg0: Expr, arg1: list[ExprGroup], arg2: list[int], arg3: DebugInfo) -> Expr:
+def subscript_with_multiple_indices(arg0: ExprCxx, arg1: list[ExprGroup], arg2: tuple[int, ...], arg3: DebugInfo) -> ExprCxx:
     ...
 def test_cpp_exception() -> None:
     ...
@@ -2911,7 +2786,7 @@ def trigger_crash() -> None:
     ...
 def trigger_sig_fpe() -> int:
     ...
-def value_cast(arg0: Expr, arg1: DataType) -> Expr:
+def value_cast(arg0: ExprCxx, arg1: DataTypeCxx) -> ExprCxx:
     ...
 def wait_for_debugger() -> None:
     ...
@@ -2921,14 +2796,8 @@ def with_amdgpu() -> bool:
     ...
 def with_cuda() -> bool:
     ...
-def with_dx11() -> bool:
-    ...
-def with_dx12() -> bool:
-    ...
 def with_metal() -> bool:
     ...
-def with_opengl(arg0: bool) -> bool:
-    ...
 def with_vulkan() -> bool:
     ...
 ADJOINT: SNodeGradType  # value = <SNodeGradType.ADJOINT: 1>
@@ -2941,20 +2810,20 @@ CLAMP: BoundaryMode  # value = <BoundaryMode.CLAMP: 1>
 CV: MeshRelationType  # value = <MeshRelationType.CV: 12>
 Cell: MeshElementType  # value = <MeshElementType.Cell: 3>
 DUAL: SNodeGradType  # value = <SNodeGradType.DUAL: 2>
-DataType_f16: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_f32: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_f64: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_gen: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_i16: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_i32: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_i64: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_i8: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_u1: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_u16: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_u32: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_u64: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_u8: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
-DataType_unknown: DataType  # value = <taichi._lib.core.taichi_python.DataType object>
+DataType_f16: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_f32: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_f64: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_gen: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_i16: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_i32: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_i64: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_i8: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_u1: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_u16: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_u32: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_u64: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_u8: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
+DataType_unknown: DataTypeCxx  # value = <gstaichi._lib.core.gstaichi_python.DataTypeCxx object>
 EC: MeshRelationType  # value = <MeshRelationType.EC: 7>
 EE: MeshRelationType  # value = <MeshRelationType.EE: 5>
 EF: MeshRelationType  # value = <MeshRelationType.EF: 6>
@@ -2966,16 +2835,11 @@ FF: MeshRelationType  # value = <MeshRelationType.FF: 10>
 FORWARD: AutodiffMode  # value = <AutodiffMode.FORWARD: 0>
 FV: MeshRelationType  # value = <MeshRelationType.FV: 8>
 Face: MeshElementType  # value = <MeshElementType.Face: 2>
-MATRIX: ArgKind  # value = <ArgKind.MATRIX: 1>
-NDARRAY: ArgKind  # value = <ArgKind.NDARRAY: 2>
 NONE: AutodiffMode  # value = <AutodiffMode.NONE: 2>
 NULL: Layout  # value = <Layout.NULL: 2>
 PRIMAL: SNodeGradType  # value = <SNodeGradType.PRIMAL: 0>
 REVERSE: AutodiffMode  # value = <AutodiffMode.REVERSE: 1>
-RWTEXTURE: ArgKind  # value = <ArgKind.RWTEXTURE: 4>
-SCALAR: ArgKind  # value = <ArgKind.SCALAR: 0>
 SOA: Layout  # value = <Layout.SOA: 1>
-TEXTURE: ArgKind  # value = <ArgKind.TEXTURE: 3>
 Tetrahedron: MeshTopology  # value = <MeshTopology.Tetrahedron: 4>
 Triangle: MeshTopology  # value = <MeshTopology.Triangle: 3>
 UNSAFE: BoundaryMode  # value = <BoundaryMode.UNSAFE: 0>
@@ -2989,7 +2853,7 @@ abs: UnaryOpType  # value = <UnaryOpType.abs: 8>
 acos: UnaryOpType  # value = <UnaryOpType.acos: 13>
 add: BinaryOpType  # value = <BinaryOpType.add: 1>
 adstack: Extension  # value = <Extension.adstack: 5>
-amdgpu: Arch  # value = <Arch.amdgpu: 9>
+amdgpu: Arch  # value = <Arch.amdgpu: 6>
 arm64: Arch  # value = <Arch.arm64: 1>
 asin: UnaryOpType  # value = <UnaryOpType.asin: 11>
 assertion: Extension  # value = <Extension.assertion: 7>
@@ -3020,8 +2884,6 @@ cuda: Arch  # value = <Arch.cuda: 3>
 data64: Extension  # value = <Extension.data64: 4>
 dense: SNodeType  # value = <SNodeType.dense: 1>
 div: BinaryOpType  # value = <BinaryOpType.div: 5>
-dx11: Arch  # value = <Arch.dx11: 6>
-dx12: Arch  # value = <Arch.dx12: 7>
 dynamic: SNodeType  # value = <SNodeType.dynamic: 2>
 exp: UnaryOpType  # value = <UnaryOpType.exp: 18>
 extfunc: Extension  # value = <Extension.extfunc: 8>
@@ -3029,7 +2891,6 @@ floor: UnaryOpType  # value = <UnaryOpType.floor: 3>
 floordiv: BinaryOpType  # value = <BinaryOpType.floordiv: 4>
 frexp: UnaryOpType  # value = <UnaryOpType.frexp: 4>
 g2r: ConvType  # value = <ConvType.g2r: 2>
-gles: Arch  # value = <Arch.gles: 11>
 hash: SNodeType  # value = <SNodeType.hash: 5>
 inv: UnaryOpType  # value = <UnaryOpType.inv: 16>
 js: Arch  # value = <Arch.js: 2>
@@ -3050,8 +2911,7 @@ min: BinaryOpType  # value = <BinaryOpType.min: 8>
 mod: BinaryOpType  # value = <BinaryOpType.mod: 6>
 mul: BinaryOpType  # value = <BinaryOpType.mul: 0>
 neg: UnaryOpType  # value = <UnaryOpType.neg: 0>
-opencl: Arch  # value = <Arch.opencl: 8>
-opengl: Arch  # value = <Arch.opengl: 5>
+opencl: Arch  # value = <Arch.opencl: 5>
 place: SNodeType  # value = <SNodeType.place: 6>
 pointer: SNodeType  # value = <SNodeType.pointer: 3>
 popcnt: UnaryOpType  # value = <UnaryOpType.popcnt: 20>
@@ -3073,5 +2933,5 @@ tan: UnaryOpType  # value = <UnaryOpType.tan: 14>
 tanh: UnaryOpType  # value = <UnaryOpType.tanh: 15>
 truediv: BinaryOpType  # value = <BinaryOpType.truediv: 3>
 undefined: UnaryOpType  # value = <UnaryOpType.undefined: 25>
-vulkan: Arch  # value = <Arch.vulkan: 10>
+vulkan: Arch  # value = <Arch.vulkan: 7>
 x64: Arch  # value = <Arch.x64: 0>