Synaptogen.jl

Synaptogen.jl

This is the Julia implementation of Synaptogen.

It runs on both CPUs and GPUs, and can easily handle millions of synaptic weights.

Installation

  import Pkg; Pkg.add("Synaptogen")

Examples

These basic examples do the following:

• Initialize a million memory cells (in their high resistance states)
• Apply -2 V to each cell, putting them into their low resistance states
• Apply a random voltage to each cell
• Make a current readout of all the cells individually (at a default of 0.2 V)
• Perform a "Vector Matrix Multiplication" by 1024×1024 crossbar readout

CPU version

using Synaptogen
M = 1024 * 1024
cells = [Cell() for m in 1:M]

applyVoltage!.(cells, -2)

voltages = randn(Float32, M)
applyVoltage!.(cells, voltages)

I = Iread.(cells)

crossbar = reshape(cells, 1024, 1024) # Reshaping optional
col_voltages = randn(Float32, 1024) * .2f0
row_currents = crossbar * col_voltages

GPU version

using Synaptogen, CUDA
M = 1024 * 1024
cells = CellArrayGPU(M)

applyVoltage!(cells, -2)

voltages = CUDA.randn(M)
applyVoltage!(cells, voltages)

I = Iread(cells)

col_voltages = CUDA.randn(1024) * .2f0
row_currents = cells * col_voltages