diff --git a/compiler/base/contact.py b/compiler/base/contact.py
index 1ffdc7bf..f4fda552 100644
--- a/compiler/base/contact.py
+++ b/compiler/base/contact.py
@@ -47,8 +47,16 @@ class contact(hierarchy_design.hierarchy_design):
         
         self.layer_stack = layer_stack
         self.dimensions = dimensions
-        if directions:
+
+        # Non-preferred directions
+        if directions == "nonpref":
+            first_dir = "H" if self.get_preferred_direction(layer_stack[0])=="V" else "V"
+            second_dir = "H" if self.get_preferred_direction(layer_stack[2])=="V" else "V"
+            self.directions = (first_dir, second_dir)
+        # User directions
+        elif directions:
             self.directions = directions
+        # Preferred directions
         else:
             self.directions = (tech.preferred_directions[layer_stack[0]],
                                tech.preferred_directions[layer_stack[2]])
diff --git a/compiler/base/design.py b/compiler/base/design.py
index 73ace38e..7a570b28 100644
--- a/compiler/base/design.py
+++ b/compiler/base/design.py
@@ -6,6 +6,7 @@
 # All rights reserved.
 #
 from hierarchy_design import hierarchy_design
+from utils import round_to_grid
 import contact
 from globals import OPTS
 import re
@@ -31,48 +32,98 @@ class design(hierarchy_design):
         in many places in the compiler.
         """
         
+        from tech import layer_indices
         import tech
-        for key in dir(tech):
-            # Single layer width rules
-            match = re.match(r".*_stack$", key)
-            if match:
+        for layer in layer_indices:
+            key = "{}_stack".format(layer)
+
+            # Set the stack as a local helper
+            try:
                 layer_stack = getattr(tech, key)
-
-                # Set the stack as a local helper
                 setattr(self, key, layer_stack)
+            except AttributeError:
+                pass
 
-                # Add the pitch
-                setattr(self,
-                        "{}_pitch".format(layer_stack[0]),
-                        self.compute_pitch(layer_stack))
-
+            # Skip computing the pitch for active
+            if layer == "active":
+                continue
+            
+            # Add the pitch
+            setattr(self,
+                    "{}_pitch".format(layer),
+                    self.compute_pitch(layer, True))
+            
+            # Add the non-preferrd pitch (which has vias in the "wrong" way)
+            setattr(self,
+                    "{}_nonpref_pitch".format(layer),
+                    self.compute_pitch(layer, False))
+                
         if False:
-            print("m1_pitch", self.m1_pitch)
-            print("m2_pitch", self.m2_pitch)
-            print("m3_pitch", self.m3_pitch)
+            from tech import preferred_directions
+            print(preferred_directions)
+            from tech import layer, layer_indices
+            for name in layer_indices:
+                if name == "active":
+                    continue
+                try:
+                    print("{0} width {1} space {2}".format(name,
+                                                           getattr(self, "{}_width".format(name)),
+                                                           getattr(self, "{}_space".format(name))))
+
+                    print("pitch {0} nonpref {1}".format(getattr(self, "{}_pitch".format(name)),
+                                                         getattr(self, "{}_nonpref_pitch".format(name))))
+                except AttributeError:
+                    pass
             import sys
             sys.exit(1)
 
-    def compute_pitch(self, layer_stack):
+    def compute_pitch(self, layer, preferred=True):
         
         """
-        This is contact direction independent pitch,
-        i.e. we take the maximum contact dimension
+        This is the preferred direction pitch
+        i.e. we take the minimum or maximum contact dimension
         """
+        # Find the layer stacks this is used in
+        from tech import layer_stacks
+        pitches = []
+        for stack in layer_stacks:
+            # Compute the pitch with both vias above and below (if they exist)
+            if stack[0] == layer:
+                pitches.append(self.compute_layer_pitch(stack, preferred))
+            if stack[2] == layer:
+                pitches.append(self.compute_layer_pitch(stack[::-1], True))
+
+        return max(pitches)
+
+    def compute_layer_pitch(self, layer_stack, preferred):
+
         (layer1, via, layer2) = layer_stack
+        try:
+            if layer1 == "poly" or layer1 == "active":
+                contact1 = getattr(contact, layer1 + "_contact")
+            else:
+                contact1 = getattr(contact, layer1 + "_via")
+        except AttributeError:
+            contact1 = getattr(contact, layer2 + "_via")
 
-        if layer1 == "poly" or layer1 == "active":
-            contact1 = getattr(contact, layer1 + "_contact")
+        if preferred:
+            if self.get_preferred_direction(layer1) == "V":
+                contact_width = contact1.first_layer_width
+            else:
+                contact_width = contact1.first_layer_height
         else:
-            contact1 = getattr(contact, layer1 + "_via")
-        max_contact = max(contact1.width, contact1.height)
-        
-        layer1_space = getattr(self, layer1 + "_space")
-        layer2_space = getattr(self, layer2 + "_space")
-        pitch = max_contact + max(layer1_space, layer2_space)
+            if self.get_preferred_direction(layer1) == "V":
+                contact_width = contact1.first_layer_height
+            else:
+                contact_width = contact1.first_layer_width
+        layer_space = getattr(self, layer1 + "_space")
 
-        return pitch
-        
+        #print(layer_stack)
+        #print(contact1)
+        pitch = contact_width + layer_space
+
+        return round_to_grid(pitch)
+    
     def setup_drc_constants(self):
         """
         These are some DRC constants used in many places
diff --git a/compiler/base/hierarchy_layout.py b/compiler/base/hierarchy_layout.py
index e6f9ac9f..df7095df 100644
--- a/compiler/base/hierarchy_layout.py
+++ b/compiler/base/hierarchy_layout.py
@@ -12,6 +12,7 @@ import debug
 from math import sqrt
 from tech import drc, GDS
 from tech import layer as techlayer
+from tech import layer_indices
 from tech import layer_stacks
 import os
 from globals import OPTS
@@ -41,12 +42,15 @@ class layout():
         self.visited = []    # List of modules we have already visited
         self.is_library_cell = False # Flag for library cells
         self.gds_read()
+        
         try:
             from tech import power_grid
             self.pwr_grid_layer = power_grid[0]
         except ImportError:
             self.pwr_grid_layer = "m3"
 
+            
+
     ############################################################
     # GDS layout
     ############################################################
@@ -444,7 +448,40 @@ class layout():
                     path=coordinates,
                     layer_widths=layer_widths)
 
-    def add_wire(self, layers, coordinates):
+    def add_zjog(self, layer, start, end, first_direction="H"):
+        """
+        Add a simple jog at the halfway point.
+        If layer is a single value, it is a path.
+        If layer is a tuple, it is a wire with preferred directions.
+        """
+
+        # vertical first
+        if first_direction == "V":
+            mid1 = vector(start.x, 0.5 * start.y + 0.5 * end.y)
+            mid2 = vector(end.x, mid1.y)
+        # horizontal first
+        elif first_direction == "H":
+            mid1 = vector(0.5 * start.x + 0.5 * end.x, start.y)
+            mid2 = vector(mid1, end.y)
+        else:
+            debug.error("Invalid direction for jog -- must be H or V.")
+            
+        if layer in layer_stacks:
+            self.add_wire(layer, [start, mid1, mid2, end])
+        elif layer in techlayer:
+            self.add_path(layer, [start, mid1, mid2, end])
+        else:
+            debug.error("Could not find layer {}".format(layer))
+
+    def add_horizontal_zjog_path(self, layer, start, end):
+        """ Add a simple jog at the halfway point """
+
+        # horizontal first
+        mid1 = vector(0.5 * start.x + 0.5 * end.x, start.y)
+        mid2 = vector(mid1, end.y)
+        self.add_path(layer, [start, mid1, mid2, end])
+        
+    def add_wire(self, layers, coordinates, widen_short_wires=True):
         """Connects a routing path on given layer,coordinates,width.
         The layers are the (horizontal, via, vertical). """
         import wire
@@ -452,7 +489,8 @@ class layout():
         # into rectangles and contacts
         wire.wire(obj=self,
                   layer_stack=layers,
-                  position_list=coordinates)
+                  position_list=coordinates,
+                  widen_short_wires=widen_short_wires)
 
     def get_preferred_direction(self, layer):
         """ Return the preferred routing directions """
@@ -461,11 +499,6 @@ class layout():
 
     def add_via(self, layers, offset, size=[1, 1], directions=None, implant_type=None, well_type=None):
         """ Add a three layer via structure. """
-
-        if not directions:
-            directions = (self.get_preferred_direction(layers[0]),
-                          self.get_preferred_direction(layers[2]))
-
         from sram_factory import factory
         via = factory.create(module_type="contact",
                              layer_stack=layers,
@@ -486,11 +519,6 @@ class layout():
         Add a three layer via structure by the center coordinate
         accounting for mirroring and rotation.
         """
-
-        if not directions:
-            directions = (self.get_preferred_direction(layers[0]),
-                          self.get_preferred_direction(layers[2]))
-
         from sram_factory import factory
         via = factory.create(module_type="contact",
                              layer_stack=layers,
@@ -513,36 +541,47 @@ class layout():
         return inst
 
     def add_via_stack(self, offset, from_layer, to_layer,
-                      direction=None,
-                      size=[1, 1]):
+                      directions=None,
+                      size=[1, 1],
+                      implant_type=None,
+                      well_type=None):
         """
         Punch a stack of vias from a start layer to a target layer.
         """
         return self.__add_via_stack_internal(offset=offset,
-                                             direction=direction,
+                                             directions=directions,
                                              from_layer=from_layer,
                                              to_layer=to_layer,
                                              via_func=self.add_via,
                                              last_via=None,
-                                             size=size)
+                                             size=size,
+                                             implant_type=implant_type,
+                                             well_type=well_type)
 
-    def add_via_stack_center(self, offset, from_layer, to_layer,
-                             direction=None,
-                             size=[1, 1]):
+    def add_via_stack_center(self,
+                             offset,
+                             from_layer,
+                             to_layer,
+                             directions=None,
+                             size=[1, 1],
+                             implant_type=None,
+                             well_type=None):
         """
         Punch a stack of vias from a start layer to a target layer by the center
         coordinate accounting for mirroring and rotation.
         """
         return self.__add_via_stack_internal(offset=offset,
-                                             direction=direction,
+                                             directions=directions,
                                              from_layer=from_layer,
                                              to_layer=to_layer,
                                              via_func=self.add_via_center,
                                              last_via=None,
-                                             size=size)
+                                             size=size,
+                                             implant_type=implant_type,
+                                             well_type=well_type)
 
-    def __add_via_stack_internal(self, offset, direction, from_layer, to_layer,
-                                 via_func, last_via, size):
+    def __add_via_stack_internal(self, offset, directions, from_layer, to_layer,
+                                 via_func, last_via, size, implant_type=None, well_type=None):
         """
         Punch a stack of vias from a start layer to a target layer. Here we
         figure out whether to punch it up or down the stack.
@@ -551,8 +590,8 @@ class layout():
         if from_layer == to_layer:
             return last_via
 
-        from_id = int(from_layer[1])
-        to_id   = int(to_layer[1])
+        from_id = layer_indices[from_layer]
+        to_id   = layer_indices[to_layer]
 
         if from_id < to_id: # grow the stack up
             search_id = 0
@@ -563,19 +602,25 @@ class layout():
 
         curr_stack = next(filter(lambda stack: stack[search_id] == from_layer, layer_stacks), None)
         if curr_stack is None:
-            raise ValueError("Cannot create via from '{0}' to '{1}'." \
-                             "Layer '{0}' not defined"
-                             .format(from_layer, to_layer))
+            raise ValueError("Cannot create via from '{0}' to '{1}'."
+                             "Layer '{0}' not defined".format(from_layer, to_layer))
 
-        via = via_func(layers=curr_stack, size=size, offset=offset, directions=direction)
-        return self.__add_via_stack_internal(offset=offset,
-                                             direction=direction,
-                                             from_layer=curr_stack[next_id],
-                                             to_layer=to_layer,
-                                             via_func=via_func,
-                                             last_via=via,
-                                             size=size)
+        via = via_func(layers=curr_stack,
+                       size=size,
+                       offset=offset,
+                       directions=directions,
+                       implant_type=implant_type,
+                       well_type=well_type)
 
+        via = self.__add_via_stack_internal(offset=offset,
+                                            directions=directions,
+                                            from_layer=curr_stack[next_id],
+                                            to_layer=to_layer,
+                                            via_func=via_func,
+                                            last_via=via,
+                                            size=size)
+        return via
+    
     def add_ptx(self, offset, mirror="R0", rotate=0, width=1, mults=1, tx_type="nmos"):
         """Adds a ptx module to the design."""
         import ptx
@@ -729,47 +774,47 @@ class layout():
 
         return blockages
 
-    def create_horizontal_pin_bus(self, layer, pitch, offset, names, length):
+    def create_horizontal_pin_bus(self, layer, offset, names, length, pitch=None):
         """ Create a horizontal bus of pins. """
         return self.create_bus(layer,
-                               pitch,
                                offset,
                                names,
                                length,
                                vertical=False,
-                               make_pins=True)
+                               make_pins=True,
+                               pitch=pitch)
 
-    def create_vertical_pin_bus(self, layer, pitch, offset, names, length):
+    def create_vertical_pin_bus(self, layer, offset, names, length, pitch=None):
         """ Create a horizontal bus of pins. """
         return self.create_bus(layer,
-                               pitch,
                                offset,
                                names,
                                length,
                                vertical=True,
-                               make_pins=True)
+                               make_pins=True,
+                               pitch=pitch)
 
-    def create_vertical_bus(self, layer, pitch, offset, names, length):
+    def create_vertical_bus(self, layer, offset, names, length, pitch=None):
         """ Create a horizontal bus. """
         return self.create_bus(layer,
-                               pitch,
                                offset,
                                names,
                                length,
                                vertical=True,
-                               make_pins=False)
+                               make_pins=False,
+                               pitch=pitch)
 
-    def create_horizontal_bus(self, layer, pitch, offset, names, length):
+    def create_horizontal_bus(self, layer, offset, names, length, pitch=None):
         """ Create a horizontal bus. """
         return self.create_bus(layer,
-                               pitch,
                                offset,
                                names,
                                length,
                                vertical=False,
-                               make_pins=False)
+                               make_pins=False,
+                               pitch=pitch)
 
-    def create_bus(self, layer, pitch, offset, names, length, vertical, make_pins):
+    def create_bus(self, layer, offset, names, length, vertical, make_pins, pitch=None):
         """
         Create a horizontal or vertical bus. It can be either just rectangles, or actual
         layout pins. It returns an map of line center line positions indexed by name.
@@ -779,11 +824,14 @@ class layout():
 
         # half minwidth so we can return the center line offsets
         half_minwidth = 0.5 * drc["minwidth_{}".format(layer)]
+        if not pitch:
+            pitch = getattr(self, "{}_pitch".format(layer))
 
         line_positions = {}
         if vertical:
             for i in range(len(names)):
-                line_offset = offset + vector(i * pitch, 0)
+                line_offset = offset + vector(i * pitch,
+                                              0)
                 if make_pins:
                     self.add_layout_pin(text=names[i],
                                         layer=layer,
@@ -848,8 +896,10 @@ class layout():
                 # left/right then up/down
                 mid_pos = vector(bus_pos.x, pin_pos.y)
 
+            # Don't widen short wires because pin_pos and mid_pos could be really close
             self.add_wire(layer_stack,
-                          [bus_pos, mid_pos, pin_pos])
+                          [bus_pos, mid_pos, pin_pos],
+                          widen_short_wires=False)
 
             # Connect to the pin on the instances with a via if it is
             # not on the right layer
@@ -866,33 +916,76 @@ class layout():
                                     offset=bus_pos,
                                     rotate=90)
 
+    def connect_vbus(self, src_pin, dest_pin, hlayer="m3", vlayer="m2"):
+        """
+        Helper routine to connect an instance to a vertical bus.
+        Routes horizontal then vertical L shape.
+        """
+
+        if src_pin.cx()<dest_pin.cx():
+            in_pos = src_pin.rc()
+        else:
+            in_pos = src_pin.lc()
+        if src_pin.cy() < dest_pin.cy():
+            out_pos = dest_pin.bc()
+        else:
+            out_pos = dest_pin.uc()
+
+        # move horizontal first on layer stack
+        mid_pos = vector(out_pos.x, in_pos.y)
+        self.add_via_stack_center(from_layer=src_pin.layer,
+                                  to_layer=hlayer,
+                                  offset=in_pos)
+        self.add_path(hlayer, [in_pos, mid_pos])
+        self.add_via_stack_center(from_layer=hlayer,
+                                  to_layer=vlayer,
+                                  offset=mid_pos)
+        self.add_path(vlayer, [mid_pos, out_pos])
+        self.add_via_stack_center(from_layer=vlayer,
+                                  to_layer=dest_pin.layer,
+                                  offset=out_pos)
+
+    def connect_hbus(self, src_pin, dest_pin, hlayer="m3", vlayer="m2"):
+        """
+        Helper routine to connect an instance to a horizontal bus.
+        Routes horizontal then vertical L shape.
+        """
+
+        if src_pin.cx()<dest_pin.cx():
+            in_pos = src_pin.rc()
+        else:
+            in_pos = src_pin.lc()
+        if src_pin.cy() < dest_pin.cy():
+            out_pos = dest_pin.lc()
+        else:
+            out_pos = dest_pin.rc()
+
+        # move horizontal first
+        mid_pos = vector(out_pos.x, in_pos.y)
+        self.add_via_stack_center(from_layer=src_pin.layer,
+                                  to_layer=hlayer,
+                                  offset=in_pos)
+        self.add_path(hlayer, [in_pos, mid_pos])
+        self.add_via_stack_center(from_layer=hlayer,
+                                  to_layer=vlayer,
+                                  offset=mid_pos)
+        self.add_path(vlayer, [mid_pos, out_pos])
+        self.add_via_stack_center(from_layer=vlayer,
+                                  to_layer=dest_pin.layer,
+                                  offset=out_pos)
+        
     def get_layer_pitch(self, layer):
         """ Return the track pitch on a given layer """
-        if layer == "m1":
-            return (self.m1_pitch,
-                    self.m1_pitch - self.m1_space,
-                    self.m1_space)
-        elif layer == "m2":
-            return (self.m2_pitch,
-                    self.m2_pitch - self.m2_space,
-                    self.m2_space)
-        elif layer == "m3":
-            return (self.m3_pitch,
-                    self.m3_pitch - self.m3_space,
-                    self.m3_space)
-        elif layer == "m4":
-            from tech import layer as tech_layer
-            if "m4" in tech_layer:
-                return (self.m3_pitch,
-                        self.m3_pitch - self.m4_space,
-                        self.m4_space)
-            else:
-                return (self.m3_pitch,
-                        self.m3_pitch - self.m3_space,
-                        self.m3_space)
-        else:
-            debug.error("Cannot find layer pitch.")
-
+        try:
+            # FIXME: Using non-pref pitch here due to overlap bug in VCG constraints.
+            # It should just result in inefficient channel width but will work.
+            pitch = getattr(self, "{}_pitch".format(layer))
+            nonpref_pitch = getattr(self, "{}_nonpref_pitch".format(layer))
+            space = getattr(self, "{}_space".format(layer))
+        except AttributeError:
+            debug.error("Cannot find layer pitch.", -1)
+        return (nonpref_pitch, pitch, pitch - space, space)
+    
     def add_horizontal_trunk_route(self,
                                    pins,
                                    trunk_offset,
@@ -906,7 +999,7 @@ class layout():
         min_x = min([pin.center().x for pin in pins])
 
         # if we are less than a pitch, just create a non-preferred layer jog
-        if max_x-min_x <= pitch:
+        if max_x - min_x <= pitch:
             half_layer_width = 0.5 * drc["minwidth_{0}".format(self.vertical_layer)]
 
             # Add the horizontal trunk on the vertical layer!
@@ -975,14 +1068,15 @@ class layout():
     def create_channel_route(self, netlist,
                              offset,
                              layer_stack,
-                             layer_dirs=None,
+                             directions=None,
                              vertical=False):
         """
-        The net list is a list of the nets. Each net is a list of pins
-        to be connected.  Offset is the lower-left of where the
+        The net list is a list of the nets with each net being a list of pins
+        to be connected. The offset is the lower-left of where the
         routing channel will start.  This does NOT try to minimize the
         number of tracks -- instead, it picks an order to avoid the
-        vertical conflicts between pins.
+        vertical conflicts between pins. The track size must be the number of
+        nets times the *nonpreferred* routing of the non-track layer pitch.
 
         """
         def remove_net_from_graph(pin, g):
@@ -999,12 +1093,17 @@ class layout():
                     g[other_pin]=conflicts
             return g
 
-        def vcg_nets_overlap(net1, net2, vertical, pitch):
+        def vcg_nets_overlap(net1, net2, vertical):
             """
             Check all the pin pairs on two nets and return a pin
             overlap if any pin overlaps.
             """
 
+            if vertical:
+                pitch = self.horizontal_nonpref_pitch
+            else:
+                pitch = self.vertical_nonpref_pitch
+            
             for pin1 in net1:
                 for pin2 in net2:
                     if vcg_pin_overlap(pin1, pin2, vertical, pitch):
@@ -1017,8 +1116,7 @@ class layout():
             
             # FIXME: If the pins are not in a row, this may break.
             # However, a top pin shouldn't overlap another top pin,
-            # for example, so the
-            # extra comparison *shouldn't* matter.
+            # for example, so the extra comparison *shouldn't* matter.
 
             # Pin 1 must be in the "BOTTOM" set
             x_overlap = pin1.by() < pin2.by() and abs(pin1.center().x - pin2.center().x) < pitch
@@ -1028,7 +1126,7 @@ class layout():
             overlaps = (not vertical and x_overlap) or (vertical and y_overlap)
             return overlaps
 
-        if not layer_dirs:
+        if not directions:
             # Use the preferred layer directions
             if self.get_preferred_direction(layer_stack[0]) == "V":
                 self.vertical_layer = layer_stack[0]
@@ -1039,8 +1137,8 @@ class layout():
         else:
             # Use the layer directions specified to the router rather than
             # the preferred directions
-            debug.check(layer_dirs[0] != layer_dirs[1], "Must have unique layer directions.")
-            if layer_dirs[0] == "V":
+            debug.check(directions[0] != directions[1], "Must have unique layer directions.")
+            if directions[0] == "V":
                 self.vertical_layer = layer_stack[0]
                 self.horizontal_layer = layer_stack[2]
             else:
@@ -1048,9 +1146,10 @@ class layout():
                 self.vertical_layer = layer_stack[2]
 
         layer_stuff = self.get_layer_pitch(self.vertical_layer)
-        (self.vertical_pitch, self.vertical_width, self.vertical_space) = layer_stuff
+        (self.vertical_nonpref_pitch, self.vertical_pitch, self.vertical_width, self.vertical_space) = layer_stuff
+        
         layer_stuff = self.get_layer_pitch(self.horizontal_layer)
-        (self.horizontal_pitch, self.horizontal_width, self.horizontal_space) = layer_stuff
+        (self.horizontal_nonpref_pitch, self.horizontal_pitch, self.horizontal_width, self.horizontal_space) = layer_stuff
 
         # FIXME: Must extend this to a horizontal conflict graph
         # too if we want to minimize the
@@ -1070,6 +1169,10 @@ class layout():
             index += 1
             nets[net_name] = pin_list
 
+        # print("Nets:")
+        # for net_name in nets:
+         #    print(net_name, [x.name for x in nets[net_name]])
+            
         # Find the vertical pin conflicts
         # FIXME: O(n^2) but who cares for now
         for net_name1 in nets:
@@ -1081,21 +1184,15 @@ class layout():
                 # Skip yourself
                 if net_name1 == net_name2:
                     continue
-                if vertical and vcg_nets_overlap(nets[net_name1],
-                                                 nets[net_name2],
-                                                 vertical,
-                                                 self.vertical_pitch):
-                    vcg[net_name2].append(net_name1)
-                elif not vertical and vcg_nets_overlap(nets[net_name1],
-                                                       nets[net_name2],
-                                                       vertical,
-                                                       self.horizontal_pitch):
+                if vcg_nets_overlap(nets[net_name1],
+                                    nets[net_name2],
+                                    vertical):
                     vcg[net_name2].append(net_name1)
 
         # list of routes to do
         while vcg:
             # from pprint import pformat
-            # print("VCG:\n",pformat(vcg))
+            # print("VCG:\n", pformat(vcg))
             # get a route from conflict graph with empty fanout set
             net_name = None
             for net_name, conflicts in vcg.items():
@@ -1119,26 +1216,28 @@ class layout():
                 self.add_vertical_trunk_route(pin_list,
                                               offset,
                                               layer_stack,
-                                              self.vertical_pitch)
-                offset += vector(self.vertical_pitch, 0)
+                                              self.vertical_nonpref_pitch)
+                # This accounts for the via-to-via spacings
+                offset += vector(self.horizontal_nonpref_pitch, 0)
             else:
                 self.add_horizontal_trunk_route(pin_list,
                                                 offset,
                                                 layer_stack,
-                                                self.horizontal_pitch)
-                offset += vector(0, self.horizontal_pitch)
+                                                self.horizontal_nonpref_pitch)
+                # This accounts for the via-to-via spacings
+                offset += vector(0, self.vertical_nonpref_pitch)
 
-    def create_vertical_channel_route(self, netlist, offset, layer_stack, layer_dirs=None):
+    def create_vertical_channel_route(self, netlist, offset, layer_stack, directions=None):
         """
         Wrapper to create a vertical channel route
         """
-        self.create_channel_route(netlist, offset, layer_stack, layer_dirs, vertical=True)
+        self.create_channel_route(netlist, offset, layer_stack, directions, vertical=True)
 
-    def create_horizontal_channel_route(self, netlist, offset, layer_stack, layer_dirs=None):
+    def create_horizontal_channel_route(self, netlist, offset, layer_stack, directions=None):
         """
         Wrapper to create a horizontal channel route
         """
-        self.create_channel_route(netlist, offset, layer_stack, layer_dirs, vertical=False)
+        self.create_channel_route(netlist, offset, layer_stack, directions, vertical=False)
 
     def add_boundary(self, ll=vector(0, 0), ur=None):
         """ Add boundary for debugging dimensions """
@@ -1200,27 +1299,18 @@ class layout():
                               "supply router."
                               .format(name, inst.name, self.pwr_grid_layer))
 
-    def add_power_pin(self, name, loc, size=[1, 1], vertical=False, start_layer="m1"):
+    def add_power_pin(self, name, loc, size=[1, 1], directions=None, start_layer="m1"):
         """
         Add a single power pin from the lowest power_grid layer down to M1 (or li) at
         the given center location. The starting layer is specified to determine
         which vias are needed.
         """
 
-        # Force vdd/gnd via stack to be vertically or horizontally oriented
-        # Default: None, uses prefered metal directions
-        if vertical:
-            direction = ("V", "V")
-        elif not vertical and vertical is not None:
-            direction = ("H", "H")
-        else:
-            direction = None
-
         via = self.add_via_stack_center(from_layer=start_layer,
                                         to_layer=self.pwr_grid_layer,
                                         size=size,
                                         offset=loc,
-                                        direction=direction)
+                                        directions=directions)
         if start_layer == self.pwr_grid_layer:
             self.add_layout_pin_rect_center(text=name,
                                             layer=self.pwr_grid_layer,
diff --git a/compiler/base/wire.py b/compiler/base/wire.py
index 10272ede..9c8dc4e6 100644
--- a/compiler/base/wire.py
+++ b/compiler/base/wire.py
@@ -19,11 +19,14 @@ class wire(wire_path):
     not, it will always go down first.
     The points are the center of the wire.
     The layer stack is the vertical, contact/via, and horizontal layers, respectively.
+    The widen option will avoid via-to-via spacing problems for really short segments
+    (added as an option so we can disable it in bus connections)
     """
-    def __init__(self, obj, layer_stack, position_list):
+    def __init__(self, obj, layer_stack, position_list, widen_short_wires=True):
         self.obj = obj
         self.layer_stack = layer_stack
         self.position_list = position_list
+        self.widen_short_wires = widen_short_wires
         self.pins = [] # used for matching parm lengths
         self.switch_pos_list = []
 
@@ -114,7 +117,7 @@ class wire(wire_path):
                 line_length = pl[index + 1][0] - pl[index][0]
                 # Make the wire wider to avoid via-to-via spacing problems
                 # But don't make it wider if it is shorter than one via
-                if abs(line_length) < self.pitch and abs(line_length) > self.horiz_layer_contact_width:
+                if self.widen_short_wires and abs(line_length) < self.pitch and abs(line_length) > self.horiz_layer_contact_width:
                     width = self.horiz_layer_contact_width
                 else:
                     width = self.horiz_layer_width
@@ -134,7 +137,7 @@ class wire(wire_path):
                 line_length = pl[index + 1][1] - pl[index][1]
                 # Make the wire wider to avoid via-to-via spacing problems
                 # But don't make it wider if it is shorter than one via
-                if abs(line_length) < self.pitch and abs(line_length) > self.vert_layer_contact_width:
+                if self.widen_short_wires and abs(line_length) < self.pitch and abs(line_length) > self.vert_layer_contact_width:
                     width = self.vert_layer_contact_width
                 else:
                     width = self.vert_layer_width
diff --git a/compiler/base/wire_path.py b/compiler/base/wire_path.py
index ebfb8a0a..31d0ae78 100644
--- a/compiler/base/wire_path.py
+++ b/compiler/base/wire_path.py
@@ -24,11 +24,12 @@ def create_rectilinear_route(my_list):
             my_list.append(vector(pl[index][0], pl[index + 1][1]))
     my_list.append(vector(pl[-1]))
     return my_list
-        
+
+
 class wire_path():
     """
     Object metal wire_path; given the layer type
-    Add a wire_path of minimium metal width between a set of points. 
+    Add a wire_path of minimium metal width between a set of points.
     The points should be rectilinear to control the bend points. If
     not, it will always go down first. The points are the center of the wire_path.
     If width is not given, it uses minimum layer width.
@@ -37,7 +38,7 @@ class wire_path():
         self.obj = obj
         self.layer_name = layer
         self.layer_id = techlayer[layer]
-        if width==None:
+        if width == None:
             self.layer_width = drc["minwidth_{0}".format(layer)]
         else:
             self.layer_width = width
@@ -46,7 +47,6 @@ class wire_path():
         self.switch_pos_list = []
         self.create_layout()
 
