tips for optimising my neural net kernel

Accelerated Computing CUDA CUDA Programming and Performance
1

I moved the thread here.