Results 1 to 2 of 2

Thread: imsl_c_fft_complex size limit?

  1. #1
    Junior Member
    Join Date
    Apr 2012
    Posts
    3

    imsl_c_fft_complex size limit?

    Hi,

    When signal length is over 64*1024*1024 data points imsl_c_fft_complex seems to produce incorrect results.

    I know that the results are not correct, because I use a pure sine waveform as input, so that I know what FFT spectrum to expect and, as an additional sanity check, I do the same FFT on GPU. GPU results match the expected spectrum, while IMSL don't. Interestingly, it looks as if every other complex point in imsl_c_fft_complex output is wrong.

    This doesn't happen when the signal length is <= 64*1024*1024. Just in case you ask, I've got 16 GB of RAM installed on my PC.

    Has anyone ever encountered a similar problem?

    Regards,
    Sergey

    PS I forgot to mention that I was calling imsl_c_fft_complex from .NET code. Upon further investigation it turns out that if I call imsl_c_fft_complex from C code, there is no corruption whatsoever. I had to increase the default reserve stack size though... Anyway, I am still investigating the problem, so any help would be appreciated)))
    Last edited by Sergey S; 06-13-2014 at 07:42 AM.

  2. #2
    Junior Member
    Join Date
    Feb 2009
    Location
    Houston, TX
    Posts
    27
    Have you tried linking to the ImslCS DLL using MKL? The default CNL library is linking to MKL. This would then give you apples to apples comparison. Do you have sample code you can supply?

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •