Abstract
We consider the problem of deterministic integer sorting on unit-cost sequential and parallel machines with a large word length and show that n integers drawn from {0,...,m-1} can be sorted using a word length of O(m log n) bits either in O(n) time on a unit-cost RAM or in O(log n) time on a unit-cost EREW PRAM with O(n/log n) processors. Spending O(log log log m) additional sequential or parallel time, we can reduce the necessary word length to O(min}n log n log m + (log m)1+ε{lunate}, mε{lunate}+log n}) bits, for any fixed ε{lunate}>0. Previous algorithms with a linear time-processor count either cannot so arbitrary integers or require a much larger word length.
| Original language | English |
|---|---|
| Pages (from-to) | 57-63 |
| Number of pages | 7 |
| Journal | Information Processing Letters |
| Volume | 36 |
| Issue number | 2 |
| DOIs | |
| Publication status | Published - 15 Oct 1990 |
| Externally published | Yes |
Keywords
- Integer sorting
- nonconservative algorithms
- parallel algorithms
- parallel sorting
- unit-cost measure
- word length
Fingerprint
Dive into the research topics of 'Improved nonconservative sequential and parallel integer sorting'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver