warp-lang 1.7.0__py3-none-manylinux_2_28_x86_64.whl

warp/tests/test_fp16.py

@@ -0,0 +1,136 @@
+# SPDX-FileCopyrightText: Copyright (c) 2023 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import unittest
+
+import numpy as np
+
+import warp as wp
+from warp.tests.unittest_utils import *
+
+
+@wp.kernel
+def load_store_half(f32: wp.array(dtype=wp.float32), f16: wp.array(dtype=wp.float16)):
+    tid = wp.tid()
+
+    # check conversion from f32->f16
+    a = wp.float16(f32[tid])
+    b = f16[tid]
+
+    wp.expect_eq(a, b)
+
+    # check stores
+    f16[tid] = a
+
+
+def test_fp16_conversion(test, device):
+    s = [1.0, 2.0, 3.0, -3.14159]
+
+    np_f32 = np.array(s, dtype=np.float32)
+    np_f16 = np.array(s, dtype=np.float16)
+
+    wp_f32 = wp.array(s, dtype=wp.float32, device=device)
+    wp_f16 = wp.array(s, dtype=wp.float16, device=device)
+
+    assert_np_equal(np_f32, wp_f32.numpy())
+    assert_np_equal(np_f16, wp_f16.numpy())
+
+    wp.launch(load_store_half, dim=len(s), inputs=[wp_f32, wp_f16], device=device)
+
+    # check that stores worked
+    assert_np_equal(np_f16, wp_f16.numpy())
+
+
+@wp.kernel
+def value_load_store_half(f16_value: wp.float16, f16_array: wp.array(dtype=wp.float16)):
+    wp.expect_eq(f16_value, f16_array[0])
+
+    # check stores
+    f16_array[0] = f16_value
+
+
+def test_fp16_kernel_parameter(test, device):
+    """Test the ability to pass in fp16 into kernels as parameters"""
+
+    s = [1.0, 2.0, 3.0, -3.14159]
+
+    for test_val in s:
+        np_f16 = np.array([test_val], dtype=np.float16)
+        wp_f16 = wp.array([test_val], dtype=wp.float16, device=device)
+
+        wp.launch(value_load_store_half, (1,), inputs=[wp.float16(test_val), wp_f16], device=device)
+
+        # check that stores worked
+        assert_np_equal(np_f16, wp_f16.numpy())
+
+        # Do the same thing but pass in test_val as a Python float to test automatic conversion
+        wp_f16 = wp.array([test_val], dtype=wp.float16, device=device)
+        wp.launch(value_load_store_half, (1,), inputs=[test_val, wp_f16], device=device)
+        assert_np_equal(np_f16, wp_f16.numpy())
+
+
+@wp.kernel
+def mul_half(input: wp.array(dtype=wp.float16), output: wp.array(dtype=wp.float16)):
+    tid = wp.tid()
+
+    # convert to compute type fp32
+    x = wp.float(input[tid]) * 2.0
+
+    # store back as fp16
+    output[tid] = wp.float16(x)
+
+
+def test_fp16_grad(test, device):
+    rng = np.random.default_rng(123)
+
+    # checks that gradients are correctly propagated for
+    # fp16 arrays, even when intermediate calculations
+    # are performed in e.g.: fp32
+
+    s = rng.random(size=15).astype(np.float16)
+
+    input = wp.array(s, dtype=wp.float16, device=device, requires_grad=True)
+    output = wp.zeros_like(input)
+
+    tape = wp.Tape()
+    with tape:
+        wp.launch(mul_half, dim=len(s), inputs=[input, output], device=device)
+
+    ones = wp.array(np.ones(len(output)), dtype=wp.float16, device=device)
+
+    tape.backward(grads={output: ones})
+
+    assert_np_equal(input.grad.numpy(), np.ones(len(s)) * 2.0)
+
+
+class TestFp16(unittest.TestCase):
+    pass
+
+
+devices = []
+if wp.is_cpu_available():
