Eagerly use all vended numba.cuda modules, Bump supported Numba-CUDA version to 0.21+ (#239)

This PR bumps Numbast compatible Numba-CUDA version to 0.21.0+. And
properly documents each version's Numba-CUDA version pinning at FAQ
section.

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **Documentation**
* Added guidance on numba-cuda compatibility and binding-generation
requirements.

* **Chores**
* Raised and broadened numba-cuda version constraints to
>=0.21.0,<0.23.0 across environment and manifest files.
* Aligned project imports and build surfaces to CUDA-targeted APIs
consistent with the updated constraint.

* **Tests**
* Updated test suites to import and use CUDA-specific types and
datamodels.

* **New Files**
  * Added a small CUDA header used by tests.

<sub>✏️ Tip: You can customize this high-level summary in your review
settings.</sub>
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

---------

Co-authored-by: Michael Wang <isVoid@users.noreply.github.com>

Author: isVoid

ci/test_conda_python.sh

@@ -23,7 +23,7 @@ rapids-mamba-retry create -n test \
   cuda-version=${RAPIDS_CUDA_VERSION%.*} \
   cuda-nvrtc \
   numba >=0.59 \
-  "numba-cuda>=0.20.1,<0.21.0" \
+  "numba-cuda>=0.21.0,<0.23.0" \
   cuda-cudart-dev \
   python=${RAPIDS_PY_VERSION} \
   cffi \

conda/environment-cu12.yaml

@@ -9,7 +9,7 @@ dependencies:
   - python>=3.10
   - cmake>=3.29 # Allow overriding CMAKE_PREFIX_PATH and CMAKE_INSTALL_PREFIX
   - clangdev >=18,<22.0
-  - numba-cuda[cu12] >=0.20.1,<0.21.0
+  - numba-cuda[cu12] >=0.21.0,<0.23.0
   - pybind11
   - pytest
   - pytest-benchmark

conda/environment-cu13.yaml

@@ -9,7 +9,7 @@ dependencies:
   - python>=3.10
   - cmake>=3.29 # Allow overriding CMAKE_PREFIX_PATH and CMAKE_INSTALL_PREFIX
   - clangdev >=18,<22.0
-  - numba-cuda[cu13] >=0.20.1,<0.21.0
+  - numba-cuda[cu13] >=0.21.0,<0.23.0
   - pybind11
   - pytest
   - pytest-benchmark

conda/environment_template.yaml

@@ -9,7 +9,7 @@ dependencies:
   - python={{ python_version }}
   - cmake>=3.29 # Allow overriding CMAKE_PREFIX_PATH and CMAKE_INSTALL_PREFIX
   - clangdev >=18,<22.0
-  - numba-cuda >=0.20.1,<0.21.0
+  - numba-cuda >=0.21.0,<0.23.0
   - pybind11
   - pytest
   - pytest-benchmark

conda/recipes/numbast/meta.yaml

@@ -52,7 +52,7 @@ requirements:
     - cuda-version >=12.5
     - numba >=0.59
     - python
-    - numba-cuda >=0.20.1,<0.21.0
+    - numba-cuda >=0.21.0,<0.23.0
     - pyyaml
     - click
     - jinja2

conda/recipes/numbast_extensions/meta.yaml

@@ -43,7 +43,7 @@ requirements:
     - cuda-cudart-dev
     - python
     - numba >=0.59
-    - numba-cuda >=0.20.1,<0.21.0
+    - numba-cuda >=0.21.0,<0.23.0
     - numbast >=0.2.0
 
   test:

docs/source/faq.rst

@@ -122,3 +122,48 @@ Notes
   libc++, ensure its headers are installed (e.g., ``/usr/include/c++/v1`` or Conda ``libcxx-devel``) and pass
   ``-stdlib=libc++``; otherwise libstdc++ is typically selected by default on Linux.
 - AST Canopy is linked against Clang 20. For host resources, please use corresponding Clang version.
+
+
+Generated bindings and Numba-CUDA version requirements
+-------------------------------------------------------
+
+**What version of numba-cuda do generated bindings require?**
+
+Bindings generated by Numbast have specific version requirements for ``numba-cuda`` at runtime. The version of
+Numbast used to generate the bindings determines the compatible ``numba-cuda`` versions.
+
+.. list-table:: Numbast to numba-cuda compatibility
+   :header-rows: 1
+   :widths: 30 70
+
+   * - Numbast Version
+     - Required numba-cuda Version
+   * - 0.6.0 (current dev)
+     - ``>=0.21.0,<0.23.0``
+   * - 0.5.x
+     - ``>=0.20.1,<0.21.0``
+
+**Why do generated bindings have version requirements?**
+
+Numbast generates Python code that uses Numba-CUDA's internal APIs. These APIs can change between releases, so
+bindings generated with a specific version of Numbast are tested against a specific range of ``numba-cuda`` versions.
+
+**How do I ensure compatibility?**
+
+*For dynamic binding generation:*
+
+- The correct ``numba-cuda`` version constraints are automatically enforced at the package dependency level
+  and managed by your package manager (pip or conda). When you install Numbast, compatible versions of
+  ``numba-cuda`` are installed automatically via the dependencies specified in ``pyproject.toml`` and Conda
+  environment files.
+
+*For static binding generation:*
+
+- When distributing generated bindings, document the required ``numba-cuda`` version range in your package
+  dependencies so users can install a compatible version.
+- Generated static bindings (see :doc:`Static binding generation <static>`) can be regenerated with newer
+  versions of Numbast if you need to support newer ``numba-cuda`` releases.
+
+.. note::
+   These version restrictions may be relaxed or removed once ``numba-cuda`` releases a stable 1.0 version with
+   stabilized public APIs. Until then, bindings are tested against specific version ranges to ensure compatibility.

numbast/benchmarks/test_arithmetic.py

@@ -1,6 +1,6 @@
 import numba.cuda as cuda
+from numba.cuda.types import float32
 import numpy as np
-from numba import float32
 
 import pytest
 

numbast/pyproject.toml

@@ -18,7 +18,7 @@ classifiers = [
 ]
 dependencies = [
   "numba>=0.59.0",
-  "numba-cuda>=0.20.1,<0.21.0",
+  "numba-cuda>=0.21.0,<0.23.0",
   "ast_canopy>=0.5.0",
   "pyyaml",
   "click",
@@ -41,12 +41,12 @@ repository = "https://github.com/NVIDIA/numbast"
 [project.optional-dependencies]
 dev = ["ruff"]
 docs = ["sphinx>=7.0", "sphinx-copybutton>=0.5.2"]
-test-cu12 = ["pytest", "cffi", "numba-cuda[cu12]>=0.20.1,<0.21.0"]
+test-cu12 = ["pytest", "cffi", "numba-cuda[cu12]>=0.21.0,<0.23.0"]
 test-cu13 = [
   "pytest",
   "cffi",
   "cuda-toolkit[cudart,crt,curand,cccl,nvcc]==13.*",
-  "numba-cuda[cu13]>=0.20.1,<0.21.0",
+  "numba-cuda[cu13]>=0.21.0,<0.23.0",
 ]
 
 

numbast/src/numbast/class_template.py

@@ -24,7 +24,7 @@
 from numba.cuda import declare_device
 from numba.cuda.cudadecl import register_global, register, register_attr
 from numba.cuda.cudaimpl import lower
-from numba.core.imputils import numba_typeref_ctor
+from numba.cuda.core.imputils import numba_typeref_ctor
 from numba.core.typing.npydecl import parse_dtype
 from numba.core.errors import RequireLiteralValue, TypingError