interpretation based value analysis is a classical approach for verifying programs with floating-point computations. However, state-of-the-art tools compute an over-approximation of the variable values that can be very coarse. In this paper, we show that constraint solvers can significantly refine the approximations computed with abstract interpretation tools. We introduce a hybrid approach that combines abstract interpretation and constraint programming techniques in a single static and

more »

... ic analysis. rAiCp, the system we developed is substantially more precise than Fluctuat, a state-of-the-art static analyser. Moreover, it could eliminate 13 false alarms generated by Fluctuat on a standard set of benchmarks. straint solvers over floating-point numbers, Constraint solvers over real number intervals, Abstract interpretation-based approximation Introduction Programs with floating-point computations control complex and critical physical systems in various domains such as transportation, nuclear energy, or medicine. Floating-point computations are an additional source of errors and famous computer bugs are due to errors in floating-point computations, e.g., the Patriot missile failure. Floating-point computations are usually derived from mathematical models on real numbers [14] . However, real and floating-point computation models are different: for the same sequence of operations, floating-point numbers do not behave identically to real numbers. For instance, with binary floatingpoint numbers, some decimal real numbers are not representable (e.g., 0.1 has no exact representation), arithmetic operators are not associative and may be subject to phenomena such as absorption (e.g., a + b is rounded to a when a is far greater than b) or cancellation (subtraction of nearly equal operands after rounding that only keeps the rounding error).