Conversion to non-managed GPU memory#5

Open

drreynolds wants to merge 17 commits intomainfrom

feature/non-managed

Commits on May 31, 2022

Initial work toward moving reaction rate tables to non-managed memory (still in progress)
drreynolds
committed

Commits on Jun 13, 2022

Converted code to allocate separate host/device reaction rate structures when RAJA is enabled. Code compiles without issue; will begin testing soon.
drreynolds
committed

Commits on Jun 14, 2022

Streamlined dual host/device reaction network data structure implementation
drreynolds
committed
Changed allocation of dual host/device reaction rate structure to generate pointer to object. Tests now run but results are incorrect; need to debug.
drreynolds
committed

Commits on Jun 20, 2022

Updates to enable HIP-based build
drreynolds
committed

Commits on Jun 21, 2022

Commits on Dec 23, 2022

Merged main into feature/non-managed branch (needs testing)
drreynolds
committed

Commits on Jan 5, 2023

Finished updating non-managed branch with changes from main. Tested thoroughly with CUDA and serial backends.
drreynolds
committed

Commits on Apr 11, 2023

Added HIP specifics to CMakeLists.txt
drreynolds
committed

Commits on Apr 12, 2023

Updated build setup for Crusher
drreynolds
committed

Commits on May 24, 2023

Merge updates from main into feature/non-managed in preparation for a PR.
drreynolds
committed