Wenkai Du e3e1c6b29c rccl-prim-test: add all-to-all benchmark (#185) ...

For gfx908, support simple detection of ring topology.
Call ReduceOrCopyMulti directly from kernel.
Also simplify code by removing kernel start synchronization option
which has no effect on throughput measurements.

[ROCm/rccl commit: ebc823e603]

2020-03-16 10:00:54 -07:00

..

rccl

rccl-prim-test: add all-to-all benchmark (#185)

2020-03-16 10:00:54 -07:00