NE10 NEON optimized floating point library.
Is there any reference output available to check the output, I am currently working on checking the output validity of functions in NE10. I am working with GNU octave to check input and output.
Is the output-input correspondence already checked by the team? I did some preliminary analysis on FFT and here are the results. https://docs.google.com/file/d/0B318pI7x2QVccXhNUWdiYlhmUWs/edit Please Comment.
My Fir filter after many trials ended up like this.... https://docs.google.com/document/d/1c5ktcQ6VYlK2im3P2yGNNxRAZonssCApdLS7qkQwq6M/edit Please tell if anyone has checked the real working of code other than simulations with random numbers. I am giving my best to get the output but its always distorted. Please help!!!!!- I need to make use of this library. Thanks
He has already posted this on the Ne10 Github issue tracker and the ARM website. I've already responded in several locations that the FFTs (both real and complex) and FIRs are fully unit tested and conformance tested. Conformance tests compare output from our functions against output from Matlab functions. I suggested, perhaps, his input structures were not properly populated and that he review how to populate the structure type from examples on stackoverflow (not re-post the same question here). I'll triple check our outputs but please keep the conversation to one thread if you're still having trouble.
https://github.com/projectNe10/Ne10/issues/46I
Thanks!