v1.2.0

• Added NeuralTextures library and NeuralTextureViewer example.
• Renamed cubic texturing entry points to align with existing naming conventions.
• Updated documentation for stochastic texture filtering.
• OptiX SDK headers are automatically downloaded from GitHub if preinstalled SDK is not found.
• Reduced Cmake configuration time overhead.
• Improved CMake install/export logic, particularly when building a subset of libraries.
• Fixed CMake 4.2 issue involving third-party dependencies with legacy CMake minimum versions.
• PyOptiX submodule changes:
  • CMake 4.x support.
  • Windows installation improvements (DLL loading for Python 3.8+)
  • Updated pybind11 to v2.13.6

v1.1.0

• Improved OptiX SDK integration: Automatically downloads OptiX headers using FetchContent when find_package(OptiX) fails, simplifying the build process for projects without a locally installed SDK
• Enhanced DDS file support: Added capability to read and write tiled DDS files, along with fixes to tile width calculations
• Added compressed texture cache: Implemented compressed texture caching with automatic conversion retry logic and a new example application demonstrating the feature
• Fixed texture filtering derivatives: Corrected derivative calculations in bilinear sampling where scale factors were incorrectly applied
• Fixed cubic texture filtering: Resolved an issue where textures with large gradients (|grad| > 1) incorrectly returned resident=false, which could cause invalid texture lookups
• Updated UDIM handling: Revised UDIM texture coordinate handling to better align with industry-standard practices
• Improved build compatibility: Fixed MSVC build issues and updated vcpkg manifest to properly handle Boost dependencies for OpenImageIO

v1.0.1

• Added support for block compressed formats (.dds files) in the demand load library.
• Made numerous improvements to cubic filtering.
• Fixed a texture tile indexing error in DemandLoading
• Fixed pixel stride calculation in PbrtAlphaMapImageSource
• Enhanced compatibility: added support for CUDA 13.0, OptiX 8.1+, and fixed various build issues (Windows, OptiX, C++ standards).
• Added support for installation of OptiX Toolkit.
• Refactored and improved debug utilities and statistics reporting.
• Improved test coverage for demand loading and geometry.
• Refactored code for better maintainability: extracted and renamed libraries, improved modularity, and cleaned up includes.
• Updated dependencies and build scripts for better compatibility with vcpkg and OpenImageIO 3.x.
• Improved logging, error handling, and documentation throughout the toolkit.
• Various bug fixes, code cleanups and enhancements.

v1.0.0

• Combined most submodules into the main repository (except otk-pyoptix).
• Fixed demand loading error in TCC mode (in cuMemGetAllocationGranularity).

v0.9.7

Added stochastic texture filtering and cubic filtering.

v0.9.6

Added READMEs for PBRT-related libraries and examples (see the examples submodule).

v0.9.5

Added demand-loaded geometry sample that supports very large PBRT scenes (see examples/DemandLoading/DemandPbrtScene).

v0.9.3

• The DemandLoader interface now provides a setPageTableEntry method, which is helpful for asynchronous resource request handling. In such an approach ResourceCallback can enqueue a request and return false, indicating that the request has not yet been satisfied. Later, when the request has been processed, setPageTableEntry can called to update the page table. Note that it's not necessary to call this method when requests are processed synchronously: if the ResourceCallback returns true, the page table is automatically updated.
• embed_cuda now generates OptiX IR by default.

v0.9.2

Overhauled vcpkg dependency management and build options.
Add CMake presets, which simplify build configuration. See README.md for details.
Fixed intermittent PTX compilation issues caused by lack of trailing zero bytes in embedded PTX.
See CHANGELOG files in submodules for additional changes.

v0.9.1

Numerous changes to DemandLoading library and examples.
See DemandLoading/CHANGELOG.md and examples/CHANGELOG.md