diff --git a/compiler/base/channel_route.py b/compiler/base/channel_route.py
index cd4bbdc1..00ad6e37 100644
--- a/compiler/base/channel_route.py
+++ b/compiler/base/channel_route.py
@@ -12,6 +12,69 @@ from vector import vector
 import design
 
 
+class channel_net():
+    def __init__(self, net_name, pins, vertical):
+        self.name = net_name
+        self.pins = pins
+        self.vertical = vertical
+        
+        # Keep track of the internval
+        if vertical:
+            self.min_value = min(i.by() for i in pins)
+            self.max_value = max(i.uy() for i in pins)
+        else:
+            self.min_value = min(i.lx() for i in pins)
+            self.max_value = max(i.rx() for i in pins)
+            
+        # Keep track of the conflicts
+        self.conflicts = []
+
+    def __str__(self):
+        return self.name
+    
+    def __repr__(self):
+        return self.name
+    
+    def __lt__(self, other):
+        return self.min_value < other.min_value
+
+    def pin_overlap(self, pin1, pin2, pitch):
+        """ Check for vertical or horizontal overlap of the two pins """
+        
+        # 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.
+        
+        # Pin 1 must be in the "BOTTOM" set
+        x_overlap = pin1.by() < pin2.by() and abs(pin1.center().x - pin2.center().x) < pitch
+        
+        # Pin 1 must be in the "LEFT" set
+        y_overlap = pin1.lx() < pin2.lx() and abs(pin1.center().y - pin2.center().y) < pitch
+        overlaps = (not self.vertical and x_overlap) or (self.vertical and y_overlap)
+        return overlaps
+        
+    def pins_overlap(self, other, pitch):
+        """
+        Check all the pin pairs on two nets and return a pin
+        overlap if any pin overlaps.
+        """
+
+        for pin1 in self.pins:
+            for pin2 in other.pins:
+                if self.pin_overlap(pin1, pin2, pitch):
+                    return True
+
+        return False
+
+    def segment_overlap(self, other):
+        """
+        Check if the horizontal span of the two nets overlaps eachother.
+        """
+        min_overlap = self.min_value >= other.min_value and self.min_value <= other.max_value
+        max_overlap = self.max_value >= other.min_value and self.max_value <= other.max_value
+        return min_overlap or max_overlap
+        
+        
 class channel_route(design.design):
 
     unique_id = 0
@@ -21,7 +84,8 @@ class channel_route(design.design):
                  offset,
                  layer_stack,
                  directions=None,
-                 vertical=False):
+                 vertical=False,
+                 parent=None):
         """
         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
@@ -40,6 +104,8 @@ class channel_route(design.design):
         self.layer_stack = layer_stack
         self.directions = directions
         self.vertical = vertical
+        # For debugging...
+        self.parent = parent
         
         if not directions or directions == "pref":
             # Use the preferred layer directions
@@ -86,114 +152,139 @@ class channel_route(design.design):
         # FIXME: This is O(n^2), so maybe optimize it.
         for other_pin, conflicts in g.items():
             if pin in conflicts:
-                conflicts.remove(pin)
-                g[other_pin]=conflicts
+                g[other_pin].remove(pin)
         return g
+            
+    def route(self):
+        # Create names for the nets for the graphs
+        nets = []
+        index = 0
+        # print(self.netlist)
+        for pin_list in self.netlist:
+            nets.append(channel_net("n{}".format(index), pin_list, self.vertical))
+            index += 1
 
-    def vcg_nets_overlap(self, net1, net2):
-        """
-        Check all the pin pairs on two nets and return a pin
-        overlap if any pin overlaps.
-        """
+        # Create the (undirected) horizontal constraint graph
+        hcg = collections.OrderedDict()
+        for net1 in nets:
+            for net2 in nets:
+                if net1.name == net2.name:
+                    continue
+                if net1.segment_overlap(net2):
+                    try:
+                        hcg[net1.name].add(net2.name)
+                    except KeyError:
+                        hcg[net1.name] = set([net2.name])
+                    try:
+                        hcg[net2.name].add(net1.name)
+                    except KeyError:
+                        hcg[net2.name] = set([net1.name])
 
+            
+        # Initialize the vertical conflict graph (vcg)
+        # and make a list of all pins
+        vcg = collections.OrderedDict()
+
+        # 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
         if self.vertical:
             pitch = self.horizontal_nonpref_pitch
         else:
             pitch = self.vertical_nonpref_pitch
 
-        for pin1 in net1:
-            for pin2 in net2:
-                if self.vcg_pin_overlap(pin1, pin2, pitch):
-                    return True
+        for net in nets:
+            vcg[net.name] = set()
 
-        return False
-            
-    def route(self):
-        # FIXME: Must extend this to a horizontal conflict graph
-        # too if we want to minimize the
-        # number of tracks!
-        # hcg = {}
-
-        # Initialize the vertical conflict graph (vcg)
-        # and make a list of all pins
-        vcg = collections.OrderedDict()
-
-        # Create names for the nets for the graphs
-        nets = collections.OrderedDict()
-        index = 0
-        # print(netlist)
-        for pin_list in self.netlist:
-            net_name = "n{}".format(index)
-            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:
-            if net_name1 not in vcg.keys():
-                vcg[net_name1] = []
-            for net_name2 in nets:
-                if net_name2 not in vcg.keys():
-                    vcg[net_name2] = []
+        for net1 in nets:
+            for net2 in nets:
                 # Skip yourself
-                if net_name1 == net_name2:
+                if net1.name == net2.name:
                     continue
-                if self.vcg_nets_overlap(nets[net_name1],
-                                         nets[net_name2]):
-                    vcg[net_name2].append(net_name1)
+                
+                if net1.pins_overlap(net2, pitch):
+                    vcg[net2.name].add(net1.name)
 
-        current_offset = self.offset
+        # Check if there are any cycles net1 <---> net2 in the VCG
         
