This work presents a comparative analysis of soft error susceptibility in a well-known embedded microprocessor ARM Cortex-A9 single core, widely used along with safety critical applications, running embedded software developed for a bare metal environment and with operating system. The soft error analysis is performed by fault injection on OVPSim-FIM simulator platform. The faults injection campaign injects thousands of bit-flips in the microprocessor register file while executing a set of benchmarks with a diverse code behavior from control flow dependence to data intensive. Results present the percentage of masking faults, the classification of the errors and the number of exceptions detected by the operating system. The proposed method and the obtained results can help guiding software developers in choosing different architectures of the code in order to improve fault masking.