3.0.2.4242e39

• Remove kernel encryption to simplify codebase ( Thanks @LunNova )
• New versioning: major.minor.patch.hash ( #44 )
• Fix OBB polygon clipping
• Fix RTIP 1.1 (HWI was disabled on RNDA 2)
• Fix packing leaves (causing a hang on Linux)
• Fix intersection filter for triangle packets
• Fix BVH refit logic

fixes made by @meistdan

3.0.4fea77f

hot fixes for the v3

3.0.9ba63f3

• RDNA 4 RT features
  • Compressed BVH8 and triangle packets
  • Intersectable instances
  • Oriented bounding boxes (for HQ build)
• BVH import API changed
  • Generic BVH2 input format
  • Internally transformed to the HW format
• Per-context logger
  • Resolving issue with a single instance
• Tighter boxes in SBVH for triangles and instances
• Instance transformation fix
• Other fixes and code cleanups

2.5.a21e075.3

hotfix of 2.5.a21e075.2
#37

2.5.a21e075.2

• minor cmake improvements
• fix embedded oro_compiled_kernels.hipfb ( #34 )
• add support for more gfx

2.5.a21e075

hotfix: Initialize identity flag

2.5.4e650d5

cmake - add option to disable cuda ( FORCE_DISABLE_CUDA )
cmake - move kernels precompilation to a custom_command/custom_target
Fix rendering artifacts

2.5.8c74270

• update from HIP 5 to HIP 6
• minor improvements in the cmake script
• perf improvement in maths
• prepare support for more GPUs

2.5.cfa5e2a.2

Merge pull request #25 from GPUOpen-LibrariesAndSDKs/next-release-3

update to 2.5

v2.5.cfa5e2a

• When you run the cmake, you should use none or one flag among this list:

-DBAKE_KERNEL=ON          --> Enable the use of encrypted and baked uncompiled kernels source code
-DBAKE_COMPILED_KERNEL=ON --> If enabled, the kernels will be compiled and embedded as a buffer inside the binary. If you do that, you should also enable the flag 'PRECOMPILE'
-DBITCODE=ON              --> Enable bitcode linking: when running the program, it's expected to find precompiled kernels as external files.

• More flexible vector types
• Unifying math into a single header
• Collapse crash fix
• Other minor fixes