|
|
Boost Interest : |
Subject: [Boost-interest] Avalanche fine-grained data flow graph library and lattice-valued Datalog DSEL released
From: Jeremiah Willcock (jewillco_at_[hidden])
Date: 2013-11-14 12:45:26
The Center for Research in Extreme Scale Technology at Indiana University
is pleased to announce the initial release of Avalanche, a library for
expressing computations as fine-grained data flow graphs (a la tbb::flow)
for both sequential and distributed memory computation. Unlike tbb::flow,
Avalanche uses C++11 and metaprogramming techniques to configure flow
graphs at compile-time and allow them to be optimized to have similar
performance to hand-written, lower-level implementations.
Distributed-memory support requires AM++, which is planned for release
soon; sequential support only requires C++11 and a recent version of
Boost.
Also included in the Avalanche release is a sequential implementation of
lattice-valued Datalog as a domain-specific embedded language in C++,
built on top of Avalanche flow graphs. Datalog is a domain-specific
language for querying databases; lattice-valued Datalog extends it with
extra constructs for greater expressiveness. The implementation can
interface to user-defined data structures, with Boost.Graph provided as an
initial example. The current implementation only supports in-memory
storage, however.
The software is released under the Boost license at
<URL:http://www.crest.iu.edu/projects/avalanche/>. Please contact
Jeremiah Willcock at <jewillco_at_[hidden]> with any questions or
requests for assistance.
-- Jeremiah Willcock and Andrew Lumsdaine