This package provides a library written in C implementing realtime digital filtering functions for multichannel signals (i.e. filtering multiple signals with the same filter). The core library implements FIR and IIR filtering for float and double data type. Additional functions are also provided to design few usual filters: Butterworth, Chebyshev, windowed sinc...

In addition, the filter functions have been especially optimized for multichannel signals. If data allows it (and if possible with the compilation flags used), the library automatically switch to optimized SIMD (Single Instruction Multiple Data) code, allowing to reduce by 3~4 the time spent in processing the data.

Any POSIX platform and windows. The SIMD code operating on real data is possible if the host CPU supports SSE2 instruction sets (Intel pentium4, AMD k8 or later processors support SSE2), and SIMD code operating on complex data is possible if host CPU supports SSE3.

This library is organized as a GNU package and can be compiled and installed in the same way (see INSTALL file for further information).

In order the library to be compiled with SIMD version of the filter functions, make sure you specify the correct architecture (add -march=target_arch to CFLAGS with gcc) or enable the necessary instruction set (add -msse2 or -msse3 to CFLAGS with gcc)

On x86 family platform, the 387 floating point coprocessor is well known to introduce some numerical error whose symptoms may be seen in the filter design functions. Therefore it is recommended to generate floating point code using SSE instructions whenever possible. By default, the configure script will specify the necessary compiler flags if gcc is used. If you use another compiler you may need set manually CFLAGS at the invokation of the configure script (see your compiler documentation to know which flags should be passed).