Open adamgreig opened 9 years ago
Explicit SIMD might not be needed. Just as a point of reference: For the FIR filters in idsp the compiler is able to automatically infer AVX-256 SIMD instructions by itself. In addition to the other tweaks this makes them some 20 times faster than the the FIR filters here, i.e. GS/s speeds on a single Skylake core.
Especially for representing IQ data!