Skip to main content

Graph decyclify algorithm implementation as in Sandnes & Sinnen paper (2004) in Python

Project description


CI codecov

Graph decyclify algorithm implementation as in Sandnes & Sinnen paper (2004) in Python.

"A new strategy for multiprocessor scheduling of cyclic task graphs", link to article in Research Gate.

See open issues for current status of the project.

Graph with cycles

Graph without cycles

decyclify algorithm

The algorithm uses two matrices, D and C.

D is the intraiteration dependencies matrix. It represents the dependencies in the graph within a cycle.

Intra-iteration matrix

C is the interiteration dependencies matrix. It represents the dependencies in the graph between cycles.

Inter-iteration matrix

Node Iterators

This is not part of the paper. Here we show how the algorithm can be used to first remove the cycles. Next, we use the matrices to decide how to traverse the graph.

Graph unrolling

The first iterator, the CycleIterator simply goes through all the tasks in the cycles and executes them in order. The decyclify function is used to avoid repeating a node due to a cycle.

The second iterator available is the TasksIterator. With this, for each cycle it returns the next tasks available, as well as any tasks in the next cycles that can be returned.

A task is considered ready to be returned when its sibling in the previous cycle has been executed, and after its inter-cycle dependency (if any) has been satisfied as well.

It should be possible to use these iterators, or create new ones, and apply it to tools such as workflow managers that support only DAG scheduling, to schedule an infinite graph, via graph-unrolling. The next cycle is simply an integer counter incremented, but could be an ISO8601 date-time function.

NOTE: this part of the project was a summer holidays project, and is in need of documentation, more tests, code review, etc. Feel free to submit pull requests.


0.1 (2020-12-29)

  • Added a couple of node iterators. With these, it is possible to iterate the graph per cycle, or per task. This latter enables a task to start as soon as its sibling in a previous cycle has been executed, as long as there are no inter-cycle dependencies.
  • #3 Implemented the algorithm to unroll a graph using the Decyclify algorithm
  • #10 Create intraiteration matrix (D) and interiteration matrix (C)
  • #2 Graph input
  • #1 Build and packaging


Licensed under the Apache License. See LICENSE for more.

Project details

Release history Release notifications | RSS feed

This version


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Files for decyclify, version 0.1
Filename, size File type Python version Upload date Hashes
Filename, size decyclify-0.1-py3-none-any.whl (13.3 kB) File type Wheel Python version py3 Upload date Hashes View
Filename, size decyclify-0.1.tar.gz (13.5 kB) File type Source Python version None Upload date Hashes View

Supported by

Pingdom Pingdom Monitoring Google Google Object Storage and Download Analytics Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page