Aparapi on amazon EC2 GPU instance

[Richard Grossman]

Hi

I've started to work on a project using GPU to accelerate processing data. 

I'm using all kind of EC2 AWS instance like p2 g3s or f1 instance.

Actually I'm just trying to see the benefit using GPU over CPU.

So I take the aparapi-example and focus on map-reduce example.

The problem is that I can't really see any improvement in processing.

On all type of machine on EC2 I'm using the map-reduce example as reference it's take between 19 and 7s for a 50 M elements array. In fact the first time I run it's take a lot of time but after it's faster but still slower than CPU

On my laptop that doesn't have a proper GPU(intel board) The same operation take no more 1-2 second

When I try the KernelManager API I get on my laptop `ALT` as device type (expected)

When I make the same on AWS the system detect the `GPU`.

I only use ubuntu 

Could you help me to find what the problem or is it something expected ?

Thanks for your help