swh:1:snp:2d869aa00591d2ac8ec8e7abacdda563d413189d
Tip revision: d23100d28be7288611eeb474d98d624750a2c56b authored by Roberto Di Cosmo on 30 November 2017, 20:59:20 UTC
Merge pull request #69 from UnixJunkie/disable_core_pinning
Merge pull request #69 from UnixJunkie/disable_core_pinning
Tip revision: d23100d
Changelog
2011/11/30 (RDC)
Version 0.9.8 contains the following major new features w.r.t. 0.9.4
- a chunksize parameter can be used to control the granularity of the
parallelism: each worker will handle a series of chunks of this size
and ask for them when ready, thus allowing the system to achieve
automatic load balancing
- very specialised versions of the map function are now available for
arrays and float arrays, allowing to obtain significant speed-up even
on relatively light computations
- autoconf and ocamlbuild harness should simplify compilation and installation.
2011/08/30 (RDC)
internally convert lists to array to avoid quadratic penalty in execution time on long lists,
thanks to Paul Vernaza <pvernaza@andrew.cmu.edu> for pointing out this issue;
added 'a sequence type to allow using efficiently the library both with lists and arrays.