trying to implement mean for large data array

[antonio-fc]

Hello team,

I am currently trying to implement the mean operation for a stream of 9216 floats. This would mean going from the 9216 floats to a single float as the stream output. I am able to transform this with 9 x 1024 float objects and reduce it to 9 float objects that represent the partial means. Unfortunately, I am unable to obtain the final result since these objects cannot be acquired simultaneously. My goal is to obtain the mean of these 9 floats and have a single float as the output of the application. Is there a way I can achieve this?

The application can be found at https://github.com/antonio-fc/my_mlir-aie/tree/testing/apps/app4. The last core description in the aie2.py file is currently just a passthrough for the 9 float objects. However, I want to make it so that this compute tile performs the last step to obtain the single float object. The application uses 'make run_py' to be executed and just prints the 9 partial means.

Thanks in advance, and if more information is needed please let me know.

Best,
Antonio

[jgmelber]

Hi Antonio, I think the link is missing from your code. I would like to review your implementation before making any uninformed suggestions! Thanks!

[antonio-fc]

done, its in the description now.

[jgmelber]

I made a quick attempt. This might work for you antonio-fc/my_mlir-aie@testing...jgmelber:mlir-aie:patch-1

[antonio-fc]

Hi, i tried this and i got this error message