The goal of this experiment is to test the scaling of our code on forest fire
samples. I used the code from 

~/dev/graph_models

as a starting point. 

The main codes are:

scaling_study_1
scaling_study_s

Neither run to completon on my machine with 24GB of RAM.
The first one failed on the largest study due to swapping 
while generating the graph,. The second one failed in a 
really strange way. It had plenty of memory, but kept
the CPU utilization at a low percentage. Not sure what
happened! We'll make due the best we can.


    end
dt = 1.841560e+03, npush = 2616175484
dt = 1.633588e+03, npush = 627257594
>> nnz(y)