-
     def create_layout(self):
         self.create_rectilinear()
         self.connect_corner()
@@ -60,9 +60,9 @@ class wire_path():
 
     def connect_corner(self):
         """ Add a corner square at every corner of the wire_path."""
-        from itertools import tee,islice
-        nwise = lambda g,n=2: zip(*(islice(g,i,None) for i,g in enumerate(tee(g,n))))
-        threewise=nwise(self.position_list,3)
+        from itertools import tee, islice
+        nwise = lambda g, n=2: zip(*(islice(g, i, None) for i, g in enumerate(tee(g, n))))
+        threewise=nwise(self.position_list, 3)
 
         for (a, offset, c) in list(threewise):
             # add a exceptions to prevent a corner when we retrace back in the same direction
@@ -74,7 +74,6 @@ class wire_path():
                              offset[1] - 0.5 * self.layer_width]
             self.draw_corner_wire(corner_offset)
 
-
     def draw_corner_wire(self, offset):
         """ This function adds the corner squares since the center
         line convention only draws to the center of the corner."""
@@ -117,7 +116,7 @@ class wire_path():
 
     def add_line(self, layer_name, length, offset, orientation, layer_width):
         """
-        straight line object with layer_minwidth 
+        straight line object with layer_minwidth
         (orientation: "vertical" or "horizontal") default is vertical
         """
 
diff --git a/compiler/bitcells/bitcell.py b/compiler/bitcells/bitcell.py
index 68e70e7e..b0e79208 100644
--- a/compiler/bitcells/bitcell.py
+++ b/compiler/bitcells/bitcell.py
@@ -50,7 +50,7 @@ class bitcell(bitcell_base.bitcell_base):
         self.pin_map = bitcell.pin_map
         self.add_pin_types(self.type_list)
         self.nets_match = self.do_nets_exist(self.storage_nets)
-
+              
     def get_all_wl_names(self):
         """ Creates a list of all wordline pin names """
         if props.bitcell.split_wl:
diff --git a/compiler/bitcells/pbitcell.py b/compiler/bitcells/pbitcell.py
index be96a129..aea7a6dc 100644
--- a/compiler/bitcells/pbitcell.py
+++ b/compiler/bitcells/pbitcell.py
@@ -264,7 +264,7 @@ class pbitcell(bitcell_base.bitcell_base):
                              + 2 * self.implant_enclose_active \
                              + 0.5*(self.inverter_pmos.active_contact.height - self.m1_width)
         metal1_constraint = max(inverter_pmos_contact_extension, 0) + self.m1_space
-        self.vdd_offset = max(implant_constraint, metal1_constraint)  + 0.5*self.m1_width
+        self.vdd_offset = max(implant_constraint, metal1_constraint)  + self.m1_width
 
         # read port dimensions
         width_reduction = self.read_nmos.active_width - self.read_nmos.get_pin("D").cx()
@@ -275,7 +275,7 @@ class pbitcell(bitcell_base.bitcell_base):
         Calculate positions that describe the edges
         and dimensions of the cell
         """
-        self.botmost_ypos = self.m1_offset - self.total_ports * self.m1_pitch
+        self.botmost_ypos = self.m1_offset - self.total_ports * self.m1_nonpref_pitch
         self.topmost_ypos = self.inverter_nmos_ypos \
                             + self.inverter_nmos.active_height \
                             + self.inverter_gap \
@@ -378,14 +378,15 @@ class pbitcell(bitcell_base.bitcell_base):
                                       + 0.5 * contact.poly_contact.height,
                                       self.cross_couple_upper_ypos)
         self.add_via_center(layers=self.poly_stack,
-                            offset=contact_offset_left)
-
+                            offset=contact_offset_left,
+                            directions=("H", "H"))
 
         contact_offset_right =  vector(self.inverter_nmos_right.get_pin("S").lc().x \
                                        - 0.5*contact.poly_contact.height,
                                        self.cross_couple_lower_ypos)
         self.add_via_center(layers=self.poly_stack,
-                            offset=contact_offset_right)
+                            offset=contact_offset_right,
+                            directions=("H", "H"))
 
         # connect contacts to gate poly (cross couple connections)
         gate_offset_right = vector(self.inverter_nmos_right.get_pin("G").lc().x,
@@ -399,12 +400,12 @@ class pbitcell(bitcell_base.bitcell_base):
     def route_rails(self):
         """ Adds gnd and vdd rails and connects them to the inverters """
         # Add rails for vdd and gnd
-        gnd_ypos = self.m1_offset - self.total_ports * self.m1_pitch
+        gnd_ypos = self.m1_offset - self.total_ports * self.m1_nonpref_pitch
         self.gnd_position = vector(0, gnd_ypos)
         self.add_rect_center(layer="m1",
                              offset=self.gnd_position,
                              width=self.width)
-        self.add_power_pin("gnd", vector(0, gnd_ypos))
+        self.add_power_pin("gnd", vector(0, gnd_ypos), directions=("H", "H"))
         
         
         vdd_ypos = self.inverter_nmos_ypos \
@@ -416,7 +417,7 @@ class pbitcell(bitcell_base.bitcell_base):
         self.add_rect_center(layer="m1",
                              offset=self.vdd_position,
                              width=self.width)
-        self.add_power_pin("vdd", vector(0, vdd_ypos))
+        self.add_power_pin("vdd", vector(0, vdd_ypos), directions=("H", "H"))
         
     def create_readwrite_ports(self):
         """
@@ -483,7 +484,7 @@ class pbitcell(bitcell_base.bitcell_base):
                                                        self.port_ypos])
 
             # add pin for RWWL
-            rwwl_ypos = self.m1_offset - k * self.m1_pitch
+            rwwl_ypos = self.m1_offset - k * self.m1_nonpref_pitch
             self.rwwl_positions[k] = vector(0, rwwl_ypos)
             self.add_layout_pin_rect_center(text=self.rw_wl_names[k],
                                             layer="m1",
@@ -579,8 +580,8 @@ class pbitcell(bitcell_base.bitcell_base):
 
             # add pin for WWL
             wwl_ypos = rwwl_ypos = self.m1_offset \
-                                   - self.num_rw_ports * self.m1_pitch \
-                                   - k * self.m1_pitch
+                                   - self.num_rw_ports * self.m1_nonpref_pitch \
+                                   - k * self.m1_nonpref_pitch
             self.wwl_positions[k] = vector(0, wwl_ypos)
             self.add_layout_pin_rect_center(text=self.w_wl_names[k],
                                             layer="m1",
@@ -705,9 +706,9 @@ class pbitcell(bitcell_base.bitcell_base):
 
             # add pin for RWL
             rwl_ypos = rwwl_ypos = self.m1_offset \
-                                   - self.num_rw_ports * self.m1_pitch \
-                                   - self.num_w_ports * self.m1_pitch \
-                                   - k * self.m1_pitch
+                                   - self.num_rw_ports * self.m1_nonpref_pitch \
+                                   - self.num_w_ports * self.m1_nonpref_pitch \
+                                   - k * self.m1_nonpref_pitch
             self.rwl_positions[k] = vector(0, rwl_ypos)
             self.add_layout_pin_rect_center(text=self.r_wl_names[k],
                                             layer="m1",
@@ -771,6 +772,7 @@ class pbitcell(bitcell_base.bitcell_base):
                 self.add_via_center(layers=self.poly_stack,
                                     offset=port_contact_offset)
 
+
                 self.add_path("poly", [gate_offset, port_contact_offset])
                 self.add_path("m1",
                               [port_contact_offset, wl_contact_offset])
@@ -821,7 +823,8 @@ class pbitcell(bitcell_base.bitcell_base):
             # Leave bitline disconnected if a dummy cell
             if not self.dummy_bitcell:
                 self.add_via_center(layers=self.m1_stack,
-                                    offset=port_contact_offest)
+                                    offset=port_contact_offest,
+                                    directions="nonpref")
 
             self.add_path("m2",
                           [port_contact_offest, bl_offset], width=contact.m1_via.height)
@@ -833,7 +836,8 @@ class pbitcell(bitcell_base.bitcell_base):
             # Leave bitline disconnected if a dummy cell
             if not self.dummy_bitcell:
                 self.add_via_center(layers=self.m1_stack,
-                                    offset=port_contact_offest)
+                                    offset=port_contact_offest,
+                                    directions="nonpref")
 
             self.add_path("m2",
                           [port_contact_offest, br_offset], width=contact.m1_via.height)
@@ -850,7 +854,9 @@ class pbitcell(bitcell_base.bitcell_base):
 
         for position in nmos_contact_positions:
             self.add_via_center(layers=self.m1_stack,
-                                offset=position)
+                                offset=position,
+                                directions=("V", "V"))
+
 
             if position.x > 0:
                 contact_correct = 0.5 * contact.m1_via.height
@@ -859,7 +865,8 @@ class pbitcell(bitcell_base.bitcell_base):
             supply_offset = vector(position.x + contact_correct,
                                    self.gnd_position.y)
             self.add_via_center(layers=self.m1_stack,
-                                offset=supply_offset)
+                                offset=supply_offset,
+                                directions=("H", "H"))
 
             self.add_path("m2", [position, supply_offset])
 
@@ -924,13 +931,16 @@ class pbitcell(bitcell_base.bitcell_base):
                                        - self.poly_to_contact - 0.5*contact.poly_contact.width,
                                        self.cross_couple_upper_ypos)
         self.add_via_center(layers=self.poly_stack,
-                            offset=left_storage_contact)
+                            offset=left_storage_contact,
+                            directions=("H", "H"))
+
 
         right_storage_contact =  vector(self.inverter_nmos_right.get_pin("G").rc().x \
                                         + self.poly_to_contact + 0.5*contact.poly_contact.width,
                                         self.cross_couple_upper_ypos)
         self.add_via_center(layers=self.poly_stack,
-                            offset=right_storage_contact)
+                            offset=right_storage_contact,
+                            directions=("H", "H"))
 
         inverter_gate_offset_left = vector(self.inverter_nmos_left.get_pin("G").lc().x, self.cross_couple_upper_ypos)
         self.add_path("poly", [left_storage_contact, inverter_gate_offset_left])
@@ -1007,8 +1017,7 @@ class pbitcell(bitcell_base.bitcell_base):
             well_height = self.vdd_position.y - inverter_well_ypos \
                           + self.nwell_enclose_active + drc["minwidth_tx"]
 
-            # FIXME fudge factor xpos
-            offset = [inverter_well_xpos + 2*self.nwell_enclose_active, inverter_well_ypos]
+            offset = [inverter_well_xpos, inverter_well_ypos]
             self.add_rect(layer="nwell",
                           offset=offset,
                           width=well_width,
diff --git a/compiler/modules/bank.py b/compiler/modules/bank.py
index 597d2fc0..40eabde6 100644
--- a/compiler/modules/bank.py
+++ b/compiler/modules/bank.py
@@ -86,10 +86,10 @@ class bank(design.design):
             self.add_pin(self.bitcell_array.get_rbl_bl_name(self.port_rbl_map[port]), "OUTPUT")
         for port in self.write_ports:
             for bit in range(self.word_size + self.num_spare_cols):
-                self.add_pin("din{0}_{1}".format(port,bit), "INPUT")
+                self.add_pin("din{0}_{1}".format(port, bit), "INPUT")
         for port in self.all_ports:
             for bit in range(self.addr_size):
-                self.add_pin("addr{0}_{1}".format(port,bit), "INPUT")
+                self.add_pin("addr{0}_{1}".format(port, bit), "INPUT")
 
         # For more than one bank, we have a bank select and name
         # the signals gated_*.
@@ -133,16 +133,14 @@ class bank(design.design):
         bl_pin = self.bitcell_array_inst.get_pin(bl_pin_name)
         # This will ensure the pin is only on the top or bottom edge
         if port % 2:
-            via_offset = bl_pin.uc() + vector(0, self.m2_pitch)
+            via_offset = bl_pin.uc() + vector(0, 1.5 * self.m2_pitch)
             left_right_offset = vector(self.max_x_offset, via_offset.y)
         else:
-            via_offset = bl_pin.bc() - vector(0, self.m2_pitch)
+            via_offset = bl_pin.bc() - vector(0, 1.5 * self.m2_pitch)
             left_right_offset = vector(self.min_x_offset, via_offset.y)
-        if bl_pin == "m1":
-            self.add_via_center(layers=self.m1_stack,
-                                offset=via_offset)
-        self.add_via_center(layers=self.m2_stack,
-                            offset=via_offset)
+        self.add_via_stack_center(from_layer=bl_pin.layer,
+                                  to_layer="m3",
+                                  offset=via_offset)
         self.add_layout_pin_segment_center(text="rbl_bl{0}".format(port),
                                            layer="m3",
                                            start=left_right_offset,
@@ -643,11 +641,10 @@ class bank(design.design):
         # Port 0
         # The bank is at (0,0), so this is to the left of the y-axis.
         # 2 pitches on the right for vias/jogs to access the inputs
-        control_bus_offset = vector(-self.m2_pitch * self.num_control_lines[0] - self.m2_pitch, self.min_y_offset)
+        control_bus_offset = vector(-self.m3_pitch * self.num_control_lines[0] - self.m3_pitch, self.min_y_offset)
         # The control bus is routed up to two pitches below the bitcell array
         control_bus_length = self.main_bitcell_array_bottom - self.min_y_offset - 2 * self.m1_pitch
         self.bus_xoffset[0] = self.create_bus(layer="m2",
-                                              pitch=self.m2_pitch,
                                               offset=control_bus_offset,
                                               names=self.control_signals[0],
                                               length=control_bus_length,
@@ -658,11 +655,10 @@ class bank(design.design):
         if len(self.all_ports)==2:
             # The other control bus is routed up to two pitches above the bitcell array
             control_bus_length = self.max_y_offset - self.main_bitcell_array_top - 2 * self.m1_pitch
-            control_bus_offset = vector(self.bitcell_array_right + self.m2_pitch,
+            control_bus_offset = vector(self.bitcell_array_right + self.m3_pitch,
                                         self.max_y_offset - control_bus_length)
             # The bus for the right port is reversed so that the rbl_wl is closest to the array
             self.bus_xoffset[1] = self.create_bus(layer="m2",
-                                                  pitch=self.m2_pitch,
                                                   offset=control_bus_offset,
                                                   names=list(reversed(self.control_signals[1])),
                                                   length=control_bus_length,
@@ -862,9 +858,9 @@ class bank(design.design):
                 self.copy_layout_pin(self.column_decoder_inst[port], decoder_name, addr_name)
 
         if port % 2:
-            offset = self.column_decoder_inst[port].ll() - vector(self.num_col_addr_lines * self.m2_pitch, 0)
+            offset = self.column_decoder_inst[port].ll() - vector(self.num_col_addr_lines * self.m2_nonpref_pitch, 0)
         else:
-            offset = self.column_decoder_inst[port].lr() + vector(self.m2_pitch, 0)
+            offset = self.column_decoder_inst[port].lr() + vector(self.m2_nonpref_pitch, 0)
 
         decode_pins = [self.column_decoder_inst[port].get_pin(x) for x in decode_names]
         
@@ -872,7 +868,9 @@ class bank(design.design):
         column_mux_pins = [self.port_data_inst[port].get_pin(x) for x in sel_names]
         
         route_map = list(zip(decode_pins, column_mux_pins))
-        self.create_vertical_channel_route(route_map, offset, self.m1_stack)
+        self.create_vertical_channel_route(route_map,
+                                           offset,
+                                           self.m1_stack)
 
     def add_lvs_correspondence_points(self):
         """
@@ -928,30 +926,39 @@ class bank(design.design):
         # pre-decoder and this connection is in metal3
         connection = []
         connection.append((self.prefix + "p_en_bar{}".format(port),
-                           self.port_data_inst[port].get_pin("p_en_bar").lc()))
+                           self.port_data_inst[port].get_pin("p_en_bar").lc(),
+                           self.port_data_inst[port].get_pin("p_en_bar").layer))
 
         rbl_wl_name = self.bitcell_array.get_rbl_wl_name(self.port_rbl_map[port])
         connection.append((self.prefix + "wl_en{}".format(port),
-                           self.bitcell_array_inst.get_pin(rbl_wl_name).lc()))
+                           self.bitcell_array_inst.get_pin(rbl_wl_name).lc(),
+                           self.bitcell_array_inst.get_pin(rbl_wl_name).layer))
             
         if port in self.write_ports:
             if port % 2:
                 connection.append((self.prefix + "w_en{}".format(port),
-                                   self.port_data_inst[port].get_pin("w_en").rc()))
+                                   self.port_data_inst[port].get_pin("w_en").rc(),
+                                   self.port_data_inst[port].get_pin("w_en").layer))
             else:
                 connection.append((self.prefix + "w_en{}".format(port),
-                                   self.port_data_inst[port].get_pin("w_en").lc()))
+                                   self.port_data_inst[port].get_pin("w_en").lc(),
+                                  self.port_data_inst[port].get_pin("w_en").layer))
                 
         if port in self.read_ports:
             connection.append((self.prefix + "s_en{}".format(port),
-                               self.port_data_inst[port].get_pin("s_en").lc()))
+                               self.port_data_inst[port].get_pin("s_en").lc(),
+                              self.port_data_inst[port].get_pin("s_en").layer))
 
-        for (control_signal, pin_pos) in connection:
-            control_mid_pos = self.bus_xoffset[port][control_signal]
-            control_pos = vector(self.bus_xoffset[port][control_signal].x, pin_pos.y)
-            self.add_wire(self.m1_stack, [control_mid_pos, control_pos, pin_pos])
-            self.add_via_center(layers=self.m1_stack,
-                                offset=control_pos)
+        for (control_signal, pin_pos, pin_layer) in connection:
+            if port==0:
+                y_offset = self.min_y_offset
+            else:
+                y_offset = self.max_y_offset
+            control_pos = vector(self.bus_xoffset[port][control_signal].x, y_offset)
+            if pin_layer == "m1":
+                self.add_wire(self.m1_stack, [control_pos, pin_pos])
+            elif pin_layer == "m3":
+                self.add_wire(self.m2_stack[::-1], [control_pos, pin_pos])
 
         # clk to wordline_driver
         control_signal = self.prefix + "wl_en{}".format(port)
diff --git a/compiler/modules/bank_select.py b/compiler/modules/bank_select.py
index 3be10d3e..b6246268 100644
--- a/compiler/modules/bank_select.py
+++ b/compiler/modules/bank_select.py
@@ -205,7 +205,7 @@ class bank_select(design.design):
         bank_sel_line_end = vector(xoffset_bank_sel, self.yoffset_maxpoint)
         self.add_path("m2", [bank_sel_line_pos, bank_sel_line_end])
         self.add_via_center(layers=self.m1_stack,
-                            offset=bank_sel_inv_pin.lc())
+                            offset=bank_sel_inv_pin.center())
 
         # Route the pin to the left edge as well
         bank_sel_pin_pos=vector(0, 0)
@@ -242,30 +242,31 @@ class bank_select(design.design):
                 
             # Connect the logic output to inverter input
             out_pin = logic_inst.get_pin("Z")
-            out_pos = out_pin.rc()
+            out_pos = out_pin.center()
             in_pin = inv_inst.get_pin("A")
-            in_pos = in_pin.lc()
+            in_pos = in_pin.center()
             mid1_pos = vector(0.5 * (out_pos.x + in_pos.x), out_pos.y)
             mid2_pos = vector(0.5 * (out_pos.x + in_pos.x), in_pos.y)
             self.add_path("m1", [out_pos, mid1_pos, mid2_pos, in_pos])
             
             # Connect the logic B input to bank_sel / bank_sel_bar
-            logic_pos = logic_inst.get_pin("B").lc() - vector(0.5 * contact.m1_via.height, 0)
+            logic_pin = logic_inst.get_pin("B")
+            logic_pos = logic_pin.center()
             input_pos = vector(xoffset_bank_signal, logic_pos.y)
-            self.add_path("m2", [logic_pos, input_pos])
-            self.add_via_center(layers=self.m1_stack,
-                                offset=logic_pos,
-                                directions=("H", "H"))
+            self.add_path("m3", [logic_pos, input_pos])
+            self.add_via_center(self.m2_stack,
+                                input_pos)
+            self.add_via_stack_center(from_layer=logic_pin.layer,
+                                      to_layer="m3",
+                                      offset=logic_pos)
 
             # Connect the logic A input to the input pin
-            logic_pos = logic_inst.get_pin("A").lc()
+            logic_pin = logic_inst.get_pin("A")
+            logic_pos = logic_pin.center()
             input_pos = vector(0, logic_pos.y)
-            self.add_via_center(layers=self.m1_stack,
-                                offset=logic_pos,
-                                directions=("H", "H"))
-            self.add_via_center(layers=self.m2_stack,
-                                offset=logic_pos,
-                                directions=("H", "H"))
+            self.add_via_stack_center(from_layer=logic_pin.layer,
+                                      to_layer="m3",
+                                      offset=logic_pos)
             self.add_layout_pin_segment_center(text=input_name,
                                                layer="m3",
                                                start=input_pos,
diff --git a/compiler/modules/bitcell_base_array.py b/compiler/modules/bitcell_base_array.py
index 8ee9336b..a8829fc3 100644
--- a/compiler/modules/bitcell_base_array.py
+++ b/compiler/modules/bitcell_base_array.py
@@ -104,9 +104,9 @@ class bitcell_base_array(design.design):
         # For non-square via stacks, vertical/horizontal direction refers to the stack orientation in 2d space
         # Default uses prefered directions for each layer; this cell property is only currently used by s8 tech (03/20)
         try:
-            force_power_pins_vertical = cell_properties.bitcell_force_power_pins_vertical
+            bitcell_power_pin_directions = cell_properties.bitcell_power_pin_directions
         except AttributeError:
-            force_power_pins_vertical = None
+            bitcell_power_pin_directions = None
 
         # Add vdd/gnd via stacks
         for row in range(self.row_size):
@@ -114,7 +114,10 @@ class bitcell_base_array(design.design):
                 inst = self.cell_inst[row,col]
                 for pin_name in ["vdd", "gnd"]:
                     for pin in inst.get_pins(pin_name):
-                        self.add_power_pin(name=pin_name, loc=pin.center(), vertical=force_power_pins_vertical, start_layer=pin.layer)
+                        self.add_power_pin(name=pin_name,
+                                           loc=pin.center(),
+                                           directions=bitcell_power_pin_directions,
+                                           start_layer=pin.layer)
 
     def _adjust_x_offset(self, xoffset, col, col_offset):
         tempx = xoffset
diff --git a/compiler/modules/control_logic.py b/compiler/modules/control_logic.py
index 5dd8182e..078dc3ae 100644
--- a/compiler/modules/control_logic.py
+++ b/compiler/modules/control_logic.py
@@ -384,7 +384,10 @@ class control_logic(design.design):
         height = self.control_logic_center.y - self.m2_pitch
         offset = vector(self.ctrl_dff_array.width, 0)
         
-        self.rail_offsets = self.create_vertical_bus("m2", self.m2_pitch, offset, self.internal_bus_list, height)
+        self.rail_offsets = self.create_vertical_bus("m2",
+                                                     offset,
+                                                     self.internal_bus_list,
+                                                     height)
 
     def create_instances(self):
         """ Create all the instances """
@@ -432,7 +435,7 @@ class control_logic(design.design):
         row += 1
         if (self.port_type == "rw") or (self.port_type == "w"):
             self.place_rbl_delay_row(row)
-            row += 1        
+            row += 1
         if (self.port_type == "rw") or (self.port_type == "r"):
             self.place_sen_row(row)
             row += 1
@@ -522,16 +525,8 @@ class control_logic(design.design):
         self.add_via_center(layers=self.m1_stack,
                             offset=clk_pos)
 
-        # Connect this at the bottom of the buffer
-        out_pos = self.clk_buf_inst.get_pin("Z").center()
-        mid1 = vector(out_pos.x, 2 * self.m2_pitch)
-        mid2 = vector(self.rail_offsets["clk_buf"].x, mid1.y)
-        bus_pos = self.rail_offsets["clk_buf"]
-        self.add_wire(self.m2_stack[::-1], [out_pos, mid1, mid2, bus_pos])
-        # The pin is on M1, so we need another via as well
-        self.add_via_center(layers=self.m1_stack,
-                            offset=self.clk_buf_inst.get_pin("Z").center())
-
+        self.route_output_to_bus_jogged(self.clk_buf_inst,
+                                        "clk_buf")
         self.connect_output(self.clk_buf_inst, "Z", "clk_buf")
 
     def create_gated_clk_bar_row(self):
@@ -541,7 +536,7 @@ class control_logic(design.design):
         
         self.gated_clk_bar_inst = self.add_inst(name="and2_gated_clk_bar",
                                                 mod=self.and2)
-        self.connect_inst(["cs", "clk_bar", "gated_clk_bar", "vdd", "gnd"])
+        self.connect_inst(["clk_bar", "cs", "gated_clk_bar", "vdd", "gnd"])
 
     def place_gated_clk_bar_row(self, row):
         x_offset = self.control_x_offset
@@ -554,31 +549,26 @@ class control_logic(design.design):
     def route_gated_clk_bar(self):
         clkbuf_map = zip(["A"], ["clk_buf"])
         self.connect_vertical_bus(clkbuf_map, self.clk_bar_inst, self.rail_offsets)
-        
+
         out_pos = self.clk_bar_inst.get_pin("Z").center()
-        in_pos = self.gated_clk_bar_inst.get_pin("B").center()
-        mid1 = vector(in_pos.x, out_pos.y)
-        self.add_path("m1", [out_pos, mid1, in_pos])
+        in_pos = self.gated_clk_bar_inst.get_pin("A").center()
+        self.add_zjog("m1", out_pos, in_pos)
 
         # This is the second gate over, so it needs to be on M3
-        clkbuf_map = zip(["A"], ["cs"])
+        clkbuf_map = zip(["B"], ["cs"])
         self.connect_vertical_bus(clkbuf_map,
                                   self.gated_clk_bar_inst,
                                   self.rail_offsets,
                                   self.m2_stack[::-1])
         # The pin is on M1, so we need another via as well
-        self.add_via_center(layers=self.m1_stack,
-                            offset=self.gated_clk_bar_inst.get_pin("A").center())
+        b_pin = self.gated_clk_bar_inst.get_pin("B")
+        self.add_via_stack_center(from_layer=b_pin.layer,
+                                  to_layer="m3",
+                                  offset=b_pin.center())
 
         # This is the second gate over, so it needs to be on M3
-        clkbuf_map = zip(["Z"], ["gated_clk_bar"])
-        self.connect_vertical_bus(clkbuf_map,
-                                  self.gated_clk_bar_inst,
-                                  self.rail_offsets,
-                                  self.m2_stack[::-1])
-        # The pin is on M1, so we need another via as well
-        self.add_via_center(layers=self.m1_stack,
-                            offset=self.gated_clk_bar_inst.get_pin("Z").center())
+        self.route_output_to_bus_jogged(self.gated_clk_bar_inst,
+                                        "gated_clk_bar")
 
     def create_gated_clk_buf_row(self):
         self.gated_clk_buf_inst = self.add_inst(name="and2_gated_clk_buf",
@@ -594,7 +584,9 @@ class control_logic(design.design):
         
     def route_gated_clk_buf(self):
         clkbuf_map = zip(["A", "B"], ["clk_buf", "cs"])
-        self.connect_vertical_bus(clkbuf_map, self.gated_clk_buf_inst, self.rail_offsets)
+        self.connect_vertical_bus(clkbuf_map,
+                                  self.gated_clk_buf_inst,
+                                  self.rail_offsets)
 
         clkbuf_map = zip(["Z"], ["gated_clk_buf"])
         self.connect_vertical_bus(clkbuf_map,
@@ -602,8 +594,10 @@ class control_logic(design.design):
                                   self.rail_offsets,
                                   self.m2_stack[::-1])
         # The pin is on M1, so we need another via as well
-        self.add_via_center(layers=self.m1_stack,
-                            offset=self.gated_clk_buf_inst.get_pin("Z").center())
+        z_pin = self.gated_clk_buf_inst.get_pin("Z")
+        self.add_via_stack_center(from_layer=z_pin.layer,
+                                  to_layer="m2",
+                                  offset=z_pin.center())
         
     def create_wlen_row(self):
         # input pre_p_en, output: wl_en
@@ -647,11 +641,16 @@ class control_logic(design.design):
         in_map = zip(["A", "B"], ["gated_clk_buf", "rbl_bl_delay"])
         self.connect_vertical_bus(in_map, self.p_en_bar_nand_inst, self.rail_offsets)
 
