Techniques for Memory-Efficient Model Checking of C and C++ Code

DiVM: Model checking with LLVM and graph memory

“…Among other things, this means that it is not a complete virtual machine, merely an intermediate representation suitable for static analysis, optimisation and native code generation. In particular, it may not always be possible to encode an entire program in LLVM 1 Intermediate Representation. The LLVM IR is used across the majority of the LLVM toolchain and is an abstract counterpart of the machine-level assembly language.…”

“…This precision, however, does not come for freemodel checking, especially in the context of software, is computationally very expensive. Nonetheless, as our previous work shows [1], a combination of state space reduction techniques, compression and of a tailored approach to test case construction makes model checking a genuinely useful programming aid. For example, we have successfully applied this approach in development of scalable concurrent data structures [2] in C++.…”