Sequence of warp and how to avoid divergence when folding shared memory in a reduction kernel

[Lian_Yunlong]

I just need some keywords and/or some links