A typical feature that limits the performance of {\em low-cost} SIMD array processors is the small amount of memory owned by each Processing Element. As a consequence, such systems can utilize only a specific processor virtualization mechanism, based on the sequential scanning of the data set stored into an external memory. This work presents some considerations on the hardware organization of the external memory and discusses a few basic criteria for the development of efficient algorithms. These criteria are illustrated with the help of a case study: a morphological filter for the measurement of the local slope of curves in binary images. The aim of this work is twofold: (i) to show that, as far as a low-cost array processor (like PAPRICA) is concerned, the data coding is a critical point in the algorithm design; and (ii) to provide a new iterative morphological algorithm for slope evaluation in binary images.