TODO.txt

Finish recorders - done
Check membrane dynamics are taken account of in spike transfer - done
Check dt is correctly applied in kernels - done for LIF
Get params from last year's work and Nicolas' papers - done
Check results with sim by nicolas (fixed synapse strengths) - done

Speed check with Nicolas sim (remove synapse updates and add timer to inner loop) - done
Is there a memory effect of not calling waitForKernel() (does it grow?) 

Consider separating currents into separate variables
Consider adding parallel spike transfer
Add delay to spike transfer
Stop recalculating local workgroup size on each enqueueNDRangeKernel() - done
Separate drives to EXC and INH populations
Separate timestepping for Synapse (ms) vs Neuron (sub-ms) [inner loop for neuron steps]
Compare Marsaglia and Random123 using DIEHARD
Change to Random123 based on card random number generator
Consider using exp() rather than simple Euler

Consider changing lif-lif dereferencing system (it's inefficient)
Event-based synapse updates