I read that Matlab is able to calculate some coefficients for FFTs that help him to do FFTs very quickly.
The first time Matlab calculates these coefficients, so the first time he does the calculations slowly, but then he uses the calculated coefficients and makes the calculations much faster.
Does the MKL library know how to do the same? Perhaps, if I don't delete the DFTI_DESCRIPTOR_HANDLE descriptor after calculations, then the following calculations will be even faster?
Thanks for reaching out to us.
Could you please explain what do you mean by the following calculations will be even faster and if possible provide us a code snippet
which helps us to get some more insights regarding the issue.
and also please do let us know your environment details
This refers to the rotational coefficients W of the FFT algorithm.
The FFT calculation the first time will be slower than subsequent times because the first calculation calculates the rotation coefficients W.
After calculating the rotational factors W, the FFT calculation becomes faster.
I want to know what MKL tools are available to me to make the FFT as fast as possible.
>>I want to know what MKL tools are available to me to make the FFT as fast as possible.
There is a tool named FFT Length-Advisor which will make FFT to perform fastest.
Please refer the below link which helps you to get more details regarding the tool
Hope the provided information might help you.
I assume that your issue is resolved. If you need any additional information, please post a new question as this thread will no longer be monitored by Intel.
Have a good day!