New version of testvalue.sh up. I added the input value to the output.txt file to make it easier to read since it starts at 0. Right now we're trying to work out the equation for the scale of certain cfq and deadline tunables on the GN, which is a bizarre one:
7, 15, 23, 31, 39, 46, 54, 62
Increases by 8s but occasionally (on the 5th jump, though sometimes 6th) it increases by 7.
Edit: Every 3rd increase of 7 occurs on the 6th instead of the 5th.
Edit 2: Stats genius of the group, @pkgnex
worked out what the trick was; dropping the decimals once calculated.
So in Excel: y=TRUNC(x*7.8125,0)
The "why" of the GN tunables working this way, with a value of 125/16 is a very good question. No ideas there, personally. Something to do with the GN having a HZ value of 128 (where on others it's 100 or even 200) I'd assume, but I don't see the direct connection.
This was very handy to know though, because the cfq+deadline source requires the number of jumps
(x) to hardcode the values, not the actual desired numbers like you'd expect.