Commit 661b9939 authored by Adrian Kuegel's avatar Adrian Kuegel Committed by TensorFlower Gardener
Browse files

Improve KeyValueSort performance on GPU (part 2) 

Use shared memory to avoid copying to/from global memory several times
when processing one tile of data. Also parallelize
processing one tile of data, and make tiles larger (64 -> 2048).

PiperOrigin-RevId: 220626790
parent c0384187
Show whitespace changes
Inline Side-by-side
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Please to comment