DigiCortex 0.99a - Support for Intel Ivy Bridge EP / EX "IvyTown" CPUs

DigiCortex v0.99a brings performance-monitoring support and optimizations for upcoming Intel Ivy Bridge EP / EX ("IvyTown") Xeon E5 2600/4600 v2 and E7 2800/4800/8800 v2 CPUs including support for memory-bandwidth and power-consumption hardware performance monitoring registers. In addition, it contains performance-fixes for systems with number of CPUs not equal to power of 2. Click on the article to read more!

DigiCortex 0.98 - Support for Intel Haswell AVX2 Instructions

DigiCortex v0.98 implements support for Intel(R) Haswell microarchitecture including optimizations using Advanced Vector Extensions 2 (AVX2), Fused Multiply-Add (FMA) and Gather instructions. Furthermore, performance monitoring in SpikeBenchmark has also been extended so that Haswell CPUs are now supported. Click on the article to read more!

DigiCortex 0.97 - Updated Model of Early Visual System

DigiCortex v0.97 adds further improvements to the model of early vision: thalamic (dLGN) processing. Signal coming from the virtual retina is now passed through a pool of neurons in dorsal Lateral Geniculate Nucleus (dLGN) and projected to primary visual cortex (V1). Cortico-thalamic feedback and inhibition through Reticular Thalamic Nucleus (RTN) is also modeled. Click on the article to read more!

DigiCortex 0.96 - Preview of the Retinal Module

DigiCortex v0.96 brings early support for visual input from the outside world. Input from a web camera or a Sony(R) PS3 Eye Cam can be fed to artificial retina processing stage (using Virtual Retina package from INRIA Neuromathcomp research-team) and then to two arrays of simulated retinal ganglion cells ('ON' and 'OFF' layers). Final destination for the retinal signal is primary visual cortex (V1). Click on the article to read more!

insideHPC interview with Ana and Ivan on GTC2013

Ana Balevic and Ivan Dimkovic from DigiCortex.net talk with Rich Brueckner from insideHPC about the DigiCortex project - video of the interview is posted here: http://insidehpc.com/2013/03/26/amazing-digicortex-engine-maps-the-brain... check it out!

SpikeFun/DigiCortex goes twitter - #digicortex

Follow news on the SpikeFun/DigiCortex project on twitter: #digicortex

16.7 Million Neurons, 3.5 Billion Synapses

NVIDIA GK110 vs. Dual Intel Xeon E5 2687W - Fight!

We tested early version of the SpikeFun v0.95 CUDA compute plug-in against the reference optimized x86 code-path. For the test, two state of the art GPU and CPU solutions were used: NVIDIA GK110 ("Big Kepler") and dual eight-core Intel Xeon E5 2687W. Click on the title to find out who gets the performance crown...

DigiCortex 0.95 - running on CUDA/GPU!

DigiCortex v0.95 brings the early preview of CUDA acceleration. Despite the early stage of the CUDA compute plug-in, simulation performance is more than 7 times faster than optimized multi-threaded CPU code running on a high-end dual 8-core Xeon E5 for large simulations. Click on the title to find out more...

SpikeFun v0.94 - Adaptive Exponential Integrate and Fire (AdEx) Neuron Model

SpikeFun v0.94 brings support for Adaptive Exponential Integrate and Fire (AdEx or Brette-Gerstner) neuron model. AdEx neuron model is qualitatively similar to Izhikevich' neuron model, undergoes the same bifurcations and can reproduce various spiking patterns of cortical neurons. However, AdEx model parameters could be directly related to biophysical quantities and its spike initiation is exponential, not quadratic. DigiCortex Engine can simulate neurons with both models in the same simulation. Click on the title to find out more...


Subscribe to www.digicortex.net RSS