```
$ ./some.sh
T,N,S,W,baseline_time,exec_time,overhead,t_inj,i_inj,bit,threshold,TP,FP,TN,FN
-,-,-,-,-------------,---------,--------,-----,-----,---,---------,--,--,--,--
1,20000,200000,1000,10000,3.105506,3.128887,0.752888,15505,107356,20,1.000000E-08,1,467,7,0
2,20000,200000,1000,10000,3.100944,3.119424,0.595950,11239,163788,24,1.000000E-08,1,470,4,0
3,20000,200000,1000,10000,3.126926,3.156333,0.940453,3937,23440,10,1.000000E-08,1,466,8,0
4,20000,200000,1000,10000,3.145935,3.165772,0.630557,14976,123857,16,1.000000E-08,1,466,8,0
5,20000,200000,1000,10000,3.164470,3.187001,0.712000,8028,70187,26,1.000000E-08,1,471,3,0
6,20000,200000,1000,10000,3.180360,3.222868,1.336582,14184,104769,8,1.000000E-08,1,466,8,0
7,20000,200000,1000,10000,3.155341,3.174633,0.611411,18174,56828,20,1.000000E-08,1,471,3,0
8,20000,200000,1000,10000,3.226369,3.230618,0.131692,13475,115903,29,1.000000E-08,1,464,10,0
9,20000,200000,1000,10000,3.222665,3.224744,0.064512,8479,112625,30,1.000000E-08,2,464,9,0
10,20000,200000,1000,10000,3.242528,3.203076,1.216707,11147,159143,2,1.000000E-08,1,464,10,0
11,20000,200000,1000,10000,3.247372,3.243506,0.119049,12779,122596,0,1.000000E-08,1,459,15,0

```

ex:
./lcc 4 32
./bin/jac.4.32.abft.inject 20 200 0
./bin/jac.4.32.abft.inject 20 200 29

next step:
- need to implement re-computation piece (otherwise flipped
  value propagates through all dependent checksum/invariants)

I hoisted the stencil sets S* out of all of the power propogation pieces

but still not ideal, something is going on
the algorithmic overhead decreases as we increase the checksum region
size, but the actual overhead increases

don't know why yet