Re:Benchmarking State-Of-the-Art Deep Learning Software Tools
by T-A on Saturday August 27, 2016 @09:27AM
(#12372250)
Context-switching is expensive. You have to swap out the data being worked on by thread #1 and swap in the data for thread #2. So you end up being bottlenecked by memory bandwidth and latency rather than by raw compute.
[ Parent
| on HN ]
There are no comments on this comment yet.