+    devices.append("cpu")
+for cuda_device in get_selected_cuda_test_devices():
+    if cuda_device.arch >= 70:
+        devices.append(cuda_device)
+
+add_function_test(TestFp16, "test_fp16_conversion", test_fp16_conversion, devices=devices)
+add_function_test(TestFp16, "test_fp16_grad", test_fp16_grad, devices=devices)
+add_function_test(TestFp16, "test_fp16_kernel_parameter", test_fp16_kernel_parameter, devices=devices)
+
+
+if __name__ == "__main__":
+    wp.clear_kernel_cache()
+    unittest.main(verbosity=2)

warp/tests/test_func.py

@@ -0,0 +1,454 @@
+# SPDX-FileCopyrightText: Copyright (c) 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import math
+import unittest
+from typing import Any, Tuple
+
+import numpy as np
+
+import warp as wp
+from warp.tests.unittest_utils import *
+
+
+@wp.func
+def sqr(x: float):
+    return x * x
+
+
+# test nested user function calls
+# and explicit return type hints
+@wp.func
+def cube(x: float) -> float:
+    return sqr(x) * x
+
+
+@wp.func
+def custom(x: int):
+    return x + 1
+
+
+@wp.func
+def custom(x: float):
+    return x + 1.0
+
+
+@wp.func
+def custom(x: wp.vec3):
+    return x + wp.vec3(1.0, 0.0, 0.0)
+
+
+@wp.func
+def noreturn(x: wp.vec3):
+    x = x + wp.vec3(0.0, 1.0, 0.0)
+
+    wp.expect_eq(x, wp.vec3(1.0, 1.0, 0.0))
+
+
+@wp.kernel
+def test_overload_func():
+    # tests overloading a custom @wp.func
+
+    i = custom(1)
+    f = custom(1.0)
+    v = custom(wp.vec3(1.0, 0.0, 0.0))
+
+    wp.expect_eq(i, 2)
+    wp.expect_eq(f, 2.0)
+    wp.expect_eq(v, wp.vec3(2.0, 0.0, 0.0))
+
+    noreturn(wp.vec3(1.0, 0.0, 0.0))
+
+
+@wp.func
+def foo(x: int):
+    # This shouldn't be picked up.
+    return x * 2
+
+
+@wp.func
+def foo(x: int):
+    return x * 3
+
+
+@wp.kernel
+def test_override_func():
+    i = foo(1)
+    wp.expect_eq(i, 3)
+
+
+def test_func_closure_capture(test, device):
+    def make_closure_kernel(func):
+        def closure_kernel_fn(data: wp.array(dtype=float), expected: float):
+            f = func(data[wp.tid()])
+            wp.expect_eq(f, expected)
+
+        return wp.Kernel(func=closure_kernel_fn)
+
+    sqr_closure = make_closure_kernel(sqr)
+    cube_closure = make_closure_kernel(cube)
+
+    data = wp.array([2.0], dtype=float, device=device)
+    expected_sqr = 4.0
+    expected_cube = 8.0
+
+    wp.launch(sqr_closure, dim=data.shape, inputs=[data, expected_sqr], device=device)
+    wp.launch(cube_closure, dim=data.shape, inputs=[data, expected_cube], device=device)
+
+
+@wp.func
+def test_func(param1: wp.int32, param2: wp.int32, param3: wp.int32) -> wp.float32:
+    return 1.0
+
+
+@wp.kernel
+def test_return_kernel(test_data: wp.array(dtype=wp.float32)):
+    tid = wp.tid()
+    test_data[tid] = wp.lerp(test_func(0, 1, 2), test_func(0, 1, 2), 0.5)
+
+
+def test_return_func(test, device):
+    test_data = wp.zeros(100, dtype=wp.float32, device=device)
+    wp.launch(kernel=test_return_kernel, dim=test_data.size, inputs=[test_data], device=device)
+
+
+@wp.func
+def multi_valued_func(a: wp.float32, b: wp.float32):
+    return a + b, a - b, a * b, a / b
+
+
+def test_multi_valued_func(test, device):
+    @wp.kernel
