v3.00

Highlights

• Officially support hip6.0+ & cuda12.2 in the main branch
• Updated WMMA and other unit tests
• Added support for additional build flags

What's Changed

• Release/hip5.7 cuda12.2 by @RichardGe in #80
• close of hiprtc DLL was missing by @RichardGe in #82
• remove hip6 DLLs in the paths by @RichardGe in #84
• Release/hip5.7 cuda12.2 by @RichardGe in #85
• Feature/custom dll by @RichardGe in #86
• fix struct size by @RichardGe in #87
• fix struct size by @RichardGe in #88
• Release/hip5.7 cuda12.2 by @RichardGe in #89
• Feature/oro 0 dll by @RichardGe in #91
• help debugging with more printf by @RichardGe in #92
• [ORO-0] ROCm 6x paths by @meistdan in #94
• Fix copy/move-and-swap and refine comments by @KaoCC in #102
• Now oroCtxSetCurrent() is needed. by @takahiroharada in #106
• improve management of build flags for the precompiled/baked use-case by @RichardGe in #113
• update DLLs to 6.1.2 by @RichardGe in #107
• Wmma update by @takahiroharada in #105
• Release/hip6.0 cuda12.2 by @KaoCC in #114

New Contributors

• @RichardGe made their first contribution in #80
• @meistdan made their first contribution in #94

Full Changelog: v2.00...v3.00

v2.00

• Support a lot more CUDA/HIP functions compared to OROCHI 1. Should be almost exhaustive.
• We will keep one branch per version of CUDA/HIP, (example of branch name: release/hip5.7_cuda12.2),
  so developers can switch on branches depending on their environment.
  If you need a combination that doesn't exist, open a Issue.
• Change compared to OROCHI 1: you need to install the CUDA SDK corresponding to the branch you are using.
  for example, if you use branch release/hip5.7_cuda12.2, install CUDA SDK 12.2.
  However CUDA will still be dynamically loaded at runtime, only includes of the SDK are used at compile time.
• New Demo for Textures
• New Demo for D3D12 interop
• Some refactoring/improvement of OrochiUtils was done.
• Orochi.h can be included in the kernel files to have the oro* names
• The binding and naming between HIP/CUDA has been improved and developed in a way it should be easier to maintain for future versions.
• Most Orochi/OrochiUtils API has not been changed so updating the project from Orochi 1.0 to 2.0 should be straightforward.

v1.00

• Bitcode linking support
• Added OrochiUtils. A wrapper for convenience
• A workaround for 22.7.1 AMD driver regression (missing RTC)
• Support more HIP and CUDA APIs
• Use only from CUDA driver apis (except for RTC)
• Proper error handling
• Unit test
• Bug fixes

Initial release

Adl changes (#3)

* Add some more API's and cuda runtime stuff

* Add missing hipMemcpy define

* hipMemcpy fixes

* Update cuew for cudart calls

* Impmenet ppGetDeviceProperties for CUDA

* Add getCurAPI

* [RPRNEXT-0] Fix a build error on linux.

* [POP-0] Fix a build error on linux.

Co-authored-by: Aaryaman Vasishta <aaryaman.vasishta@amd.com>
Co-authored-by: Sho Ikeda <Sho.Ikeda@amd.com>