Hello everybody,
im using the Math.h file to do a rfft with 1024 points und q15 datatype.
So far everything works fine, and I get good results. It needs a computation time of ~720us on my CortexM4 @100Mhz.
This is very Impressive, but to long for my application. I only need 80 Values of the Result so my idea was to optimize the function, so that it only calculates useful values.
But there is no source-code behind the math.h functions. Is it possible to get the code? in C? or is it not available.
Otherwise I need to do it from scratch, with means a lot of work for me.
Thanks everybody for reading
Joe
Joe,
Have you looked at the CMSIS DSP implementation: RealFFT
Best regards
Simon.
Hi Simon,
thank you for your answer! When I click to your Link, i only get some basic Information about the function "arm_rfft_q15".
I used exactly this function to do my fft in 720us. But my question is, if there is the C-Code of arm_rfft_q15 is available?
If so, i could try to make a pruned fft out of it.
As per the bottom of the CMSIS DSP Software Library page, you should find the C source for the CMSIS-DSP under CMSIS\DSP_Lib\Source.
In this particular case (and assuming you are using uVision), you are likely looking for:
C:\Keil<version>\ARM\Pack\ARM\CMSIS\<version>\CMSIS\DSP_Lib\Source\TransformFunctions\arm_rfft_q15.c
Thank You!