- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
This is a comment after readingRezaur Rahman'sIntel Software Network article entitledThe Intel Math Kernel Library and its Fast Fourier Transform Routine.
While running FFT routines (CFFT1D) on a Xeon MP machine, I get a MKL INTERNAL ERROR : Condition 0 detected in function CFFT1D sometimes. The number of errors vary across runs for the same executable, which is not understood.
Link Copied
- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
Hi,
If you are trying to use the FFT directly from MKL, note that there are two sets of interfaces that MKL supports. One of them is deprecated and the interface you are using (CFFT1D) falls into that area. We would suggest you to use the newer interfaces called Dfti interfaces.
However, if you must use the deprecated interface for some backward compatibility reason, please provide us with the sequence of FFT calls you are making that reproduces the error. It may be a bug in our library.
Also note that ifyou aretrying to port FFTW based code to MKL, there are some FFTW wrappers available with latestMKL 8.1 release that you can use to ease the migration process. Please look athttp://www.intel.com/software/products/mkl/docs/fftw_mkl_user_notes_2.htmon how to use the wrappers.
-Regards
Rezaur Rahman
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page