ADiJaC
Summary:
ADiJaC uses source code transformation to generate derivative codes in both the forward and the reverse modes of automatic differentiation.
URL: http://adijac.cs.pub.ro
Developers:
- Emil-Ioan Slusanschi, Vlad Dumitrel, Silvia Stegaru, Cristina Ilie, Alex Teaca, Daniel Mahu, Computer Science and Engineering, University Politehnica of Bucharest (http://cs.pub.ro/)
- Christian Bischof, Institute for Scientific Computing, Technische Universität Darmstadt (http://www.sc.informatik.tu-darmstadt.de/)
| Mode: |
Forward Reverse |
| Method: |
Source transformation |
| Supported Language: |
Java |
Reference:
E. Slusanschi
Algorithmic differentiation of Java programs
Ph.D. thesis, Department of Computer Science, RWTH Aachen University, 2008
Emil I. Slusanschi, Vlad Dumitrel
ADiJaC -- Automatic Differentiation of Java Classfiles
Article in ACM Transaction on Mathematical Software, ACM, 2016
Algorithmic differentiation of Java programs
Ph.D. thesis, Department of Computer Science, RWTH Aachen University, 2008
Emil I. Slusanschi, Vlad Dumitrel
ADiJaC -- Automatic Differentiation of Java Classfiles
Article in ACM Transaction on Mathematical Software, ACM, 2016
Features:
- Internal Representations - Jimple and Grimp from the Soot Framework
- Interprocedural Activity Analysis
- Dependency Analysis & Instruction Reordering
- Vector mode for the Forward Mode Implementation
- TBR Analysis in the Reverse Mode Implementation
Supported Platforms:
- Windows
- Unix/Linux
- Mac
Licensing: license
Entries in our publication database that actually use ADiJaC in the numerical experiments: 2
The following diagram shows these entries versus the year of the publication.
|
|
|
|||
| '08 | '16 | ||||
| Year | |||||