Good morning All,
I would need to compare the computational times required by a NN to process a fixed length signal (3 channels- length: 516 elements) onboard of a STM microcontroller.
For this reason, I created a model with:
a fromworkspace block to collect three fixed length signals,
a stateful-regression-predict block( that recalls the NNs stored in a .mat file)
a simout block (please check the attached figure).

As I need to specify timestamps for each time-step of the sequence, I provide data as 516×4 array (first column is the specified time, the other three columns are the three channels of the input signal). This results in the model predicting the output at the corresponding timestamps with real-time. However, I would like to know the time that the NN takes to process the entire signals not in real-time, just as it was a simulation run on the STM board.
Do you have any clue on how to achieve this result?
I tried to check the "Measure task execution time" option in Hardware Settings/Code Generation/Verification. However, an error "Code generation information file does not exist." displayed while deployng the model to the STM Nucloeo board. It came out the this is a bug of Versione 2022 and 2023. Therefore, I switched to Version 2021, however, even if the model was deployed, once I opened the Data Inspector but no information about execution time is reported. Do you know any viable workaround?
Many thanks in advance
Best regards,
GiovanniGood morning All,
I would need to compare the computational times required by a NN to process a fixed length signal (3 channels- length: 516 elements) onboard of a STM microcontroller.
For this reason, I created a model with:
a fromworkspace block to collect three fixed length signals,
a stateful-regression-predict block( that recalls the NNs stored in a .mat file)
a simout block (please check the attached figure).

As I need to specify timestamps for each time-step of the sequence, I provide data as 516×4 array (first column is the specified time, the other three columns are the three channels of the input signal). This results in the model predicting the output at the corresponding timestamps with real-time. However, I would like to know the time that the NN takes to process the entire signals not in real-time, just as it was a simulation run on the STM board.
Do you have any clue on how to achieve this result?
I tried to check the "Measure task execution time" option in Hardware Settings/Code Generation/Verification. However, an error "Code generation information file does not exist." displayed while deployng the model to the STM Nucloeo board. It came out the this is a bug of Versione 2022 and 2023. Therefore, I switched to Version 2021, however, even if the model was deployed, once I opened the Data Inspector but no information about execution time is reported. Do you know any viable workaround?
Many thanks in advance
Best regards,
Giovanni Good morning All,
I would need to compare the computational times required by a NN to process a fixed length signal (3 channels- length: 516 elements) onboard of a STM microcontroller.
For this reason, I created a model with:
a fromworkspace block to collect three fixed length signals,
a stateful-regression-predict block( that recalls the NNs stored in a .mat file)
a simout block (please check the attached figure).

As I need to specify timestamps for each time-step of the sequence, I provide data as 516×4 array (first column is the specified time, the other three columns are the three channels of the input signal). This results in the model predicting the output at the corresponding timestamps with real-time. However, I would like to know the time that the NN takes to process the entire signals not in real-time, just as it was a simulation run on the STM board.
Do you have any clue on how to achieve this result?
I tried to check the "Measure task execution time" option in Hardware Settings/Code Generation/Verification. However, an error "Code generation information file does not exist." displayed while deployng the model to the STM Nucloeo board. It came out the this is a bug of Versione 2022 and 2023. Therefore, I switched to Version 2021, however, even if the model was deployed, once I opened the Data Inspector but no information about execution time is reported. Do you know any viable workaround?
Many thanks in advance
Best regards,
Giovanni time profiler, stm32 nucleo board MATLAB Answers — New Questions

​