C++ AMP: Determining the capabilities of the system's GPU hardware

Question

I've had a bit of experience in developing with CUDA, and I was wondering: does C++ AMP have a way of determining the capabilities of the system it's running on at run-time, to tune the kernels to the existing hardware? (To maximize the utilization of the GPU(s) installed on the system)

Answer 1

Hi cjstevenson

C++ AMP works on a variety of hardware from a variety of vendors.

The hardware abstraction in C++ AMP is called an accelerator, and you can query hardware properties through the accelerator interface (and also the accelerator_view):
http://www.danielmoth.com/Blog/concurrencyaccelerator.aspx

If you don’t see there what you are looking for, it is probably not available, sorry.

Cheers
Daniel

---

http://www.danielmoth.com/Blog/