- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
I would like to know how IPP routines are coded.
I am especially interested in the FFT routine.
The reason is FFT consists of mainly complex multiplications.
My experience with Intel C++ compiler is that it cannot vectorize
complex operations properly.
The performance level of FFT on a 2.8C GHz P4 is about 6GFlops.
Now it is impossible that this CPU reaches that performance without full vectorization.
Since the Intel C++ compiler cannot vectorize this code efficiently, I conclude that IPP FFT is coded in assembly.
Can somebody confirm this?
If yes do we have access to the FFT code?
Thanks.
Link Copied
1 Reply
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
You are right, we have good implementation of FFT in IPP. It was written by hand, usingassembler.
Currently our policy is not to provide IPP sources.
Regards,
Vladimir

Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page