-        out_pos = self.p_en_bar_nand_inst.get_pin("Z").rc()
-        in_pos = self.p_en_bar_driver_inst.get_pin("A").lc()
-        mid1 = vector(out_pos.x, in_pos.y)
-        self.add_wire(self.m1_stack, [out_pos, mid1, in_pos])
-        
+        out_pin = self.p_en_bar_nand_inst.get_pin("Z")
+        out_pos = out_pin.center()
+        in_pin = self.p_en_bar_driver_inst.get_pin("A")
+        in_pos = in_pin.center()
+        mid1 = vector(in_pos.x, out_pos.y)
+        self.add_path(out_pin.layer, [out_pos, mid1, in_pos])
+        self.add_via_stack_center(from_layer=out_pin.layer,
+                                  to_layer=in_pin.layer,
+                                  offset=in_pin.center())
+
         self.connect_output(self.p_en_bar_driver_inst, "Z", "p_en_bar")
         
     def create_sen_row(self):
@@ -704,11 +703,7 @@ class control_logic(design.design):
         # Connect from delay line
         # Connect to rail
 
-        rbl_map = zip(["Z"], ["rbl_bl_delay_bar"])
-        self.connect_vertical_bus(rbl_map, self.rbl_bl_delay_inv_inst, self.rail_offsets, ("m3", "via2", "m2"))
-        # The pin is on M1, so we need another via as well
-        self.add_via_center(layers=self.m1_stack,
-                            offset=self.rbl_bl_delay_inv_inst.get_pin("Z").center())
+        self.route_output_to_bus_jogged(self.rbl_bl_delay_inv_inst, "rbl_bl_delay_bar")
         
         rbl_map = zip(["A"], ["rbl_bl_delay"])
         self.connect_vertical_bus(rbl_map, self.rbl_bl_delay_inv_inst, self.rail_offsets)
@@ -766,11 +761,12 @@ class control_logic(design.design):
             dff_out_map = zip(["dout_bar_0", "dout_0"], ["cs", "cs_bar"])
         else:
             dff_out_map = zip(["dout_bar_0"], ["cs"])
-        self.connect_vertical_bus(dff_out_map, self.ctrl_dff_inst, self.rail_offsets, ("m3", "via2", "m2"))
+        self.connect_vertical_bus(dff_out_map, self.ctrl_dff_inst, self.rail_offsets, self.m2_stack[::-1])
         
         # Connect the clock rail to the other clock rail
+        # by routing in the supply rail track to avoid channel conflicts
         in_pos = self.ctrl_dff_inst.get_pin("clk").uc()
-        mid_pos = in_pos + vector(0, 2 * self.m2_pitch)
+        mid_pos = in_pos + vector(0, self.and2.height)
         rail_pos = vector(self.rail_offsets["clk_buf"].x, mid_pos.y)
         self.add_wire(self.m1_stack, [in_pos, mid_pos, rail_pos])
         self.add_via_center(layers=self.m1_stack,
@@ -823,10 +819,10 @@ class control_logic(design.design):
                     self.add_path("m1", [row_loc, pin_loc])
             
         self.copy_layout_pin(self.delay_inst, "gnd")
-        self.copy_layout_pin(self.delay_inst, "vdd")        
+        self.copy_layout_pin(self.delay_inst, "vdd")
 
         self.copy_layout_pin(self.ctrl_dff_inst, "gnd")
-        self.copy_layout_pin(self.ctrl_dff_inst, "vdd")        
+        self.copy_layout_pin(self.ctrl_dff_inst, "vdd")
         
     def add_lvs_correspondence_points(self):
         """ This adds some points for easier debugging if LVS goes wrong.
@@ -1000,3 +996,15 @@ class control_logic(design.design):
         offset = vector(x_offset, y_offset)
         inst.place(offset, mirror)
         return x_offset + inst.width
+
+    def route_output_to_bus_jogged(self, inst, name):
+        # Connect this at the bottom of the buffer
+        out_pos = inst.get_pin("Z").center()
+        mid1 = vector(out_pos.x, out_pos.y - 0.25 * inst.mod.height)
+        mid2 = vector(self.rail_offsets[name].x, mid1.y)
+        bus_pos = self.rail_offsets[name]
+        self.add_wire(self.m2_stack[::-1], [out_pos, mid1, mid2, bus_pos])
+        # The pin is on M1, so we need another via as well
+        self.add_via_center(layers=self.m1_stack,
+                            offset=out_pos)
+    
diff --git a/compiler/modules/delay_chain.py b/compiler/modules/delay_chain.py
index b34974e5..ad9bc1cc 100644
--- a/compiler/modules/delay_chain.py
+++ b/compiler/modules/delay_chain.py
@@ -62,7 +62,7 @@ class delay_chain(design.design):
         self.add_pin("gnd", "GROUND")
 
     def add_modules(self):
-        self.inv = factory.create(module_type="pinv", route_output=False)
+        self.inv = factory.create(module_type="pinv")
         self.add_mod(self.inv)
 
     def create_inverters(self):
diff --git a/compiler/modules/dff_buf.py b/compiler/modules/dff_buf.py
index 500767d8..0366b10b 100644
--- a/compiler/modules/dff_buf.py
+++ b/compiler/modules/dff_buf.py
@@ -180,7 +180,7 @@ class dff_buf(design.design):
                             height=din_pin.height())
 
         dout_pin = self.inv2_inst.get_pin("Z")
-        mid_pos = dout_pin.center() + vector(self.m1_pitch, 0)
+        mid_pos = dout_pin.center() + vector(self.m1_nonpref_pitch, 0)
         q_pos = mid_pos - vector(0, self.m2_pitch)
         self.add_layout_pin_rect_center(text="Q",
                                         layer="m2",
diff --git a/compiler/modules/hierarchical_decoder.py b/compiler/modules/hierarchical_decoder.py
index 69822149..c1593932 100644
--- a/compiler/modules/hierarchical_decoder.py
+++ b/compiler/modules/hierarchical_decoder.py
@@ -12,7 +12,7 @@ from sram_factory import factory
 from vector import vector
 from globals import OPTS
 from errors import drc_error
-from tech import cell_properties
+from tech import cell_properties, layer
 
 
 class hierarchical_decoder(design.design):
@@ -89,6 +89,7 @@ class hierarchical_decoder(design.design):
         self.place_pre_decoder()
         self.place_row_decoder()
         self.route_inputs()
+        self.route_outputs()
         self.route_decoder_bus()
         self.route_vdd_gnd()
         self.offset_all_coordinates()
@@ -103,7 +104,7 @@ class hierarchical_decoder(design.design):
                                    height=self.cell_height)
         self.add_mod(self.and2)
         self.and3 = factory.create(module_type="pand3",
-                                    height=self.cell_height)
+                                   height=self.cell_height)
         self.add_mod(self.and3)
         
         self.add_decoders()
@@ -186,19 +187,36 @@ class hierarchical_decoder(design.design):
         self.num_rows = math.ceil(self.num_outputs / self.cell_multiple)
         # We will place this many final decoders per row
         self.decoders_per_row = math.ceil(self.num_outputs / self.num_rows)
-        
+        # We will need to use M2 and M3 in the vertical bus if we have multiple decoders per row
+        if self.decoders_per_row == 1:
+            self.decoder_bus_pitch = self.m2_pitch
+        elif self.decoders_per_row == 2:
+            self.decoder_bus_pitch = self.m3_pitch
+        else:
+            debug.error("Insufficient layers for multi-bit height decoder.", -1)
+
         # Calculates height and width of row-decoder
         if (self.num_inputs == 4 or self.num_inputs == 5):
             nand_width = self.and2.width
+            nand_inputs = 2
         else:
             nand_width = self.and3.width
-        self.internal_routing_width = self.m2_pitch * (self.total_number_of_predecoder_outputs + 1)
+            nand_inputs = 3
+        self.internal_routing_width = self.decoder_bus_pitch * (self.total_number_of_predecoder_outputs + 1)
         self.row_decoder_height = self.inv.height * self.num_rows
+        
+        decoder_input_wire_height = self.decoders_per_row * nand_inputs * self.m2_pitch
+        # print(self.decoders_per_row, nand_inputs)
+        # print(decoder_input_wire_height, self.cell_height)
+        if decoder_input_wire_height > self.cell_height:
+            debug.warning("Cannot fit multi-bit decoder routes per row.")
+        # debug.check(decoder_input_wire_height < self.cell_height, "Cannot fit multi-bit decoder routes per row.")
 
         self.input_routing_width = (self.num_inputs + 1) * self.m2_pitch
+
         # Calculates height and width of hierarchical decoder
         # Add extra pitch for good measure
-        self.height = max(self.predecoder_height, self.row_decoder_height) + self.m3_pitch
+        self.height = max(self.predecoder_height, self.row_decoder_height) + self.m2_pitch
         self.width = self.input_routing_width + self.predecoder_width \
                      + self.internal_routing_width \
                      + self.decoders_per_row * nand_width + self.inv.width
@@ -218,7 +236,6 @@ class hierarchical_decoder(design.design):
 
         input_bus_names = ["addr_{0}".format(i) for i in range(self.num_inputs)]
         self.input_bus = self.create_vertical_pin_bus(layer="m2",
-                                                      pitch=self.m2_pitch,
                                                       offset=input_offset,
                                                       names=input_bus_names,
                                                       length=input_height)
@@ -238,7 +255,7 @@ class hierarchical_decoder(design.design):
 
                 # To prevent conflicts, we will offset each input connect so
                 # that it aligns with the vdd/gnd rails
-                decoder_offset = decoder_pin.bc() + vector(0, (i + 1) * self.inv.height)
+                decoder_offset = decoder_pin.bc() + vector(0, (i + 1) * (self.inv.height + self.m1_pitch))
                 input_offset = input_pos.scale(1, 0) + decoder_offset.scale(0, 1)
                 
                 self.route_input_bus(decoder_offset, input_offset)
@@ -254,7 +271,7 @@ class hierarchical_decoder(design.design):
 
                 # To prevent conflicts, we will offset each input connect so
                 # that it aligns with the vdd/gnd rails
-                decoder_offset = decoder_pin.bc() + vector(0, (i + 1) * self.inv.height)
+                decoder_offset = decoder_pin.bc() + vector(0, (i + 1) * (self.inv.height + self.m1_pitch))
                 input_offset = input_pos.scale(1, 0) + decoder_offset.scale(0, 1)
                 
                 self.route_input_bus(decoder_offset, input_offset)
@@ -265,11 +282,13 @@ class hierarchical_decoder(design.design):
         vertical M2 coordinate to the predecode inputs
         """
         
-        self.add_via_center(layers=self.m2_stack,
-                            offset=input_offset)
-        self.add_via_center(layers=self.m2_stack,
-                            offset=output_offset)
-        self.add_path(("m3"), [input_offset, output_offset])
+        self.add_via_stack_center(from_layer="m2",
+                                  to_layer="m3",
+                                  offset=input_offset)
+        self.add_via_stack_center(from_layer="m2",
+                                  to_layer="m3",
+                                  offset=output_offset)
+        self.add_path("m3", [input_offset, output_offset])
     
     def add_pins(self):
         """ Add the module pins """
@@ -411,7 +430,7 @@ class hierarchical_decoder(design.design):
         """
         if (self.num_inputs >= 4):
             self.place_decoder_and_array()
-            self.route_decoder()
+
 
     def place_decoder_and_array(self):
         """
@@ -447,16 +466,28 @@ class hierarchical_decoder(design.design):
             self.and_inst[inst_index].place(offset=vector(x_off, y_off),
                                             mirror=mirror)
 
-    def route_decoder(self):
+    def route_outputs(self):
         """ Add the pins. """
 
-        for output in range(self.num_outputs):
-            z_pin = self.and_inst[output].get_pin("Z")
-            self.add_layout_pin(text="decode_{0}".format(output),
-                                layer="m1",
-                                offset=z_pin.ll(),
-                                width=z_pin.width(),
-                                height=z_pin.height())
+        max_xoffset = max(x.rx() for x in self.and_inst)
+        
+        for output_index in range(self.num_outputs):
+            row_remainder = (output_index % self.decoders_per_row)
+
+            and_inst = self.and_inst[output_index]
+            z_pin = and_inst.get_pin("Z")
+            if row_remainder == 0 and self.decoders_per_row > 1:
+                layer = "m3"
+                self.add_via_stack_center(from_layer=z_pin.layer,
+                                          to_layer="m3",
+                                          offset=z_pin.center())
+            else:
+                layer = z_pin.layer
+
+            self.add_layout_pin_segment_center(text="decode_{0}".format(output_index),
+                                               layer=layer,
+                                               start=z_pin.center(),
+                                               end=vector(max_xoffset, z_pin.cy()))
         
     def route_decoder_bus(self):
         """
@@ -468,7 +499,7 @@ class hierarchical_decoder(design.design):
             # This leaves an offset for the predecoder output jogs
             input_bus_names = ["predecode_{0}".format(i) for i in range(self.total_number_of_predecoder_outputs)]
             self.predecode_bus = self.create_vertical_pin_bus(layer="m2",
-                                                              pitch=self.m2_pitch,
+                                                              pitch=self.decoder_bus_pitch,
                                                               offset=vector(0, 0),
                                                               names=input_bus_names,
                                                               length=self.height)
@@ -512,22 +543,27 @@ class hierarchical_decoder(design.design):
         """
         output_index = 0
 
+        if "li" in layer:
+            self.decoder_layers = [self.m1_stack, self.m2_stack[::-1]]
+        else:
+            self.decoder_layers = [self.m2_stack[::-1]]
+        debug.check(self.decoders_per_row <= len(self.decoder_layers), "Must have more layers for multi-height decoder.")
+            
         if (self.num_inputs == 4 or self.num_inputs == 5):
             for index_B in self.predec_groups[1]:
                 for index_A in self.predec_groups[0]:
                     # FIXME: convert to connect_bus?
                     if (output_index < self.num_outputs):
-                        row_index = math.floor(output_index / self.decoders_per_row)
-                        row_remainder = (output_index % self.decoders_per_row)
-                        row_offset = row_index * self.and_inst[0].height + (2 * row_remainder + 1) * self.m3_pitch
                         predecode_name = "predecode_{}".format(index_A)
                         self.route_predecode_bus_outputs(predecode_name,
                                                          self.and_inst[output_index].get_pin("A"),
-                                                         row_offset)
+                                                         output_index,
+                                                         0)
                         predecode_name = "predecode_{}".format(index_B)
                         self.route_predecode_bus_outputs(predecode_name,
                                                          self.and_inst[output_index].get_pin("B"),
-                                                         row_offset + self.m3_pitch)
+                                                         output_index,
+                                                         1)
                     output_index = output_index + 1
 
         elif (self.num_inputs > 5):
@@ -536,21 +572,21 @@ class hierarchical_decoder(design.design):
                     for index_A in self.predec_groups[0]:
                         # FIXME: convert to connect_bus?
                         if (output_index < self.num_outputs):
-                            row_index = math.floor(output_index / self.decoders_per_row)
-                            row_remainder = (output_index % self.decoders_per_row)
-                            row_offset = row_index * self.and_inst[0].height + (3 * row_remainder + 1) * self.m3_pitch
                             predecode_name = "predecode_{}".format(index_A)
                             self.route_predecode_bus_outputs(predecode_name,
                                                              self.and_inst[output_index].get_pin("A"),
-                                                             row_offset)
+                                                             output_index,
+                                                             0)
                             predecode_name = "predecode_{}".format(index_B)
                             self.route_predecode_bus_outputs(predecode_name,
                                                              self.and_inst[output_index].get_pin("B"),
-                                                             row_offset + self.m3_pitch)
+                                                             output_index,
+                                                             1)
                             predecode_name = "predecode_{}".format(index_C)
                             self.route_predecode_bus_outputs(predecode_name,
                                                              self.and_inst[output_index].get_pin("C"),
-                                                             row_offset + 2 * self.m3_pitch)
+                                                             output_index,
+                                                             2)
                         output_index = output_index + 1
 
     def route_vdd_gnd(self):
@@ -570,39 +606,56 @@ class hierarchical_decoder(design.design):
                 # The nand and inv are the same height rows...
                 supply_pin = self.and_inst[num].get_pin(pin_name)
                 pin_pos = vector(xoffset, supply_pin.cy())
-                self.add_path("m1",
+                self.add_path(supply_pin.layer,
                               [supply_pin.lc(), vector(xoffset, supply_pin.cy())])
                 self.add_power_pin(name=pin_name,
-                                   loc=pin_pos)
+                                   loc=pin_pos,
+                                   start_layer=supply_pin.layer)
                 
         # Copy the pins from the predecoders
         for pre in self.pre2x4_inst + self.pre3x8_inst:
             self.copy_layout_pin(pre, "vdd")
             self.copy_layout_pin(pre, "gnd")
         
-    def route_predecode_bus_outputs(self, rail_name, pin, y_offset):
+    def route_predecode_bus_outputs(self, rail_name, pin, output_index, pin_index):
         """
         Connect the routing rail to the given metal1 pin
         using a routing track at the given y_offset
-        
         """
-        pin_pos = pin.center()
-        # If we have a single decoder per row, we can route on M1
-        if self.decoders_per_row == 1:
-            rail_pos = vector(self.predecode_bus[rail_name].x, pin_pos.y)
-            self.add_path("m1", [rail_pos, pin_pos])
-            self.add_via_center(layers=self.m1_stack,
-                                offset=rail_pos)
-        # If not, we must route over the decoder cells on M3
-        else:
-            rail_pos = vector(self.predecode_bus[rail_name].x, y_offset)
-            mid_pos = vector(pin_pos.x, rail_pos.y)
-            self.add_wire(self.m2_stack[::-1], [rail_pos, mid_pos, pin_pos])
-            self.add_via_center(layers=self.m2_stack,
-                                offset=rail_pos)
-            self.add_via_center(layers=self.m1_stack,
-                                offset=pin_pos)
 
+        row_index = math.floor(output_index / self.decoders_per_row)
+        row_remainder = (output_index % self.decoders_per_row)
+        row_offset = row_index * self.and_inst[0].height
+        
+        pin_pos = pin.center()
+
+        # y_offset is the same for both the M2 and M4 routes so that the rail
+        # contacts align and don't cause problems
+        if pin_index == 0:
+            # Bottom pitch
+            y_offset = row_offset
+        elif pin_index == 1:
+            # One pitch from top
+            y_offset = row_offset + self.and_inst[0].height - self.m3_pitch
+        elif pin_index == 2:
+            # One pitch from bottom
+            y_offset = row_offset + self.m3_pitch
+        else:
+            debug.error("Invalid decoder pitch.")
+            
+        rail_pos = vector(self.predecode_bus[rail_name].x, y_offset)
+        mid_pos = vector(pin_pos.x, rail_pos.y)
+        self.add_wire(self.decoder_layers[row_remainder], [rail_pos, mid_pos, pin_pos])
+        
+        self.add_via_stack_center(from_layer="m2",
+                                  to_layer=self.decoder_layers[row_remainder][0],
+                                  offset=rail_pos)
+        
+        self.add_via_stack_center(from_layer=pin.layer,
+                                  to_layer=self.decoder_layers[row_remainder][2],
+                                  offset=pin_pos,
+                                  directions=("H", "H"))
+        
     def route_predecode_bus_inputs(self, rail_name, pin, x_offset):
         """
         Connect the routing rail to the given metal1 pin using a jog
@@ -610,13 +663,21 @@ class hierarchical_decoder(design.design):
         """
         # This routes the pin up to the rail, basically, to avoid conflicts.
         # It would be fixed with a channel router.
+        # pin_pos = pin.center()
+        # mid_point1 = vector(x_offset, pin_pos.y)
+        # mid_point2 = vector(x_offset, pin_pos.y + self.inv.height / 2)
+        # rail_pos = vector(self.predecode_bus[rail_name].x, mid_point2.y)
+        # self.add_path("m1", [pin_pos, mid_point1, mid_point2, rail_pos])
+
         pin_pos = pin.center()
-        mid_point1 = vector(x_offset, pin_pos.y)
-        mid_point2 = vector(x_offset, pin_pos.y + self.inv.height / 2)
-        rail_pos = vector(self.predecode_bus[rail_name].x, mid_point2.y)
-        self.add_wire(self.m1_stack, [pin_pos, mid_point1, mid_point2, rail_pos])
-        self.add_via_center(layers=self.m1_stack,
-                            offset=rail_pos)
+        rail_pos = vector(self.predecode_bus[rail_name].x, pin_pos.y)
+        self.add_path("m1", [pin_pos, rail_pos])
+        self.add_via_stack_center(from_layer=pin.layer,
+                                  to_layer="m1",
+                                  offset=pin_pos)
+        self.add_via_stack_center(from_layer="m1",
+                                  to_layer="m2",
+                                  offset=rail_pos)
 
     def input_load(self):
         if self.determine_predecodes(self.num_inputs)[1]==0:
diff --git a/compiler/modules/hierarchical_predecode.py b/compiler/modules/hierarchical_predecode.py
index ccee34a9..9a658546 100644
--- a/compiler/modules/hierarchical_predecode.py
+++ b/compiler/modules/hierarchical_predecode.py
@@ -11,6 +11,7 @@ import math
 import contact
 from vector import vector
 from sram_factory import factory
+from tech import cell_properties
 
 
 class hierarchical_predecode(design.design):
@@ -19,7 +20,17 @@ class hierarchical_predecode(design.design):
     """
     def __init__(self, name, input_number, height=None):
         self.number_of_inputs = input_number
-        self.cell_height = height
+
+        if not height:
+            b = factory.create(module_type="bitcell")
+            try:
+                self.cell_multiple = cell_properties.bitcell.decoder_bitcell_multiple
+            except AttributeError:
+                self.cell_multiple = 1
+            self.cell_height = self.cell_multiple * b.height
+        else:
+            self.cell_height = height
+        
         self.number_of_outputs = int(math.pow(2, self.number_of_inputs))
         design.design.__init__(self, name)
     
@@ -57,10 +68,10 @@ class hierarchical_predecode(design.design):
         self.height = self.number_of_outputs * self.and_mod.height
 
         # x offset for input inverters
-        self.x_off_inv_1 = self.number_of_inputs*self.m2_pitch 
+        self.x_off_inv_1 = self.number_of_inputs * self.m2_pitch + self.m2_space
 
         # x offset to AND decoder includes the left rails, mid rails and inverters, plus two extra m2 pitches
-        self.x_off_and = self.x_off_inv_1 + self.inv.width + (2*self.number_of_inputs + 2) * self.m2_pitch
+        self.x_off_and = self.x_off_inv_1 + self.inv.width + (2 * self.number_of_inputs + 2) * self.m2_pitch
 
         # x offset to output inverters
         self.width = self.x_off_and + self.and_mod.width
@@ -68,9 +79,8 @@ class hierarchical_predecode(design.design):
     def route_rails(self):
         """ Create all of the rails for the inputs and vdd/gnd/inputs_bar/inputs """
         input_names = ["in_{}".format(x) for x in range(self.number_of_inputs)]
-        offset = vector(0.5 * self.m2_width, self.m1_pitch)
+        offset = vector(0.5 * self.m2_width, self.m3_pitch)
         self.input_rails = self.create_vertical_pin_bus(layer="m2",
-                                                        pitch=self.m2_pitch,
                                                         offset=offset,
                                                         names=input_names,
                                                         length=self.height - 2 * self.m1_pitch)
@@ -78,9 +88,8 @@ class hierarchical_predecode(design.design):
         invert_names = ["Abar_{}".format(x) for x in range(self.number_of_inputs)]
         non_invert_names = ["A_{}".format(x) for x in range(self.number_of_inputs)]
         decode_names = invert_names + non_invert_names
-        offset = vector(self.x_off_inv_1 + self.inv.width + 2 * self.m2_pitch, self.m1_pitch)
+        offset = vector(self.x_off_inv_1 + self.inv.width + 2 * self.m2_pitch, self.m3_pitch)
         self.decode_rails = self.create_vertical_bus(layer="m2",
-                                                     pitch=self.m2_pitch,
                                                      offset=offset,
                                                      names=decode_names,
                                                      length=self.height - 2 * self.m1_pitch)
@@ -146,16 +155,18 @@ class hierarchical_predecode(design.design):
             # route one signal next to each vdd/gnd rail since this is
             # typically where the p/n devices are and there are no
             # pins in the and gates.
-            y_offset = (num + self.number_of_inputs) * self.inv.height + contact.m1_via.width + self.m1_space
+            y_offset = (num + self.number_of_inputs) * self.inv.height + contact.m2_via.width + self.m2_space
             in_pin = "in_{}".format(num)
             a_pin = "A_{}".format(num)
             in_pos = vector(self.input_rails[in_pin].x, y_offset)
             a_pos = vector(self.decode_rails[a_pin].x, y_offset)
             self.add_path("m1", [in_pos, a_pos])
-            self.add_via_center(layers=self.m1_stack,
-                                offset=[self.input_rails[in_pin].x, y_offset])
-            self.add_via_center(layers=self.m1_stack,
-                                offset=[self.decode_rails[a_pin].x, y_offset])
+            self.add_via_stack_center(from_layer="m1",
+                                      to_layer="m2",
+                                      offset=[self.input_rails[in_pin].x, y_offset])
+            self.add_via_stack_center(from_layer="m1",
+                                      to_layer="m2",
+                                      offset=[self.decode_rails[a_pin].x, y_offset])
 
     def route_output_and(self):
         """
@@ -165,37 +176,43 @@ class hierarchical_predecode(design.design):
 
             z_pin = self.and_inst[num].get_pin("Z")
             self.add_layout_pin(text="out_{}".format(num),
-                                layer="m1",
+                                layer=z_pin.layer,
                                 offset=z_pin.ll(),
                                 height=z_pin.height(),
                                 width=z_pin.width())
     
     def route_input_inverters(self):
         """
-        Route all conections of the inputs inverters [Inputs, outputs, vdd, gnd] 
+        Route all conections of the inputs inverters [Inputs, outputs, vdd, gnd]
         """
         for inv_num in range(self.number_of_inputs):
             out_pin = "Abar_{}".format(inv_num)
             in_pin = "in_{}".format(inv_num)
             
-            #add output so that it is just below the vdd or gnd rail
+            # add output so that it is just below the vdd or gnd rail
             # since this is where the p/n devices are and there are no
             # pins in the and gates.
-            y_offset = (inv_num+1) * self.inv.height - 3*self.m1_space
-            inv_out_pos = self.in_inst[inv_num].get_pin("Z").rc()
-            right_pos = inv_out_pos + vector(self.inv.width - self.inv.get_pin("Z").lx(),0)
-            rail_pos = vector(self.decode_rails[out_pin].x,y_offset)
-            self.add_path("m1", [inv_out_pos, right_pos, vector(right_pos.x, y_offset), rail_pos])
-            self.add_via_center(layers = self.m1_stack,
-                                offset=rail_pos)
-
+            y_offset = (inv_num + 1) * self.inv.height - 3 * self.m1_space
+            inv_out_pin = self.in_inst[inv_num].get_pin("Z")
+            inv_out_pos = inv_out_pin.rc()
+            right_pos = inv_out_pos + vector(self.inv.width - self.inv.get_pin("Z").lx(), 0)
+            rail_pos = vector(self.decode_rails[out_pin].x, y_offset)
+            self.add_path(inv_out_pin.layer, [inv_out_pos, right_pos, vector(right_pos.x, y_offset), rail_pos])
+            self.add_via_stack_center(from_layer=inv_out_pin.layer,
+                                      to_layer="m2",
+                                      offset=rail_pos)
             
-            #route input
-            inv_in_pos = self.in_inst[inv_num].get_pin("A").lc()
-            in_pos = vector(self.input_rails[in_pin].x,inv_in_pos.y)
+            # route input
+            pin = self.in_inst[inv_num].get_pin("A")
+            inv_in_pos = pin.lc()
+            in_pos = vector(self.input_rails[in_pin].x, inv_in_pos.y)
             self.add_path("m1", [in_pos, inv_in_pos])
-            self.add_via_center(layers=self.m1_stack,
-                                offset=in_pos)
+            self.add_via_stack_center(from_layer=pin.layer,
+                                      to_layer="m1",
+                                      offset=inv_in_pos)
+            self.add_via_stack_center(from_layer="m1",
+                                      to_layer="m2",
+                                      offset=in_pos)
             
     def route_and_to_rails(self):
         # This 2D array defines the connection mapping
@@ -205,17 +222,28 @@ class hierarchical_predecode(design.design):
             index_lst= and_input_line_combination[k]
 
             if self.number_of_inputs == 2:
-                gate_lst = ["A","B"]
+                gate_lst = ["A", "B"]
             else:
-                gate_lst = ["A","B","C"]
+                gate_lst = ["A", "B", "C"]
 
             # this will connect pins A,B or A,B,C
-            for rail_pin,gate_pin in zip(index_lst,gate_lst):
-                pin_pos = self.and_inst[k].get_pin(gate_pin).lc()
+            for rail_pin, gate_pin in zip(index_lst, gate_lst):
+                pin = self.and_inst[k].get_pin(gate_pin)
+                pin_pos = pin.center()
                 rail_pos = vector(self.decode_rails[rail_pin].x, pin_pos.y)
                 self.add_path("m1", [rail_pos, pin_pos])
-                self.add_via_center(layers=self.m1_stack,
-                                    offset=rail_pos)
+                self.add_via_stack_center(from_layer="m1",
+                                          to_layer="m2",
+                                          offset=rail_pos)
+                if gate_pin == "A":
+                    direction = None
+                else:
+                    direction = ("H", "H")
+                    
+                self.add_via_stack_center(from_layer=pin.layer,
+                                          to_layer="m1",
+                                          offset=pin_pos,
+                                          directions=direction)
 
     def route_vdd_gnd(self):
         """ Add a pin for each row of vdd/gnd which are must-connects next level up. """
@@ -230,14 +258,16 @@ class hierarchical_predecode(design.design):
             for n in ["vdd", "gnd"]:
                 and_pin = self.and_inst[num].get_pin(n)
                 supply_offset = and_pin.ll().scale(0, 1)
-                self.add_rect(layer="m1",
+                self.add_rect(layer=and_pin.layer,
                               offset=supply_offset,
                               width=self.and_inst[num].rx())
 
