Performing an FFT using C++ AMP

Question

In Direct3D we have the ability to perform 1, 2 or 3D FFTs using the D3DX11CreateFFTxxx etc.

Is there a way to access this functionality when using C++ AMP or do I have to write my own code in a kernal to implement the calculation?

 

Thanks in Advance

Mark

 

Answer 1

Ah the FFT, yes there is some calls in Direct3D, but I probably could conjure up something in a few weeks

I recall Fourier analysis from an old honors math class I took, it focused on the wave equation

Read the manuals for DirectX if that API is all you need

 

---

Windows MVP 2010-11, XP, Vista, 7. Expanding into Windows Server 2008 R2, SQL Server, SharePoint, Cloud, Virtualization etc. etc.

Hardcore Games, Legendary is the only Way to Play

Developer | Windows IT | Chess | Economics | Vegan Advocate | PC Reviews

Answer 2

Hi Mark,

The D3D FFT functions can't be directly used from a C++ AMP kernel.

I haven't tried this, but since ID3DDX11FFT interface operates on unordered access views, it should be possible to perform these FFT transformations outside of a kernel, and then use these UAV's to create C++ AMP arrays to do further computations on.  If you try this, let us know how it works.

Thanks.

---

++don;

Answer 3

Don,

Thanks for the suggestion - it didn't occur to me to mix DirectCompute and C++ AMP in the same application.

Kind Regards

Mark

Answer 4

Don,

I am not aware of mixing DirectCompute & AMP either. The application I am working on involves lots of Texture2D and UAVs, which are not directly supported in C++ AMP. Could you elaborate on the approach of mixing C++ AMP & DirectCompute, or even show some code examples?

Thanks!

lakeeast

Answer 5

Hi Mark & lakeeast,

If you look at discussion here: http://social.msdn.microsoft.com/Forums/en-US/parallelcppnative/thread/b6e8f31e-723b-4eeb-8b4e-6a6ee5e756f3

It talks a little bit about the interop support between C++ AMP and Direct3D (not just DirectCompute). The main requirement is that the C++ Array and ID3D11Buffer can interop only if the buffer is a structured buffer with stride of 4 bytes.  In the preview bits, there are some other requirements that we are looking to relax.

 

Thanks,

Weirong

 

 

Answer 6

Mark, we now have a wrapper for FFT which means you do not need to do the interop yourself, so please check it out:

http://blogs.msdn.com/b/nativeconcurrency/archive/2011/11/15/the-c-amp-fft-library.aspx

---

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

Answer 7

1 question I have is how many data points are there that need to be transformed

 

---

Windows MVP 2010-11, XP, Vista, 7. Expanding into Windows Server 2008 R2, SQL Server, SharePoint, Cloud, Virtualization etc. etc.

Hardcore Games, Legendary is the only Way to Play

Developer | Windows IT | Chess | Economics | Vegan Advocate | PC Reviews