Getting GPU(Graphics Processing Unit)

First of all, why do you want to get what GPU they are using? There really shouldn't be be any reason for this - there is no telling if the user has over/underclocked their GPU. Besides, even specific GPU vendors will have different drivers...

My point is that you should be writing your code so that it runs on pretty much any (newer) GPU. Writing code for specific GPUs is generally not a Good Thing(tm).

I am not writing a code for any specific GPU, I just want to retrieve the information related to it, like GPU weightage, when one of my application is running.

By "weightage", I assume you mean the GPU load (i.e.: How much of the GPU is busy, as a percentage)?

Yes exactly, thats what I am looking for

I can't say I know how to do this. I have really only messed with DirectX.

Why exactly do you need to know this, anyway? Can't you measure the framerate of your application to get a rough idea of how bady the GPU is being stressed?

The act of GPU profiling is quite intensive operation in itself; one has to take into account command latencies, pixel pipeline bottlenecks (affecting fillrate), geometry pipeline bottlenecks (affecting effective polygon counts), memory bandwidth and latency, and all sorts of other stuff.

The point is, it would considerably slow down your application if you would profile each frame in it. Have you considered using special profiling tools before you ship your executable to your "clients"? I recommend using the NVidia NVPerfKit if you have a NVidia card.

http://developer.nvidia.com/page/tools.html