Stack Data Management for Limited Local Memory (LLM) Multi-core Processors

Abstract—Limited Local Memory (LLM) architectures are power-efficient, scalable memory multi-core architectures, in which cores have a scratch-pad like local memory that is software controlled. Any data transfers between the main memory and the local memory must be explicitly present as Direct Memory Access (DMA) commands in the application. Stack data management of the cores is an important problem in LLM architecture, and our previous work outlined a promising scheme for that [1]. In this paper, we improve the previous approach, and now can i) manage limitless stack data, ii) increase the applicability of stack management, and iii) perform stack management with smaller footprint on the local memory. We demonstrate these by executing benchmarks from the MiBench suite on the IBM Cell processor. Index Terms—Stack, local memory, scratch pad memory, embedded systems, multi-core processor, IBM Cell, MPI I