navatala_gpu
Generated Python bindings and runtime facade for Navatala GPU operations.
Scope
- This package contains Python modules, a nanobind extension scaffold, machine-readable API metadata, and smoke tests for the curated public surface.
- The package links against the public Navatala GPU C ABI exposed by
navatala/navatala_ffi.h. - Raw kernel wrappers under
navatala_gpu.kernelsare an internal/advanced surface; most users should prefer the curated modules such asnavatala_gpu.runtimeandnavatala_gpu.linalg.
Build
The default package build uses the bundled public runtime in stub mode, so
pip install . works on hosts without GPU SDKs. Advanced users may link
against an already-built runtime by passing the public runtime include
directory and shared library:
cmake -S . -B build \
-DNAVATALA_FFI_INCLUDE_DIR=/path/to/gpu_runtime/include \
-DNAVATALA_FFI_LIBRARY=/path/to/libgpu_runtime.so
cmake --build build -j
For Python packaging, use a build frontend compatible with scikit-build-core.
Python Usage
import navatala_gpu
caps = navatala_gpu.get_capabilities()
assert "extension_loaded" in caps
assert isinstance(navatala_gpu.supports("axpy", backend="opencl", dtype="float32"), bool)
Runtime Notes
import navatala_gpuis designed to be safe on hosts without GPU drivers.- Backend availability is reported through
navatala_gpu.get_capabilities()andnavatala_gpu.supports(...). - NumPy interop uses explicit host/device copies through
Tensor.from_numpy(...)andTensor.to_numpy(...). - DLPack interop is zero-copy only where the runtime can prove backend compatibility.
Manifest id: pyabi6-2675513039
1 maintainer
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
navatala_gpu-0.1.4.tar.gz
(4.3 MB
view details)
Built Distributions
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