-        # list of routes to do
-        while vcg:
+
+        # Some of the pins may be to the left/below the channel offset,
+        # so adjust if this is the case
+        self.min_value = min([n.min_value for n in nets])
+        self.max_value = min([n.max_value for n in nets])
+        if self.vertical:
+            real_channel_offset = vector(self.offset.x, min(self.min_value, self.offset.y))
+        else:
+            real_channel_offset = vector(min(self.min_value, self.offset.x), self.offset.y)
+        current_offset = real_channel_offset
+
+        # Sort nets by left edge value
+        nets.sort()
+        while len(nets) > 0:
+
+            current_offset_value = current_offset.y if self.vertical else current_offset.x
+            
             # from pprint import pformat
             # print("VCG:\n", pformat(vcg))
+            # for name,net in vcg.items():
+            #    print(name, net.min_value, net.max_value, net.conflicts)
+            # print(current_offset)
             # get a route from conflict graph with empty fanout set
-            net_name = None
-            for net_name, conflicts in vcg.items():
-                if len(conflicts) == 0:
-                    vcg = self.remove_net_from_graph(net_name, vcg)
+            for net in nets:
+                # If it has no conflicts and the interval is to the right of the current offset in the track
+                if net.min_value >= current_offset_value and len(vcg[net.name]) == 0:
+                    # print("Routing {}".format(net.name))
+                    # Add the trunk routes from the bottom up for
+                    # horizontal or the left to right for vertical
+                    if self.vertical:
+                        self.add_vertical_trunk_route(net.pins,
+                                                      current_offset,
+                                                      self.vertical_nonpref_pitch)
+                        current_offset = vector(current_offset.x, net.max_value + self.horizontal_nonpref_pitch)
+                    else:
+                        self.add_horizontal_trunk_route(net.pins,
+                                                        current_offset,
+                                                        self.horizontal_nonpref_pitch)
+                        current_offset = vector(net.max_value + self.vertical_nonpref_pitch, current_offset.y)
+
+                    # Remove the net from other constriants in the VCG
+                    vcg = self.remove_net_from_graph(net.name, vcg)
+                    nets.remove(net)
+                    
                     break
             else:
-                # FIXME: We don't support cyclic VCGs right now.
-                debug.error("Cyclic VCG in channel router.", -1)
-                
-            # These are the pins we'll have to connect
-            pin_list = nets[net_name]
-            # print("Routing:", net_name, [x.name for x in pin_list])
-
-            # Remove the net from other constriants in the VCG
-            vcg = self.remove_net_from_graph(net_name, vcg)
-
-            # Add the trunk routes from the bottom up for
-            # horizontal or the left to right for vertical
-            if self.vertical:
-                self.add_vertical_trunk_route(pin_list,
-                                              current_offset,
-                                              self.vertical_nonpref_pitch)
-                # This accounts for the via-to-via spacings
-                current_offset += vector(self.horizontal_nonpref_pitch, 0)
-            else:
-                self.add_horizontal_trunk_route(pin_list,
-                                                current_offset,
-                                                self.horizontal_nonpref_pitch)
-                # This accounts for the via-to-via spacings
-                current_offset += vector(0, self.vertical_nonpref_pitch)
+                # If we made a full pass and the offset didn't change...
+                current_offset_value = current_offset.y if self.vertical else current_offset.x
+                initial_offset_value = real_channel_offset.y if self.vertical else real_channel_offset.x
+                if current_offset_value == initial_offset_value:
+                    debug.info(0, "Channel offset: {}".format(real_channel_offset))
+                    debug.info(0, "Current offset: {}".format(current_offset))
+                    debug.info(0, "VCG {}".format(str(vcg)))
+                    debug.info(0, "HCG {}".format(str(hcg)))
+                    for net in nets:
+                        debug.info(0, "{0} pin: {1}".format(net.name, str(net.pins)))
+                    if self.parent:
+                        debug.info(0, "Saving vcg.gds")
+                        self.parent.gds_write("vcg.gds")
+                    debug.error("Cyclic VCG in channel router.", -1)
 
+                # Increment the track and reset the offset to the start (like a typewriter)
+                if self.vertical:
+                    current_offset = vector(current_offset.x + self.horizontal_nonpref_pitch, real_channel_offset.y)
+                else:
+                    current_offset = vector(real_channel_offset.x, current_offset.y + self.vertical_nonpref_pitch)
+                    
         # Return the size of the channel
         if self.vertical:
-            self.width = 0
-            self.height = current_offset.y
-            return current_offset.y + self.vertical_nonpref_pitch - self.offset.y
+            self.width = current_offset.x + self.horizontal_nonpref_pitch - self.offset.x
+            self.height = self.max_value + self.vertical_nonpref_pitch - self.offset.y
         else:
-            self.width = current_offset.x
-            self.height = 0
-            return current_offset.x + self.horizontal_nonpref_pitch - self.offset.x
-
+            self.width = self.max_value + self.horizontal_nonpref_pitch - self.offset.x
+            self.height = current_offset.y + self.vertical_nonpref_pitch - self.offset.y
+            
     def get_layer_pitch(self, layer):
         """ Return the track pitch on a given layer """
         try:
@@ -226,6 +317,17 @@ class channel_route(design.design):
             self.add_path(self.vertical_layer,
                           [vector(min_x - half_layer_width, trunk_offset.y),
                            vector(max_x + half_layer_width, trunk_offset.y)])
+
+            # Route each pin to the trunk
+            for pin in pins:
+                if pin.cy() < trunk_offset.y:
+                    pin_pos = pin.uc()
+                else:
+                    pin_pos = pin.bc()
+                
+                # No bend needed here
+                mid = vector(pin_pos.x, trunk_offset.y)
+                self.add_path(self.vertical_layer, [pin_pos, mid])
         else:
             # Add the horizontal trunk
             self.add_path(self.horizontal_layer,
@@ -269,6 +371,17 @@ class channel_route(design.design):
             self.add_path(self.horizontal_layer,
                           [vector(trunk_offset.x, min_y - half_layer_width),
                            vector(trunk_offset.x, max_y + half_layer_width)])
+
+            # Route each pin to the trunk
+            for pin in pins:
+                # Find the correct side of the pin
+                if pin.cx() < trunk_offset.x:
+                    pin_pos = pin.rc()
+                else:
+                    pin_pos = pin.lc()
+                # No bend needed here
+                mid = vector(trunk_offset.x, pin_pos.y)
+                self.add_path(self.horizontal_layer, [pin_pos, mid])
         else:
             # Add the vertical trunk
             self.add_path(self.vertical_layer,
@@ -292,18 +405,4 @@ class channel_route(design.design):
                                       to_layer=self.horizontal_layer,
                                       offset=pin_pos)
 
-    def vcg_pin_overlap(self, pin1, pin2, pitch):
-        """ Check for vertical or horizontal overlap of the two pins """
-        
-        # 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.
-        
-        # Pin 1 must be in the "BOTTOM" set
-        x_overlap = pin1.by() < pin2.by() and abs(pin1.center().x - pin2.center().x) < pitch
-        
-        # Pin 1 must be in the "LEFT" set
-        y_overlap = pin1.lx() < pin2.lx() and abs(pin1.center().y - pin2.center().y) < pitch
-        overlaps = (not self.vertical and x_overlap) or (self.vertical and y_overlap)
-        return overlaps
 
diff --git a/compiler/base/hierarchy_layout.py b/compiler/base/hierarchy_layout.py
index 8995488a..22826054 100644
--- a/compiler/base/hierarchy_layout.py
+++ b/compiler/base/hierarchy_layout.py
@@ -1018,7 +1018,7 @@ class layout():
         Wrapper to create a vertical channel route
         """
         import channel_route
-        cr = channel_route.channel_route(netlist, offset, layer_stack, directions, vertical=True)
+        cr = channel_route.channel_route(netlist, offset, layer_stack, directions, vertical=True, parent=self)
         self.add_inst("vc", cr)
         self.connect_inst([])
 
