Commit f1acb3bd authored by Benoit Steiner's avatar Benoit Steiner Committed by TensorFlower Gardener
Browse files

Make sure that kernels that are meant to run on GPU are scheduled inline. This 

saves a context switch.
As a result, when running on GPU, the inception and translation models ares 2 to 3% faster, and the ptb_word_lm model is 10 to 20% faster while using up to 10% less CPU time.
Change: 128756911
parent 0105863a
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