+    def test_multi_valued_kernel(test_data1: wp.array(dtype=wp.float32), test_data2: wp.array(dtype=wp.float32)):
+        tid = wp.tid()
+        d1, d2 = test_data1[tid], test_data2[tid]
+        a, b, c, d = multi_valued_func(d1, d2)
+        wp.expect_eq(a, d1 + d2)
+        wp.expect_eq(b, d1 - d2)
+        wp.expect_eq(c, d1 * d2)
+        wp.expect_eq(d, d1 / d2)
+
+    test_data1 = wp.array(np.arange(100), dtype=wp.float32, device=device)
+    test_data2 = wp.array(np.arange(100, 0, -1), dtype=wp.float32, device=device)
+    wp.launch(kernel=test_multi_valued_kernel, dim=test_data1.size, inputs=[test_data1, test_data2], device=device)
+
+
+@wp.kernel
+def test_func_defaults():
+    # test default as expected
+    wp.expect_near(1.0, 1.0 + 1.0e-6)
+
+    # test that changing tolerance still works
+    wp.expect_near(1.0, 1.1, 0.5)
+
+
+@wp.func
+def sign(x: float):
+    return 123.0
+
+
+@wp.kernel
+def test_builtin_shadowing():
+    wp.expect_eq(sign(1.23), 123.0)
+
+
+@wp.func
+def user_func_with_defaults(a: int = 123, b: int = 234) -> int:
+    return a + b
+
+
+@wp.kernel
+def user_func_with_defaults_kernel():
+    a = user_func_with_defaults()
+    wp.expect_eq(a, 357)
+
+    b = user_func_with_defaults(111)
+    wp.expect_eq(b, 345)
+
+    c = user_func_with_defaults(111, 222)
+    wp.expect_eq(c, 333)
+
+    d = user_func_with_defaults(a=111)
+    wp.expect_eq(d, 345)
+
+    e = user_func_with_defaults(b=111)
+    wp.expect_eq(e, 234)
+
+
+def test_user_func_with_defaults(test, device):
+    wp.launch(user_func_with_defaults_kernel, dim=1, device=device)
+
+    a = user_func_with_defaults()
+    assert a == 357
+
+    b = user_func_with_defaults(111)
+    assert b == 345
+
+    c = user_func_with_defaults(111, 222)
+    assert c == 333
+
+    d = user_func_with_defaults(a=111)
+    assert d == 345
+
+    e = user_func_with_defaults(b=111)
+    assert e == 234
+
+
+@wp.func
+def user_func_return_multiple_values(a: int, b: float) -> Tuple[int, float]:
+    return a + a, b * b
+
+
+@wp.kernel
+def test_user_func_return_multiple_values():
+    a, b = user_func_return_multiple_values(123, 234.0)
+    wp.expect_eq(a, 246)
+    wp.expect_eq(b, 54756.0)
+
+
+@wp.func
+def user_func_overload(
+    b: wp.array(dtype=Any),
+    i: int,
+):
+    return b[i] * 2.0
+
+
+@wp.kernel
+def user_func_overload_resolution_kernel(
+    a: wp.array(dtype=Any),
+    b: wp.array(dtype=Any),
+):
+    i = wp.tid()
+    a[i] = user_func_overload(b, i)
+
+
+def test_user_func_overload_resolution(test, device):
+    a0 = wp.array((1, 2, 3), dtype=wp.vec3)
+    b0 = wp.array((2, 3, 4), dtype=wp.vec3)
+
+    a1 = wp.array((5,), dtype=float)
+    b1 = wp.array((6,), dtype=float)
+
+    wp.launch(user_func_overload_resolution_kernel, a0.shape, (a0, b0))
+    wp.launch(user_func_overload_resolution_kernel, a1.shape, (a1, b1))
+
+    assert_np_equal(a0.numpy()[0], (4, 6, 8))
+    assert a1.numpy()[0] == 12
+
+
+@wp.func
+def user_func_return_none() -> None:
+    pass
+
+
+@wp.kernel
+def test_return_annotation_none() -> None:
+    user_func_return_none()
+
+
+devices = get_test_devices()
+
+
+class TestFunc(unittest.TestCase):
+    def test_user_func_export(self):