@@ -1027,7 +1027,7 @@ class layout():
         Wrapper to create a horizontal channel route
         """
         import channel_route
-        cr = channel_route.channel_route(netlist, offset, layer_stack, directions, vertical=False)
+        cr = channel_route.channel_route(netlist, offset, layer_stack, directions, vertical=False, parent=self)
         self.add_inst("hc", cr)
         self.connect_inst([])
         
diff --git a/compiler/characterizer/delay.py b/compiler/characterizer/delay.py
index b7faaec3..3f5d61af 100644
--- a/compiler/characterizer/delay.py
+++ b/compiler/characterizer/delay.py
@@ -185,9 +185,8 @@ class delay(simulation):
         self.sen_meas = delay_measure("delay_sen", self.clk_frmt, self.sen_name+"{}", "FALL", "RISE", measure_scale=1e9)
         self.sen_meas.meta_str = sram_op.READ_ZERO        
         self.sen_meas.meta_add_delay = True
-        self.dout_volt_meas.append(self.sen_meas)
-        
-        return self.dout_volt_meas
+
+        return self.dout_volt_meas + [self.sen_meas]
      
     def create_read_bit_measures(self):
         """ Adds bit measurements for read0 and read1 cycles """
@@ -1351,7 +1350,7 @@ class delay(simulation):
         Return the analytical model results for the SRAM. 
         """
         if OPTS.num_rw_ports > 1 or OPTS.num_w_ports > 0 and OPTS.num_r_ports > 0:
-            debug.warning("Analytical characterization results are not supported for multiport.")
+            debug.warning("In analytical mode, all ports have the timing of the first read port.")
         
         # Probe set to 0th bit, does not matter for analytical delay.
         self.set_probe('0'*self.addr_size, 0)
diff --git a/compiler/example_configs/example_config_1w_1r_scn4m_subm.py b/compiler/example_configs/example_config_1w_1r_scn4m_subm.py
index 55ac4016..7698f1a3 100644
--- a/compiler/example_configs/example_config_1w_1r_scn4m_subm.py
+++ b/compiler/example_configs/example_config_1w_1r_scn4m_subm.py
@@ -1,9 +1,9 @@
 word_size = 2
 num_words = 16
 
-num_rw_ports = 1
+num_rw_ports = 0
 num_r_ports = 1
-num_w_ports = 0
+num_w_ports = 1
 
 tech_name = "scn4m_subm"
 nominal_corners_only = False
diff --git a/compiler/modules/bank.py b/compiler/modules/bank.py
index 158ac37b..b0707edb 100644
--- a/compiler/modules/bank.py
+++ b/compiler/modules/bank.py
@@ -211,10 +211,11 @@ class bank(design.design):
         self.port_data_offsets[port] = vector(self.main_bitcell_array_left - self.bitcell_array.cell.width, 0)
 
         # UPPER LEFT QUADRANT
-        # To the left of the bitcell array
+        # To the left of the bitcell array above the predecoders and control logic
         x_offset = self.m2_gap + self.port_address.width
         self.port_address_offsets[port] = vector(-x_offset,
                                                  self.main_bitcell_array_bottom)
+        self.predecoder_height = self.port_address.predecoder_height + self.port_address_offsets[port].y
 
         # LOWER LEFT QUADRANT
         # Place the col decoder left aligned with wordline driver
diff --git a/compiler/modules/hierarchical_decoder.py b/compiler/modules/hierarchical_decoder.py
index 3233bdc8..dbacd051 100644
--- a/compiler/modules/hierarchical_decoder.py
+++ b/compiler/modules/hierarchical_decoder.py
@@ -314,24 +314,23 @@ class hierarchical_decoder(design.design):
         for i in range(self.no_of_pre3x8):
             self.place_pre3x8(i)
 
+        self.predecode_height = 0
+        if self.no_of_pre2x4 > 0:
+            self.predecode_height = self.pre2x4_inst[-1].uy()
+        if self.no_of_pre3x8 > 0:
+            self.predecode_height = self.pre3x8_inst[-1].uy()
+
     def place_pre2x4(self, num):
         """ Place 2x4 predecoder to the left of the origin """
         
-        if (self.num_inputs == 2):
-            base = vector(-self.pre2_4.width, 0)
-        else:
-            base= vector(-self.pre2_4.width, num * (self.pre2_4.height + self.predecoder_spacing))
-
+        base= vector(-self.pre2_4.width, num * (self.pre2_4.height + self.predecoder_spacing))
         self.pre2x4_inst[num].place(base)
         
     def place_pre3x8(self, num):
         """ Place 3x8 predecoder to the left of the origin and above any 2x4 decoders """
-        if (self.num_inputs == 3):
-            offset = vector(-self.pre_3_8.width, 0)
-        else:
-            height = self.no_of_pre2x4 * (self.pre2_4.height + self.predecoder_spacing) + num * (self.pre3_8.height + self.predecoder_spacing)
-            offset = vector(-self.pre3_8.width, height)
-
+        height = self.no_of_pre2x4 * (self.pre2_4.height + self.predecoder_spacing) \
+                 + num * (self.pre3_8.height + self.predecoder_spacing)
+        offset = vector(-self.pre3_8.width, height)
         self.pre3x8_inst[num].place(offset)
 
     def create_row_decoder(self):
diff --git a/compiler/modules/port_address.py b/compiler/modules/port_address.py
index a73e536c..980c9d96 100644
--- a/compiler/modules/port_address.py
+++ b/compiler/modules/port_address.py
@@ -153,6 +153,8 @@ class port_address(design.design):
         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)
+        # Pass this up
+        self.predecoder_height = self.row_decoder.predecoder_height
 
         self.height = self.row_decoder.height
         self.width = self.wordline_driver_inst.rx()
diff --git a/compiler/pgates/pgate.py b/compiler/pgates/pgate.py
index 985c8a86..1e55d5fb 100644
--- a/compiler/pgates/pgate.py
+++ b/compiler/pgates/pgate.py
@@ -372,66 +372,78 @@ class pgate(design.design):
         self.width = width
 
     @staticmethod
-    def bin_width(tx_type, target_width):
+    def best_bin(tx_type, target_width):
+        """
+        Determine the width transistor that meets the accuracy requirement and is larger than target_width.
+        """
+
+        # Find all of the relavent scaled bins and multiples
+        scaled_bins = pgate.scaled_bins(tx_type, target_width)
         
