Add cutlass 3xTF32,DMMA based L2/cosine distance kernels for SM 8.0 or higher#939

Merged

rapids-bot[bot] merged 28 commits intorapidsai:branch-22.12from mdoijade:cutlass_dist_kernels

Nov 16, 2022

+1,798-144

Commits on Oct 20, 2022

cutlass based euclidean expanded, cosine kernels
mdoijade
committed

Commits on Oct 21, 2022

add prior ampere pairwisedistmat kernel to prevent redundant kernel compilation, fix clang formating and correct line endings
mdoijade
committed
add noexcept to the functor methods
mdoijade
committed
merge branch 22.12 and resolve conflicts
mdoijade
committed

Commits on Oct 27, 2022

fix comments, remove redundant code and fix formatting issues
mdoijade
committed

Commits on Oct 28, 2022

Commits on Nov 3, 2022

fix the cutlass_include_dir path in cmake
mdoijade
committed

Commits on Nov 4, 2022

Commits on Nov 7, 2022

merge branch-22.12 latest changes
mdoijade
committed

Commits on Nov 11, 2022

temp fix to check if python build works
mdoijade
committed

Commits on Nov 15, 2022

add raft-exports instead of raft-distance-exports as other raft components also use distance headers
mdoijade
committed

Commits on Nov 16, 2022