Release notes ============= Being an experimental research framework, Dr.Jit does not strictly follow the `Semantic Versioning `_ convention. That said, we will strive to document breaking API changes in the release notes below. Incoming release ---------------- - ... Dr.Jit 0.2.2 ---------------- *September 12, 2022* - Add bindings for ``dr.llvm_version()`` `[07e9da8] `_ - Fix ``dr.make_opaque`` for diff TensorXf `[f6bde89] `_ - Change behavior of ``dr.diff_array_t`` to always return a type `[e0172fc] `_ - Compilation fixes for stub files generation `[bf770d4] `_ - Add missing operations on Tensor types `[15d490c] `_ - Fix dr.shape for ragged arrays `[a026b56] `_ - Add the ``dr.wrap_ad()`` function decorator for interoperability between AD-aware frameworks (e.g. Dr.Jit and PyTorch) `[4a1528e] `_ - ``dr.device`` handles the case where the input was allocated by another framework `[9e993a6] `_ Dr.Jit 0.2.1 ------------- *July 20, 2022* - Fix upsampling of multichannel textures `[53dd605] `_ Dr.Jit 0.2.0 ------------- *July 19, 2022* - Change cubic texture gradient and hessian functions to also return the texture values `[1d50efe] `_ - Add support for non-array types in ``dr.CustomOp`` inputs - Various minor fixes in C++ test suite `[dcaf69a] `_ Dr.Jit 0.1.0 ------------- *July 18, 2022* - Initial release