+        for (scaled_width, multiple) in scaled_bins:
+            if abs(target_width - scaled_width) / target_width <= 1 - OPTS.accuracy_requirement:
+                break
+        else:
+            debug.error("failed to bin tx size {}, try reducing accuracy requirement".format(target_width), 1)
+
+        debug.info(2, "binning {0} tx, target: {4}, found {1} x {2} = {3}".format(tx_type,
+                                                                                  multiple,
+                                                                                  scaled_width / multiple,
+                                                                                  scaled_width,
+                                                                                  target_width))
+
+        return(scaled_width / multiple, multiple)
+
+    @staticmethod
+    def scaled_bins(tx_type, target_width):
+        """
+        Determine a set of widths and multiples that could be close to the right size
+        sorted by the fewest number of fingers.
+        """
         if tx_type == "nmos":
             bins = nmos_bins[drc("minwidth_poly")]
         elif tx_type == "pmos":
             bins = pmos_bins[drc("minwidth_poly")]
         else:
             debug.error("invalid tx type")
-        
+
+        # Prune out bins that are too big, except for one bigger
         bins = bins[0:bisect_left(bins, target_width) + 1]
+
+        # Determine multiple of target width for each bin
         if len(bins) == 1:
-            selected_bin = bins[0]
-            scaling_factor = math.ceil(target_width / selected_bin)
-            scaled_bin = bins[0] * scaling_factor
-            
+            scaled_bins = [(bins[0], math.ceil(target_width / bins[0]))]
         else:
-            base_bins = []
             scaled_bins = []
-            scaling_factors = []
+            # Add the biggest size as 1x multiple
+            scaled_bins.append((bins[-1], 1))
+            # Compute discrete multiple of other sizes
+            for width in reversed(bins[:-1]):
+                multiple = math.ceil(target_width / width)
+                scaled_bins.append((multiple * width, multiple))
 
-            for width in bins:
-                m = math.ceil(target_width / width)
-                base_bins.append(width)
-                scaling_factors.append(m)
-                scaled_bins.append(m * width)
-                
-            select = -1
-            for i in reversed(range(0, len(scaled_bins))):
-                if abs(target_width - scaled_bins[i])/target_width <= 1-OPTS.accuracy_requirement:
-                    select = i
-                    break
-            if select == -1:
-                debug.error("failed to bin tx size {}, try reducing accuracy requirement".format(target_width), 1)
-            scaling_factor = scaling_factors[select]
-            scaled_bin = scaled_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, selected_bin * scaling_factor, target_width))
-
-        return(selected_bin, scaling_factor)
-
-    def permute_widths(self, tx_type, target_width):
+        return(scaled_bins)
 
+    @staticmethod
+    def nearest_bin(tx_type, target_width):
+        """
+        Determine the nearest width to the given target_width
+        while assuming a single multiple.
+        """
         if tx_type == "nmos":
             bins = nmos_bins[drc("minwidth_poly")]
         elif tx_type == "pmos":
             bins = pmos_bins[drc("minwidth_poly")]
         else:
-            debug.error("invalid tx type")       
-        bins = bins[0:bisect_left(bins, target_width) + 1]
-        if len(bins) == 1:
-            scaled_bins = [(bins[0], math.ceil(target_width / bins[0]))]
-        else:
-            scaled_bins = []
-            scaled_bins.append((bins[-1], 1))
-            for width in bins[:-1]:
-                m = math.ceil(target_width / width)
-                scaled_bins.append((m * width, m))
+            debug.error("invalid tx type")
 
-        return(scaled_bins)
-        
-    def bin_accuracy(self, ideal_width, width):
-        return 1-abs((ideal_width - width)/ideal_width)
+        # Find the next larger bin
+        bin_loc = bisect_left(bins, target_width)
+        if bin_loc < len(bins):
+            return bins[bin_loc]
+        else:
+            return bins[-1]
+
+    @staticmethod
+    def bin_accuracy(ideal_width, width):
+        return 1 - abs((ideal_width - width) / ideal_width)
diff --git a/compiler/pgates/pinv.py b/compiler/pgates/pinv.py
index aa0f7d9a..4caf2a18 100644
--- a/compiler/pgates/pinv.py
+++ b/compiler/pgates/pinv.py
@@ -14,15 +14,11 @@ from vector import vector
 from math import ceil
 from globals import OPTS
 from utils import round_to_grid
-from bisect import bisect_left
 import logical_effort
 from sram_factory import factory
 from errors import drc_error
 
-if(OPTS.tech_name == "sky130"):
-    from tech import nmos_bins, pmos_bins
 
-    
 class pinv(pgate.pgate):
     """
     Pinv generates gds of a parametrically sized inverter. The
@@ -164,8 +160,8 @@ class pinv(pgate.pgate):
         else:
             self.nmos_width = self.nmos_size * drc("minwidth_tx")
             self.pmos_width = self.pmos_size * drc("minwidth_tx")
-            nmos_bins = self.permute_widths("nmos", self.nmos_width)
-            pmos_bins = self.permute_widths("pmos", self.pmos_width)
+            nmos_bins = self.scaled_bins("nmos", self.nmos_width)
+            pmos_bins = self.scaled_bins("pmos", self.pmos_width)
 
             valid_pmos = []
             for bin in pmos_bins:
diff --git a/compiler/pgates/pinv_dec.py b/compiler/pgates/pinv_dec.py
index 4d4f2229..672bde2d 100644
--- a/compiler/pgates/pinv_dec.py
+++ b/compiler/pgates/pinv_dec.py
@@ -13,6 +13,7 @@ from vector import vector
 from globals import OPTS
 from sram_factory import factory
 
+
 class pinv_dec(pinv.pinv):
     """
     This is another version of pinv but with layout for the decoder.
@@ -50,9 +51,8 @@ class pinv_dec(pinv.pinv):
         self.nmos_width = self.nmos_size * drc("minwidth_tx")
         self.pmos_width = self.pmos_size * drc("minwidth_tx")
         if OPTS.tech_name == "sky130":
-            (self.nmos_width, self.tx_mults) = self.bin_width("nmos", self.nmos_width)
-            (self.pmos_width, self.tx_mults) = self.bin_width("pmos", self.pmos_width)
-            return
+            self.nmos_width = self.nearest_bin("nmos", self.nmos_width)
+            self.pmos_width = self.nearest_bin("pmos", self.pmos_width)
         
     # Over-ride the route input gate to call the horizontal version.
     # Other top-level netlist and layout functions are not changed.
diff --git a/compiler/pgates/pnand2.py b/compiler/pgates/pnand2.py
index 3974d95b..fb6bb210 100644
--- a/compiler/pgates/pnand2.py
+++ b/compiler/pgates/pnand2.py
@@ -39,8 +39,8 @@ class pnand2(pgate.pgate):
         self.tx_mults = 1
 
         if OPTS.tech_name == "sky130":
-            (self.nmos_width, self.tx_mults) = self.bin_width("nmos", self.nmos_width)
-            (self.pmos_width, self.tx_mults) = self.bin_width("pmos", self.pmos_width)
+            self.nmos_width = self.nearest_bin("nmos", self.nmos_width)
+            self.pmos_width = self.nearest_bin("pmos", self.pmos_width)
             
         # Creates the netlist and layout
         pgate.pgate.__init__(self, name, height, add_wells)
diff --git a/compiler/pgates/pnand3.py b/compiler/pgates/pnand3.py
index ff988d88..e4e71e61 100644
--- a/compiler/pgates/pnand3.py
+++ b/compiler/pgates/pnand3.py
@@ -42,8 +42,8 @@ class pnand3(pgate.pgate):
         self.tx_mults = 1
 
         if OPTS.tech_name == "sky130":
-            (self.nmos_width, self.tx_mults) = self.bin_width("nmos", self.nmos_width)
-            (self.pmos_width, self.tx_mults) = self.bin_width("pmos", self.pmos_width)
+            self.nmos_width = self.nearest_bin("nmos", self.nmos_width)
+            self.pmos_width = self.nearest_bin("pmos", self.pmos_width)
 
