Closed smu160 closed 8 months ago
Sure thing! The only issue with benchmarking fftw
is measuring memory allocations. This project uses the divan
library to track this, but it can't track allocations in C code. So, for fftw
, I will only report time taken and not total memory allocated.
Benchmarks of fftw
have been added as of this commit!
Hi,
Thank you for creating and running these benchmarks! It's always great to have the implementation tested on yet another chip, and to have results independently verified. If you are willing/able to add another benchmark with the rust bindings for
fftw
, I'd replace the current benchmarks we use with what you created here.Let me know your thoughts. Looking forward!
Best, Saveliy