2018-01-20 11:48 AM
Hello engineers,
I would like to ask you about your computing time of your iir. Ive tried cmsis but ive measured avg. 50us computing time for 32 samples using direct form biquad single stage filter.
Many thanks for your answers.
#stm32l4 #iir2018-01-20 11:58 AM
1) How are the samples and coefficients expressed?
2) What speed (MHz) are you running your stm32l4?
Remember that stm32l4 only has a single-precision floating-point unit built into the arm processor.
If your numbers are different (by default C and C++ numbers and calculations are double-precision) then you won't make use of the floating-point unit and calculations will take many more processor cycles.
And you might need to tell your development system / compiler to use the floating-point unit. And you might also need to tell your compiler to allow variables to be passed within the fpu (because that's incompatible with the call-standard for non-fpu subroutine calls).
Hope this helps,
Danish
2018-01-21 07:48 AM
Hello Danish,
first of all my question was simple. what is your iir computing time but newermind. i newer get an exact answer on this forum. So i will try it once again. What is your computing time of IIR on STM32L4?
But lets look on your usless comment.
1, if anyone is talking about speeed of computing on some cpu especcially about filters which logically you want to compute as fast as possible you set up CPU frequency as high as possible (for stm32L4 is 80MHz)
2, ive wrote avg time so its logic that ive tried different methods and different type of coefficients and samples.
I gave simple question. WHAT IS YOUR IIR COMPUTING TIME. i have no much skills about filters so logically iam looking for way.
2018-01-21 02:42 PM
But, without knowing the details that
Ali.Danish
mentioned, any comparison could be pointless - you would have no idea whether you were comparing apples with M16 nuts.And,no - it is not necessarily the case that one will automatically run a filter at the CPU's maximum speed.