                 # Add pins in two locations
                 for xoffset in [in_xoffset]:
                     pin_pos = vector(xoffset, and_pin.cy())
-                    self.add_power_pin(n, pin_pos)
+                    self.add_power_pin(name=n,
+                                       loc=pin_pos,
+                                       start_layer=and_pin.layer)
             
 
 
diff --git a/compiler/modules/port_address.py b/compiler/modules/port_address.py
index bc204370..f9befc3f 100644
--- a/compiler/modules/port_address.py
+++ b/compiler/modules/port_address.py
@@ -3,8 +3,6 @@
 # Copyright (c) 2016-2019 Regents of the University of California
 # All rights reserved.
 #
-import sys
-from tech import drc, parameter
 from math import log, ceil
 import debug
 import design
@@ -13,6 +11,7 @@ from vector import vector
 
 from globals import OPTS
 
+
 class port_address(design.design):
     """
     Create the address port (row decoder and wordline driver)..
@@ -25,17 +24,16 @@ class port_address(design.design):
         self.addr_size = ceil(log(self.num_rows, 2))
         
         if name == "":
-            name = "port_address_{0}_{1}".format(cols,rows)
+            name = "port_address_{0}_{1}".format(cols, rows)
         design.design.__init__(self, name)
-        debug.info(2, "create data port of cols {0} rows {1}".format(cols,rows))
+        debug.info(2, "create data port of cols {0} rows {1}".format(cols, rows))
 
         self.create_netlist()
         if not OPTS.netlist_only:
-            debug.check(len(self.all_ports)<=2,"Bank layout cannot handle more than two ports.")
+            debug.check(len(self.all_ports) <= 2, "Bank layout cannot handle more than two ports.")
             self.create_layout()
             self.add_boundary()
 
-
     def create_netlist(self):
         self.add_pins()
         self.add_modules()
@@ -51,16 +49,15 @@ class port_address(design.design):
         """ Adding pins for port address module"""
 
         for bit in range(self.addr_size):
-            self.add_pin("addr_{0}".format(bit),"INPUT")
+            self.add_pin("addr_{0}".format(bit), "INPUT")
         
         self.add_pin("wl_en", "INPUT")
 
         for bit in range(self.num_rows):
-            self.add_pin("wl_{0}".format(bit),"OUTPUT")
+            self.add_pin("wl_{0}".format(bit), "OUTPUT")
         
-        self.add_pin("vdd","POWER")
-        self.add_pin("gnd","GROUND")
-
+        self.add_pin("vdd", "POWER")
+        self.add_pin("gnd", "GROUND")
         
     def route_layout(self):
         """ Create routing amoung the modules """
@@ -71,8 +68,8 @@ class port_address(design.design):
     def route_supplies(self):
         """ Propagate all vdd/gnd pins up to this level for all modules """
         for inst in self.insts:
-            self.copy_power_pins(inst,"vdd")
-            self.copy_power_pins(inst,"gnd")
+            self.copy_power_pins(inst, "vdd")
+            self.copy_power_pins(inst, "gnd")
 
     def route_pins(self):
         for row in range(self.addr_size):
@@ -119,12 +116,10 @@ class port_address(design.design):
         temp.extend(["vdd", "gnd"])
         self.connect_inst(temp)
 
-
-            
     def create_wordline_driver(self):
         """ Create the Wordline Driver """
 
-        self.wordline_driver_inst = self.add_inst(name="wordline_driver", 
+        self.wordline_driver_inst = self.add_inst(name="wordline_driver",
                                                   mod=self.wordline_driver)
 
         temp = []
@@ -137,19 +132,13 @@ class port_address(design.design):
         temp.append("gnd")
         self.connect_inst(temp)
 
-            
-
     def place_instances(self):
         """
         Compute the offsets and place the instances.
         """
 
-        # A space for wells or jogging m2
-        self.m2_gap = max(2*drc("pwell_to_nwell") + drc("nwell_enclose_active"),
-                          3*self.m2_pitch)
-        
-        row_decoder_offset = vector(0,0)
-        wordline_driver_offset = vector(self.row_decoder.width + self.m2_gap,0)
+        row_decoder_offset = vector(0, 0)
+        wordline_driver_offset = vector(self.row_decoder.width, 0)
         
         self.wordline_driver_inst.place(wordline_driver_offset)
         self.row_decoder_inst.place(row_decoder_offset)
diff --git a/compiler/modules/port_data.py b/compiler/modules/port_data.py
index 1d434929..9ff99bdd 100644
--- a/compiler/modules/port_data.py
+++ b/compiler/modules/port_data.py
@@ -789,7 +789,7 @@ class port_data(design.design):
 
         # Channel route each mux separately since we don't minimize the number
         # of tracks in teh channel router yet. If we did, we could route all the bits at once!
-        offset = bot_inst_group.inst.ul() + vector(0, self.m1_pitch)
+        offset = bot_inst_group.inst.ul() + vector(0, self.m1_nonpref_pitch)
         for bit in range(num_bits):
             bottom_names = self._get_bitline_pins(bot_inst_group, bit)
             top_names = self._get_bitline_pins(top_inst_group, bit)
diff --git a/compiler/modules/replica_bitcell_array.py b/compiler/modules/replica_bitcell_array.py
index e405401a..349655c4 100644
--- a/compiler/modules/replica_bitcell_array.py
+++ b/compiler/modules/replica_bitcell_array.py
@@ -355,32 +355,30 @@ class replica_bitcell_array(design.design):
                                         width=self.width,
                                         height=pin.height())
 
-
         # Replica bitlines
-        for port in range(self.left_rbl+self.right_rbl):
+        for port in range(self.left_rbl + self.right_rbl):
             inst = self.replica_col_inst[port]
-            for (pin_name, bl_name) in zip(self.cell.get_all_bitline_names(),self.replica_bl_names[port]):
+            for (pin_name, bl_name) in zip(self.cell.get_all_bitline_names(), self.replica_bl_names[port]):
                 pin = inst.get_pin(pin_name)
                 if bl_name in self.rbl_bl_names or bl_name in self.rbl_br_names:
                     name = bl_name
                 else:
-                    name = "rbl_{0}_{1}".format(pin_name,port)
+                    name = "rbl_{0}_{1}".format(pin_name, port)
                 self.add_layout_pin(text=name,
                                     layer=pin.layer,
-                                    offset=pin.ll().scale(1,0),
+                                    offset=pin.ll().scale(1, 0),
                                     width=pin.width(),
                                     height=self.height)
-
                     
-        for pin_name in ["vdd","gnd"]:
+        for pin_name in ["vdd", "gnd"]:
             for inst in self.insts:
                 pin_list = inst.get_pins(pin_name)
                 for pin in pin_list:
-                    self.add_power_pin(name=pin_name, loc=pin.center(), vertical=True, start_layer=pin.layer)
+                    self.add_power_pin(name=pin_name,
+                                       loc=pin.center(),
+                                       directions=("V", "V"),
+                                       start_layer=pin.layer)
 
-
-
-                        
     def get_rbl_wl_name(self, port):
         """ Return the WL for the given RBL port """
         return self.rbl_wl_names[port]
@@ -399,7 +397,7 @@ class replica_bitcell_array(design.design):
         
         # Dynamic Power from Bitline
         bl_wire = self.gen_bl_wire()
-        cell_load = 2 * bl_wire.return_input_cap() 
+        cell_load = 2 * bl_wire.return_input_cap()
         bl_swing = OPTS.rbl_delay_percentage
         freq = spice["default_event_frequency"]
         bitline_dynamic = self.calc_dynamic_power(corner, cell_load, freq, swing=bl_swing)
diff --git a/compiler/modules/sense_amp_array.py b/compiler/modules/sense_amp_array.py
index 837bf4b4..db6dfe22 100644
--- a/compiler/modules/sense_amp_array.py
+++ b/compiler/modules/sense_amp_array.py
@@ -151,13 +151,13 @@ class sense_amp_array(design.design):
             self.add_power_pin(name="gnd",
                                loc=gnd_pin.center(),
                                start_layer=gnd_pin.layer,
-                               vertical=True)
+                               directions=("V", "V"))
             
             vdd_pin = inst.get_pin("vdd")
             self.add_power_pin(name="vdd",
                                loc=vdd_pin.center(),
                                start_layer=vdd_pin.layer,
-                               vertical=True)
+                               directions=("V", "V"))
 
             bl_pin = inst.get_pin(inst.mod.get_bl_names())
             br_pin = inst.get_pin(inst.mod.get_br_names())
diff --git a/compiler/modules/wordline_driver.py b/compiler/modules/wordline_driver.py
index 76a31074..55f5e707 100644
--- a/compiler/modules/wordline_driver.py
+++ b/compiler/modules/wordline_driver.py
@@ -8,7 +8,7 @@
 import debug
 import design
 import math
-import contact
+from tech import drc
 from vector import vector
 from sram_factory import factory
 from globals import OPTS
@@ -26,8 +26,8 @@ class wordline_driver(design.design):
         debug.info(1, "Creating {0}".format(self.name))
         self.add_comment("rows: {0} cols: {1}".format(rows, cols))
         
-        self.rows = rows
-        self.cols = cols
+        self.bitcell_rows = rows
+        self.bitcell_cols = cols
 
         b = factory.create(module_type="bitcell")
         try:
@@ -36,6 +36,11 @@ class wordline_driver(design.design):
             self.cell_multiple = 1
         self.cell_height = self.cell_multiple * b.height
         
+        # We may have more than one bitcell per decoder row
+        self.num_rows = math.ceil(self.bitcell_rows / self.cell_multiple)
+        # We will place this many final decoders per row
+        self.decoders_per_row = math.ceil(self.bitcell_rows / self.num_rows)
+        
         self.create_netlist()
         if not OPTS.netlist_only:
             self.create_layout()
@@ -56,10 +61,10 @@ class wordline_driver(design.design):
         
     def add_pins(self):
         # inputs to wordline_driver.
-        for i in range(self.rows):
+        for i in range(self.bitcell_rows):
             self.add_pin("in_{0}".format(i), "INPUT")
         # Outputs from wordline_driver.
-        for i in range(self.rows):
+        for i in range(self.bitcell_rows):
             self.add_pin("wl_{0}".format(i), "OUTPUT")
         self.add_pin("en", "INPUT")
         self.add_pin("vdd", "POWER")
@@ -68,7 +73,7 @@ class wordline_driver(design.design):
     def add_modules(self):
         self.and2 = factory.create(module_type="pand2",
                                    height=self.cell_height,
-                                   size=self.cols)
+                                   size=self.bitcell_cols)
         self.add_mod(self.and2)
         
     def route_vdd_gnd(self):
@@ -79,7 +84,7 @@ class wordline_driver(design.design):
 
         # Find the x offsets for where the vias/pins should be placed
         xoffset_list = [self.and_inst[0].lx()]
-        for num in range(self.rows):
+        for num in range(self.bitcell_rows):
             # this will result in duplicate polygons for rails, but who cares
             
             # use the inverter offset even though it will be the and's too
@@ -97,32 +102,34 @@ class wordline_driver(design.design):
             
     def create_drivers(self):
         self.and_inst = []
-        for row in range(self.rows):
+        for row in range(self.bitcell_rows):
             name_and = "wl_driver_and{}".format(row)
 
             # add and2
             self.and_inst.append(self.add_inst(name=name_and,
                                                 mod=self.and2))
-            self.connect_inst(["en",
-                               "in_{0}".format(row),
+            self.connect_inst(["in_{0}".format(row),
+                               "en",
                                "wl_{0}".format(row),
                                "vdd", "gnd"])
 
     def setup_layout_constants(self):
         # We may have more than one bitcell per decoder row
-        self.num_rows = math.ceil(self.rows / self.cell_multiple)
+        self.driver_rows = math.ceil(self.bitcell_rows / self.cell_multiple)
         # We will place this many final decoders per row
-        self.decoders_per_row = math.ceil(self.rows / self.num_rows)
+        self.decoders_per_row = math.ceil(self.bitcell_rows / self.driver_rows)
             
     def place_drivers(self):
-        and2_xoffset = 2 * self.m1_width + 5 * self.m1_space
+
+        # Leave a well gap to separate the bitcell array well from this well
+        well_gap = 2 * drc("pwell_to_nwell") + drc("nwell_enclose_active")
         
-        self.width = and2_xoffset + self.and2.width
-        self.height = self.and2.height * self.num_rows
+        self.width = self.decoders_per_row * self.and2.width + well_gap
+        self.height = self.and2.height * self.driver_rows
         
-        for row in range(self.rows):
-            #row = math.floor(inst_index / self.decoders_per_row)
-            #dec = inst_index % self.decoders_per_row
+        for inst_index in range(self.bitcell_rows):
+            row = math.floor(inst_index / self.decoders_per_row)
+            dec = inst_index % self.decoders_per_row
             
             if (row % 2):
                 y_offset = self.and2.height * (row + 1)
@@ -131,63 +138,42 @@ class wordline_driver(design.design):
                 y_offset = self.and2.height * row
                 inst_mirror = "R0"
 
-            # x_off = self.internal_routing_width + dec * and_mod.width
-            and2_offset = [and2_xoffset, y_offset]
+            x_offset = dec * self.and2.width
+            and2_offset = [x_offset, y_offset]
             
             # add and2
-            self.and_inst[row].place(offset=and2_offset,
-                                      mirror=inst_mirror)
+            self.and_inst[inst_index].place(offset=and2_offset,
+                                            mirror=inst_mirror)
 
     def route_layout(self):
         """ Route all of the signals """
 
         # Wordline enable connection
-        en_offset = [self.m1_width + 2 * self.m1_space, 0]
+        en_pin = self.and_inst[0].get_pin("B")
+        en_bottom_pos = vector(en_pin.lx(), 0)
         en_pin = self.add_layout_pin(text="en",
                                      layer="m2",
-                                     offset=en_offset,
-                                     width=self.m2_width,
+                                     offset=en_bottom_pos,
                                      height=self.height)
         
-        for row in range(self.rows):
-            and_inst = self.and_inst[row]
-            
-            # en connection
-            a_pin = and_inst.get_pin("A")
-            a_pos = a_pin.lc()
-            clk_offset = vector(en_pin.bc().x, a_pos.y)
-            self.add_segment_center(layer="m1",
-                                    start=clk_offset,
-                                    end=a_pos)
-            self.add_via_center(layers=self.m1_stack,
-                                offset=clk_offset)
+        for inst_index in range(self.bitcell_rows):
+            and_inst = self.and_inst[inst_index]
+            row = math.floor(inst_index / self.decoders_per_row)
 
-            # connect the decoder input pin to and2 B
+            # Drop a via
             b_pin = and_inst.get_pin("B")
-            b_pos = b_pin.lc()
-            # needs to move down since B and input is
-            # nearly aligned with A inv input
-            up_or_down = self.m2_space if row % 2 else -self.m2_space
-            input_offset = vector(0, b_pos.y + up_or_down)
-            base_offset = vector(clk_offset.x, input_offset.y)
-            contact_offset = vector(0.5 * self.m2_width + self.m2_space + 0.5 * contact.m1_via.width, 0)
-            mid_via_offset = base_offset + contact_offset
-
+            self.add_via_stack_center(from_layer=b_pin.layer,
+                                      to_layer="m2",
+                                      offset=b_pin.center())
+            
+            # connect the decoder input pin to and2 A
+            a_pin = and_inst.get_pin("A")
+            a_pos = a_pin.center()
             # must under the clk line in M1
             self.add_layout_pin_segment_center(text="in_{0}".format(row),
                                                layer="m1",
-                                               start=input_offset,
-                                               end=mid_via_offset)
-            self.add_via_center(layers=self.m1_stack,
-                                offset=mid_via_offset,
-                                directions=("V", "V"))
-
-            # now connect to the and2 B
-            self.add_path("m2", [mid_via_offset, b_pos])
-            contact_offset = b_pos - vector(0.5 * contact.m1_via.height, 0)
-            self.add_via_center(layers=self.m1_stack,
-                                offset=contact_offset,
-                                directions=("H", "H"))
+                                               start=vector(0, a_pos.y),
+                                               end=a_pos)
 
             # output each WL on the right
             wl_offset = and_inst.get_pin("Z").rc()
diff --git a/compiler/modules/write_driver_array.py b/compiler/modules/write_driver_array.py
index a7ec61f0..640db97e 100644
--- a/compiler/modules/write_driver_array.py
+++ b/compiler/modules/write_driver_array.py
@@ -209,7 +209,7 @@ class write_driver_array(design.design):
                 for pin in pin_list:
                     self.add_power_pin(name=n,
                                        loc=pin.center(),
-                                       vertical=True,
+                                       directions=("V", "V"),
                                        start_layer=pin.layer)
         if self.write_size:
             for bit in range(self.num_wmasks):
diff --git a/compiler/modules/write_mask_and_array.py b/compiler/modules/write_mask_and_array.py
index 25acb324..4d4e0c50 100644
--- a/compiler/modules/write_mask_and_array.py
+++ b/compiler/modules/write_mask_and_array.py
@@ -5,9 +5,7 @@
 # (acting for and on behalf of Oklahoma State University)
 # All rights reserved.
 #
-from math import log
 import design
-from tech import drc
 import debug
 from sram_factory import factory
 from vector import vector
@@ -43,33 +41,28 @@ class write_mask_and_array(design.design):
         self.add_pins()
         self.create_and2_array()
 
-
     def create_layout(self):
         self.place_and2_array()
-        spacing = self.wmask_en_len - self.and2.width
-        self.width = (self.num_wmasks*self.and2.width) + ((self.num_wmasks-1)*spacing)
-        self.height = self.and2.height
         self.add_layout_pins()
         self.add_boundary()
         self.DRC_LVS()
 
     def add_pins(self):
         for bit in range(self.num_wmasks):
-            self.add_pin("wmask_in_{}".format(bit),"INPUT")
+            self.add_pin("wmask_in_{}".format(bit), "INPUT")
         self.add_pin("en", "INPUT")
         for bit in range(self.num_wmasks):
-            self.add_pin("wmask_out_{}".format(bit),"OUTPUT")
-        self.add_pin("vdd","POWER")
-        self.add_pin("gnd","GROUND")
+            self.add_pin("wmask_out_{}".format(bit), "OUTPUT")
+        self.add_pin("vdd", "POWER")
+        self.add_pin("gnd", "GROUND")
 
     def add_modules(self):
         # Size the AND gate for the number of write drivers it drives, which is equal to the write size.
         # Assume stage effort of 3 to compute the size
         self.and2 = factory.create(module_type="pand2",
-                                   size=self.write_size/4.0)
+                                   size=self.write_size / 4.0)
         self.add_mod(self.and2)
 
-
     def create_and2_array(self):
         self.and2_insts = {}
         for bit in range(self.num_wmasks):
@@ -81,7 +74,6 @@ class write_mask_and_array(design.design):
                                "wmask_out_{}".format(bit),
                                "vdd", "gnd"])
 
-
     def place_and2_array(self):
         # Place the write mask AND array at the start of each write driver enable length.
         # This ensures the write mask AND array will be directly under the corresponding write driver enable wire.
@@ -96,56 +88,45 @@ class write_mask_and_array(design.design):
 
         self.wmask_en_len = self.words_per_row * (self.write_size * self.driver_spacing)
         debug.check(self.wmask_en_len >= self.and2.width,
-                    "Write mask AND is wider than the corresponding write drivers {0} vs {1}.".format(self.and2.width,self.wmask_en_len))
+                    "Write mask AND is wider than the corresponding write drivers {0} vs {1}.".format(self.and2.width,
+                                                                                                      self.wmask_en_len))
 
+        self.width = self.bitcell.width * self.columns
+        self.height = self.and2.height
+        
         for i in range(self.num_wmasks):
             base = vector(i * self.wmask_en_len, 0)
             self.and2_insts[i].place(base)
 
-
     def add_layout_pins(self):
 
         # Create the enable pin that connects all write mask AND array's B pins
-        beg_en_pin = self.and2_insts[0].get_pin("B")
-        end_en_pin = self.and2_insts[self.num_wmasks-1].get_pin("B")
-        if self.port % 2:
-            # Extend metal3 to edge of AND array in multiport
-            en_to_edge = self.and2.width - beg_en_pin.cx()
-            self.add_layout_pin(text="en",
-                                layer="m3",
-                                offset=beg_en_pin.bc(),
-                                width=end_en_pin.cx() - beg_en_pin.cx() + en_to_edge)
-            self.add_via_center(layers=self.m1_stack,
-                                offset=vector(end_en_pin.cx() + en_to_edge, end_en_pin.cy()))
-            self.add_via_center(layers=self.m2_stack,
-                                offset=vector(end_en_pin.cx() + en_to_edge, end_en_pin.cy()))
-        else:
-            self.add_layout_pin(text="en",
-                                layer="m3",
-                                offset=beg_en_pin.bc(),
-                                width=end_en_pin.cx() - beg_en_pin.cx())
+        en_pin = self.and2_insts[0].get_pin("B")
+        self.add_layout_pin_segment_center(text="en",
+                                           layer="m3",
+                                           start=vector(0, en_pin.cy()),
+                                           end=vector(self.width, en_pin.cy()))
 
         for i in range(self.num_wmasks):
             # Copy remaining layout pins
-            self.copy_layout_pin(self.and2_insts[i],"A","wmask_in_{0}".format(i))
-            self.copy_layout_pin(self.and2_insts[i],"Z","wmask_out_{0}".format(i))
+            self.copy_layout_pin(self.and2_insts[i], "A", "wmask_in_{0}".format(i))
+            self.copy_layout_pin(self.and2_insts[i], "Z", "wmask_out_{0}".format(i))
 
             # Add via connections to metal3 for AND array's B pin
             en_pin = self.and2_insts[i].get_pin("B")
-            self.add_via_center(layers=self.m1_stack,
-                                offset=en_pin.center())
-            self.add_via_center(layers=self.m2_stack,
-                                offset=en_pin.center())
+            en_pos = en_pin.center()
+            self.add_via_stack_center(from_layer=en_pin.layer,
+                                      to_layer="m3",
+                                      offset=en_pos)
 
             for supply in ["gnd", "vdd"]:
                 supply_pin=self.and2_insts[i].get_pin(supply)
                 self.add_power_pin(supply, supply_pin.center())
 
-
         for supply in ["gnd", "vdd"]:
             supply_pin_left = self.and2_insts[0].get_pin(supply)
-            supply_pin_right = self.and2_insts[self.num_wmasks-1].get_pin(supply)
-            self.add_path("m1",[supply_pin_left.lc(), supply_pin_right.rc()])
+            supply_pin_right = self.and2_insts[self.num_wmasks - 1].get_pin(supply)
+            self.add_path("m1", [supply_pin_left.lc(), supply_pin_right.rc()])
             
     def get_cin(self):
         """Get the relative capacitance of all the input connections in the bank"""
diff --git a/compiler/pgates/pand2.py b/compiler/pgates/pand2.py
index 4c044f1c..69af1e62 100644
--- a/compiler/pgates/pand2.py
+++ b/compiler/pgates/pand2.py
@@ -44,6 +44,7 @@ class pand2(pgate.pgate):
         self.place_insts()
         self.add_wires()
         self.add_layout_pins()
+        self.route_supply_rails()
         self.add_boundary()
         self.DRC_LVS()
         
@@ -76,26 +77,10 @@ class pand2(pgate.pgate):
         a2_pin = self.inv_inst.get_pin("A")
         mid1_point = vector(0.5 * (z1_pin.cx() + a2_pin.cx()), z1_pin.cy())
         mid2_point = vector(mid1_point, a2_pin.cy())
-        self.add_path("m1",
+        self.add_path(self.route_layer,
                       [z1_pin.center(), mid1_point, mid2_point, a2_pin.center()])
         
     def add_layout_pins(self):
-        # Continous vdd rail along with label.
-        vdd_pin = self.inv_inst.get_pin("vdd")
-        self.add_layout_pin(text="vdd",
-                            layer="m1",
-                            offset=vdd_pin.ll().scale(0, 1),
-                            width=self.width,
-                            height=vdd_pin.height())
-        
-        # Continous gnd rail along with label.
-        gnd_pin = self.inv_inst.get_pin("gnd")
-        self.add_layout_pin(text="gnd",
-                            layer="m1",
-                            offset=gnd_pin.ll().scale(0, 1),
-                            width=self.width,
-                            height=vdd_pin.height())
-            
         pin = self.inv_inst.get_pin("Z")
         self.add_layout_pin_rect_center(text="Z",
                                         layer=pin.layer,
diff --git a/compiler/pgates/pand3.py b/compiler/pgates/pand3.py
index f8cc2ac3..841ac69d 100644
--- a/compiler/pgates/pand3.py
+++ b/compiler/pgates/pand3.py
@@ -44,6 +44,7 @@ class pand3(pgate.pgate):
         self.place_insts()
         self.add_wires()
         self.add_layout_pins()
+        self.route_supply_rails()
         self.add_boundary()
         self.DRC_LVS()
         
@@ -77,26 +78,10 @@ class pand3(pgate.pgate):
         a2_pin = self.inv_inst.get_pin("A")
         mid1_point = vector(0.5 * (z1_pin.cx()+a2_pin.cx()), z1_pin.cy())
         mid2_point = vector(mid1_point, a2_pin.cy())
-        self.add_path("m1",
+        self.add_path(z1_pin.layer,
                       [z1_pin.center(), mid1_point, mid2_point, a2_pin.center()])
         
     def add_layout_pins(self):
-        # Continous vdd rail along with label.
-        vdd_pin = self.inv_inst.get_pin("vdd")
-        self.add_layout_pin(text="vdd",
-                            layer="m1",
-                            offset=vdd_pin.ll().scale(0, 1),
-                            width=self.width,
-                            height=vdd_pin.height())
-        
-        # Continous gnd rail along with label.
-        gnd_pin = self.inv_inst.get_pin("gnd")
-        self.add_layout_pin(text="gnd",
-                            layer="m1",
-                            offset=gnd_pin.ll().scale(0, 1),
-                            width=self.width,
-                            height=vdd_pin.height())
-            
         pin = self.inv_inst.get_pin("Z")
         self.add_layout_pin_rect_center(text="Z",
                                         layer=pin.layer,
diff --git a/compiler/pgates/pbuf.py b/compiler/pgates/pbuf.py
index 6f9719eb..4d90286d 100644
--- a/compiler/pgates/pbuf.py
+++ b/compiler/pgates/pbuf.py
@@ -37,6 +37,7 @@ class pbuf(pgate.pgate):
         self.place_insts()
         self.add_wires()
         self.add_layout_pins()
+        self.route_supply_rails()
         self.add_boundary()
         
     def add_pins(self):
@@ -78,26 +79,9 @@ class pbuf(pgate.pgate):
         # inv1 Z to inv2 A
         z1_pin = self.inv1_inst.get_pin("Z")
         a2_pin = self.inv2_inst.get_pin("A")
-        mid_point = vector(z1_pin.cx(), a2_pin.cy())
-        self.add_path("m1", [z1_pin.center(), mid_point, a2_pin.center()])
+        self.add_zjog(self.route_layer, z1_pin.center(), a2_pin.center())
         
     def add_layout_pins(self):
-        # Continous vdd rail along with label.
-        vdd_pin = self.inv1_inst.get_pin("vdd")
-        self.add_layout_pin(text="vdd",
-                            layer="m1",
-                            offset=vdd_pin.ll().scale(0, 1),
-                            width=self.width,
-                            height=vdd_pin.height())
-        
-        # Continous gnd rail along with label.
-        gnd_pin = self.inv1_inst.get_pin("gnd")
-        self.add_layout_pin(text="gnd",
-                            layer="m1",
-                            offset=gnd_pin.ll().scale(0, 1),
-                            width=self.width,
-                            height=vdd_pin.height())
-            
         z_pin = self.inv2_inst.get_pin("Z")
         self.add_layout_pin_rect_center(text="Z",
                                         layer=z_pin.layer,
diff --git a/compiler/pgates/pdriver.py b/compiler/pgates/pdriver.py
index 4bf654a4..9bba7aee 100644
--- a/compiler/pgates/pdriver.py
+++ b/compiler/pgates/pdriver.py
@@ -76,6 +76,7 @@ class pdriver(pgate.pgate):
 
         self.width = self.inv_inst_list[-1].rx()
         self.height = self.inv_inst_list[0].height
+        self.route_supply_rails()
         self.add_boundary()
         
     def add_pins(self):
@@ -141,26 +142,11 @@ class pdriver(pgate.pgate):
             z_inst_list.append(self.inv_inst_list[x].get_pin("Z"))
             a_inst_list.append(self.inv_inst_list[x + 1].get_pin("A"))
             mid_point = vector(z_inst_list[x].cx(), a_inst_list[x].cy())
-            self.add_path("m1",
+            self.add_path(self.route_layer,
                           [z_inst_list[x].center(), mid_point,
                            a_inst_list[x].center()])
 
     def add_layout_pins(self):
-        # Continous vdd rail along with label.
-        vdd_pin = self.inv_inst_list[0].get_pin("vdd")
-        self.add_layout_pin(text="vdd",
-                            layer="m1",
-                            offset=vdd_pin.ll().scale(0, 1),
-                            width=self.width,
-                            height=vdd_pin.height())
-        
-        # Continous gnd rail along with label.
-        gnd_pin = self.inv_inst_list[0].get_pin("gnd")
-        self.add_layout_pin(text="gnd",
-                            layer="m1",
-                            offset=gnd_pin.ll().scale(0, 1),
-                            width=self.width,
-                            height=vdd_pin.height())
 
         z_pin = self.inv_inst_list[len(self.inv_inst_list) - 1].get_pin("Z")
         self.add_layout_pin_rect_center(text="Z",
diff --git a/compiler/pgates/pgate.py b/compiler/pgates/pgate.py
index a942b35f..5351be07 100644
--- a/compiler/pgates/pgate.py
+++ b/compiler/pgates/pgate.py
@@ -17,6 +17,7 @@ from globals import OPTS
 if(OPTS.tech_name == "s8"):
     from tech import nmos_bins, pmos_bins, accuracy_requirement
 
