arm_cfft_q15 slower than deprecated arm_cfft_radix2_q15

[cchuravy]

I am migrating a piece of signal processing code from one M0 based MCU to another, and noticed that the arm_cfft_radix2_q15 function has been deprecated in favor of the more general purpose arm_cfft_q15 with the promise of being faster.

However, bench marking between the two files, I discovered that the new function takes >3x as long as the older one. I'm running on an STM32F091RC at 48MHz. arm_cfft_q15 takes approximately 17ms, where arm_cfft_radix2_q15 takes approximately 5ms to yield the same result.

Takes ~17ms
q15_t input_buffer[1024];
arm_cfft_radix2_instance_q15 S;
arm_cfft_radix2_init_q15(&S, 512, 0, 1);
arm_cfft_radix2_q15(&S, input_buffer);

Takes ~5ms
q15_t input_buffer[1024];
arm_cfft_q15(&arm_cfft_sR_q15_len512, input_buffer, 0, 1);

[llefaucheur]

Hi cchuravy. Thank you for this valid comment. We reproduce it : the radix4 implementation is slower than the radix2 one for M0, which is not the case for all processors except M0/M3. We take your point for the next revision of the library. Regards, Laurent.

[christophe0606]

We are moving CMSIS-DSP to a new repository : https://github.com/ARM-software/CMSIS-DSP

I have recreated the github issues on the new repository (sometimes merging a few related ones). You can see the link just above this comment (... mentioned this issue ...).

So, I am closing the github issue here.