+        # tests calling overloaded user-defined functions from Python
+        i = custom(1)
+        f = custom(1.0)
+        v = custom(wp.vec3(1.0, 0.0, 0.0))
+
+        self.assertEqual(i, 2)
+        self.assertEqual(f, 2.0)
+        assert_np_equal(np.array([*v]), np.array([2.0, 0.0, 0.0]))
+
+    def test_native_func_export(self):
+        # tests calling native functions from Python
+
+        q = wp.quat(0.0, 0.0, 0.0, 1.0)
+        assert_np_equal(np.array([*q]), np.array([0.0, 0.0, 0.0, 1.0]))
+
+        r = wp.quat_from_axis_angle(wp.vec3(1.0, 0.0, 0.0), 2.0)
+        assert_np_equal(np.array([*r]), np.array([0.8414709568023682, 0.0, 0.0, 0.5403022170066833]), tol=1.0e-3)
+
+        q = wp.quat(1.0, 2.0, 3.0, 4.0)
+        q = wp.normalize(q) * 2.0
+        assert_np_equal(
+            np.array([*q]),
+            np.array([0.18257418274879456, 0.3651483654975891, 0.547722578048706, 0.7302967309951782]) * 2.0,
+            tol=1.0e-3,
+        )
+
+        v2 = wp.vec2(1.0, 2.0)
+        v2 = wp.normalize(v2) * 2.0
+        assert_np_equal(np.array([*v2]), np.array([0.4472135901451111, 0.8944271802902222]) * 2.0, tol=1.0e-3)
+
+        v3 = wp.vec3(1.0, 2.0, 3.0)
+        v3 = wp.normalize(v3) * 2.0
+        assert_np_equal(
+            np.array([*v3]), np.array([0.26726123690605164, 0.5345224738121033, 0.8017836809158325]) * 2.0, tol=1.0e-3
+        )
+
+        v4 = wp.vec4(1.0, 2.0, 3.0, 4.0)
+        v4 = wp.normalize(v4) * 2.0
+        assert_np_equal(
+            np.array([*v4]),
+            np.array([0.18257418274879456, 0.3651483654975891, 0.547722578048706, 0.7302967309951782]) * 2.0,
+            tol=1.0e-3,
+        )
+
+        v = wp.vec2(0.0)
+        v += wp.vec2(1.0, 1.0)
+        assert v == wp.vec2(1.0, 1.0)
+        v -= wp.vec2(1.0, 1.0)
+        assert v == wp.vec2(0.0, 0.0)
+        v = wp.vec2(2.0, 2.0) - wp.vec2(1.0, 1.0)
+        assert v == wp.vec2(1.0, 1.0)
+        v *= 2.0
+        assert v == wp.vec2(2.0, 2.0)
+        v = v * 2.0
+        assert v == wp.vec2(4.0, 4.0)
+        v = v / 2.0
+        assert v == wp.vec2(2.0, 2.0)
+        v /= 2.0
+        assert v == wp.vec2(1.0, 1.0)
+        v = -v
+        assert v == wp.vec2(-1.0, -1.0)
+        v = +v
+        assert v == wp.vec2(-1.0, -1.0)
+
+        m22 = wp.mat22(1.0, 2.0, 3.0, 4.0)
+        m22 = m22 + m22
+
+        self.assertEqual(m22[1, 1], 8.0)
+        self.assertEqual(str(m22), "[[2.0, 4.0],\n [6.0, 8.0]]")
+
+        t = wp.transform(
+            wp.vec3(1.0, 2.0, 3.0),
+            wp.quat(4.0, 5.0, 6.0, 7.0),
+        )
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+        self.assertSequenceEqual(
+            t * wp.transform(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0), (396.0, 432.0, 720.0, 56.0, 70.0, 84.0, -28.0)
+        )
+        self.assertSequenceEqual(
+            t * wp.transform((1.0, 2.0, 3.0), (4.0, 5.0, 6.0, 7.0)), (396.0, 432.0, 720.0, 56.0, 70.0, 84.0, -28.0)
+        )
+
+        t = wp.transform()
+        self.assertSequenceEqual(t, (0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0))
+
+        t = wp.transform(p=(1.0, 2.0, 3.0), q=(4.0, 5.0, 6.0, 7.0))
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+