+    
 class pgate(design.design):
     """
     This is a module that implements some shared
@@ -30,9 +31,25 @@ class pgate(design.design):
         if height:
             self.height = height
         elif not height:
-            # By default, we make it 10 M1 pitch tall
-            self.height = 10*self.m1_pitch
+            # By default, something simple
+            self.height = 14 * self.m1_pitch
             
+        if "li" in layer:
+            self.route_layer = "li"
+        else:
+            self.route_layer = "m1"
+        self.route_layer_width = getattr(self, "{}_width".format(self.route_layer))
+        self.route_layer_space = getattr(self, "{}_space".format(self.route_layer))
+        self.route_layer_pitch = getattr(self, "{}_pitch".format(self.route_layer))
+
+        # This is the space from a S/D contact to the supply rail
+        # Assume the contact starts at the active edge
+        contact_to_vdd_rail_space = 0.5 * self.m1_width + self.m1_space
+        # This is a poly-to-poly of a flipped cell
+        poly_to_poly_gate_space = self.poly_extend_active + self.poly_space
+        self.top_bottom_space = max(contact_to_vdd_rail_space,
+                                    poly_to_poly_gate_space)
+        
         self.create_netlist()
         if not OPTS.netlist_only:
             self.create_layout()
@@ -47,27 +64,27 @@ class pgate(design.design):
         """ Pure virtual function """
         debug.error("Must over-ride create_layout.", -1)
 
-    def connect_pin_to_rail(self, inst, pin, supply):
+    def connect_pin_to_rail(self, inst, pin_name, supply_name):
         """ Connects a ptx pin to a supply rail. """
-        source_pin = inst.get_pin(pin)
-        supply_pin = self.get_pin(supply)
-        if supply_pin.overlaps(source_pin):
-            return
-            
-        if supply == "gnd":
-            height = supply_pin.by() - source_pin.by()
-        elif supply == "vdd":
-            height = supply_pin.uy() - source_pin.by()
-        else:
-            debug.error("Invalid supply name.", -1)
+        supply_pin = self.get_pin(supply_name)
         
-        if abs(height) > 0:
-            self.add_rect(layer="m1",
+        source_pins = inst.get_pins(pin_name)
+        for source_pin in source_pins:
+            
+            if supply_name == "gnd":
+                height = supply_pin.by() - source_pin.by()
+            elif supply_name == "vdd":
+                height = supply_pin.uy() - source_pin.by()
+            else:
+                debug.error("Invalid supply name.", -1)
+
+            debug.check(supply_pin.layer == source_pin.layer, "Supply pin is not on correct layer.")
+            self.add_rect(layer=source_pin.layer,
                           offset=source_pin.ll(),
                           height=height,
                           width=source_pin.width())
     
-    def route_input_gate(self, pmos_inst, nmos_inst, ypos, name, position="left"):
+    def route_input_gate(self, pmos_inst, nmos_inst, ypos, name, position="left", directions=None):
         """
         Route the input gate to the left side of the cell for access.
         Position specifies to place the contact the left, center, or
@@ -93,8 +110,6 @@ class pgate(design.design):
 
         # Center is completely symmetric.
         contact_width = contact.poly_contact.width
-        contact_m1_width = contact.poly_contact.second_layer_width
-        contact_m1_height = contact.poly_contact.second_layer_height
             
         if position == "center":
             contact_offset = left_gate_offset \
@@ -111,18 +126,16 @@ class pgate(design.design):
         else:
             debug.error("Invalid contact placement option.", -1)
 
-        if hasattr(self, "li_stack"):
-            self.add_via_center(layers=self.li_stack,
-                                offset=contact_offset)
-            
-        self.add_via_center(layers=self.poly_stack,
-                            offset=contact_offset)
+        via = self.add_via_stack_center(from_layer="poly",
+                                        to_layer=self.route_layer,
+                                        offset=contact_offset,
+                                        directions=directions)
 
         self.add_layout_pin_rect_center(text=name,
-                                        layer="m1",
+                                        layer=self.route_layer,
                                         offset=contact_offset,
-                                        width=contact_m1_width,
-                                        height=contact_m1_height)
+                                        width=via.mod.second_layer_width,
+                                        height=via.mod.second_layer_height)
         # This is to ensure that the contact is
         # connected to the gate
         mid_point = contact_offset.scale(0.5, 1) \
@@ -194,12 +207,10 @@ class pgate(design.design):
         self.nwell_contact = self.add_via_center(layers=layer_stack,
                                                  offset=contact_offset,
                                                  implant_type="n",
-                                                 well_type="n")
-        if hasattr(self, "li_stack"):
-            self.add_via_center(layers=self.li_stack,
-                                offset=contact_offset)
+                                                 well_type="n",
+                                                 directions=("V", "V"))
         
-        self.add_rect_center(layer="m1",
+        self.add_rect_center(layer=self.route_layer,
                              offset=contact_offset + vector(0, 0.5 * (self.height - contact_offset.y)),
                              width=self.nwell_contact.mod.second_layer_width,
                              height=self.height - contact_offset.y)
@@ -249,13 +260,10 @@ class pgate(design.design):
         self.pwell_contact= self.add_via_center(layers=layer_stack,
                                                 offset=contact_offset,
                                                 implant_type="p",
-                                                well_type="p")
-
-        if hasattr(self, "li_stack"):
-            self.add_via_center(layers=self.li_stack,
-                                offset=contact_offset)
-
-        self.add_rect_center(layer="m1",
+                                                well_type="p",
+                                                directions=("V", "V"))
+            
+        self.add_rect_center(layer=self.route_layer,
                              offset=contact_offset.scale(1, 0.5),
                              width=self.pwell_contact.mod.second_layer_width,
                              height=contact_offset.y)
@@ -279,6 +287,18 @@ class pgate(design.design):
         #               offset=implant_offset,
         #               width=implant_width,
         #               height=implant_height)
+        
+    def route_supply_rails(self):
+        """ Add vdd/gnd rails to the top and bottom. """
+        self.add_layout_pin_rect_center(text="gnd",
+                                        layer=self.route_layer,
+                                        offset=vector(0.5 * self.width, 0),
+                                        width=self.width)
+        
+        self.add_layout_pin_rect_center(text="vdd",
+                                        layer=self.route_layer,
+                                        offset=vector(0.5 * self.width, self.height),
+                                        width=self.width)
 
     def determine_width(self):
         """ Determine the width based on the well contacts (assumed to be on the right side) """
@@ -286,9 +306,10 @@ class pgate(design.design):
         self.width = max(self.nwell_contact.rx(), self.pwell_contact.rx()) + self.m1_space + 0.5 * contact.m1_via.width
         self.well_width = self.width + 2 * self.nwell_enclose_active
         # Height is an input parameter, so it is not recomputed.
-    
-    def bin_width(self, tx_type, target_width):
 
+    @staticmethod
+    def bin_width(tx_type, target_width):
+        
         if tx_type == "nmos":
             bins = nmos_bins[drc("minwidth_poly")]
         elif tx_type == "pmos":
@@ -303,22 +324,24 @@ class pgate(design.design):
             scaled_bin = bins[0] * scaling_factor
             
         else:
+            base_bins = []
             scaled_bins = []
             scaling_factors = []
             scaled_bins.append(bins[-1])
+            base_bins.append(bins[-1])
             scaling_factors.append(1)
             for width in bins[0:-1]:
                 m = math.ceil(target_width / width)
+                base_bins.append(width)
                 scaling_factors.append(m)
                 scaled_bins.append(m * width)
 
             select = bisect_left(scaled_bins, target_width)
             scaling_factor = scaling_factors[select]
             scaled_bin = scaled_bins[select]
-            select = (select + 1) % len(scaled_bins)
-            selected_bin = bins[select]
+            selected_bin = base_bins[select]
 
-        debug.info(2, "binning {0} tx, target: {4}, found {1} x {2} = {3}".format(tx_type, selected_bin, scaling_factor, scaled_bin, target_width))
+        debug.info(2, "binning {0} tx, target: {4}, found {1} x {2} = {3}".format(tx_type, selected_bin, scaling_factor, selected_bin * scaling_factor, target_width))
 
         return(selected_bin, scaling_factor)
 
@@ -343,4 +366,4 @@ class pgate(design.design):
         return(scaled_bins)
         
     def bin_accuracy(self, ideal_width, width):
-        return abs(1-(ideal_width - width)/ideal_width)
\ No newline at end of file
+        return abs(1-(ideal_width - width)/ideal_width)
diff --git a/compiler/pgates/pinv.py b/compiler/pgates/pinv.py
index 0a26a2fd..e0e1ce81 100644
--- a/compiler/pgates/pinv.py
+++ b/compiler/pgates/pinv.py
@@ -22,17 +22,17 @@ from errors import drc_error
 if(OPTS.tech_name == "s8"):
     from tech import nmos_bins, pmos_bins, accuracy_requirement
 
+    
 class pinv(pgate.pgate):
     """
     Pinv generates gds of a parametrically sized inverter. The
     size is specified as the drive size (relative to minimum NMOS) and
     a beta value for choosing the pmos size.  The inverter's cell
     height is usually the same as the 6t library cell and is measured
-    from center of rail to rail..  The route_output will route the
-    output to the right side of the cell for easier access.
+    from center of rail to rail.
     """
 
-    def __init__(self, name, size=1, beta=parameter["beta"], height=None, route_output=True):
+    def __init__(self, name, size=1, beta=parameter["beta"], height=None):
 
         debug.info(2,
                    "creating pinv structure {0} with size of {1}".format(name,
@@ -43,7 +43,6 @@ class pinv(pgate.pgate):
         self.nmos_size = size
         self.pmos_size = beta * size
         self.beta = beta
-        self.route_output = False
         
         pgate.pgate.__init__(self, name, height)
 
@@ -108,13 +107,6 @@ class pinv(pgate.pgate):
         min_channel = max(contact.poly_contact.width + self.m1_space,
                           contact.poly_contact.width + 2 * self.poly_to_active)
         
-        # This is the extra space needed to ensure DRC rules
-        # to the active contacts
-        extra_contact_space = max(-nmos.get_pin("D").by(), 0)
-        # This is a poly-to-poly of a flipped cell
-        self.top_bottom_space = max(0.5*self.m1_width + self.m1_space + extra_contact_space, 
-                                    self.poly_extend_active + self.poly_space)
-
         total_height = tx_height + min_channel + 2 * self.top_bottom_space
         # debug.check(self.height > total_height,
         #             "Cell height {0} too small for simple min height {1}.".format(self.height,
@@ -202,30 +194,22 @@ class pinv(pgate.pgate):
                                    width=self.nmos_width,
                                    mults=self.tx_mults,
                                    tx_type="nmos",
+                                   add_source_contact=self.route_layer,
+                                   add_drain_contact=self.route_layer,
                                    connect_poly=True,
-                                   connect_active=True)
+                                   connect_drain_active=True)
         self.add_mod(self.nmos)
         
         self.pmos = factory.create(module_type="ptx",
                                    width=self.pmos_width,
                                    mults=self.tx_mults,
                                    tx_type="pmos",
+                                   add_source_contact=self.route_layer,
+                                   add_drain_contact=self.route_layer,
                                    connect_poly=True,
-                                   connect_active=True)
+                                   connect_drain_active=True)
         self.add_mod(self.pmos)
         
-    def route_supply_rails(self):
-        """ Add vdd/gnd rails to the top and bottom. """
-        self.add_layout_pin_rect_center(text="gnd",
-                                        layer="m1",
-                                        offset=vector(0.5 * self.width, 0),
-                                        width=self.width)
-
-        self.add_layout_pin_rect_center(text="vdd",
-                                        layer="m1",
-                                        offset=vector(0.5 * self.width, self.height),
-                                        width=self.width)
-
     def create_ptx(self):
         """
         Create the PMOS and NMOS netlist.
@@ -266,7 +250,7 @@ class pinv(pgate.pgate):
         Route the output (drains) together.
         Optionally, routes output to edge.
         """
-
+            
         # Get the drain pins
         nmos_drain_pin = self.nmos_inst.get_pin("D")
         pmos_drain_pin = self.pmos_inst.get_pin("D")
@@ -274,24 +258,16 @@ class pinv(pgate.pgate):
         # Pick point at right most of NMOS and connect down to PMOS
         nmos_drain_pos = nmos_drain_pin.bc()
         pmos_drain_pos = vector(nmos_drain_pos.x, pmos_drain_pin.uc().y)
-        self.add_path("m1", [nmos_drain_pos, pmos_drain_pos])
+        self.add_path(self.route_layer, [nmos_drain_pos, pmos_drain_pos])
 
         # Remember the mid for the output
         mid_drain_offset = vector(nmos_drain_pos.x, self.output_pos.y)
 
-        if self.route_output:
-            # This extends the output to the edge of the cell
-            output_offset = mid_drain_offset.scale(0, 1) + vector(self.width, 0)
-            self.add_layout_pin_segment_center(text="Z",
-                                               layer="m1",
-                                               start=mid_drain_offset,
-                                               end=output_offset)
-        else:
-            # This leaves the output as an internal pin (min sized)
-            self.add_layout_pin_rect_center(text="Z",
-                                            layer="m1",
-                                            offset=mid_drain_offset \
-                                            + vector(0.5 * self.m1_width, 0))
+        # This leaves the output as an internal pin (min sized)
+        output_offset = mid_drain_offset + vector(0.5 * self.route_layer_width, 0)
+        self.add_layout_pin_rect_center(text="Z",
+                                        layer=self.route_layer,
+                                        offset=output_offset)
 
     def add_well_contacts(self):
         """ Add n/p well taps to the layout and connect to supplies """
diff --git a/compiler/pgates/pnand2.py b/compiler/pgates/pnand2.py
index 08024cea..5d20201e 100644
--- a/compiler/pgates/pnand2.py
+++ b/compiler/pgates/pnand2.py
@@ -72,65 +72,45 @@ class pnand2(pgate.pgate):
 
     def add_ptx(self):
         """ Create the PMOS and NMOS transistors. """
-        self.nmos_nd = factory.create(module_type="ptx",
-                                      width=self.nmos_width,
-                                      mults=self.tx_mults,
-                                      tx_type="nmos",
-                                      add_drain_contact=False,
-                                      connect_poly=True,
-                                      connect_active=True)
-        self.add_mod(self.nmos_nd)
+        self.nmos_left = factory.create(module_type="ptx",
+                                        width=self.nmos_width,
+                                        mults=self.tx_mults,
+                                        tx_type="nmos",
+                                        add_source_contact=self.route_layer,
+                                        add_drain_contact="active")
+        self.add_mod(self.nmos_left)
         
-        self.nmos_ns = factory.create(module_type="ptx",
-                                      width=self.nmos_width,
-                                      mults=self.tx_mults,
-                                      tx_type="nmos",
-                                      add_source_contact=False,
-                                      connect_poly=True,
-                                      connect_active=True)
-        self.add_mod(self.nmos_ns)
+        self.nmos_right = factory.create(module_type="ptx",
+                                         width=self.nmos_width,
+                                         mults=self.tx_mults,
+                                         tx_type="nmos",
+                                         add_source_contact="active",
+                                         add_drain_contact=self.route_layer)
+        self.add_mod(self.nmos_right)
 
-        self.pmos = factory.create(module_type="ptx",
-                                   width=self.pmos_width,
-                                   mults=self.tx_mults,
-                                   tx_type="pmos",
-                                   connect_poly=True,
-                                   connect_active=True)
-        self.add_mod(self.pmos)
+        self.pmos_left = factory.create(module_type="ptx",
+                                        width=self.pmos_width,
+                                        mults=self.tx_mults,
+                                        tx_type="pmos",
+                                        add_source_contact=self.route_layer,
+                                        add_drain_contact=self.route_layer)
+        self.add_mod(self.pmos_left)
 
+        self.pmos_right = factory.create(module_type="ptx",
+                                         width=self.pmos_width,
+                                         mults=self.tx_mults,
+                                         tx_type="pmos",
+                                         add_source_contact=self.route_layer,
+                                         add_drain_contact=self.route_layer)
+        self.add_mod(self.pmos_right)
+        
     def setup_layout_constants(self):
         """ Pre-compute some handy layout parameters. """
 
-        # metal spacing to allow contacts on any layer
-        self.input_spacing = max(self.poly_space + contact.poly_contact.first_layer_width,
-                                 self.m1_space + contact.m1_via.first_layer_width,
-                                 self.m2_space + contact.m2_via.first_layer_width,
-                                 self.m3_space + contact.m2_via.second_layer_width)
-
-        
         # Compute the other pmos2 location,
         # but determining offset to overlap the
         # source and drain pins
-        self.overlap_offset = self.pmos.get_pin("D").ll() - self.pmos.get_pin("S").ll()
-
-        # This is the extra space needed to ensure DRC rules
-        # to the active contacts
-        extra_contact_space = max(-self.nmos_nd.get_pin("D").by(), 0)
-        # This is a poly-to-poly of a flipped cell
-        self.top_bottom_space = max(0.5 * self.m1_width + self.m1_space + extra_contact_space,
-                                    self.poly_extend_active + self.poly_space)
-        
-    def route_supply_rails(self):
-        """ Add vdd/gnd rails to the top and bottom. """
-        self.add_layout_pin_rect_center(text="gnd",
-                                        layer="m1",
-                                        offset=vector(0.5*self.width, 0),
-                                        width=self.width)
-
-        self.add_layout_pin_rect_center(text="vdd",
-                                        layer="m1",
-                                        offset=vector(0.5 * self.width, self.height),
-                                        width=self.width)
+        self.overlap_offset = self.pmos_left.get_pin("D").center() - self.pmos_left.get_pin("S").center()
 
     def create_ptx(self):
         """
@@ -138,19 +118,19 @@ class pnand2(pgate.pgate):
         """
 
         self.pmos1_inst = self.add_inst(name="pnand2_pmos1",
-                                        mod=self.pmos)
+                                        mod=self.pmos_left)
         self.connect_inst(["vdd", "A", "Z", "vdd"])
 
         self.pmos2_inst = self.add_inst(name="pnand2_pmos2",
-                                        mod=self.pmos)
+                                        mod=self.pmos_right)
         self.connect_inst(["Z", "B", "vdd", "vdd"])
 
         self.nmos1_inst = self.add_inst(name="pnand2_nmos1",
-                                        mod=self.nmos_nd)
+                                        mod=self.nmos_left)
         self.connect_inst(["Z", "B", "net1", "gnd"])
 
         self.nmos2_inst = self.add_inst(name="pnand2_nmos2",
-                                        mod=self.nmos_ns)
+                                        mod=self.nmos_right)
         self.connect_inst(["net1", "A", "gnd", "gnd"])
 
     def place_ptx(self):
@@ -159,35 +139,29 @@ class pnand2(pgate.pgate):
         to provide maximum routing in channel
         """
 
-        pmos1_pos = vector(self.pmos.active_offset.x,
-                           self.height - self.pmos.active_height \
+        pmos1_pos = vector(self.pmos_left.active_offset.x,
+                           self.height - self.pmos_left.active_height \
                            - self.top_bottom_space)
         self.pmos1_inst.place(pmos1_pos)
 
         self.pmos2_pos = pmos1_pos + self.overlap_offset
         self.pmos2_inst.place(self.pmos2_pos)
 
-        nmos1_pos = vector(self.pmos.active_offset.x,
+        nmos1_pos = vector(self.pmos_left.active_offset.x,
                            self.top_bottom_space)
         self.nmos1_inst.place(nmos1_pos)
 
         self.nmos2_pos = nmos1_pos + self.overlap_offset
         self.nmos2_inst.place(self.nmos2_pos)
 
-        # Output position will be in between the PMOS and NMOS
-        self.output_pos = vector(0,
-                                 0.5 * (pmos1_pos.y + nmos1_pos.y + self.nmos_nd.active_height))
-
     def add_well_contacts(self):
         """
         Add n/p well taps to the layout and connect to supplies
         AFTER the wells are created
         """
 
-        self.add_nwell_contact(self.pmos,
-                               self.pmos2_pos + vector(self.m1_pitch, 0))
-        self.add_pwell_contact(self.nmos_nd,
-                               self.nmos2_pos + vector(self.m1_pitch, 0))
+        self.add_nwell_contact(self.pmos_right, self.pmos2_pos)
+        self.add_pwell_contact(self.nmos_left, self.nmos2_pos)
         
     def connect_rails(self):
         """ Connect the nmos and pmos to its respective power rails """
@@ -200,35 +174,46 @@ class pnand2(pgate.pgate):
 
     def route_inputs(self):
         """ Route the A and B inputs """
-        inputB_yoffset = self.nmos2_inst.uy() + 0.5 * contact.poly_contact.height
+
+
+        # Top of NMOS drain
+        nmos_pin = self.nmos2_inst.get_pin("D")
+        bottom_pin_offset = nmos_pin.uy()
+        self.inputA_yoffset = bottom_pin_offset + self.m1_pitch
+
+        self.inputB_yoffset = self.inputA_yoffset + self.m3_pitch
+
+        # This will help with the wells and the input/output placement
         self.route_input_gate(self.pmos2_inst,
                               self.nmos2_inst,
-                              inputB_yoffset,
+                              self.inputB_yoffset,
                               "B",
                               position="center")
         
-        # This will help with the wells and the input/output placement
-        self.inputA_yoffset = self.pmos2_inst.by() - self.poly_extend_active \
-                              - contact.poly_contact.height
         self.route_input_gate(self.pmos1_inst,
                               self.nmos1_inst,
                               self.inputA_yoffset,
-                              "A")
+                              "A",
+                              position="center")
 
     def route_output(self):
         """ Route the Z output """
+        
+        # One routing track layer below the PMOS contacts
+        route_layer_offset = 0.5 * self.route_layer_width + self.route_layer_space
+        output_yoffset = self.pmos1_inst.get_pin("D").by() - route_layer_offset
+
 
         # PMOS1 drain
         pmos_pin = self.pmos1_inst.get_pin("D")
-        top_pin_offset = pmos_pin.center()
+        top_pin_offset = pmos_pin.bc()
         # NMOS2 drain
         nmos_pin = self.nmos2_inst.get_pin("D")
-        bottom_pin_offset = nmos_pin.center()
+        bottom_pin_offset = nmos_pin.uc()
         
         # Output pin
-        c_pin = self.get_pin("B")
-        out_offset = vector(c_pin.cx() + self.m1_pitch,
-                            self.inputA_yoffset)
+        out_offset = vector(nmos_pin.cx() + self.route_layer_pitch,
+                            output_yoffset)
 
         # This routes on M2
         # # Midpoints of the L routes go horizontal first then vertical
@@ -251,27 +236,22 @@ class pnand2(pgate.pgate):
         #               [top_pin_offset, mid1_offset, out_offset,
         #                mid2_offset, bottom_pin_offset])
 
-        # This routes on M1
+        # This routes on route_layer
         # Midpoints of the L routes goes vertical first then horizontal
-        mid1_offset = vector(top_pin_offset.x, out_offset.y)
-        # Midpoints of the L routes goes horizontal first then vertical
-        mid2_offset = vector(out_offset.x, bottom_pin_offset.y)
+        top_mid_offset = vector(top_pin_offset.x, out_offset.y)
+        # Top transistors
+        self.add_path(self.route_layer,
+                      [top_pin_offset, top_mid_offset, out_offset])
 
-        self.add_path("m1",
-                      [top_pin_offset, mid1_offset, out_offset])
-        # Route in two segments to have the width rule
-        self.add_path("m1",
-                      [bottom_pin_offset, mid2_offset + vector(0.5 * self.m1_width, 0)],
-                      width=nmos_pin.height())
-        self.add_path("m1",
-                      [mid2_offset, out_offset])
+        bottom_mid_offset = bottom_pin_offset + vector(0, self.route_layer_pitch)
+        # Bottom transistors
+        self.add_path(self.route_layer,
+                      [out_offset, bottom_mid_offset, bottom_pin_offset])
         
         # This extends the output to the edge of the cell
         self.add_layout_pin_rect_center(text="Z",
-                                        layer="m1",
-                                        offset=out_offset,
-                                        width=contact.m1_via.first_layer_width,
-                                        height=contact.m1_via.first_layer_height)
+                                        layer=self.route_layer,
+                                        offset=out_offset)
 
     def analytical_power(self, corner, load):
         """Returns dynamic and leakage power. Results in nW"""
diff --git a/compiler/pgates/pnand3.py b/compiler/pgates/pnand3.py
index 5f047fc4..454b75dd 100644
--- a/compiler/pgates/pnand3.py
+++ b/compiler/pgates/pnand3.py
@@ -5,7 +5,6 @@
 # (acting for and on behalf of Oklahoma State University)
 # All rights reserved.
 #
-import contact
 import pgate
 import debug
 from tech import drc, parameter, spice
@@ -14,6 +13,7 @@ import logical_effort
 from sram_factory import factory
 from globals import OPTS
 
+
 class pnand3(pgate.pgate):
     """
     This module generates gds of a parametrically sized 2-input nand.
@@ -74,68 +74,64 @@ class pnand3(pgate.pgate):
         
     def add_ptx(self):
         """ Create the PMOS and NMOS transistors. """
-        self.nmos_nsnd = factory.create(module_type="ptx",
+        self.nmos_center = factory.create(module_type="ptx",
+                                          width=self.nmos_width,
+                                          mults=self.tx_mults,
+                                          tx_type="nmos",
+                                          add_source_contact="active",
+                                          add_drain_contact="active")
+        self.add_mod(self.nmos_center)
+
+        self.nmos_right = factory.create(module_type="ptx",
+                                         width=self.nmos_width,
+                                         mults=self.tx_mults,
+                                         tx_type="nmos",
+                                         add_source_contact="active",
+                                         add_drain_contact=self.route_layer)
+        self.add_mod(self.nmos_right)
+
+        self.nmos_left = factory.create(module_type="ptx",
                                         width=self.nmos_width,
                                         mults=self.tx_mults,
                                         tx_type="nmos",
-                                        add_source_contact=False,
-                                        add_drain_contact=False,
-                                        connect_poly=True,
-                                        connect_active=True)
-        self.add_mod(self.nmos_nsnd)
-
-        self.nmos_ns = factory.create(module_type="ptx",
-                                      width=self.nmos_width,
-                                      mults=self.tx_mults,
-                                      tx_type="nmos",
-                                      add_source_contact=False,
-                                      connect_poly=True,
-                                      connect_active=True)
-        self.add_mod(self.nmos_ns)
-
-        self.nmos_nd = factory.create(module_type="ptx",
-                                      width=self.nmos_width,
-                                      mults=self.tx_mults,
-                                      tx_type="nmos",
-                                      add_drain_contact=False,
-                                      connect_poly=True,
-                                      connect_active=True)
-        self.add_mod(self.nmos_nd)
+                                        add_source_contact=self.route_layer,
+                                        add_drain_contact="active")
+        self.add_mod(self.nmos_left)
         
-        self.pmos = factory.create(module_type="ptx",
-                                   width=self.pmos_width,
-                                   mults=self.tx_mults,
-                                   tx_type="pmos",
-                                   connect_poly=True,
-                                   connect_active=True)
-        self.add_mod(self.pmos)
+        self.pmos_left = factory.create(module_type="ptx",
+                                        width=self.pmos_width,
+                                        mults=self.tx_mults,
+                                        tx_type="pmos",
+                                        add_source_contact=self.route_layer,
+                                        add_drain_contact=self.route_layer)
+        self.add_mod(self.pmos_left)
 
+        self.pmos_center = factory.create(module_type="ptx",
+                                          width=self.pmos_width,
+                                          mults=self.tx_mults,
+                                          tx_type="pmos",
+                                          add_source_contact=self.route_layer,
+                                          add_drain_contact=self.route_layer)
+        self.add_mod(self.pmos_center)
+
+        self.pmos_right = factory.create(module_type="ptx",
+                                         width=self.pmos_width,
+                                         mults=self.tx_mults,
+                                         tx_type="pmos",
+                                         add_source_contact=self.route_layer,
+                                         add_drain_contact=self.route_layer)
+        self.add_mod(self.pmos_right)
+        
     def setup_layout_constants(self):
         """ Pre-compute some handy layout parameters. """
         
         # Compute the overlap of the source and drain pins
-        overlap_xoffset = self.pmos.get_pin("D").ll().x - self.pmos.get_pin("S").ll().x
-        self.ptx_offset = vector(overlap_xoffset, 0)
+        self.ptx_offset = self.pmos_left.get_pin("D").center() - self.pmos_left.get_pin("S").center()
         
