Scratchpad Memory Size Optimization for Real-Time Multiprocess Embedded Applications

  • Jude Angelo Ambrose
  • Ben Juurlink
  • Sandra Irobi
Part of the Lecture Notes in Electrical Engineering book series (LNEE, volume 127)


The size of the on-chip memory affects the chip area, cost, as well as the power and energy dissipation, which are of prime concerns for embedded systems. For real-time applications, scratchpad memories (SPMs) are preferable to caches and, furthermore, sufficient scratchpad memory should be provided in order to guarantee real-time performance. In order to avoid runtime surprises, however, embedded system designers tend to over-budget the memory size, thereby increasing its associated costs. We consider the problem of determining the minimum memory size required to fulfill the real-time requirements of a set of recurring tasks, given the tasks’ execution times as a function of the amount of memory it is allocated. Three SPM partitioning strategies are considered: private, where each task is allocated a private partition, shared, where all tasks share the SPM and therefore a context switch incurs the cost of bringing in its data from the next memory level, and hybrid, where some tasks are allocated a private partition and others share a partition. For the private strategy a polynomial-time algorithm is presented that determines the smallest amount of SPM needed to fulfill the real-time requirements. The hybrid partitioning strategy is NP-complete and an exhaustive search algorithm is presented that determines the optimal solution. In addition, two heuristics for the hybrid strategy are proposed and the algorithms are evaluated using synthetic task models with different characteristics. Results show that the heuristics produce solutions close to the exhaustive search while performing orders of magnitude better in terms of search time.


Embed System Memory Size Context Switch Private Strategy Exhaustive Search Algorithm 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Banakar, R., Steinke, S., Lee, B.-S., Balakrishnan, M., Marwedel, P.: Scratchpad memory: design alternative for cache on-chip memory in embedded systems. In: CODES, pp. 73–78 (2002)Google Scholar
  2. 2.
    Keitel-Schulz, D., Wehn, N.: Embedded dram development: Technology, physical design, and application issues. In: IEEE Des. and Test of Comp. (2001)Google Scholar
  3. 3.
    Puaut, I., Pais, C.: Scratchpad memories vs locked caches in hard real-time systems: a quantitative comparison. In: DATE, pp. 1484–1489 (2007)Google Scholar

Copyright information

© Springer-Verlag GmbH Berlin Heidelberg 2012

Authors and Affiliations

  • Jude Angelo Ambrose
    • 1
  • Ben Juurlink
    • 2
  • Sandra Irobi
    • 1
  1. 1.Computer Engineering, EWIDelft University of TechnologyDelftThe Netherlands
  2. 2.Computer EngineeringUniversity of BerlinBerlinGermany

Personalised recommendations