Optimizing sorting algorithms by using sorting networks

Michael Codish, Luís Cruz-Filipe, Markus Nebel, Peter Schneider-Kamp

Research output: Contribution to journalArticlepeer-review

4 Scopus citations


In this paper, we show how the theory of sorting networks can be applied to synthesize optimized general-purpose sorting libraries. Standard sorting libraries are often based on combinations of the classic Quicksort algorithm, with insertion sort applied as base case for small, fixed, numbers of inputs. Unrolling the code for the base case by ignoring loop conditions eliminates branching, resulting in code equivalent to a sorting network. By replacing it with faster sorting networks, we can improve the performance of these algorithms. We show that by considering the number of comparisons and swaps alone we are not able to predict any real advantage of this approach. However, significant speed-ups are obtained when taking advantage of instruction level parallelism and non-branching conditional assignment instructions, both of which are common in modern CPU architectures. Furthermore, a close control of how often registers have to be spilled to memory gives us a complete explanation of the performance of different sorting networks, allowing us to choose an optimal one for each particular architecture. Our experimental results show that using code synthesized from these efficient sorting networks as the base case for Quicksort libraries results in significant real-world speed-ups.

Original languageEnglish
Pages (from-to)559-579
Number of pages21
JournalFormal Aspects of Computing
Issue number3
StatePublished - 1 May 2017


  • Instruction-level parallelism
  • Out-of-order execution
  • Sorting algorithms
  • Sorting networks

ASJC Scopus subject areas

  • Software
  • Theoretical Computer Science


Dive into the research topics of 'Optimizing sorting algorithms by using sorting networks'. Together they form a unique fingerprint.

Cite this