Variational data assimilation consists in estimating control parameters of a numerical model in order to minimize the misfit between the forecast values and the actual observations. The YAO framework is a code generator that facilitates, especially for the adjoint model, the writing and the generation of a variational data assimilation program for a given numerical application. In this paper we present how the modular graph specific to YAO enables the automatic and efficient parallelization of the generated code with OpenMP on shared memory architectures. Thanks to this modular graph we are also able to completely avoid the data race conditions (write/write conflicts). Performance tests with actual applications demonstrates good speedups on a multicore CPU.