This paper presents LLBMC, a tool for finding bugs and runtime errors in sequential C/C++ programs. LLBMC employs bounded model checking using an SMT-solver for the theory of bitvectors and arrays and thus achieves precision down to the level of single bits. The two main features of LLBMC that distinguish it from other bounded model checking tools for C/C++ are (i) its bit-precise memory model, which makes it possible to support arbitrary type conversions via stores and loads; and (ii) that it operates on a compiler intermediate representation and not directly on the source code.

Modeling the semantics of programming languages like C for the automated termination analysis of programs is a challenge if complete coverage of all language features should be achieved. On the other hand, low-level intermediate languages that occur during the compilation of C programs to machine code have a much simpler semantics since most of the intricacies of C are taken care of by the compiler frontend. It is thus a promising approach to use these intermediate languages for the automated termination analysis of C programs. In this paper we present the tool KITTeL based on this approach. For this, programs in the compiler intermediate language are translated into term rewrite systems (TRSs), and the termination proof itself is then performed on the automatically generated TRS. An evaluation on a large collection of C programs shows the effectiveness and practicality of KITTeL on "typical" examples.