The Auburn Home Page
A Kit for Benchmarking Functional Data Structures
Version 2.4 of Auburn conforms to the Haskell'98 standard.
Version 2.3 of Auburn improves on version 1.0 by featuring:
- A decision tree inducer that produces a tree classifying
the best data structure according to how it is used
- The ability to extract a profile of how any
application uses a data structure
- More automation of benchmarking procedure
- Improved profile information, including shadow
profiling, and abstract phasing of profiles
- More general handling of dug files
This includes an INSTALL file that gives a
list of what is needed or recommended for building and using
For the best results, you will need a Haskell compiler with the
ability to call Haskell from C (an extension to the Green Card
extension!). The York nhc13 compiler does this, see the York FP web page for
details of release.
1.0 and Version
2.0a of Auburn are also available.
The thesis Benchmarking Purely Functional Data Structures
uses Auburn version 2.3, some
implementations , and some real
The following papers also describe Auburn:
- IFL'97 workshop, Auburn: A Kit for Benchmarking Functional
Data Structures. The paper appears in
LNCS Volume 1467, and is
available in gzip'd
(© Springer-Verlag). This paper uses Auburn version 1.0.
- IFL'98 workshop, Exploring Datatype Usage Space. The
paper only appears in the draft proceedings (printed by University
College London), but is available here in gzip'd Postscript.
This paper does not directly use any particular version of Auburn.
- PADL'99 workshop, Automated Benchmarking of Functional Data
Structures. The paper appears in LNCS Volume 1551, and is
available in gzip'd Postscript
(© Springer-Verlag). This paper uses Auburn version 2.0a.
Known problems will be kept up-to-date.
All problems that will not vanish with future releases (because the
problem is say, compiler specific) are dealt with in more detail and
either patches are provided in the Auburn distribution or manual
work-arounds are given.