File flow.txt will run synthesis, logic simulation, place and route,
LVS, DRC and creation of CIF and GDS files.

The vasy, boom, boog and loon synthesis steps are all looped to
try to find the fastest netlist.

vasy produces two netlists, coming from two different sets of
     parameters.
boom produces in excess of 200 netlists.
boog produces around 50 netlists which have different critical
     path delays. Boog only considers the minimum drive strength
     cells.
loon speeds up each of the netlists by optimising in turn with
     0fF and 6fF wireload libraries until no more improvement is
     possible. This step is quite slow, so a time limit has been
     set. The input netlists from boog have been sorted so that
     the fastest ones are optimised first.

The P&R uses customised versions of OCP and NERO. Scripts extend
the width by two tracks left and right for good well and substrate
termination. Routing maps metal-1 to ALU2 so that metal-1 connections
between adjacent cells are possible. Two runs give layout with
horizontal and vertical metal-2.

After routing the netlist, a script adds pieces of metal around the
vias to give a 2 lambda end overlap of metal. There are no DRC
violations with this technology because of the 10 lambda track pitch.

Critical path delay is 589 ps    (these numbers same as ssxlib013)
Area is                496 um^2
Gate count              42
Occupancy               77%

The metal length numbers below are indicative only as segments are
not merged. The entire additional segment added around each via is
included, rather than the 1 lambda extra overlap. The number of vias
though is a true indication of the quality of the route.

                 stxlib013     ssxlib013
               m2-ver m2-hor m2-ver m2-hor
length metal-1  4050   3910
       metal-2  2420   3870   2860   3960
       metal-3  3620   2320   3820   2760
number vias      147    109    182    102

The vertical metal-2 makes the best use of horizontal metal-1
routing in the stxlib013. For horizontal metal-2, metal-1 routing
between adjacent cells would be wrong-way.
One can see that the horizontal metal-2 has 80 fewer vias than
vertical metal-2 for the ssxlib013. This is an indication that
horizontal metal-2 is a better solution for this library.
For vertical metal-2, allowing horizontal metal-1 connections
reduces the number of vias from 182 to 147 or by 19%.
