This paper presents the software tools that has been designed to improve the programming efficiency of the PAPRICA-3 massively parallel system. PAPRICA-3 Assembly code is automatically generated starting from C++ programs, using a library that features new classes representing portions of images. The underlying structure of the system (memory addresses, register mapping, handling of multi-bit data,...) is invisible to the user, who can now focus on the development of the applications rather than on the exploitation of the architectural characteristics of the hardware. A system-level simulator has been developed to allow both the testing of applications and the determination of system performance. The specific solutions adopted in the design of the software simulator allow to achieve a high simulation speed (1000:1 with respect to the real PAPRICA-3 hardware) even on low-cost systems such as PCs.