Large network simulations put special demands on simulation
software. With our roots in traditional computer science, it is
natural for us to engage also in simulator development. The SPLIT
program developed some years ago [Hammarlund, 1996; Hammarlund and
Ekeberg, 1998] is a parallelizing simulator which handles very large
network simulations. Our largest network model simulated so far had
more than 50000 neurons and 4 million synapses. The See simulator
[Djurfeldt et al, 1999] represents the continuation of this
development. Efficient neural network simulators are important for
another reason as well: they serve as models for how to perform
neural network operations efficiently, e.g. in on-board embedded
systems.
Large network simulations put special demands on simulation
software. With our roots in traditional computer science, it is
natural for us to engage also in simulator development. The SPLIT
program developed some years ago [Hammarlund, 1996; Hammarlund and
Ekeberg, 1998] is a parallelizing simulator which handles very large
network simulations. Our largest network model simulated so far had
more than 50000 neurons and 4 million synapses. The See simulator
[Djurfeldt et al, 1999] represents the continuation of this
development. Efficient neural network simulators are important for
another reason as well: they serve as models for how to perform
neural network operations efficiently, e.g. in on-board embedded
systems.