topic Why does the mkl fft routine return unreasonably large number? in Intel® oneAPI Math Kernel Library

Why does the mkl fft routine return unreasonably large number?

hadesmajesty — Tue, 02 Nov 2010 13:16:48 GMT

The code is like this

Status = DftiCreateDescriptor(FFT_Handle32, DFTI_DOUBLE, DFTI_REAL, 1, k32+1)

Status = DftiCommitDescriptor(FFT_Handle32)

Status = DftiComputeBackward(FFT_Handle32, temptv32)

End do

I have initialized the FFT_Handle32, and run the loop for many times for different arrays temptv32. One of the arrays has exactly zero, but the backward FFT routine returns numbers like 6.187317420301585E+195. Why is that?

Thank you in advance.

Why does the mkl fft routine return unreasonably large number?

barragan_villanueva_ — Tue, 02 Nov 2010 14:49:27 GMT

Hi,

Please send us small testcase to analyze your problem together with information: MKL-version, architecture, OS, etc.

Why does the mkl fft routine return unreasonably large number?

Gennady_F_Intel — Tue, 14 Dec 2010 08:55:21 GMT

Hellohadesmajesty,

how about the test case for reproducing the problem? The existing cases don't show the problem on our side.

--Gennady