Reduction for int8 and bfloat16#125

Merged

asroy merged 41 commits into

ck_reduce_int8_bp16

Mar 22, 2022

Commits on Mar 6, 2022

Use thread cluster descriptor and explicit M_K 2d descriptor to simply Blockwise Reduction
qianfengz
committed
Change by replacing ReduceDims by NumReduceDims as Device Reduce interface template parameter
qianfengz
committed

Commits on Mar 8, 2022

Commits on Mar 10, 2022

Commits on Mar 13, 2022

Refine and fix (in GetWorkspaceSizeInBytes of MultiBlockPartialReduce) to make int8 completely pass
qianfengz
committed
Tiny renaming in gridwise_2d_reduction_multiblock_partial_reduce.hpp
qianfengz
committed
Tiny fix in script/profile_reduce_no_index.sh
qianfengz
committed
Refine in DeviceReduce layer with regard to using NumInvariantDim/NumReduceDim or InvariantDims/ReduceDims
qianfengz
committed
Generic renaming in host reduction and DeviceReduce layer
qianfengz
committed
Add support for 4-d all dimension reduction in the profiler and add_device_reduce_xxx instances
qianfengz
committed
Merge branch 'develop' into ck_reduce_int8_bp16
qianfengz
committed

Commits on Mar 15, 2022

Merge branch 'add_more_static_assert_to_threadwise_copy' into ck_reduce_int8_bp16
qianfengz
committed

Commits on Mar 19, 2022

Use multi-thread and simplification for host Reduction implementation
qianfengz
committed

Commits on Mar 21, 2022

Commits on Mar 22, 2022