+        t = wp.transform(q=(4.0, 5.0, 6.0, 7.0), p=(1.0, 2.0, 3.0))
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+
+        t = wp.transform((1.0, 2.0, 3.0), q=(4.0, 5.0, 6.0, 7.0))
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+
+        t = wp.transform(p=(1.0, 2.0, 3.0))
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 0.0, 0.0, 0.0, 1.0))
+
+        t = wp.transform(q=(4.0, 5.0, 6.0, 7.0))
+        self.assertSequenceEqual(t, (0.0, 0.0, 0.0, 4.0, 5.0, 6.0, 7.0))
+
+        t = wp.transform((1.0, 2.0, 3.0), (4.0, 5.0, 6.0, 7.0))
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+
+        t = wp.transform(p=wp.vec3(1.0, 2.0, 3.0), q=wp.quat(4.0, 5.0, 6.0, 7.0))
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+
+        t = wp.transform(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0)
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+
+        t = wp.transform(wp.transform(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+
+        t = wp.transform(*wp.transform(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+        self.assertSequenceEqual(t, (1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0))
+
+        transformf = wp.types.transformation(dtype=float)
+
+        t = wp.transformf((1.0, 2.0, 3.0), (4.0, 5.0, 6.0, 7.0))
+        self.assertSequenceEqual(
+            t + transformf((2.0, 3.0, 4.0), (5.0, 6.0, 7.0, 8.0)),
+            (3.0, 5.0, 7.0, 9.0, 11.0, 13.0, 15.0),
+        )
+        self.assertSequenceEqual(
+            t - transformf((2.0, 3.0, 4.0), (5.0, 6.0, 7.0, 8.0)),
+            (-1.0, -1.0, -1.0, -1.0, -1.0, -1.0, -1.0),
+        )
+
+        f = wp.sin(math.pi * 0.5)
+        self.assertAlmostEqual(f, 1.0, places=3)
+
+        m = wp.mat22(0.0, 0.0, 0.0, 0.0)
+        m += wp.mat22(1.0, 1.0, 1.0, 1.0)
+        assert m == wp.mat22(1.0, 1.0, 1.0, 1.0)
+        m -= wp.mat22(1.0, 1.0, 1.0, 1.0)
+        assert m == wp.mat22(0.0, 0.0, 0.0, 0.0)
+        m = wp.mat22(2.0, 2.0, 2.0, 2.0) - wp.mat22(1.0, 1.0, 1.0, 1.0)
+        assert m == wp.mat22(1.0, 1.0, 1.0, 1.0)
+        m *= 2.0
+        assert m == wp.mat22(2.0, 2.0, 2.0, 2.0)
+        m = m * 2.0
+        assert m == wp.mat22(4.0, 4.0, 4.0, 4.0)
+        m = m / 2.0
+        assert m == wp.mat22(2.0, 2.0, 2.0, 2.0)
+        m /= 2.0
+        assert m == wp.mat22(1.0, 1.0, 1.0, 1.0)
+        m = -m
+        assert m == wp.mat22(-1.0, -1.0, -1.0, -1.0)
+        m = +m
+        assert m == wp.mat22(-1.0, -1.0, -1.0, -1.0)
+        m = m * m
+        assert m == wp.mat22(2.0, 2.0, 2.0, 2.0)
+
+    def test_native_function_error_resolution(self):
+        a = wp.mat22f(1.0, 2.0, 3.0, 4.0)
+        b = wp.mat22d(1.0, 2.0, 3.0, 4.0)
+        with self.assertRaisesRegex(
+            RuntimeError,
+            r"^Couldn't find a function 'mul' compatible with " r"the arguments 'mat22f, mat22d'$",
+        ):
+            a * b
+
+
+add_kernel_test(TestFunc, kernel=test_overload_func, name="test_overload_func", dim=1, devices=devices)
+add_function_test(TestFunc, func=test_return_func, name="test_return_func", devices=devices)
+add_kernel_test(TestFunc, kernel=test_override_func, name="test_override_func", dim=1, devices=devices)
