CMSIS DSP

[cbenson]

Posted on October 30, 2015 at 21:05

I am trying to implement an FFT on Nucleo-L053R8. I am using System Workbench and STM32CubeMX, on Windows 7 64. Adding a function like arm_rfft_init_q15, causes the program size to become so large as to no longer fit in the STM32L053. Is this a consequence of having to load the floating point libraries from ARM_MATH? I believe I have the linker set to only load referenced code. May I am missing something. I could switch to a CPU with a FPU, but this is not a high performance instrument, I would like to avoid overkill.

Regards, Corky #stm32l0-dsp-cmsis

[Tesla DeLorean]

Posted on October 30, 2015 at 21:14

The source to the DSP library is available. You should be able to look at the size of the object files. The .MAP should be able to provide you with a break-down of the contribution of each function pulled into the image, at least Keil does this.

The Cortex-M0 has the smallest instruction set, objects are going to be some what bigger than those for the M3/M4 parts as more instructions are needed to perform the same tasks.