         # This is the extra space needed to ensure DRC rules
         # to the active contacts
         nmos = factory.create(module_type="ptx", tx_type="nmos")
         extra_contact_space = max(-nmos.get_pin("D").by(), 0)
-        # This is a poly-to-poly of a flipped cell
-        self.top_bottom_space = max(0.5 * self.m1_width + self.m1_space + extra_contact_space,
-                                    self.poly_extend_active + self.poly_space)
-        
-    def route_supply_rails(self):
-        """ Add vdd/gnd rails to the top and bottom. """
-        self.add_layout_pin_rect_center(text="gnd",
-                                        layer="m1",
-                                        offset=vector(0.5 * self.width, 0),
-                                        width=self.width)
-
-        self.add_layout_pin_rect_center(text="vdd",
-                                        layer="m1",
-                                        offset=vector(0.5 * self.width, self.height),
-                                        width=self.width)
 
     def create_ptx(self):
         """
@@ -143,27 +139,27 @@ class pnand3(pgate.pgate):
         """
 
         self.pmos1_inst = self.add_inst(name="pnand3_pmos1",
-                                        mod=self.pmos)
+                                        mod=self.pmos_left)
         self.connect_inst(["vdd", "A", "Z", "vdd"])
 
         self.pmos2_inst = self.add_inst(name="pnand3_pmos2",
-                                        mod=self.pmos)
+                                        mod=self.pmos_center)
         self.connect_inst(["Z", "B", "vdd", "vdd"])
 
         self.pmos3_inst = self.add_inst(name="pnand3_pmos3",
-                                        mod=self.pmos)
+                                        mod=self.pmos_right)
         self.connect_inst(["Z", "C", "vdd", "vdd"])
         
         self.nmos1_inst = self.add_inst(name="pnand3_nmos1",
-                                        mod=self.nmos_nd)
+                                        mod=self.nmos_left)
         self.connect_inst(["Z", "C", "net1", "gnd"])
 
         self.nmos2_inst = self.add_inst(name="pnand3_nmos2",
-                                        mod=self.nmos_nsnd)
+                                        mod=self.nmos_center)
         self.connect_inst(["net1", "B", "net2", "gnd"])
         
         self.nmos3_inst = self.add_inst(name="pnand3_nmos3",
-                                        mod=self.nmos_ns)
+                                        mod=self.nmos_right)
         self.connect_inst(["net2", "A", "gnd", "gnd"])
 
     def place_ptx(self):
@@ -172,8 +168,8 @@ class pnand3(pgate.pgate):
         and lowest position to provide maximum routing in channel
         """
 
-        pmos1_pos = vector(self.pmos.active_offset.x,
-                           self.height - self.pmos.active_height - self.top_bottom_space)
+        pmos1_pos = vector(self.pmos_left.active_offset.x,
+                           self.height - self.pmos_left.active_height - self.top_bottom_space)
         self.pmos1_inst.place(pmos1_pos)
 
         pmos2_pos = pmos1_pos + self.ptx_offset
@@ -182,7 +178,7 @@ class pnand3(pgate.pgate):
         self.pmos3_pos = pmos2_pos + self.ptx_offset
         self.pmos3_inst.place(self.pmos3_pos)
         
-        nmos1_pos = vector(self.pmos.active_offset.x,
+        nmos1_pos = vector(self.pmos_left.active_offset.x,
                            self.top_bottom_space)
         self.nmos1_inst.place(nmos1_pos)
 
@@ -195,9 +191,9 @@ class pnand3(pgate.pgate):
     def add_well_contacts(self):
         """ Add n/p well taps to the layout and connect to supplies """
 
-        self.add_nwell_contact(self.pmos,
+        self.add_nwell_contact(self.pmos_right,
                                self.pmos3_pos + vector(self.m1_pitch, 0))
-        self.add_pwell_contact(self.nmos_ns,
+        self.add_pwell_contact(self.nmos_right,
                                self.nmos3_pos + vector(self.m1_pitch, 0))
 
     def connect_rails(self):
@@ -212,37 +208,34 @@ class pnand3(pgate.pgate):
     def route_inputs(self):
         """ Route the A and B and C inputs """
 
-        m1_pitch = self.m1_space + contact.m1_via.first_layer_height
-        # Put B right on the well line
-        self.inputB_yoffset = self.nwell_y_offset
-        self.route_input_gate(self.pmos2_inst,
-                              self.nmos2_inst,
-                              self.inputB_yoffset,
-                              "B",
-                              position="center")
-        
-        # FIXME: constant hack
-        self.inputC_yoffset = self.inputB_yoffset - 1.15 * m1_pitch
-        self.route_input_gate(self.pmos3_inst,
-                              self.nmos3_inst,
-                              self.inputC_yoffset,
-                              "C",
-                              position="right")
+        pmos_drain_bottom = self.pmos1_inst.get_pin("D").by()
+        self.output_yoffset = pmos_drain_bottom - 0.5 * self.route_layer_width - self.route_layer_space
 
-        # FIXME: constant hack
-        if OPTS.tech_name == "s8":
-            self.inputA_yoffset = self.inputB_yoffset + 1.15 * m1_pitch
-        else:
-            self.inputA_yoffset = self.inputB_yoffset + 1.12 * m1_pitch
+        self.inputA_yoffset = self.output_yoffset - 0.5 * self.route_layer_width - self.route_layer_space
         self.route_input_gate(self.pmos1_inst,
                               self.nmos1_inst,
                               self.inputA_yoffset,
                               "A",
                               position="left")
+
+        # Put B right on the well line
+        self.inputB_yoffset = self.inputA_yoffset - self.m1_pitch
+        self.route_input_gate(self.pmos2_inst,
+                              self.nmos2_inst,
+                              self.inputB_yoffset,
+                              "B",
+                              position="center")
+
+        self.inputC_yoffset = self.inputB_yoffset - self.m1_pitch
+        self.route_input_gate(self.pmos3_inst,
+                              self.nmos3_inst,
+                              self.inputC_yoffset,
+                              "C",
+                              position="right")
         
     def route_output(self):
         """ Route the Z output """
-        
+
         # PMOS1 drain
         pmos1_pin = self.pmos1_inst.get_pin("D")
         # PMOS3 drain
@@ -250,14 +243,9 @@ class pnand3(pgate.pgate):
         # NMOS3 drain
         nmos3_pin = self.nmos3_inst.get_pin("D")
 
-        # midpoint for routing
-        mid_offset = vector(nmos3_pin.cx() + self.m1_pitch,
-                            self.inputA_yoffset)
+        out_offset = vector(nmos3_pin.cx() + self.route_layer_pitch,
+                            self.output_yoffset)
 
-        # Aligned with the well taps
-        out_offset = vector(self.nwell_contact.cx(),
-                            self.inputA_yoffset)
-        
         # Go up to metal2 for ease on all output pins
         # self.add_via_center(layers=self.m1_stack,
         #                     offset=pmos1_pin.center(),
@@ -282,26 +270,24 @@ class pnand3(pgate.pgate):
         bottom_pin_offset = nmos3_pin.center()
         
         # PMOS1 to output
-        self.add_path("m1", [top_left_pin_offset,
-                             vector(top_left_pin_offset.x, out_offset.y),
-                             out_offset])
+        self.add_path(self.route_layer, [top_left_pin_offset,
+                                           vector(top_left_pin_offset.x, out_offset.y),
+                                           out_offset])
         # PMOS3 to output
-        self.add_path("m1", [top_right_pin_offset,
-                             vector(top_right_pin_offset.x, mid_offset.y),
-                             mid_offset])
+        self.add_path(self.route_layer, [top_right_pin_offset,
+                                           vector(top_right_pin_offset.x, out_offset.y),
+                                           out_offset])
         # NMOS3 to output
-        mid2_offset = vector(mid_offset.x, bottom_pin_offset.y)
-        self.add_path("m1",
+        mid2_offset = vector(out_offset.x, bottom_pin_offset.y)
+        self.add_path(self.route_layer,
                       [bottom_pin_offset, mid2_offset],
                       width=nmos3_pin.height())
-        mid3_offset = vector(mid_offset.x, nmos3_pin.by())
-        self.add_path("m1", [mid3_offset, mid_offset])
+        mid3_offset = vector(out_offset.x, nmos3_pin.by())
+        self.add_path(self.route_layer, [mid3_offset, out_offset])
         
         self.add_layout_pin_rect_center(text="Z",
-                                        layer="m1",
-                                        offset=out_offset,
-                                        width=contact.m1_via.first_layer_width,
-                                        height=contact.m1_via.first_layer_height)
+                                        layer=self.route_layer,
+                                        offset=out_offset)
 
     def analytical_power(self, corner, load):
         """Returns dynamic and leakage power. Results in nW"""
diff --git a/compiler/pgates/pnor2.py b/compiler/pgates/pnor2.py
index 225a795c..f522578e 100644
--- a/compiler/pgates/pnor2.py
+++ b/compiler/pgates/pnor2.py
@@ -71,74 +71,54 @@ class pnor2(pgate.pgate):
 
     def add_ptx(self):
         """ Create the PMOS and NMOS transistors. """
-        self.nmos = factory.create(module_type="ptx",
-                                   width=self.nmos_width,
-                                   mults=self.tx_mults,
-                                   tx_type="nmos",
-                                   connect_poly=True,
-                                   connect_active=True)
-        self.add_mod(self.nmos)
+        self.nmos_left = factory.create(module_type="ptx",
+                                        width=self.nmos_width,
+                                        mults=self.tx_mults,
+                                        tx_type="nmos",
+                                        add_source_contact=self.route_layer,
+                                        add_drain_contact=self.route_layer)
+        self.add_mod(self.nmos_left)
 
-        self.pmos_nd = factory.create(module_type="ptx",
-                                      width=self.pmos_width,
-                                      mults=self.tx_mults,
-                                      tx_type="pmos",
-                                      add_drain_contact=False,
-                                      connect_poly=True,
-                                      connect_active=True)
-        self.add_mod(self.pmos_nd)
+        self.nmos_right = factory.create(module_type="ptx",
+                                         width=self.nmos_width,
+                                         mults=self.tx_mults,
+                                         tx_type="nmos",
+                                         add_source_contact=self.route_layer,
+                                         add_drain_contact=self.route_layer)
+        self.add_mod(self.nmos_right)
+        
+        self.pmos_left = factory.create(module_type="ptx",
+                                        width=self.pmos_width,
+                                        mults=self.tx_mults,
+                                        tx_type="pmos",
+                                        add_source_contact=self.route_layer,
+                                        add_drain_contact="active")
+        self.add_mod(self.pmos_left)
 
-        self.pmos_ns = factory.create(module_type="ptx",
-                                      width=self.pmos_width,
-                                      mults=self.tx_mults,
-                                      tx_type="pmos",
-                                      add_source_contact=False,
-                                      connect_poly=True,
-                                      connect_active=True)
-        self.add_mod(self.pmos_ns)
+        self.pmos_right = factory.create(module_type="ptx",
+                                         width=self.pmos_width,
+                                         mults=self.tx_mults,
+                                         tx_type="pmos",
+                                         add_source_contact="active",
+                                         add_drain_contact=self.route_layer)
+        self.add_mod(self.pmos_right)
         
     def setup_layout_constants(self):
         """ Pre-compute some handy layout parameters. """
 
-        # metal spacing to allow contacts on any layer
-        self.input_spacing = max(self.poly_space + contact.poly_contact.first_layer_width,
-                                 self.m1_space + contact.m1_via.first_layer_width,
-                                 self.m2_space + contact.m2_via.first_layer_width,
-                                 self.m3_space + contact.m2_via.second_layer_width)
-        
         # Compute the other pmos2 location, but determining
         # offset to overlap the source and drain pins
-        self.overlap_offset = self.pmos_ns.get_pin("D").ll() - self.pmos_nd.get_pin("S").ll()
+        self.overlap_offset = self.pmos_right.get_pin("D").center() - self.pmos_left.get_pin("S").center()
 
         # Two PMOS devices and a well contact. Separation between each.
         # Enclosure space on the sides.
-        self.width = 2 * self.pmos_ns.active_width \
-                     + self.pmos_ns.active_contact.width \
+        self.width = 2 * self.pmos_right.active_width \
+                     + self.pmos_right.active_contact.width \
                      + 2 * self.active_space \
                      + 0.5 * self.nwell_enclose_active
         self.well_width = self.width + 2 * self.nwell_enclose_active
         # Height is an input parameter, so it is not recomputed.
-
-        # This is the extra space needed to ensure DRC rules
-        # to the active contacts
-        extra_contact_space = max(-self.nmos.get_pin("D").by(), 0)
-        # This is a poly-to-poly of a flipped cell
-        self.top_bottom_space = max(0.5 * self.m1_width + self.m1_space + extra_contact_space,
-                                    self.poly_extend_active,
-                                    self.poly_space)
         
-    def route_supply_rails(self):
-        """ Add vdd/gnd rails to the top and bottom. """
-        self.add_layout_pin_rect_center(text="gnd",
-                                        layer="m1",
-                                        offset=vector(0.5 * self.width, 0),
-                                        width=self.width)
-
-        self.add_layout_pin_rect_center(text="vdd",
-                                        layer="m1",
-                                        offset=vector(0.5 * self.width, self.height),
-                                        width=self.width)
-
     def create_ptx(self):
         """
         Add PMOS and NMOS to the layout at the upper-most and lowest position
@@ -146,19 +126,19 @@ class pnor2(pgate.pgate):
         """
 
         self.pmos1_inst = self.add_inst(name="pnor2_pmos1",
-                                        mod=self.pmos_nd)
+                                        mod=self.pmos_left)
         self.connect_inst(["vdd", "A", "net1", "vdd"])
 
         self.pmos2_inst = self.add_inst(name="pnor2_pmos2",
-                                        mod=self.pmos_ns)
+                                        mod=self.pmos_right)
         self.connect_inst(["net1", "B", "Z", "vdd"])
 
         self.nmos1_inst = self.add_inst(name="pnor2_nmos1",
-                                        mod=self.nmos)
+                                        mod=self.nmos_left)
         self.connect_inst(["Z", "A", "gnd", "gnd"])
 
         self.nmos2_inst = self.add_inst(name="pnor2_nmos2",
-                                        mod=self.nmos)
+                                        mod=self.nmos_right)
         self.connect_inst(["Z", "B", "gnd", "gnd"])
         
     def place_ptx(self):
@@ -166,30 +146,35 @@ class pnor2(pgate.pgate):
         Add PMOS and NMOS to the layout at the upper-most and lowest position
         to provide maximum routing in channel
         """
+        # Some of the S/D contacts may extend beyond the active,
+        # but this needs to be done in the gate itself
+        contact_extend_active_space = max(-self.nmos_right.get_pin("D").by(), 0)
+        # Assume the contact starts at the active edge
+        contact_to_vdd_rail_space = 0.5 * self.m1_width + self.m1_space + contact_extend_active_space
+        # This is a poly-to-poly of a flipped cell
+        poly_to_poly_gate_space = self.poly_extend_active + self.poly_space
+        # Recompute this since it has a small txwith the added contact extend active spacing
+        self.top_bottom_space = max(contact_to_vdd_rail_space,
+                                    poly_to_poly_gate_space)
 
-        pmos1_pos = vector(self.pmos_ns.active_offset.x,
-                           self.height - self.pmos_ns.active_height \
-                           - self.top_bottom_space)
+        pmos1_pos = vector(self.pmos_right.active_offset.x,
+                           self.height - self.pmos_right.active_height - self.top_bottom_space)
         self.pmos1_inst.place(pmos1_pos)
 
         self.pmos2_pos = pmos1_pos + self.overlap_offset
         self.pmos2_inst.place(self.pmos2_pos)
         
-        nmos1_pos = vector(self.pmos_ns.active_offset.x, self.top_bottom_space)
+        nmos1_pos = vector(self.pmos_right.active_offset.x, self.top_bottom_space)
         self.nmos1_inst.place(nmos1_pos)
         
         self.nmos2_pos = nmos1_pos + self.overlap_offset
         self.nmos2_inst.place(self.nmos2_pos)
-        
-        # Output position will be in between the PMOS and NMOS
-        self.output_pos = vector(0,
-                                 0.5 * (pmos1_pos.y + nmos1_pos.y + self.nmos.active_height))
 
     def add_well_contacts(self):
         """ Add n/p well taps to the layout and connect to supplies """
 
-        self.add_nwell_contact(self.pmos_ns, self.pmos2_pos)
-        self.add_pwell_contact(self.nmos, self.nmos2_pos)
+        self.add_nwell_contact(self.pmos_right, self.pmos2_pos)
+        self.add_pwell_contact(self.nmos_right, self.nmos2_pos)
         
     def connect_rails(self):
         """ Connect the nmos and pmos to its respective power rails """
@@ -202,53 +187,53 @@ class pnor2(pgate.pgate):
 
     def route_inputs(self):
         """ Route the A and B inputs """
-        # Use M2 spaces so we can drop vias on the pins later!
-        inputB_yoffset = self.nmos2_inst.uy() + contact.poly_contact.height
+
+        # Top of NMOS drain
+        nmos_pin = self.nmos2_inst.get_pin("D")
+        bottom_pin_offset = nmos_pin.uy()
+        self.inputB_yoffset = bottom_pin_offset + self.m1_nonpref_pitch
+        self.inputA_yoffset = self.inputB_yoffset + self.m1_nonpref_pitch
+        
         self.route_input_gate(self.pmos2_inst,
                               self.nmos2_inst,
-                              inputB_yoffset,
+                              self.inputB_yoffset,
                               "B",
-                              position="center")
+                              position="right",
+                              directions=("V", "V"))
         
         # This will help with the wells and the input/output placement
-        self.inputA_yoffset = inputB_yoffset + self.input_spacing
         self.route_input_gate(self.pmos1_inst,
                               self.nmos1_inst,
                               self.inputA_yoffset,
-                              "A")
+                              "A",
+                              directions=("V", "V"))
+
+        self.output_yoffset = self.inputA_yoffset + self.m1_nonpref_pitch
 
     def route_output(self):
         """ Route the Z output """
-        # PMOS2 drain
+        # PMOS2 (right) drain
         pmos_pin = self.pmos2_inst.get_pin("D")
-        # NMOS1 drain
+        # NMOS1 (left) drain
         nmos_pin = self.nmos1_inst.get_pin("D")
-        # NMOS2 drain (for output via placement)
+        # NMOS2 (right) drain (for output via placement)
         nmos2_pin = self.nmos2_inst.get_pin("D")
 
         # Go up to metal2 for ease on all output pins
-        self.add_via_center(layers=self.m1_stack,
-                            offset=pmos_pin.center())
-        m1m2_contact = self.add_via_center(layers=self.m1_stack,
-                                           offset=nmos_pin.center())
-        
-        mid1_offset = vector(pmos_pin.center().x, nmos2_pin.center().y)
-        mid2_offset = vector(pmos_pin.center().x, self.inputA_yoffset)
-        mid3_offset = mid2_offset + vector(self.m2_width, 0)
+        # self.add_via_center(layers=self.m1_stack,
+        #                     offset=pmos_pin.center())
+        # m1m2_contact = self.add_via_center(layers=self.m1_stack,
+        #                                    offset=nmos_pin.center())
         
+        mid1_offset = vector(nmos_pin.center().x, self.output_yoffset)
+        mid2_offset = vector(pmos_pin.center().x, self.output_yoffset)
+
         # PMOS1 to mid-drain to NMOS2 drain
-        self.add_path("m2",
-                      [pmos_pin.center(), mid2_offset, mid3_offset])
-        self.add_path("m2",
-                      [nmos_pin.rc(), mid1_offset, mid2_offset])
-        # This extends the output to the edge of the cell
-        self.add_via_center(layers=self.m1_stack,
-                            offset=mid3_offset)
+        self.add_path(self.route_layer,
+                      [nmos_pin.center(), mid1_offset, mid2_offset, pmos_pin.center()])
         self.add_layout_pin_rect_center(text="Z",
-                                        layer="m1",
-                                        offset=mid3_offset,
-                                        width=contact.m1_via.first_layer_height,
-                                        height=contact.m1_via.first_layer_width)
+                                        layer=self.route_layer,
+                                        offset=mid2_offset)
 
     def analytical_power(self, corner, load):
         """Returns dynamic and leakage power. Results in nW"""
diff --git a/compiler/pgates/precharge.py b/compiler/pgates/precharge.py
index 6985efeb..d47e445f 100644
--- a/compiler/pgates/precharge.py
+++ b/compiler/pgates/precharge.py
@@ -81,7 +81,7 @@ class precharge(design.design):
         Initializes the upper and lower pmos
         """
         if(OPTS.tech_name == "s8"):
-            (self.ptx_width, self.ptx_mults) = pgate.bin_width(self, "pmos", self.ptx_width)
+            (self.ptx_width, self.ptx_mults) = pgate.bin_width("pmos", self.ptx_width)
         self.pmos = factory.create(module_type="ptx",
                                    width=self.ptx_width,
                                    mults=self.ptx_mults,
@@ -114,7 +114,7 @@ class precharge(design.design):
 
         self.add_power_pin("vdd",
                            self.well_contact_pos,
-                           vertical=True)
+                           directions=("V", "V"))
         
         # Hack for li layers
         if hasattr(self, "li_stack"):
@@ -159,7 +159,7 @@ class precharge(design.design):
         self.lower_pmos_inst.place(self.lower_pmos_position)
 
         # adds the upper pmos(s) to layout with 2 M2 tracks
-        ydiff = self.pmos.height + self.m2_pitch
+        ydiff = self.pmos.height + 2 * self.m2_pitch
         self.upper_pmos1_pos = self.lower_pmos_position + vector(0, ydiff)
         self.upper_pmos1_inst.place(self.upper_pmos1_pos)
 
diff --git a/compiler/pgates/ptristate_inv.py b/compiler/pgates/ptristate_inv.py
index 22f6b164..9fd5f8b6 100644
--- a/compiler/pgates/ptristate_inv.py
+++ b/compiler/pgates/ptristate_inv.py
@@ -79,9 +79,6 @@ class ptristate_inv(pgate.pgate):
         self.width = self.well_width + 0.5 * self.m1_space
         # Height is an input parameter, so it is not recomputed.
         
-        # Make sure we can put a well above and below
-        self.top_bottom_space = max(contact.active_contact.width, contact.active_contact.height)
-        
     def add_ptx(self):
         """ Create the PMOS and NMOS transistors. """
         self.nmos = factory.create(module_type="ptx",
diff --git a/compiler/pgates/ptx.py b/compiler/pgates/ptx.py
index b4b20381..c584e70b 100644
--- a/compiler/pgates/ptx.py
+++ b/compiler/pgates/ptx.py
@@ -14,7 +14,7 @@ import contact
 import logical_effort
 import os
 from globals import OPTS
-
+from pgate import pgate
 
 class ptx(design.design):
     """
@@ -23,32 +23,47 @@ class ptx(design.design):
     the transistor width. Mults is the number of transistors of the
     given width. Total width is therefore mults*width.  Options allow
     you to connect the fingered gates and active for parallel devices.
-
+    The add_*_contact option tells which layer to bring source/drain up to.
     """
     def __init__(self,
                  name="",
                  width=drc("minwidth_tx"),
                  mults=1,
                  tx_type="nmos",
-                 add_source_contact=True,
-                 add_drain_contact=True,
+                 add_source_contact=None,
+                 add_drain_contact=None,
                  series_devices=False,
-                 connect_active=False,
+                 connect_drain_active=False,
+                 connect_source_active=False,
                  connect_poly=False,
                  num_contacts=None):
+
+        if "li" in layer:
+            self.route_layer = "li"
+        else:
+            self.route_layer = "m1"
+
+        # Default contacts are the lowest layer
+        if not add_source_contact:
+            add_source_contact = self.route_layer
+
+        # Default contacts are the lowest layer
+        if not add_drain_contact:
+            add_drain_contact = self.route_layer
+
         # We need to keep unique names because outputting to GDSII
         # will use the last record with a given name. I.e., you will
         # over-write a design in GDS if one has and the other doesn't
         # have poly connected, for example.
         name = "{0}_m{1}_w{2:.3f}".format(tx_type, mults, width)
-        if not add_source_contact:
-            name += "_ns"
-        if not add_drain_contact:
-            name += "_nd"
+        name += "_s{}".format(add_source_contact)
+        name += "_d{}".format(add_drain_contact)
         if series_devices:
             name += "_sd"
-        if connect_active:
-            name += "_a"
+        if connect_drain_active:
+            name += "_da"
+        if connect_source_active:
+            name += "_sa"
         if connect_poly:
             name += "_p"
         if num_contacts:
@@ -61,13 +76,17 @@ class ptx(design.design):
         self.tx_type = tx_type
         self.mults = mults
         self.tx_width = width
-        self.connect_active = connect_active
+        self.connect_drain_active = connect_drain_active
+        self.connect_source_active = connect_source_active
         self.connect_poly = connect_poly
         self.add_source_contact = add_source_contact
         self.add_drain_contact = add_drain_contact
         self.series_devices = series_devices
         self.num_contacts = num_contacts
 
+        self.route_layer_width = drc("minwidth_{}".format(self.route_layer))
+        self.route_layer_space = drc("{0}_to_{0}".format(self.route_layer))
+        
         # Since it has variable height, it is not a pgate.
         self.create_netlist()
         # We must always create ptx layout for pbitcell
@@ -109,6 +128,7 @@ class ptx(design.design):
         perimeter_sd = 2 * self.poly_width + 2 * self.tx_width
         if OPTS.tech_name == "s8":
             # s8 technology is in microns
+            (self.tx_width, self.mults) = pgate.bin_width(self.tx_type, self.tx_width)
             main_str = "M{{0}} {{1}} {0} m={1} w={2} l={3} ".format(spice[self.tx_type],
                                                                       self.mults,
                                                                       self.tx_width,
@@ -265,55 +285,42 @@ class ptx(design.design):
                             width=poly_width,
                             height=self.poly_width)
 
-    def connect_fingered_active(self, drain_positions, source_positions):
+    def connect_fingered_active(self, positions, pin_name, top):
         """
         Connect each contact  up/down to a source or drain pin
         """
-        
+            
+        if len(positions) <= 1:
+            return
+
+        layer_space = getattr(self, "{}_space".format(self.route_layer))
+        layer_width = getattr(self, "{}_width".format(self.route_layer))
+
         # This is the distance that we must route up or down from the center
         # of the contacts to avoid DRC violations to the other contacts
         pin_offset = vector(0,
-                            0.5 * self.active_contact.second_layer_height + self.m1_space + 0.5 * self.m1_width)
+                            0.5 * self.active_contact.second_layer_height + layer_space + 0.5 * layer_width)
         # This is the width of a m1 extend the ends of the pin
-        end_offset = vector(self.m1_width / 2.0, 0)
+        end_offset = vector(layer_width / 2.0, 0)
 
-        # drains always go to the MIDDLE of the cell,
-        # so top of NMOS, bottom of PMOS
-        # so reverse the directions for NMOS compared to PMOS.
-        if self.tx_type == "pmos":
-            drain_dir = -1
-            source_dir = 1
+        # We move the opposite direction from the bottom
+        if not top:
+            offset = pin_offset.scale(-1, -1)
         else:
-            drain_dir = 1
-            source_dir = -1
-            
-        if len(source_positions) > 1:
-            source_offset = pin_offset.scale(source_dir, source_dir)
-            # remove the individual connections
-            self.remove_layout_pin("S")
-            # Add each vertical segment
-            for a in source_positions:
-                self.add_path(("m1"),
-                              [a, a + pin_offset.scale(source_dir,
-                                                       source_dir)])
-            # Add a single horizontal pin
-            self.add_layout_pin_segment_center(text="S",
-                                               layer="m1",
-                                               start=source_positions[0] + source_offset - end_offset,
-                                               end=source_positions[-1] + source_offset + end_offset)
+            offset = pin_offset
+
+        # remove the individual connections
+        self.remove_layout_pin(pin_name)
+        # Add each vertical segment
+        for a in positions:
+            self.add_path(self.route_layer,
+                          [a, a + offset])
+        # Add a single horizontal pin
+        self.add_layout_pin_segment_center(text=pin_name,
+                                           layer=self.route_layer,
+                                           start=positions[0] + offset - end_offset,
+                                           end=positions[-1] + offset + end_offset)
 
-        if len(drain_positions)>1:
-            drain_offset = pin_offset.scale(drain_dir,drain_dir)
-            self.remove_layout_pin("D") # remove the individual connections
-            # Add each vertical segment
-            for a in drain_positions:
-                self.add_path(("m1"), [a,a+drain_offset])
-            # Add a single horizontal pin
-            self.add_layout_pin_segment_center(text="D",
-                                               layer="m1",
-                                               start=drain_positions[0] + drain_offset - end_offset,
-                                               end=drain_positions[-1] + drain_offset + end_offset)
-            
     def add_poly(self):
         """
         Add the poly gates(s) and (optionally) connect them.
@@ -432,12 +439,12 @@ class ptx(design.design):
                     label = "S"
                     source_positions.append(pos)
                     
-                if (label=="S" and self.add_source_contact) or (label=="D" and self.add_drain_contact):
+                if (label=="S" and self.add_source_contact):
                     contact = self.add_diff_contact(label, pos)
-                    if label == "S":
-                        self.source_contacts.append(contact)
-                    else:
-                        self.drain_contacts.append(contact)
+                    self.source_contacts.append(contact)
+                elif (label=="D" and self.add_drain_contact):
+                    contact = self.add_diff_contact(label, pos)
+                    self.drain_contacts.append(contact)
                 else:
                     self.add_layout_pin_rect_center(text=label,
                                                     layer="active",
@@ -453,19 +460,22 @@ class ptx(design.design):
             label = "S"
             source_positions.append(pos)
 