         # Creates the netlist and layout
         pgate.pgate.__init__(self, name, height, add_wells)
diff --git a/compiler/pgates/pnor2.py b/compiler/pgates/pnor2.py
index 908bba82..aad405e8 100644
--- a/compiler/pgates/pnor2.py
+++ b/compiler/pgates/pnor2.py
@@ -38,8 +38,8 @@ class pnor2(pgate.pgate):
         self.tx_mults = 1
 
         if OPTS.tech_name == "sky130":
-            (self.nmos_width, self.tx_mults) = self.bin_width("nmos", self.nmos_width)
-            (self.pmos_width, self.tx_mults) = self.bin_width("pmos", self.pmos_width)
+            self.nmos_width = self.nearest_bin("nmos", self.nmos_width)
+            self.pmos_width = self.nearest_bin("pmos", self.pmos_width)
 
         # Creates the netlist and layout
         pgate.pgate.__init__(self, name, height, add_wells)
diff --git a/compiler/pgates/precharge.py b/compiler/pgates/precharge.py
index b3d25865..52d24390 100644
--- a/compiler/pgates/precharge.py
+++ b/compiler/pgates/precharge.py
@@ -9,11 +9,10 @@ import contact
 import design
 import debug
 from pgate import pgate
-from tech import parameter
+from tech import parameter, drc
 from vector import vector
 from globals import OPTS
 from sram_factory import factory
-from tech import drc, layer
 
 
 class precharge(design.design):
@@ -81,7 +80,7 @@ class precharge(design.design):
         Initializes the upper and lower pmos
         """
         if(OPTS.tech_name == "sky130"):
-            (self.ptx_width, self.ptx_mults) = pgate.bin_width("pmos", self.ptx_width)
+            self.ptx_width = pgate.nearest_bin("pmos", self.ptx_width)
         self.pmos = factory.create(module_type="ptx",
                                    width=self.ptx_width,
                                    mults=self.ptx_mults,
diff --git a/compiler/pgates/ptx.py b/compiler/pgates/ptx.py
index ce8591fd..b6f839f3 100644
--- a/compiler/pgates/ptx.py
+++ b/compiler/pgates/ptx.py
@@ -131,7 +131,7 @@ class ptx(design.design):
         perimeter_sd = 2 * self.poly_width + 2 * self.tx_width
         if OPTS.tech_name == "sky130" and OPTS.lvs_exe and OPTS.lvs_exe[0] == "calibre":
             # sky130 simulation cannot use the mult parameter in simulation
-            (self.tx_width, self.mults) = pgate.bin_width(self.tx_type, self.tx_width)
+            (self.tx_width, self.mults) = pgate.best_bin(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,
diff --git a/compiler/sram/sram.py b/compiler/sram/sram.py
index b2ab1124..686877da 100644
--- a/compiler/sram/sram.py
+++ b/compiler/sram/sram.py
@@ -100,9 +100,9 @@ class sram():
             import verify
             start_time = datetime.datetime.now()
             # Output the extracted design if requested
-            sp_file = OPTS.output_path + "temp_pex.sp"
+            pexname = OPTS.output_path + self.s.name + ".pex.sp"
             spname = OPTS.output_path + self.s.name + ".sp"
-            verify.run_pex(self.s.name, gdsname, spname, output=sp_file)
+            verify.run_pex(self.s.name, gdsname, spname, output=pexname)
             print_time("Extraction", datetime.datetime.now(), start_time)
         else:
             # Use generated spice file for characterization
diff --git a/compiler/sram/sram_1bank.py b/compiler/sram/sram_1bank.py
index a9731ed5..fc390f37 100644
--- a/compiler/sram/sram_1bank.py
+++ b/compiler/sram/sram_1bank.py
@@ -10,6 +10,8 @@ from vector import vector
 from sram_base import sram_base
 from contact import m2_via
 from globals import OPTS
+import channel_route
+
 
 class sram_1bank(sram_base):
     """
@@ -58,12 +60,14 @@ class sram_1bank(sram_base):
         # the sense amps/column mux and cell array)
         # The x-coordinate is placed to allow a single clock wire (plus an extra pitch)
         # up to the row address DFFs.
-        control_pos = [None] * len(self.all_ports)
-        row_addr_pos = [None] * len(self.all_ports)
-        col_addr_pos = [None] * len(self.all_ports)
-        wmask_pos = [None] * len(self.all_ports)
-        spare_wen_pos = [None] * len(self.all_ports)
-        data_pos = [None] * len(self.all_ports)
+        self.control_pos = [None] * len(self.all_ports)
+        self.row_addr_pos = [None] * len(self.all_ports)
+
+        # DFFs are placd on their own
+        self.col_addr_pos = [None] * len(self.all_ports)
+        self.wmask_pos = [None] * len(self.all_ports)
+        self.spare_wen_pos = [None] * len(self.all_ports)
+        self.data_pos = [None] * len(self.all_ports)
 
         # These positions utilize the channel route sizes.
         # FIXME: Auto-compute these rather than manual computation.
@@ -75,9 +79,11 @@ class sram_1bank(sram_base):
         # Spare wen are on a separate layer so not included
         # Start with 1 track minimum
         self.data_bus_size = [1] * len(self.all_ports)
+        self.col_addr_bus_size = [1] * len(self.all_ports)
         for port in self.all_ports:
+            # The column address wires are routed separately from the data bus and will always be smaller.
             # All ports need the col addr flops
-            self.data_bus_size[port] += self.col_addr_size
+            self.col_addr_bus_size[port] = self.col_addr_size * self.m4_nonpref_pitch
             # Write ports need the data input flops and write mask flops
             if port in self.write_ports:
                 self.data_bus_size[port] += self.num_wmasks + self.word_size
@@ -89,119 +95,152 @@ class sram_1bank(sram_base):
             self.data_bus_size[port] *= self.m4_nonpref_pitch
             # Add the gap in unit length
             self.data_bus_size[port] += self.data_bus_gap
-        
-        # Port 0
-        port = 0
 
-        # This includes 2 M2 pitches for the row addr clock line.
-        # The delay line is aligned with the bitcell array while the control logic is aligned with the port_data
-        # using the control_logic_center value.
-        control_pos[port] = vector(-self.control_logic_insts[port].width - 2 * self.m2_pitch,
-                                   self.bank.bank_array_ll.y - self.control_logic_insts[port].mod.control_logic_center.y)
-        self.control_logic_insts[port].place(control_pos[port])
-        
+        # The control and row addr flops are independent of any bus widths.
+        self.place_control()
+        self.place_row_addr_dffs()
+
+        # Place with an initial wide channel (from above)
+        self.place_dffs()
+        # Route the channel and set to the new data bus size
+        self.route_dffs(add_routes=False)
+        # Re-place with the new channel size
+        self.place_dffs()
+        # Now route the channel
+        self.route_dffs()
+
+    def place_row_addr_dffs(self):
+        """
+        Must be run after place control logic.
+        """
+        port = 0
         # The row address bits are placed above the control logic aligned on the right.
         x_offset = self.control_logic_insts[port].rx() - self.row_addr_dff_insts[port].width
         # It is above the control logic but below the top of the bitcell array
-        y_offset = max(self.control_logic_insts[port].uy(), self.bank_inst.uy() - self.row_addr_dff_insts[port].height)
-        row_addr_pos[port] = vector(x_offset, y_offset)
-        self.row_addr_dff_insts[port].place(row_addr_pos[port])
+        y_offset = max(self.control_logic_insts[port].uy(), self.bank.predecoder_height)
+        self.row_addr_pos[port] = vector(x_offset, y_offset)
+        self.row_addr_dff_insts[port].place(self.row_addr_pos[port])
 
