2 ms·
A few questions: Is this just vectorization? Because FFT libraries often implement this using either asm or intrinsics. Why compile with -O1? Why not let the
by andrepd 10y ago
A few questions:
Is this just vectorization? Because FFT libraries often implement this using either asm or intrinsics.
Why compile with -O1? Why not let the gcc optimize with -O2 or -O3? In recent versions I believe this enables vectorization attempts which could produce similar speedups to your handwritten asm.