-        if (label=="S" and self.add_source_contact) or (label=="D" and self.add_drain_contact):
+        if (label=="S" and self.add_source_contact):
             contact = self.add_diff_contact(label, pos)
-            if label == "S":
-                self.source_contacts.append(contact)
-            else:
-                self.drain_contacts.append(contact)
+            self.source_contacts.append(contact)
+        elif (label=="D" and self.add_drain_contact):
+            contact = self.add_diff_contact(label, pos)
+            self.drain_contacts.append(contact)
         else:
             self.add_layout_pin_rect_center(text=label,
                                             layer="active",
                                             offset=pos)
                 
-        if self.connect_active:
-            self.connect_fingered_active(drain_positions, source_positions)
+        if self.connect_source_active:
+            self.connect_fingered_active(source_positions, "S", top=(self.tx_type=="pmos"))
+
+        if self.connect_drain_active:
+            self.connect_fingered_active(drain_positions, "D", top=(self.tx_type=="nmos"))
             
     def get_stage_effort(self, cout):
         """Returns an object representing the parameters for delay in tau units."""
@@ -488,34 +498,39 @@ class ptx(design.design):
         return self.mults * self.tx_width / drc("minwidth_tx")
 
     def add_diff_contact(self, label, pos):
-        contact=self.add_via_center(layers=self.active_stack,
-                                    offset=pos,
-                                    size=(1, self.num_contacts),
-                                    directions=("V", "V"),
-                                    implant_type=self.implant_type,
-                                    well_type=self.well_type)
-        
-        if hasattr(self, "li_stack"):
-            contact=self.add_via_center(layers=self.li_stack,
-                                        offset=pos,
-                                        directions=("V", "V"))
 
-        # contact_area = contact.mod.second_layer_width * contact.mod.second_layer_height
-        # min_area = drc("minarea_m1")
-        # width = contact.mod.second_layer_width
-        # if contact_area < min_area:
-        #     height = min_area / width
-        # else:
-        #     height = contact.mod.second_layer_height
-        width = contact.mod.second_layer_width
-        height = contact.mod.second_layer_height
+        if label == "S":
+            layer = self.add_source_contact
+        elif label == "D":
+            layer = self.add_drain_contact
+        else:
+            debug.error("Invalid source drain name.")
+
+        if layer != "active":
+            via=self.add_via_stack_center(offset=pos,
+                                          from_layer="active",
+                                          to_layer=layer,
+                                          size=(1, self.num_contacts),
+                                          directions=("V", "V"),
+                                          implant_type=self.implant_type,
+                                          well_type=self.well_type)
+
+            pin_height = via.mod.second_layer_height
+            pin_width = via.mod.second_layer_width
+        else:
+            via = None
+            
+            pin_height = None
+            pin_width = None
+            
+        # Source drain vias are all vertical
         self.add_layout_pin_rect_center(text=label,
-                                        layer="m1",
+                                        layer=layer,
                                         offset=pos,
-                                        width=width,
-                                        height=height)
+                                        width=pin_width,
+                                        height=pin_height)
 
-        return(contact)
+        return(via)
         
     def get_cin(self):
         """Returns the relative gate cin of the tx"""
diff --git a/compiler/router/router_tech.py b/compiler/router/router_tech.py
index e0f277e9..27156eeb 100644
--- a/compiler/router/router_tech.py
+++ b/compiler/router/router_tech.py
@@ -18,17 +18,13 @@ class router_tech:
     """
     def __init__(self, layers, rail_track_width):
         """
-        Allows us to change the layers that we are routing on. First layer
-        is always horizontal, middle is via, and last is always
-        vertical.
+        Allows us to change the layers that we are routing on.
+        This uses the preferreed directions.
         """
         self.layers = layers
         self.rail_track_width = rail_track_width
 
         if len(self.layers) == 1:
-            if preferred_directions[self.layers[0]] != "H":
-                debug.warning("Using '{}' for horizontal routing, but it " \
-                              "prefers vertical routing".format(self.layers[0]))
             self.horiz_layer_name = self.vert_layer_name = self.layers[0]
             self.horiz_lpp = self.vert_lpp = layer[self.layers[0]]
             
@@ -42,13 +38,21 @@ class router_tech:
 
             # figure out wich of the two layers prefers horizontal/vertical
             # routing
-            if preferred_directions[try_horiz_layer] == "H" and preferred_directions[try_vert_layer] == "V":
+            self.horiz_layer_name = None
+            self.vert_layer_name = None
+            
+            if preferred_directions[try_horiz_layer] == "H":
                 self.horiz_layer_name = try_horiz_layer
+            else:
+                self.horiz_layer_name = try_vert_layer
+            if preferred_directions[try_vert_layer] == "V":
                 self.vert_layer_name = try_vert_layer
             else:
-                raise ValueError("Layer '{}' and '{}' are using the wrong " \
-                            "preferred_directions '{}' and '{}'. Only "\
-                            "('H', 'V') are supported")
+                self.vert_layer_name = try_horiz_layer
+
+            if not self.horiz_layer_name or not self.vert_layer_name:
+                raise ValueError("Layer '{}' and '{}' are using the wrong "
+                                 "preferred_directions '{}' and '{}'.")
 
             via_connect = contact(self.layers, (1, 1))
             max_via_size = max(via_connect.width,via_connect.height)
diff --git a/compiler/router/supply_grid_router.py b/compiler/router/supply_grid_router.py
index b28f875e..cd7b6b7b 100644
--- a/compiler/router/supply_grid_router.py
+++ b/compiler/router/supply_grid_router.py
@@ -5,21 +5,15 @@
 # (acting for and on behalf of Oklahoma State University)
 # All rights reserved.
 #
-import gdsMill
-import tech
-import math
 import debug
-from globals import OPTS,print_time
-from contact import contact
-from pin_group import pin_group
-from pin_layout import pin_layout
-from vector3d import vector3d 
+from globals import print_time
+from vector3d import vector3d
 from router import router
 from direction import direction
 from datetime import datetime
-import grid
 import grid_utils
 
+
 class supply_grid_router(router):
     """
     A router class to read an obstruction map from a gds and
@@ -44,14 +38,13 @@ class supply_grid_router(router):
         self.supply_rail_tracks = {}
 
         print_time("Init supply router", datetime.now(), start_time, 3)
-
         
     def create_routing_grid(self):
         """ 
         Create a sprase routing grid with A* expansion functions.
         """
         size = self.ur - self.ll
-        debug.info(1,"Size: {0} x {1}".format(size.x,size.y))
+        debug.info(1, "Size: {0} x {1}".format(size.x, size.y))
 
         import supply_grid
         self.rg = supply_grid.supply_grid(self.ll, self.ur, self.track_width)
@@ -60,12 +53,12 @@ class supply_grid_router(router):
         """
         Add power supply rails and connect all pins to these rails.
         """
-        debug.info(1,"Running supply router on {0} and {1}...".format(vdd_name, gnd_name))
+        debug.info(1, "Running supply router on {0} and {1}...".format(vdd_name, gnd_name))
         self.vdd_name = vdd_name
         self.gnd_name = gnd_name
 
         # Clear the pins if we have previously routed
-        if (hasattr(self,'rg')):
+        if (hasattr(self, 'rg')):
             self.clear_pins()
         else:
             # Creat a routing grid over the entire area
@@ -76,32 +69,32 @@ class supply_grid_router(router):
         # Get the pin shapes
         start_time = datetime.now()
         self.find_pins_and_blockages([self.vdd_name, self.gnd_name])
-        print_time("Finding pins and blockages",datetime.now(), start_time, 3)
+        print_time("Finding pins and blockages", datetime.now(), start_time, 3)
         # Add the supply rails in a mesh network and connect H/V with vias
         start_time = datetime.now()
         # Block everything
         self.prepare_blockages(self.gnd_name)
         # Determine the rail locations
-        self.route_supply_rails(self.gnd_name,0)
+        self.route_supply_rails(self.gnd_name, 0)
         
         # Block everything
         self.prepare_blockages(self.vdd_name)
         # Determine the rail locations
-        self.route_supply_rails(self.vdd_name,1)
-        print_time("Routing supply rails",datetime.now(), start_time, 3)
+        self.route_supply_rails(self.vdd_name, 1)
+        print_time("Routing supply rails", datetime.now(), start_time, 3)
         
         start_time = datetime.now()
         self.route_simple_overlaps(vdd_name)
         self.route_simple_overlaps(gnd_name)
-        print_time("Simple overlap routing",datetime.now(), start_time, 3)
+        print_time("Simple overlap routing", datetime.now(), start_time, 3)
         
         # Route the supply pins to the supply rails
         # Route vdd first since we want it to be shorter
         start_time = datetime.now()
         self.route_pins_to_rails(vdd_name)
         self.route_pins_to_rails(gnd_name)
-        print_time("Maze routing supplies",datetime.now(), start_time, 3)
-        #self.write_debug_gds("final.gds",False)  
+        print_time("Maze routing supplies", datetime.now(), start_time, 3)
+        # self.write_debug_gds("final.gds", False)
 
         # Did we route everything??
         if not self.check_all_routed(vdd_name):
@@ -111,9 +104,8 @@ class supply_grid_router(router):
         
         return True
 
-
     def check_all_routed(self, pin_name):
-        """ 
+        """
         Check that all pin groups are routed.
         """
         for pg in self.pin_groups[pin_name]:
@@ -125,7 +117,7 @@ class supply_grid_router(router):
         This checks for simple cases where a pin component already overlaps a supply rail.
         It will add an enclosure to ensure the overlap in wide DRC rule cases.
         """
-        debug.info(1,"Routing simple overlap pins for {0}".format(pin_name))
+        debug.info(1, "Routing simple overlap pins for {0}".format(pin_name))
 
         # These are the wire tracks
         wire_tracks = self.supply_rail_tracks[pin_name]
@@ -142,10 +134,10 @@ class supply_grid_router(router):
                 continue
             
             # Else, if we overlap some of the space track, we can patch it with an enclosure
-            #pg.create_simple_overlap_enclosure(pg.grids)
-            #pg.add_enclosure(self.cell)
+            # pg.create_simple_overlap_enclosure(pg.grids)
+            # pg.add_enclosure(self.cell)
 
-        debug.info(1,"Routed {} simple overlap pins".format(routed_count))
+        debug.info(1, "Routed {} simple overlap pins".format(routed_count))
     
     def finalize_supply_rails(self, name):
         """
@@ -158,7 +150,7 @@ class supply_grid_router(router):
 
         connections = set()
         via_areas = []
-        for i1,r1 in enumerate(all_rails):
+        for i1, r1 in enumerate(all_rails):
             # Only consider r1 horizontal rails
             e = next(iter(r1))
             if e.z==1:
@@ -166,9 +158,9 @@ class supply_grid_router(router):
 
             # We need to move this rail to the other layer for the z indices to match
             # during the intersection. This also makes a copy.
-            new_r1 = {vector3d(i.x,i.y,1) for i in r1}
+            new_r1 = {vector3d(i.x, i.y, 1) for i in r1}
 
-            for i2,r2 in enumerate(all_rails):
+            for i2, r2 in enumerate(all_rails):
                 # Never compare to yourself
                 if i1==i2:
                     continue
@@ -184,16 +176,16 @@ class supply_grid_router(router):
                 # the overlap area for placement of a via
                 overlap = new_r1 & r2
                 if len(overlap) >= 1:
-                    debug.info(3,"Via overlap {0} {1}".format(len(overlap),overlap))
-                    connections.update([i1,i2])
+                    debug.info(3, "Via overlap {0} {1}".format(len(overlap),overlap))
+                    connections.update([i1, i2])
                     via_areas.append(overlap)
                 
         # Go through and add the vias at the center of the intersection
         for area in via_areas:
             ll = grid_utils.get_lower_left(area)
             ur = grid_utils.get_upper_right(area)
-            center = (ll + ur).scale(0.5,0.5,0)
-            self.add_via(center,1)
+            center = (ll + ur).scale(0.5, 0.5, 0)
+            self.add_via(center, 1)
 
         # Determien which indices were not connected to anything above
         missing_indices = set([x for x in range(len(self.supply_rails[name]))])
@@ -204,13 +196,12 @@ class supply_grid_router(router):
         for rail_index in sorted(missing_indices, reverse=True):
             ll = grid_utils.get_lower_left(all_rails[rail_index])
             ur = grid_utils.get_upper_right(all_rails[rail_index])
-            debug.info(1,"Removing disconnected supply rail {0} .. {1}".format(ll,ur))
+            debug.info(1, "Removing disconnected supply rail {0} .. {1}".format(ll, ur))
             self.supply_rails[name].pop(rail_index)
 
         # Make the supply rails into a big giant set of grids for easy blockages.
         # Must be done after we determine which ones are connected.
         self.create_supply_track_set(name)
-        
             
     def add_supply_rails(self, name):
         """
@@ -223,7 +214,7 @@ class supply_grid_router(router):
             ur = grid_utils.get_upper_right(rail)        
             z = ll.z
             pin = self.compute_pin_enclosure(ll, ur, z, name)
-            debug.info(3,"Adding supply rail {0} {1}->{2} {3}".format(name,ll,ur,pin))
+            debug.info(3, "Adding supply rail {0} {1}->{2} {3}".format(name, ll, ur, pin))
             self.cell.add_layout_pin(text=name,
                                      layer=pin.layer,
                                      offset=pin.ll(),
@@ -243,19 +234,18 @@ class supply_grid_router(router):
         max_xoffset = self.rg.ur.x
         min_yoffset = self.rg.ll.y
         min_xoffset = self.rg.ll.x
-
         
         # Horizontal supply rails
         start_offset = min_yoffset + supply_number
         for offset in range(start_offset, max_yoffset, 2):
             # Seed the function at the location with the given width
-            wave = [vector3d(min_xoffset,offset,0)]
+            wave = [vector3d(min_xoffset, offset, 0)]
             # While we can keep expanding east in this horizontal track
             while wave and wave[0].x < max_xoffset:
                 added_rail = self.find_supply_rail(name, wave, direction.EAST)
                 if not added_rail:
                     # Just seed with the next one
-                    wave = [x+vector3d(1,0,0) for x in wave]
+                    wave = [x+vector3d(1, 0, 0) for x in wave]
                 else:
                     # Seed with the neighbor of the end of the last rail
                     wave = added_rail.neighbor(direction.EAST)
@@ -264,15 +254,15 @@ class supply_grid_router(router):
         start_offset = min_xoffset + supply_number
         for offset in range(start_offset, max_xoffset, 2):
             # Seed the function at the location with the given width
-            wave = [vector3d(offset,min_yoffset,1)]
+            wave = [vector3d(offset, min_yoffset, 1)]
             # While we can keep expanding north in this vertical track
             while wave and wave[0].y < max_yoffset:
                 added_rail = self.find_supply_rail(name, wave, direction.NORTH)
                 if not added_rail:
                     # Just seed with the next one
-                    wave = [x+vector3d(0,1,0) for x in wave]
+                    wave = [x + vector3d(0, 1, 0) for x in wave]
                 else:
-                    # Seed with the neighbor of the end of the last rail                    
+                    # Seed with the neighbor of the end of the last rail
                     wave = added_rail.neighbor(direction.NORTH)
 
     def find_supply_rail(self, name, seed_wave, direct):
@@ -294,7 +284,6 @@ class supply_grid_router(router):
         # Return the rail whether we approved it or not,
         # as it will be used to find the next start location
         return wave_path
-
     
     def probe_supply_rail(self, name, start_wave, direct):
         """
@@ -328,23 +317,19 @@ class supply_grid_router(router):
         data structure. Return whether it was added or not.
         """
         # We must have at least 2 tracks to drop plus 2 tracks for a via
-        if len(wave_path)>=4*self.rail_track_width:
+        if len(wave_path) >= 4 * self.rail_track_width:
             grid_set = wave_path.get_grids()
             self.supply_rails[name].append(grid_set)
             return True
         
         return False
 
-    
-
-                    
-                
     def route_supply_rails(self, name, supply_number):
         """
         Route the horizontal and vertical supply rails across the entire design.
         Must be done with lower left at 0,0
         """
-        debug.info(1,"Routing supply rail {0}.".format(name))
+        debug.info(1, "Routing supply rail {0}.".format(name))
 
         # Compute the grid locations of the supply rails
         self.compute_supply_rails(name, supply_number)
@@ -355,7 +340,6 @@ class supply_grid_router(router):
         # Add the rails themselves
         self.add_supply_rails(name)
 
-        
     def create_supply_track_set(self, pin_name):
         """
         Make a single set of all the tracks for the rail and wire itself.
@@ -364,24 +348,22 @@ class supply_grid_router(router):
         for rail in self.supply_rails[pin_name]:
             rail_set.update(rail)
         self.supply_rail_tracks[pin_name] = rail_set
-
-
         
     def route_pins_to_rails(self, pin_name):
         """
-        This will route each of the remaining pin components to the supply rails. 
+        This will route each of the remaining pin components to the supply rails.
         After it is done, the cells are added to the pin blockage list.
         """
 
         remaining_components = sum(not x.is_routed() for x in self.pin_groups[pin_name])
-        debug.info(1,"Maze routing {0} with {1} pin components to connect.".format(pin_name,
-                                                                                   remaining_components))
+        debug.info(1, "Maze routing {0} with {1} pin components to connect.".format(pin_name,
+                                                                                    remaining_components))
 
-        for index,pg in enumerate(self.pin_groups[pin_name]):
+        for index, pg in enumerate(self.pin_groups[pin_name]):
             if pg.is_routed():
                 continue
             
-            debug.info(3,"Routing component {0} {1}".format(pin_name, index))
+            debug.info(3, "Routing component {0} {1}".format(pin_name, index))
 
             # Clear everything in the routing grid.
             self.rg.reinit()
@@ -400,28 +382,26 @@ class supply_grid_router(router):
 
             # Actually run the A* router
             if not self.run_router(detour_scale=5):
-                self.write_debug_gds("debug_route.gds",False)
+                self.write_debug_gds("debug_route.gds", False)
                 
-            #if index==3 and pin_name=="vdd":
-            #    self.write_debug_gds("route.gds",False)
+            # if index==3 and pin_name=="vdd":
+            #     self.write_debug_gds("route.gds",False)
 
-    
     def add_supply_rail_target(self, pin_name):
         """
         Add the supply rails of given name as a routing target.
         """
-        debug.info(4,"Add supply rail target {}".format(pin_name))
+        debug.info(4, "Add supply rail target {}".format(pin_name))
         # Add the wire itself as the target
         self.rg.set_target(self.supply_rail_tracks[pin_name])
         # But unblock all the rail tracks including the space
-        self.rg.set_blocked(self.supply_rail_tracks[pin_name],False)
-
+        self.rg.set_blocked(self.supply_rail_tracks[pin_name], False)
                 
     def set_supply_rail_blocked(self, value=True):
         """
         Add the supply rails of given name as a routing target.
         """
-        debug.info(4,"Blocking supply rail")        
+        debug.info(4, "Blocking supply rail")
         for rail_name in self.supply_rail_tracks:
             self.rg.set_blocked(self.supply_rail_tracks[rail_name])
                 
diff --git a/compiler/sram/sram_1bank.py b/compiler/sram/sram_1bank.py
index 76e9f805..8dbcfaf4 100644
--- a/compiler/sram/sram_1bank.py
+++ b/compiler/sram/sram_1bank.py
@@ -59,14 +59,23 @@ class sram_1bank(sram_base):
         wmask_pos = [None] * len(self.all_ports)
         data_pos = [None] * len(self.all_ports)
 
+        # These positions utilize the channel route sizes.
+        # FIXME: Auto-compute these rather than manual computation.
+        # If a horizontal channel, they rely on the vertical channel non-preferred (contacted) pitch.
+        # If a vertical channel, they rely on the horizontal channel non-preferred (contacted) pitch.
+        # So, m3 non-pref pitch means that this is routed on the m2 layer.
         if self.write_size:
-            max_gap_size = self.m3_pitch * self.word_size + 2 * self.m1_pitch
-            max_gap_size_wmask = self.m2_pitch * max(self.num_wmasks + 1, self.col_addr_size + 1) + 2 * self.m1_pitch
+            self.data_bus_gap = self.m4_nonpref_pitch * 2
+            self.data_bus_size = self.m4_nonpref_pitch * (self.word_size) + self.data_bus_gap
+            self.wmask_bus_gap = self.m2_nonpref_pitch * 2
+            self.wmask_bus_size = self.m2_nonpref_pitch * (max(self.num_wmasks + 1, self.col_addr_size + 1)) + self.wmask_bus_gap
         else:
-            # This is M2 pitch even though it is on M1 to help stem via spacings on the trunk
-            # The M1 pitch is for supply rail spacings
-            max_gap_size = self.m2_pitch * max(self.word_size + 1,self.col_addr_size + 1) + 2 * self.m1_pitch
+            self.data_bus_gap = self.m3_nonpref_pitch * 2
+            self.data_bus_size = self.m3_nonpref_pitch * (max(self.word_size + 1, self.col_addr_size + 1)) + self.data_bus_gap
 
+        self.col_addr_bus_gap = self.m2_nonpref_pitch * 2
+        self.col_addr_bus_size = self.m2_nonpref_pitch * (self.col_addr_size) + self.col_addr_bus_gap
+        
         # Port 0
         port = 0
 
@@ -74,12 +83,12 @@ class sram_1bank(sram_base):
             if self.write_size:
                 # Add the write mask flops below the write mask AND array.
                 wmask_pos[port] = vector(self.bank.bank_array_ll.x,
-                                         -max_gap_size_wmask - self.dff.height)
+                                         - self.wmask_bus_size - self.dff.height)
                 self.wmask_dff_insts[port].place(wmask_pos[port])
 
                 # Add the data flops below the write mask flops.
                 data_pos[port] = vector(self.bank.bank_array_ll.x,
-                                        -max_gap_size - max_gap_size_wmask - 2 * self.dff.height)
+                                        - self.data_bus_size - self.wmask_bus_size - 2 * self.dff.height)
                 self.data_dff_insts[port].place(data_pos[port])
             else:
                 # Add the data flops below the bank to the right of the lower-left of bank array
@@ -89,7 +98,7 @@ class sram_1bank(sram_base):
                 # sense amps.
                 if port in self.write_ports:
                     data_pos[port] = vector(self.bank.bank_array_ll.x,
-                                            -max_gap_size - self.dff.height)
+                                            -self.data_bus_size - self.dff.height)
                     self.data_dff_insts[port].place(data_pos[port])
         else:
             wmask_pos[port] = vector(self.bank.bank_array_ll.x, 0)
@@ -99,10 +108,10 @@ class sram_1bank(sram_base):
         if self.col_addr_dff:
             if self.write_size:
                 col_addr_pos[port] = vector(self.bank.bank_array_ll.x - self.col_addr_dff_insts[port].width - self.bank.m2_gap,
-                                            -max_gap_size_wmask - self.col_addr_dff_insts[port].height)
+                                            -self.wmask_bus_size - self.col_addr_dff_insts[port].height)
             else:
                 col_addr_pos[port] = vector(self.bank.bank_array_ll.x - self.col_addr_dff_insts[port].width - self.bank.m2_gap,
-                                            -max_gap_size - self.col_addr_dff_insts[port].height)
+                                            -self.data_bus_size - self.col_addr_dff_insts[port].height)
             self.col_addr_dff_insts[port].place(col_addr_pos[port])
         else:
             col_addr_pos[port] = vector(self.bank.bank_array_ll.x, 0)
@@ -127,12 +136,12 @@ class sram_1bank(sram_base):
                 if self.write_size:
                     # Add the write mask flops below the write mask AND array.
                     wmask_pos[port] = vector(self.bank.bank_array_ur.x - self.wmask_dff_insts[port].width,
-                                             self.bank.height + max_gap_size_wmask + self.dff.height)
+                                             self.bank.height + self.wmask_bus_size + self.dff.height)
                     self.wmask_dff_insts[port].place(wmask_pos[port], mirror="MX")
 
                     # Add the data flops below the write mask flops
                     data_pos[port] = vector(self.bank.bank_array_ur.x - self.data_dff_insts[port].width,
-                                            self.bank.height + max_gap_size_wmask + max_gap_size + 2 * self.dff.height)
+                                            self.bank.height + self.wmask_bus_size + self.data_bus_size + 2 * self.dff.height)
                     self.data_dff_insts[port].place(data_pos[port], mirror="MX")
                 else:
                     # Add the data flops above the bank to the left of the upper-right of bank array
@@ -141,17 +150,17 @@ class sram_1bank(sram_base):
                     # These flops go below the sensing and leave a gap to channel route to the
                     # sense amps.
                     data_pos[port] = vector(self.bank.bank_array_ur.x - self.data_dff_insts[port].width,
-                                            self.bank.height + max_gap_size + self.dff.height)
+                                            self.bank.height + self.data_bus_size + self.dff.height)
                     self.data_dff_insts[port].place(data_pos[port], mirror="MX")
 
             # Add the col address flops above the bank to the right of the upper-right of bank array
             if self.col_addr_dff:
                 if self.write_size:
                     col_addr_pos[port] = vector(self.bank.bank_array_ur.x + self.bank.m2_gap,
-                                                self.bank.height + max_gap_size_wmask + self.dff.height)
+                                                self.bank.height + self.wmask_bus_size + self.dff.height)
                 else:
                     col_addr_pos[port] = vector(self.bank.bank_array_ur.x + self.bank.m2_gap,
-                                                    self.bank.height + max_gap_size + self.dff.height)
+                                                    self.bank.height + self.data_bus_size + self.dff.height)
                 self.col_addr_dff_insts[port].place(col_addr_pos[port], mirror="MX")
             else:
                 col_addr_pos[port] = self.bank_inst.ur()
@@ -245,7 +254,7 @@ class sram_1bank(sram_base):
             # This uses a metal2 track to the right (for port0) of the control/row addr DFF
             # to route vertically. For port1, it is to the left.
             row_addr_clk_pin = self.row_addr_dff_insts[port].get_pin("clk")
-            if port%2:
+            if port % 2:
                 control_clk_buf_pos = control_clk_buf_pin.lc()
                 row_addr_clk_pos = row_addr_clk_pin.lc()
                 mid1_pos = vector(self.row_addr_dff_insts[port].lx() - self.m2_pitch,
@@ -258,19 +267,20 @@ class sram_1bank(sram_base):
 
             # This is the steiner point where the net branches out
             clk_steiner_pos = vector(mid1_pos.x, control_clk_buf_pos.y)
-            self.add_path("m1", [control_clk_buf_pos, clk_steiner_pos])
-            self.add_via_center(layers=self.m1_stack,
-                                offset=clk_steiner_pos)
+            self.add_path(control_clk_buf_pin.layer, [control_clk_buf_pos, clk_steiner_pos])
+            self.add_via_stack_center(from_layer=control_clk_buf_pin.layer,
+                                      to_layer="m2",
+                                      offset=clk_steiner_pos)
             