+        if len(self.all_ports)>1:
+            port = 1
+            # The row address bits are placed above the control logic aligned on the left.
+            x_offset = self.control_pos[port].x - self.control_logic_insts[port].width + self.row_addr_dff_insts[port].width
+            # If it can be placed above the predecoder and below the control logic, do it
+            y_offset = self.bank.bank_array_ll.y
+            self.row_addr_pos[port] = vector(x_offset, y_offset)
+            self.row_addr_dff_insts[port].place(self.row_addr_pos[port], mirror="XY")
+        
+    def place_control(self):
+        port = 0
+        
+        # This includes 2 M2 pitches for the row addr clock line.
+        # The delay line is aligned with the bitcell array while the control logic is aligned with the port_data
+        # using the control_logic_center value.
+        self.control_pos[port] = vector(-self.control_logic_insts[port].width - 2 * self.m2_pitch,
+                                        self.bank.bank_array_ll.y - self.control_logic_insts[port].mod.control_logic_center.y)
+        self.control_logic_insts[port].place(self.control_pos[port])
+        if len(self.all_ports) > 1:
+            port = 1
+            # This includes 2 M2 pitches for the row addr clock line
+            # The delay line is aligned with the bitcell array while the control logic is aligned with the port_data
+            # using the control_logic_center value.
+            self.control_pos[port] = vector(self.bank_inst.rx() + self.control_logic_insts[port].width + 2 * self.m2_pitch,
+                                            self.bank.bank_array_ur.y
+                                            + self.control_logic_insts[port].height
+                                            - self.control_logic_insts[port].height
+                                            + self.control_logic_insts[port].mod.control_logic_center.y)
+            self.control_logic_insts[port].place(self.control_pos[port], mirror="XY")
+
+    def place_dffs(self):
+        """
+        Place the col addr, data, wmask, and spare data DFFs.
+        This can be run more than once after we recompute the channel width.
+        """
+
+        port = 0
         # Add the col address flops below the bank to the right of the control logic
         x_offset = self.control_logic_insts[port].rx() + self.dff.width
-        y_offset = - self.data_bus_size[port] - self.dff.height
+        # Place it a data bus below the x-axis, but at least as low as the control logic to not block
+        # the control logic signals
+        y_offset = min(-self.data_bus_size[port] - self.dff.height,
+                       self.control_logic_insts[port].by())
         if self.col_addr_dff:
-            col_addr_pos[port] = vector(x_offset,
-                                        y_offset)
-            self.col_addr_dff_insts[port].place(col_addr_pos[port])
+            self.col_addr_pos[port] = vector(x_offset,
+                                             y_offset)
+            self.col_addr_dff_insts[port].place(self.col_addr_pos[port])
             x_offset = self.col_addr_dff_insts[port].rx()
         else:
-            col_addr_pos[port] = vector(x_offset, 0)
+            self.col_addr_pos[port] = vector(x_offset, 0)
             
         if port in self.write_ports:
             if self.write_size:
                 # Add the write mask flops below the write mask AND array.
-                wmask_pos[port] = vector(x_offset,
-                                         y_offset)
-                self.wmask_dff_insts[port].place(wmask_pos[port])
+                self.wmask_pos[port] = vector(x_offset,
+                                              y_offset)
+                self.wmask_dff_insts[port].place(self.wmask_pos[port])
                 x_offset = self.wmask_dff_insts[port].rx()
 
             # Add the data flops below the write mask flops.
-            data_pos[port] = vector(x_offset,
-                                    y_offset)
-            self.data_dff_insts[port].place(data_pos[port])
+            self.data_pos[port] = vector(x_offset,
+                                         y_offset)
+            self.data_dff_insts[port].place(self.data_pos[port])
             x_offset = self.data_dff_insts[port].rx()
 
             # Add spare write enable flops to the right of data flops since the spare columns
             # will be on the right
             if self.num_spare_cols:
-                spare_wen_pos[port] = vector(x_offset,
-                                             y_offset)
-                self.spare_wen_dff_insts[port].place(spare_wen_pos[port])
+                self.spare_wen_pos[port] = vector(x_offset,
+                                                  y_offset)
+                self.spare_wen_dff_insts[port].place(self.spare_wen_pos[port])
                 x_offset = self.spare_wen_dff_insts[port].rx()
 
         else:
-            wmask_pos[port] = vector(x_offset, y_offset)
-            data_pos[port] = vector(x_offset, y_offset)
-            spare_wen_pos[port] = vector(x_offset, y_offset)
+            self.wmask_pos[port] = vector(x_offset, y_offset)
+            self.data_pos[port] = vector(x_offset, y_offset)
+            self.spare_wen_pos[port] = vector(x_offset, y_offset)
 
-        if len(self.all_ports)>1:
-            # Port 1
+        if len(self.all_ports) > 1:
             port = 1
-        
-            # This includes 2 M2 pitches for the row addr clock line
-            # The delay line is aligned with the bitcell array while the control logic is aligned with the port_data
-            # using the control_logic_center value.
-            control_pos[port] = vector(self.bank_inst.rx() + self.control_logic_insts[port].width + 2 * self.m2_pitch,
-                                       self.bank.bank_array_ur.y
-                                       + self.control_logic_insts[port].height
-                                       - self.control_logic_insts[port].height
-                                       + self.control_logic_insts[port].mod.control_logic_center.y)
-            self.control_logic_insts[port].place(control_pos[port], mirror="XY")
-        
-            # The row address bits are placed above the control logic aligned on the left.
-            x_offset = control_pos[port].x - self.control_logic_insts[port].width + self.row_addr_dff_insts[port].width
-            # It is below the control logic but below the bottom of the bitcell array
-            y_offset = min(self.control_logic_insts[port].by(), self.bank_inst.by() + self.row_addr_dff_insts[port].height)
-            row_addr_pos[port] = vector(x_offset, y_offset)
-            self.row_addr_dff_insts[port].place(row_addr_pos[port], mirror="XY")
-
+            
             # Add the col address flops below the bank to the right of the control logic
             x_offset = self.control_logic_insts[port].lx() - 2 * self.dff.width
-            y_offset = self.bank.height + self.data_bus_size[port] + self.dff.height
+            # Place it a data bus below the x-axis, but at least as high as the control logic to not block
+            # the control logic signals
+            y_offset = max(self.bank.height + self.data_bus_size[port] + self.dff.height,
+                           self.control_logic_insts[port].uy() - self.dff.height)
             if self.col_addr_dff:
-                col_addr_pos[port] = vector(x_offset - self.col_addr_dff_insts[port].width,
-                                            y_offset)
-                self.col_addr_dff_insts[port].place(col_addr_pos[port], mirror="MX")
+                self.col_addr_pos[port] = vector(x_offset,
+                                                 y_offset)
+                self.col_addr_dff_insts[port].place(self.col_addr_pos[port], mirror="XY")
                 x_offset = self.col_addr_dff_insts[port].lx()
             else:
-                col_addr_pos[port] = vector(x_offset, y_offset)
+                self.col_addr_pos[port] = vector(x_offset, y_offset)
             
