Adding two arrays is much slower on the GPU

[ghost]

Hi, I am adding two large arrays, for some reason (probably thread/block/grid size games) the CPU version is much faster.

I thought it is because the array size is too SMALL for the GPU to show any advantage, however, I increased the array size until I got CUDA out of memory error.

The full code is here: https://github.com/QuantScientist/Data-Science-ArrayFire-GPU/blob/master/PyCUDA/02%20add%20with%20PyCUDA.ipynb

My device info is as follows:

1 device(s) found.
Device #0: GeForce GTX 1080

Thanks,

[inducer]

This issue tracker is for bugs, not technical support. Please send a message to the mailing list for tech support.

Look up PCIe bandwidth to help figure out why your code behaves the way it does.