             # Note, the via to the control logic is taken care of above
-            self.add_wire(("m3", "via2", "m2"),
+            self.add_wire(self.m2_stack[::-1],
                           [row_addr_clk_pos, mid1_pos, clk_steiner_pos])
         
             if self.col_addr_dff:
                 dff_clk_pin = self.col_addr_dff_insts[port].get_pin("clk")
                 dff_clk_pos = dff_clk_pin.center()
                 mid_pos = vector(clk_steiner_pos.x, dff_clk_pos.y)
-                self.add_wire(("m3", "via2", "m2"),
+                self.add_wire(self.m2_stack[::-1],
                               [dff_clk_pos, mid_pos, clk_steiner_pos])
 
             if port in self.write_ports:
@@ -282,7 +292,7 @@ class sram_1bank(sram_base):
                 self.add_path("m2",
                               [mid_pos, clk_steiner_pos],
                               width=max(m2_via.width, m2_via.height))
-                self.add_wire(("m3", "via2", "m2"),
+                self.add_wire(self.m2_stack[::-1],
                               [data_dff_clk_pos, mid_pos, clk_steiner_pos])
 
                 if self.write_size:
@@ -292,7 +302,7 @@ class sram_1bank(sram_base):
                     # In some designs, the steiner via will be too close to the mid_pos via
                     # so make the wire as wide as the contacts
                     self.add_path("m2", [mid_pos, clk_steiner_pos], width=max(m2_via.width, m2_via.height))
-                    self.add_wire(("m3", "via2", "m2"), [wmask_dff_clk_pos, mid_pos, clk_steiner_pos])
+                    self.add_wire(self.m2_stack[::-1], [wmask_dff_clk_pos, mid_pos, clk_steiner_pos])
             
     def route_control_logic(self):
         """ Route the control logic pins that are not inputs """
@@ -304,14 +314,13 @@ class sram_1bank(sram_base):
                     continue
                 src_pin = self.control_logic_insts[port].get_pin(signal)
                 dest_pin = self.bank_inst.get_pin(signal + "{}".format(port))
-                self.connect_vbus_m2m3(src_pin, dest_pin)
+                self.connect_vbus(src_pin, dest_pin)
 
         for port in self.all_ports:
             # Only input (besides pins) is the replica bitline
             src_pin = self.control_logic_insts[port].get_pin("rbl_bl")
             dest_pin = self.bank_inst.get_pin("rbl_bl{}".format(port))
-            
-            self.connect_hbus_m2m3(src_pin, dest_pin)
+            self.connect_hbus(src_pin, dest_pin)
             
     def route_row_addr_dff(self):
         """ Connect the output of the row flops to the bank pins """
@@ -324,33 +333,37 @@ class sram_1bank(sram_base):
                 flop_pos = flop_pin.center()
                 bank_pos = bank_pin.center()
                 mid_pos = vector(bank_pos.x, flop_pos.y)
-                self.add_wire(("m3", "via2", "m2"),
+                self.add_wire(self.m2_stack[::-1],
                               [flop_pos, mid_pos, bank_pos])
-                self.add_via_center(layers=self.m2_stack,
-                                    offset=flop_pos)
+                self.add_via_stack_center(from_layer=flop_pin.layer,
+                                          to_layer="m3",
+                                          offset=flop_pos)
 
     def route_col_addr_dff(self):
         """ Connect the output of the col flops to the bank pins """
         for port in self.all_ports:
-            if port%2:
-                offset = self.col_addr_dff_insts[port].ll() - vector(0, (self.col_addr_size + 2) * self.m1_pitch)
+            if port % 2:
+                offset = self.col_addr_dff_insts[port].ll() - vector(0, self.col_addr_bus_size)
             else:
-                offset = self.col_addr_dff_insts[port].ul() + vector(0, 2 * self.m1_pitch)
+                offset = self.col_addr_dff_insts[port].ul() + vector(0, self.col_addr_bus_gap)
 
             bus_names = ["addr_{}".format(x) for x in range(self.col_addr_size)]
             col_addr_bus_offsets = self.create_horizontal_bus(layer="m1",
-                                                              pitch=self.m1_pitch,
                                                               offset=offset,
                                                               names=bus_names,
                                                               length=self.col_addr_dff_insts[port].width)
 
             dff_names = ["dout_{}".format(x) for x in range(self.col_addr_size)]
             data_dff_map = zip(dff_names, bus_names)
-            self.connect_horizontal_bus(data_dff_map, self.col_addr_dff_insts[port], col_addr_bus_offsets)
+            self.connect_horizontal_bus(data_dff_map,
+                                        self.col_addr_dff_insts[port],
+                                        col_addr_bus_offsets)
             
             bank_names = ["addr{0}_{1}".format(port, x) for x in range(self.col_addr_size)]
             data_bank_map = zip(bank_names, bus_names)
-            self.connect_horizontal_bus(data_bank_map, self.bank_inst, col_addr_bus_offsets)
+            self.connect_horizontal_bus(data_bank_map,
+                                        self.bank_inst,
+                                        col_addr_bus_offsets)
 
     def route_data_dff(self):
         """ Connect the output of the data flops to the write driver """
@@ -358,48 +371,47 @@ class sram_1bank(sram_base):
         for port in self.write_ports:
             if self.write_size:
                 if port % 2:
-                    offset = self.data_dff_insts[port].ll() - vector(0, (self.word_size + 2) * self.m3_pitch)
+                    offset = self.data_dff_insts[port].ll() - vector(0, self.data_bus_size)
                 else:
-                    offset = self.data_dff_insts[port].ul() + vector(0, 2 * self.m3_pitch)
+                    offset = self.data_dff_insts[port].ul() + vector(0, self.data_bus_gap)
             else:
                 if port % 2:
-                    offset = self.data_dff_insts[port].ll() - vector(0, (self.word_size + 2) * self.m1_pitch)
+                    offset = self.data_dff_insts[port].ll() - vector(0, self.data_bus_size)
                 else:
-                    offset = self.data_dff_insts[port].ul() + vector(0, 2 * self.m1_pitch)
+                    offset = self.data_dff_insts[port].ul() + vector(0, self.data_bus_gap)
 
             dff_names = ["dout_{}".format(x) for x in range(self.word_size)]
             dff_pins = [self.data_dff_insts[port].get_pin(x) for x in dff_names]
             if self.write_size:
                 for x in dff_names:
-                    pin_offset = self.data_dff_insts[port].get_pin(x).center()
+                    pin = self.data_dff_insts[port].get_pin(x)
+                    pin_offset = pin.center()
                     self.add_via_center(layers=self.m1_stack,
                                         offset=pin_offset,
                                         directions=("V", "V"))
-                    self.add_via_center(layers=self.m2_stack,
-                                        offset=pin_offset)
-                    self.add_via_center(layers=self.m3_stack,
-                                        offset=pin_offset)
+                    self.add_via_stack_center(from_layer="m2",
+                                              to_layer="m4",
+                                              offset=pin_offset)
             
             bank_names = ["din{0}_{1}".format(port, x) for x in range(self.word_size)]
             bank_pins = [self.bank_inst.get_pin(x) for x in bank_names]
             if self.write_size:
                 for x in bank_names:
+                    pin = self.bank_inst.get_pin(x)
                     if port % 2:
-                        pin_offset = self.bank_inst.get_pin(x).uc()
+                        pin_offset = pin.uc()
                     else:
-                        pin_offset = self.bank_inst.get_pin(x).bc()
-                    self.add_via_center(layers=self.m1_stack,
-                                        offset=pin_offset)
-                    self.add_via_center(layers=self.m2_stack,
-                                        offset=pin_offset)
-                    self.add_via_center(layers=self.m3_stack,
-                                        offset=pin_offset)
+                        pin_offset = pin.bc()
+                    self.add_via_stack_center(from_layer=pin.layer,
+                                              to_layer="m4",
+                                              offset=pin_offset)
 
             route_map = list(zip(bank_pins, dff_pins))
             if self.write_size:
                 layer_stack = self.m3_stack
             else:
                 layer_stack = self.m1_stack
+                
             self.create_horizontal_channel_route(netlist=route_map,
                                                  offset=offset,
                                                  layer_stack=layer_stack)
@@ -409,9 +421,9 @@ class sram_1bank(sram_base):
         # This is where the channel will start (y-dimension at least)
         for port in self.write_ports:
             if port % 2:
-                offset = self.wmask_dff_insts[port].ll() - vector(0, (self.num_wmasks + 2) * self.m1_pitch)
+                offset = self.wmask_dff_insts[port].ll() - vector(0, self.wmask_bus_size)
             else:
-                offset = self.wmask_dff_insts[port].ul() + vector(0, 2 * self.m1_pitch)
+                offset = self.wmask_dff_insts[port].ul() + vector(0, self.wmask_bus_gap)
 
             dff_names = ["dout_{}".format(x) for x in range(self.num_wmasks)]
             dff_pins = [self.wmask_dff_insts[port].get_pin(x) for x in dff_names]
diff --git a/compiler/sram/sram_base.py b/compiler/sram/sram_base.py
index b9c4c909..bcb85308 100644
--- a/compiler/sram/sram_base.py
+++ b/compiler/sram/sram_base.py
@@ -7,6 +7,7 @@
 #
 import datetime
 import debug
+from math import log
 from importlib import reload
 from vector import vector
 from globals import OPTS, print_time
@@ -137,7 +138,6 @@ class sram_base(design, verilog, lef):
             self.copy_power_pins(inst, "vdd")
             self.copy_power_pins(inst, "gnd")
             
-        import tech
         if not OPTS.route_supplies:
             # Do not route the power supply (leave as must-connect pins)
             return
@@ -148,6 +148,7 @@ class sram_base(design, verilog, lef):
             grid_stack = power_grid
         except ImportError:
             # if no power_grid is specified by tech we use sensible defaults
+            import tech
             if "m4" in tech.layer:
                 # Route a M3/M4 grid
                 grid_stack = self.m3_stack
@@ -496,70 +497,6 @@ class sram_base(design, verilog, lef):
             self.connect_inst(temp)
         
         return insts
-
-    def connect_vbus_m2m3(self, src_pin, dest_pin):
-        """
-        Helper routine to connect an instance to a vertical bus.
-        Routes horizontal then vertical L shape.
-        Dest pin is assumed to be on M2.
-        Src pin can be on M1/M2/M3.
-        """
-        
-        if src_pin.cx()<dest_pin.cx():
-            in_pos = src_pin.rc()
-        else:
-            in_pos = src_pin.lc()
-        if src_pin.cy() < dest_pin.cy():
-            out_pos = dest_pin.bc()
-        else:
-            out_pos = dest_pin.uc()
-
-        # move horizontal first
-        self.add_wire(("m3", "via2", "m2"),
-                      [in_pos,
-                       vector(out_pos.x, in_pos.y),
-                       out_pos])
-        if src_pin.layer=="m1":
-            self.add_via_center(layers=self.m1_stack,
-                                offset=in_pos)
-        if src_pin.layer in ["m1", "m2"]:
-            self.add_via_center(layers=self.m2_stack,
-                                offset=in_pos)
-
-    def connect_hbus_m2m3(self, src_pin, dest_pin):
-        """
-        Helper routine to connect an instance to a horizontal bus.
-        Routes horizontal then vertical L shape.
-        Dest pin is on M1/M2/M3.
-        Src pin can be on M1/M2/M3.
-        """
-        
-        if src_pin.cx()<dest_pin.cx():
-            in_pos = src_pin.rc()
-        else:
-            in_pos = src_pin.lc()
-        if src_pin.cy() < dest_pin.cy():
-            out_pos = dest_pin.lc()
-        else:
-            out_pos = dest_pin.rc()
-
-        # move horizontal first
-        self.add_wire(("m3", "via2", "m2"),
-                      [in_pos,
-                       vector(out_pos.x, in_pos.y),
-                       out_pos])
-        if src_pin.layer=="m1":
-            self.add_via_center(layers=self.m1_stack,
-                                offset=in_pos)
-        if src_pin.layer in ["m1", "m2"]:
-            self.add_via_center(layers=self.m2_stack,
-                                offset=in_pos)
-        if dest_pin.layer=="m1":
-            self.add_via_center(layers=self.m1_stack,
-                                offset=out_pos)
-        if dest_pin.layer=="m3":
-            self.add_via_center(layers=self.m2_stack,
-                                offset=out_pos)
             
     def sp_write(self, sp_name, lvs_netlist=False):
         # Write the entire spice of the object to the file
diff --git a/compiler/tests/03_ptx_3finger_nmos_test.py b/compiler/tests/03_ptx_3finger_nmos_test.py
index 95700e8f..0b24ffd5 100755
--- a/compiler/tests/03_ptx_3finger_nmos_test.py
+++ b/compiler/tests/03_ptx_3finger_nmos_test.py
@@ -27,7 +27,8 @@ class ptx_3finger_nmos_test(openram_test):
                              width=tech.drc["minwidth_tx"],
                              mults=3,
                              tx_type="nmos",
-                             connect_active=True,
+                             connect_source_active=True,
+                             connect_drain_active=True,
                              connect_poly=True)
         self.local_drc_check(fet)
 
diff --git a/compiler/tests/03_ptx_3finger_pmos_test.py b/compiler/tests/03_ptx_3finger_pmos_test.py
index 8c09cf9b..ccb8b586 100755
--- a/compiler/tests/03_ptx_3finger_pmos_test.py
+++ b/compiler/tests/03_ptx_3finger_pmos_test.py
@@ -27,7 +27,8 @@ class ptx_3finger_pmos_test(openram_test):
                              width=tech.drc["minwidth_tx"],
                              mults=3,
                              tx_type="pmos",
-                             connect_active=True,
+                             connect_source_active=True,
+                             connect_drain_active=True,
                              connect_poly=True)
         self.local_drc_check(fet)
 
diff --git a/compiler/tests/03_ptx_4finger_nmos_test.py b/compiler/tests/03_ptx_4finger_nmos_test.py
index da9e438d..f7f3db78 100755
--- a/compiler/tests/03_ptx_4finger_nmos_test.py
+++ b/compiler/tests/03_ptx_4finger_nmos_test.py
@@ -23,11 +23,12 @@ class ptx_4finger_nmos_test(openram_test):
         import tech
 
         debug.info(2, "Checking three fingers NMOS")
-        fet = factory.create(module_type="ptx",        
+        fet = factory.create(module_type="ptx",
                              width= tech.drc["minwidth_tx"],
                              mults=4,
                              tx_type="nmos",
-                             connect_active=True,
+                             connect_source_active=True,
+                             connect_drain_active=True,
                              connect_poly=True)
         self.local_drc_check(fet)
 
diff --git a/compiler/tests/03_ptx_4finger_pmos_test.py b/compiler/tests/03_ptx_4finger_pmos_test.py
index 4fc60d5c..0bcdf909 100755
--- a/compiler/tests/03_ptx_4finger_pmos_test.py
+++ b/compiler/tests/03_ptx_4finger_pmos_test.py
@@ -23,11 +23,12 @@ class ptx_test(openram_test):
         import tech
 
         debug.info(2, "Checking three fingers PMOS")
-        fet = factory.create(module_type="ptx",        
+        fet = factory.create(module_type="ptx",
                              width=tech.drc["minwidth_tx"],
                              mults=4,
                              tx_type="pmos",
-                             connect_active=True,
+                             connect_source_active=True,
+                             connect_drain_active=True,
                              connect_poly=True)
         self.local_drc_check(fet)
 
diff --git a/compiler/tests/03_ptx_no_contacts_test.py b/compiler/tests/03_ptx_no_contacts_test.py
new file mode 100755
index 00000000..ff39da79
--- /dev/null
+++ b/compiler/tests/03_ptx_no_contacts_test.py
@@ -0,0 +1,60 @@
+#!/usr/bin/env python3
+# See LICENSE for licensing information.
+#
+# Copyright (c) 2016-2019 Regents of the University of California and The Board
+# of Regents for the Oklahoma Agricultural and Mechanical College
+# (acting for and on behalf of Oklahoma State University)
+# All rights reserved.
+#
+import unittest
+from testutils import *
+import sys,os
+sys.path.append(os.getenv("OPENRAM_HOME"))
+import globals
+from globals import OPTS
+from sram_factory import factory
+import debug
+
+
+class ptx_no_contacts_test(openram_test):
+
+    def runTest(self):
+        config_file = "{}/tests/configs/config".format(os.getenv("OPENRAM_HOME"))
+        globals.init_openram(config_file)
+        import tech
+
+        debug.info(2, "Checking single finger no source/drain")
+        fet = factory.create(module_type="ptx",
+                             width=tech.drc["minwidth_tx"],
+                             mults=1,
+                             add_source_contact=False,
+                             add_drain_contact=False,
+                             tx_type="nmos")
+        self.local_drc_check(fet)
+
+        debug.info(2, "Checking multifinger no source/drain")
+        fet = factory.create(module_type="ptx",
+                             width=tech.drc["minwidth_tx"],
+                             mults=4,
+                             add_source_contact=False,
+                             add_drain_contact=False,
+                             tx_type="nmos")
+        self.local_drc_check(fet)
+
+        debug.info(2, "Checking series ptx")
+        fet = factory.create(module_type="ptx",
+                             width=tech.drc["minwidth_tx"],
+                             mults=4,
+                             series_devices=True,
+                             tx_type="nmos")
+        self.local_drc_check(fet)
+        
+        globals.end_openram()
+
+
+# run the test from the command line
+if __name__ == "__main__":
+    (OPTS, args) = globals.parse_args()
+    del sys.argv[1:]
+    header(__file__, OPTS.tech_name)
+    unittest.main(testRunner=debugTestRunner())
diff --git a/compiler/tests/04_pinv_100x_test.py b/compiler/tests/04_pinv_100x_test.py
new file mode 100755
index 00000000..91b55d5d
--- /dev/null
+++ b/compiler/tests/04_pinv_100x_test.py
@@ -0,0 +1,36 @@
+#!/usr/bin/env python3
+# See LICENSE for licensing information.
+#
+# Copyright (c) 2016-2019 Regents of the University of California and The Board
+# of Regents for the Oklahoma Agricultural and Mechanical College
+# (acting for and on behalf of Oklahoma State University)
+# All rights reserved.
+#
+import unittest
+from testutils import *
+import sys,os
+sys.path.append(os.getenv("OPENRAM_HOME"))
+import globals
+from globals import OPTS
+from sram_factory import factory
+import debug
+
+class pinv_test(openram_test):
+
+    def runTest(self):
+        config_file = "{}/tests/configs/config".format(os.getenv("OPENRAM_HOME"))
+        globals.init_openram(config_file)
+
+        debug.info(2, "Checking 100x inverter")
+        tx = factory.create(module_type="pinv", size=100)
+        self.local_check(tx)
+
+        globals.end_openram()
+
+
+# run the test from the command line
+if __name__ == "__main__":
+    (OPTS, args) = globals.parse_args()
+    del sys.argv[1:]
+    header(__file__, OPTS.tech_name)
+    unittest.main(testRunner=debugTestRunner())
diff --git a/compiler/tests/04_pinv_1x_test.py b/compiler/tests/04_pinv_1x_test.py
index e6c0bff2..39704975 100755
--- a/compiler/tests/04_pinv_1x_test.py
+++ b/compiler/tests/04_pinv_1x_test.py
@@ -25,7 +25,7 @@ class pinv_test(openram_test):
         tx = factory.create(module_type="pinv", size=1)
         self.local_check(tx)
 
-        globals.end_openram()        
+        globals.end_openram()
 
 # run the test from the command line
 if __name__ == "__main__":
diff --git a/compiler/tests/06_hierarchical_decoder_test.py b/compiler/tests/06_hierarchical_decoder_test.py
index 7b77c6d9..07d995f9 100755
--- a/compiler/tests/06_hierarchical_decoder_test.py
+++ b/compiler/tests/06_hierarchical_decoder_test.py
@@ -21,35 +21,32 @@ class hierarchical_decoder_test(openram_test):
         config_file = "{}/tests/configs/config".format(os.getenv("OPENRAM_HOME"))
         globals.init_openram(config_file)
 
-        # check hierarchical decoder for single port
+        # Checks 2x4 and 2-input NAND decoder
         debug.info(1, "Testing 16 row sample for hierarchical_decoder")
         a = factory.create(module_type="hierarchical_decoder", num_outputs=16)
         self.local_check(a)
-
+        
+        # Checks 2x4 and 2-input NAND decoder with non-power-of-two
         debug.info(1, "Testing 17 row sample for hierarchical_decoder")
         a = factory.create(module_type="hierarchical_decoder", num_outputs=17)
         self.local_check(a)
 
-        debug.info(1, "Testing 23 row sample for hierarchical_decoder")
-        a = factory.create(module_type="hierarchical_decoder", num_outputs=23)
-        self.local_check(a)
-
+        # Checks 2x4 with 3x8 and 2-input NAND decoder
         debug.info(1, "Testing 32 row sample for hierarchical_decoder")
         a = factory.create(module_type="hierarchical_decoder", num_outputs=32)
         self.local_check(a)
 
-        debug.info(1, "Testing 65 row sample for hierarchical_decoder")
-        a = factory.create(module_type="hierarchical_decoder", num_outputs=65)
+        # Checks 3 x 2x4 and 3-input NAND decoder
+        debug.info(1, "Testing 64 row sample for hierarchical_decoder")
+        a = factory.create(module_type="hierarchical_decoder", num_outputs=64)
         self.local_check(a)
 
-        debug.info(1, "Testing 128 row sample for hierarchical_decoder")
-        a = factory.create(module_type="hierarchical_decoder", num_outputs=128)
-        self.local_check(a)
-
-        debug.info(1, "Testing 341 row sample for hierarchical_decoder")
-        a = factory.create(module_type="hierarchical_decoder", num_outputs=341)
+        # Checks 2x4 and 2 x 3x8 and 3-input NAND with non-power-of-two
+        debug.info(1, "Testing 132 row sample for hierarchical_decoder")
+        a = factory.create(module_type="hierarchical_decoder", num_outputs=132)
         self.local_check(a)
 
+        # Checks 3 x 3x8 and 3-input NAND decoder
         debug.info(1, "Testing 512 row sample for hierarchical_decoder")
         a = factory.create(module_type="hierarchical_decoder", num_outputs=512)
         self.local_check(a)
diff --git a/compiler/tests/16_control_logic_r_test.py b/compiler/tests/16_control_logic_r_test.py
new file mode 100755
index 00000000..b695f9c9
--- /dev/null
+++ b/compiler/tests/16_control_logic_r_test.py
@@ -0,0 +1,36 @@
+#!/usr/bin/env python3
+# See LICENSE for licensing information.
+#
+# Copyright (c) 2016-2019 Regents of the University of California and The Board
+# of Regents for the Oklahoma Agricultural and Mechanical College
+# (acting for and on behalf of Oklahoma State University)
+# All rights reserved.
+#
+import unittest
+from testutils import *
+import sys, os
+sys.path.append(os.getenv("OPENRAM_HOME"))
+import globals
+from globals import OPTS
+from sram_factory import factory
+import debug
+
+
+class control_logic_test(openram_test):
+
+    def runTest(self):
+        config_file = "{}/tests/configs/config".format(os.getenv("OPENRAM_HOME"))
+        globals.init_openram(config_file)
+
+        debug.info(1, "Testing sample for control_logic_r")
+        a = factory.create(module_type="control_logic", num_rows=128, words_per_row=1, word_size=32, port_type="r")
+        self.local_check(a)
+
+        globals.end_openram()
+        
+# run the test from the command line
+if __name__ == "__main__":
+    (OPTS, args) = globals.parse_args()
+    del sys.argv[1:]
+    header(__file__, OPTS.tech_name)
+    unittest.main(testRunner=debugTestRunner())
diff --git a/compiler/tests/16_control_logic_rw_test.py b/compiler/tests/16_control_logic_rw_test.py
new file mode 100755
index 00000000..8d12085f
--- /dev/null
+++ b/compiler/tests/16_control_logic_rw_test.py
@@ -0,0 +1,36 @@
+#!/usr/bin/env python3
+# See LICENSE for licensing information.
+#
+# Copyright (c) 2016-2019 Regents of the University of California and The Board
+# of Regents for the Oklahoma Agricultural and Mechanical College
+# (acting for and on behalf of Oklahoma State University)
+# All rights reserved.
+#
+import unittest
+from testutils import *
+import sys, os
+sys.path.append(os.getenv("OPENRAM_HOME"))
+import globals
+from globals import OPTS
+from sram_factory import factory
+import debug
+
+
+class control_logic_test(openram_test):
+
+    def runTest(self):
+        config_file = "{}/tests/configs/config".format(os.getenv("OPENRAM_HOME"))
+        globals.init_openram(config_file)
+
+        debug.info(1, "Testing sample for control_logic_rw")
+        a = factory.create(module_type="control_logic", num_rows=128, words_per_row=1, word_size=32)
+        self.local_check(a)
+        
+        globals.end_openram()
+        
+# run the test from the command line
+if __name__ == "__main__":
+    (OPTS, args) = globals.parse_args()
+    del sys.argv[1:]
+    header(__file__, OPTS.tech_name)
+    unittest.main(testRunner=debugTestRunner())
diff --git a/compiler/tests/16_control_logic_test.py b/compiler/tests/16_control_logic_w_test.py
similarity index 71%
rename from compiler/tests/16_control_logic_test.py
rename to compiler/tests/16_control_logic_w_test.py
index ecc6008a..8407d1e8 100755
--- a/compiler/tests/16_control_logic_test.py
+++ b/compiler/tests/16_control_logic_w_test.py
@@ -20,20 +20,12 @@ class control_logic_test(openram_test):
     def runTest(self):
         config_file = "{}/tests/configs/config".format(os.getenv("OPENRAM_HOME"))
         globals.init_openram(config_file)
-        import control_logic
-        import tech
-
-        debug.info(1, "Testing sample for control_logic_rw")
-        a = factory.create(module_type="control_logic", num_rows=128, words_per_row=1, word_size=32)
-        self.local_check(a)
-
-        debug.info(1, "Testing sample for control_logic_r")
-        a = factory.create(module_type="control_logic", num_rows=128, words_per_row=1, word_size=32, port_type="r")
-        self.local_check(a)
 
         debug.info(1, "Testing sample for control_logic_w")
         a = factory.create(module_type="control_logic", num_rows=128, words_per_row=1, word_size=32, port_type="w")
         self.local_check(a)
+
+        globals.end_openram()
         
 # run the test from the command line
 if __name__ == "__main__":
diff --git a/compiler/tests/19_single_bank_test.py b/compiler/tests/19_single_bank_test.py
index 6cff481b..38a847a9 100755
--- a/compiler/tests/19_single_bank_test.py
+++ b/compiler/tests/19_single_bank_test.py
@@ -25,28 +25,28 @@ class single_bank_test(openram_test):
         c = sram_config(word_size=4,
                         num_words=16)
 
-        c.words_per_row=1
-        factory.reset()
-        c.recompute_sizes()
-        debug.info(1, "No column mux")
-        a = factory.create("bank", sram_config=c)
-        self.local_check(a)
+        # c.words_per_row=1
+        # factory.reset()
+        # c.recompute_sizes()
+        # debug.info(1, "No column mux")
+        # a = factory.create("bank", sram_config=c)
+        # self.local_check(a)
 
-        c.num_words=32
-        c.words_per_row=2
-        factory.reset()
-        c.recompute_sizes()
-        debug.info(1, "Two way column mux")
-        a = factory.create("bank", sram_config=c)
-        self.local_check(a)
+        # c.num_words=32
+        # c.words_per_row=2
+        # factory.reset()
+        # c.recompute_sizes()
+        # debug.info(1, "Two way column mux")
+        # a = factory.create("bank", sram_config=c)
+        # self.local_check(a)
 
-        c.num_words=64
-        c.words_per_row=4
-        factory.reset()
-        c.recompute_sizes()
-        debug.info(1, "Four way column mux")
-        a = factory.create("bank", sram_config=c)
-        self.local_check(a)
+        # c.num_words=64
+        # c.words_per_row=4
+        # factory.reset()
+        # c.recompute_sizes()
+        # debug.info(1, "Four way column mux")
+        # a = factory.create("bank", sram_config=c)
+        # self.local_check(a)
 
         c.word_size=2
         c.num_words=128
diff --git a/compiler/tests/21_ngspice_delay_test.py b/compiler/tests/21_ngspice_delay_test.py
index c944b3ce..92b8eac0 100755
--- a/compiler/tests/21_ngspice_delay_test.py
+++ b/compiler/tests/21_ngspice_delay_test.py
@@ -70,20 +70,20 @@ class timing_sram_test(openram_test):
                            'min_period': 0.898,
                            'disabled_write1_power': [0.201411]}
         elif OPTS.tech_name == "scn4m_subm":
-            golden_data = {'read1_power': [12.11658],
-                           'write1_power': [10.52653],
-                           'read0_power': [11.956710000000001],
-                           'disabled_write0_power': [7.673665],
-                           'disabled_write1_power': [7.981922000000001],
-                           'slew_lh': [1.868836],
-                           'slew_hl': [1.868836],
-                           'delay_hl': [1.8598510000000001],
-                           'delay_lh': [1.8598510000000001],
-                           'leakage_power': 0.008613619,
-                           'disabled_read0_power': [5.904712],
-                           'min_period': 6.875,
-                           'disabled_read1_power': [7.132159],
-                           'write0_power': [13.406400000000001]}
+            golden_data =  {'delay_hl': [1.8435739999999998],
+                            'delay_lh': [1.8435739999999998],
+                            'disabled_read0_power': [5.917947],
+                            'disabled_read1_power': [7.154297],
+                            'disabled_write0_power': [7.696351],
+                            'disabled_write1_power': [7.999409000000001],
+                            'leakage_power': 0.004809726,
+                            'min_period': 6.875,
+                            'read0_power': [11.833079999999999],
+                            'read1_power': [11.99236],
+                            'slew_hl': [1.8668490000000002],
+                            'slew_lh': [1.8668490000000002],
+                            'write0_power': [13.287510000000001],
+                            'write1_power': [10.416369999999999]}
         else:
             self.assertTrue(False) # other techs fail
 
diff --git a/technology/freepdk45/tech/tech.py b/technology/freepdk45/tech/tech.py
index c2bf135f..c0379e44 100644
--- a/technology/freepdk45/tech/tech.py
+++ b/technology/freepdk45/tech/tech.py
@@ -32,7 +32,7 @@ tech_modules = module_type()
 cell_properties = cell_properties()
 cell_properties.bitcell.mirror.x = True
 cell_properties.bitcell.mirror.y = False
-
+cell_properties.bitcell_power_pin_directions = ("V", "V")
 
 ###################################################
 # GDS file info
@@ -63,6 +63,13 @@ m1_stack = ("m1", "via1", "m2")
 m2_stack = ("m2", "via2", "m3")
 m3_stack = ("m3", "via3", "m4")
 
+layer_indices = {"poly": 0,
+                 "active": 0,
+                 "m1": 1,
+                 "m2": 2,
+                 "m3": 3,
+                 "m4": 4}
+
 # The FEOL stacks get us up to m1
 feol_stacks = [poly_stack,
                active_stack]
diff --git a/technology/scn4m_subm/tech/tech.py b/technology/scn4m_subm/tech/tech.py
index d8384d7f..745b381e 100644
--- a/technology/scn4m_subm/tech/tech.py
+++ b/technology/scn4m_subm/tech/tech.py
@@ -60,6 +60,13 @@ m1_stack = ("m1", "via1", "m2")
 m2_stack = ("m2", "via2", "m3")
 m3_stack = ("m3", "via3", "m4")
 
+layer_indices = {"poly": 0,
+                 "active": 0,
+                 "m1": 1,
+                 "m2": 2,
+                 "m3": 3,
+                 "m4": 4}
+
 # The FEOL stacks get us up to m1
 feol_stacks = [poly_stack,
                active_stack]