             if port in self.write_ports:
                 # Add spare write enable flops to the right of the data flops since the spare
                 # columns will be on the left
                 if self.num_spare_cols:
-                    spare_wen_pos[port] = vector(x_offset - self.spare_wen_dff_insts[port].width,
-                                                 y_offset)
-                    self.spare_wen_dff_insts[port].place(spare_wen_pos[port], mirror="MX")
+                    self.spare_wen_pos[port] = vector(x_offset - self.spare_wen_dff_insts[port].width,
+                                                      y_offset)
+                    self.spare_wen_dff_insts[port].place(self.spare_wen_pos[port], mirror="MX")
                     x_offset = self.spare_wen_dff_insts[port].lx()
 
                 if self.write_size:
                     # Add the write mask flops below the write mask AND array.
-                    wmask_pos[port] = vector(x_offset - self.wmask_dff_insts[port].width,
-                                             y_offset)
-                    self.wmask_dff_insts[port].place(wmask_pos[port], mirror="MX")
+                    self.wmask_pos[port] = vector(x_offset - self.wmask_dff_insts[port].width,
+                                                  y_offset)
+                    self.wmask_dff_insts[port].place(self.wmask_pos[port], mirror="MX")
                     x_offset = self.wmask_dff_insts[port].lx()
 
                 # Add the data flops below the write mask flops.
-                data_pos[port] = vector(x_offset - self.data_dff_insts[port].width,
-                                        y_offset)
-                self.data_dff_insts[port].place(data_pos[port], mirror="MX")
+                self.data_pos[port] = vector(x_offset - self.data_dff_insts[port].width,
+                                             y_offset)
+                self.data_dff_insts[port].place(self.data_pos[port], mirror="MX")
         else:
-            wmask_pos[port] = vector(x_offset, y_offset)
-            data_pos[port] = vector(x_offset, y_offset)
-            spare_wen_pos[port] = vector(x_offset, y_offset)
-            
+            self.wmask_pos[port] = vector(x_offset, y_offset)
+            self.data_pos[port] = vector(x_offset, y_offset)
+            self.spare_wen_pos[port] = vector(x_offset, y_offset)
+
     def add_layout_pins(self):
         """
         Add the top-level pins for a single bank SRAM with control.
@@ -343,20 +382,42 @@ class sram_1bank(sram_base):
         
         self.route_row_addr_dff()
 
+    def route_dffs(self, add_routes=True):
+        
         for port in self.all_ports:
-            self.route_dff(port)
+            self.route_dff(port, add_routes)
 
-    def route_dff(self, port):
+    def route_dff(self, port, add_routes):
 
         route_map = []
         
-        # column mux dff
+        # column mux dff is routed on it's own since it is to the far end
+        # decoder inputs are min pitch M2, so need to use lower layer stack
         if self.col_addr_size > 0:
             dff_names = ["dout_{}".format(x) for x in range(self.col_addr_size)]
             dff_pins = [self.col_addr_dff_insts[port].get_pin(x) for x in dff_names]
             bank_names = ["addr{0}_{1}".format(port, x) for x in range(self.col_addr_size)]
             bank_pins = [self.bank_inst.get_pin(x) for x in bank_names]
             route_map.extend(list(zip(bank_pins, dff_pins)))
+
+            if port == 0:
+                offset = vector(self.control_logic_insts[port].rx() + self.dff.width,
+                                - self.data_bus_size[port] + 2 * self.m1_pitch)
+            else:
+                offset = vector(0,
+                                self.bank.height + 2 * self.m1_space)
+            
+            cr = channel_route.channel_route(netlist=route_map,
+                                             offset=offset,
+                                             layer_stack=self.m1_stack,
+                                             parent=self)
+            if add_routes:
+                self.add_inst("hc", cr)
+                self.connect_inst([])
+            else:
+                self.col_addr_bus_size[port] = cr.height
+
+        route_map = []
         
         # wmask dff
         if self.num_wmasks > 0 and port in self.write_ports:
@@ -384,35 +445,46 @@ class sram_1bank(sram_base):
             bank_pins = [self.bank_inst.get_pin(x) for x in bank_names]
             route_map.extend(list(zip(bank_pins, sram_pins)))
 
-        if self.num_wmasks > 0 and port in self.write_ports:
-            layer_stack = self.m3_stack
-        else:
-            layer_stack = self.m1_stack
-                
-        if port == 0:
-            offset = vector(self.control_logic_insts[port].rx() + self.dff.width,
-                            - self.data_bus_size[port] + 2 * self.m1_pitch)
-        else:
-            offset = vector(0,
-                            self.bank.height + 2 * self.m1_space)
-
-        if len(route_map) > 0:
-            self.create_horizontal_channel_route(netlist=route_map,
-                                                 offset=offset,
-                                                 layer_stack=layer_stack)
-
-        # Route these separately because sometimes the pin pitch on the write driver is too narrow for M3 (FreePDK45)
         # spare wen dff
         if self.num_spare_cols > 0 and port in self.write_ports:
             dff_names = ["dout_{}".format(x) for x in range(self.num_spare_cols)]
             dff_pins = [self.spare_wen_dff_insts[port].get_pin(x) for x in dff_names]
             bank_names = ["bank_spare_wen{0}_{1}".format(port, x) for x in range(self.num_spare_cols)]
             bank_pins = [self.bank_inst.get_pin(x) for x in bank_names]
-            route_map = zip(bank_pins, dff_pins)
-            self.create_horizontal_channel_route(netlist=route_map,
+            route_map.extend(list(zip(bank_pins, dff_pins)))
+                
+        if len(route_map) > 0:
+            
+            if self.num_wmasks > 0 and port in self.write_ports:
+                layer_stack = self.m3_stack
+            else:
+                layer_stack = self.m1_stack
+                
+            if port == 0:
+                offset = vector(self.control_logic_insts[port].rx() + self.dff.width,
+                                - self.data_bus_size[port] + 2 * self.m1_pitch)
+                cr = channel_route.channel_route(netlist=route_map,
                                                  offset=offset,
-                                                 layer_stack=self.m1_stack)
-        
+                                                 layer_stack=layer_stack,
+                                                 parent=self)
+                if add_routes:
+                    self.add_inst("hc", cr)
+                    self.connect_inst([])
+                else:
+                    self.data_bus_size[port] = max(cr.height, self.col_addr_bus_size[port]) + self.data_bus_gap
+            else:
+                offset = vector(0,
+                                self.bank.height + 2 * self.m1_space)
+                cr = channel_route.channel_route(netlist=route_map,
+                                                 offset=offset,
+                                                 layer_stack=layer_stack,
+                                                 parent=self)
+                if add_routes:
+                    self.add_inst("hc", cr)
+                    self.connect_inst([])
+                else:
+                    self.data_bus_size[port] = max(cr.height, self.col_addr_bus_size[port]) + self.data_bus_gap
+                    
     def route_clk(self):
         """ Route the clock network """
 
