Amygdala spiking neural network Amygdala is an AI Foundry member project
Home
Understanding Amygdala
What are SNNs?
What is Amygdala?
Demonstrations
Frequently-asked questions
Supporting research
Advanced features
Distributed operation
Network evolution (GA)
GUI network editor
Using Amygdala
Platforms supported
Download & install
Documentation
Sample networks
Bugs & feedback
The future of Amygdala
Roadmap for development
Getting involved
Contacts
Development site

 

Advanced features : Distributed operation

Running Amygdala on multiple CPUs

The basic idea of Amygdala MP is to partition a network into sub networks and let them run independently, yet synchronized.

Currently, only SMP is supported through multi-threading. This includes machines with the new chips with 2 or more CPUs such as the POWER4 from IBM or the Pentium 4 with hyperthreading. Don't expect much from the first Pentium4 with hyper threading. Benchmarks show that these features are not quite mature yet. Classical SMP systems or POWER4 architectures will exhibit a considerable performance boost, though.

The upcoming versions of Amygdala will provide tools to facilitate network partitioning. Running a network on a cluster of workstations will also be possible.

Hosted by
SourceForge.net
Amygdala : spiking neural network Site last updated: Feb 28, 2003
Send feedback about this site