+add_function_test(TestFunc, func=test_func_closure_capture, name="test_func_closure_capture", devices=devices)
+add_function_test(TestFunc, func=test_multi_valued_func, name="test_multi_valued_func", devices=devices)
+add_kernel_test(TestFunc, kernel=test_func_defaults, name="test_func_defaults", dim=1, devices=devices)
+add_kernel_test(TestFunc, kernel=test_builtin_shadowing, name="test_builtin_shadowing", dim=1, devices=devices)
+add_function_test(TestFunc, func=test_user_func_with_defaults, name="test_user_func_with_defaults", devices=devices)
+add_kernel_test(
+    TestFunc,
+    kernel=test_user_func_return_multiple_values,
+    name="test_user_func_return_multiple_values",
+    dim=1,
+    devices=devices,
+)
+add_function_test(
+    TestFunc, func=test_user_func_overload_resolution, name="test_user_func_overload_resolution", devices=devices
+)
+add_kernel_test(
+    TestFunc, kernel=test_return_annotation_none, name="test_return_annotation_none", dim=1, devices=devices
+)
+
+
+if __name__ == "__main__":
+    wp.clear_kernel_cache()
+    unittest.main(verbosity=2)

warp/tests/test_future_annotations.py

@@ -0,0 +1,98 @@
+# SPDX-FileCopyrightText: Copyright (c) 2024 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+# http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# This is what we are actually testing.
+from __future__ import annotations
+
+import unittest
+
+import warp as wp
+from warp.tests.unittest_utils import *
+
+
+@wp.struct
+class FooData:
+    x: float
+    y: float
+
+
+class Foo:
+    Data = FooData
+
+    @wp.func
+    def compute():
+        pass
+
+
+@wp.kernel
+def kernel_1(
+    out: wp.array(dtype=float),
+):
+    tid = wp.tid()
+
+
+@wp.kernel
+def kernel_2(
+    out: wp.array(dtype=float),
+):
+    tid = wp.tid()
+    out[tid] = 1.23
+
+
+def create_kernel_3(foo: Foo):
+    def fn(
+        data: foo.Data,
+        out: wp.array(dtype=float),
+    ):
+        tid = wp.tid()
+
+        # Referencing a variable in a type hint like `foo.Data` isn't officially
+        # accepted by Python but it's still being used in some places (e.g.: `warp.fem`)
+        # where it works only because the variable being referenced within the function,
+        # which causes it to be promoted to a closure variable. Without that,
+        # it wouldn't be possible to resolve `foo` and to evaluate the `foo.Data`
+        # string to its corresponding type.
+        foo.compute()
+
+        out[tid] = data.x + data.y
+
+    return wp.Kernel(func=fn)
+
+
+def test_future_annotations(test, device):
+    foo = Foo()
+    foo_data = FooData()
+    foo_data.x = 1.23
+    foo_data.y = 2.34
+
+    out = wp.empty(1, dtype=float)
+
+    kernel_3 = create_kernel_3(foo)
+
+    wp.launch(kernel_1, dim=out.shape, outputs=(out,))
+    wp.launch(kernel_2, dim=out.shape, outputs=(out,))
+    wp.launch(kernel_3, dim=out.shape, inputs=(foo_data,), outputs=(out,))
+
+
+class TestFutureAnnotations(unittest.TestCase):
+    pass
+
+
+add_function_test(TestFutureAnnotations, "test_future_annotations", test_future_annotations)
+
+
+if __name__ == "__main__":
+    wp.clear_kernel_cache()
+    unittest.main(verbosity=2)