diff --git a/compiler/sram/sram_base.py b/compiler/sram/sram_base.py
index b16c28fa..12af7cf6 100644
--- a/compiler/sram/sram_base.py
+++ b/compiler/sram/sram_base.py
@@ -129,7 +129,7 @@ class sram_base(design, verilog, lef):
 
         start_time = datetime.datetime.now()
         # We only enable final verification if we have routed the design
-        self.DRC_LVS(final_verification=OPTS.route_supplies, force_check=True)
+        self.DRC_LVS(final_verification=OPTS.route_supplies, force_check=OPTS.check_lvsdrc)
         if not OPTS.is_unit_test:
             print_time("Verification", datetime.datetime.now(), start_time)
 
diff --git a/compiler/verify/calibre.py b/compiler/verify/calibre.py
index 6ee41e5f..965e3557 100644
--- a/compiler/verify/calibre.py
+++ b/compiler/verify/calibre.py
@@ -135,7 +135,7 @@ def write_calibre_lvs_script(cell_name, final_verification, gds_name, sp_name):
 def write_calibre_pex_script(cell_name, extract, output, final_verification):
     """ Write a pex script that can either just extract the netlist or the netlist+parasitics """
     if output == None:
-        output = cell_name + ".pex.netlist"
+        output = cell_name + ".pex.sp"
 
     # check if lvs report has been done
     # if not run drc and lvs
@@ -195,11 +195,14 @@ def run_drc(cell_name, gds_name, extract=False, final_verification=False):
         filter_gds(cell_name, OPTS.openram_temp + "temp.gds", OPTS.openram_temp + cell_name + ".gds")
     else:
         # Copy file to local dir if it isn't already
-        if not os.path.isfile(gds_name):
-            shutil.copy(OPTS.output_path+os.path.basename(gds_name),gds_name)
+        if not os.path.isfile(OPTS.openram_temp + os.path.basename(gds_name)):
+            shutil.copy(gds_name, OPTS.openram_temp)
 
     drc_runset = write_calibre_drc_script(cell_name, extract, final_verification, gds_name)
 
+    if not os.path.isfile(OPTS.openram_temp + os.path.basename(gds_name)):
+        shutil.copy(gds_name, OPTS.openram_temp + os.path.basename(gds_name))
+                                
     (outfile, errfile, resultsfile) = run_script(cell_name, "drc")
 
     # check the result for these lines in the summary:
@@ -241,14 +244,10 @@ def run_lvs(cell_name, gds_name, sp_name, final_verification=False):
     lvs_runset = write_calibre_lvs_script(cell_name, final_verification, gds_name, sp_name)
 
     # Copy file to local dir if it isn't already
-#    if os.path.dirname(gds_name)!=OPTS.openram_temp.rstrip('/'):
-#        shutil.copy(gds_name, OPTS.openram_temp)
-#    if os.path.dirname(sp_name)!=OPTS.openram_temp.rstrip('/'):
-#        shutil.copy(sp_name, OPTS.openram_temp)
-    if not os.path.isfile(gds_name):
-        shutil.copy(OPTS.output_path+os.path.basename(gds_name), gds_name)
-    if not os.path.isfile(sp_name):
-        shutil.copy(OPTS.output_path+os.path.basename(sp_name), sp_name)
+    if not os.path.isfile(OPTS.openram_temp + os.path.basename(gds_name)):
+        shutil.copy(gds_name, OPTS.openram_temp)
+    if not os.path.isfile(OPTS.openram_temp + os.path.basename(sp_name)):
+        shutil.copy(sp_name, OPTS.openram_temp)
 
     (outfile, errfile, resultsfile) = run_script(cell_name, "lvs")
 
@@ -329,13 +328,14 @@ def run_pex(cell_name, gds_name, sp_name, output=None, final_verification=False)
     global num_pex_runs
     num_pex_runs += 1
 
-    write_calibre_pex_script(cell_name,True,output,final_verification)
+    write_calibre_pex_script(cell_name, True, output, final_verification)
 
+    # Copy file to local dir if it isn't already
     if not os.path.isfile(OPTS.openram_temp + os.path.basename(gds_name)):
-        shutil.copy(gds_name, OPTS.openram_temp + os.path.basename(gds_name))
+        shutil.copy(gds_name, OPTS.openram_temp)
     if not os.path.isfile(OPTS.openram_temp + os.path.basename(sp_name)):
-        shutil.copy(sp_name, OPTS.openram_temp + os.path.basename(sp_name))
-
+        shutil.copy(sp_name, OPTS.openram_temp)
+    
     (outfile, errfile, resultsfile) = run_script(cell_name, "pex")
 
 
diff --git a/compiler/verify/none.py b/compiler/verify/none.py
index 41e5780e..f82d59ae 100644
--- a/compiler/verify/none.py
+++ b/compiler/verify/none.py
@@ -20,7 +20,7 @@ pex_warned = False
 def run_drc(cell_name, gds_name, extract=False, final_verification=False):
     global drc_warned
     if not drc_warned:
-        debug.warning("DRC unable to run.")
+        debug.error("DRC unable to run.", -1)
         drc_warned=True
     # Since we warned, return a failing test.
     return 1
@@ -29,7 +29,7 @@ def run_drc(cell_name, gds_name, extract=False, final_verification=False):
 def run_lvs(cell_name, gds_name, sp_name, final_verification=False):
     global lvs_warned
     if not lvs_warned:
-        debug.warning("LVS unable to run.")
+        debug.error("LVS unable to run.", -1)
         lvs_warned=True
     # Since we warned, return a failing test.
     return 1
@@ -38,7 +38,7 @@ def run_lvs(cell_name, gds_name, sp_name, final_verification=False):
 def run_pex(name, gds_name, sp_name, output=None, final_verification=False):
     global pex_warned
     if not pex_warned:
-        debug.warning("PEX unable to run.")
+        debug.error("PEX unable to run.", -1)
         pex_warned=True
     # Since we warned, return a failing test.
     return 1
diff --git a/technology/freepdk45/gds_lib/write_driver.gds b/technology/freepdk45/gds_lib/write_driver.gds
index 86015e7a..5e4fb15a 100644
Binary files a/technology/freepdk45/gds_lib/write_driver.gds and b/technology/freepdk45/gds_lib/write_driver.gds differ