Open Dooez opened 6 months ago
At the moment par_fft uses each sample-range once per butterfly, without reusing already cached data again. Performing par_fft in chunks of cache line size instead of full range length should improve cache usage and reduce memory load.
At the moment par_fft uses each sample-range once per butterfly, without reusing already cached data again. Performing par_fft in chunks of cache line size instead of full range length should improve cache usage and reduce memory load.