FPCS is a floating point constraint solver. Floating point constraints have been introduced in  to allow to reason correctly on floating point computations. Therefore, floating point constraint are dedicated to the test and verification of programs that contain floating point computation. Such a tool is required to correctly handle floating point constraints: for instance, a constraint like x*x == 2.0 with x < 0 has not solution over the floats with a rounding mode set to near while it has a solution over the reals.
FPCS mainly relies on  (see Publications for other related papers) to filter constraints over the floats with respect to the IEEE 754 standard.
FPCS is a C++ library (~50k lines, ~30k being generated with macros). It supports 4 different rounding modes, basic arithmetic operations, comparisons, float/integer conversions and some usual functions like the square root. 2B consistency filtering is complemented with stronger kB consistency filtering.
FPCS targets C programs running under linux (32 or 64 bits) and compiled with gcc without any optimization.