diff --git a/compiler/base/lef.py b/compiler/base/lef.py
index f0d6dda0..9db18ab1 100644
--- a/compiler/base/lef.py
+++ b/compiler/base/lef.py
@@ -112,18 +112,22 @@ class lef:
         for pin_name in self.pins:
             pin = self.get_pin(pin_name)
             inflated_pin = pin.inflated_pin(multiple=1)
-            for blockage in self.blockages[pin.layer]:
-                if blockage.overlaps(inflated_pin):
-                    intersection_shape = blockage.intersection(inflated_pin)
-                    # If it is zero area, don't add the pin
-                    if intersection_shape[0][0]==intersection_shape[1][0] or intersection_shape[0][1]==intersection_shape[1][1]:
-                        continue
-                    # Remove the old blockage and add the new ones
-                    self.blockages[pin.layer].remove(blockage)
-                    intersection_pin = pin_layout("", intersection_shape, inflated_pin.layer)
-                    new_blockages = blockage.cut(intersection_pin)
-                    self.blockages[pin.layer].extend(new_blockages)
-
+            another_iteration_needed = True
+            while another_iteration_needed:
+                another_iteration_needed = False
+                old_blockages = list(self.blockages[pin.layer])
+                for blockage in old_blockages:
+                    if blockage.overlaps(inflated_pin):
+                        intersection_shape = blockage.intersection(inflated_pin)
+                        # If it is zero area, don't add the pin
+                        if intersection_shape[0][0]==intersection_shape[1][0] or intersection_shape[0][1]==intersection_shape[1][1]:
+                            continue
+                        another_iteration_needed = True
+                        # Remove the old blockage and add the new ones
+                        self.blockages[pin.layer].remove(blockage)
+                        intersection_pin = pin_layout("", intersection_shape, inflated_pin.layer)
+                        new_blockages = blockage.cut(intersection_pin)
+                        self.blockages[pin.layer].extend(new_blockages)
 
     def lef_write_header(self):
         """ Header of LEF file """
diff --git a/compiler/base/verilog.py b/compiler/base/verilog.py
index 0405649d..7886615f 100644
--- a/compiler/base/verilog.py
+++ b/compiler/base/verilog.py
@@ -29,6 +29,11 @@ class verilog:
             self.vf.write("\n")
 
         self.vf.write("module {0}(\n".format(self.name))
+        self.vf.write("`ifdef USE_POWER_PINS\n")
+        self.vf.write("    vdd,\n")
+        self.vf.write("    gnd,\n")
+        self.vf.write("`endif\n")
+        
         for port in self.all_ports:
             if port in self.readwrite_ports:
                 self.vf.write("// Port {0}: RW\n".format(port))
@@ -65,6 +70,11 @@ class verilog:
         self.vf.write("  parameter T_HOLD = 1 ; //Delay to hold dout value after posedge. Value is arbitrary\n")
         self.vf.write("\n")
 
+        self.vf.write("`ifdef USE_POWER_PINS\n")
+        self.vf.write("    inout vdd;\n")
+        self.vf.write("    inout gnd;\n")
+        self.vf.write("`endif\n")
+        
         for port in self.all_ports:
             self.add_inputs_outputs(port)
 
diff --git a/compiler/example_configs/example_config_scn4m_subm.py b/compiler/example_configs/example_config_scn4m_subm.py
index d331c1fc..8fc92169 100644
--- a/compiler/example_configs/example_config_scn4m_subm.py
+++ b/compiler/example_configs/example_config_scn4m_subm.py
@@ -11,7 +11,7 @@ process_corners = ["TT"]
 supply_voltages = [5.0]
 temperatures = [25]
 
-route_supplies = True
+route_supplies = "side"
 check_lvsdrc = True
 
 output_name = "sram_{0}rw{1}r{2}w_{3}_{4}_{5}".format(num_rw_ports,
diff --git a/compiler/gdsMill/gdsMill/vlsiLayout.py b/compiler/gdsMill/gdsMill/vlsiLayout.py
index 2592d251..056c01c0 100644
--- a/compiler/gdsMill/gdsMill/vlsiLayout.py
+++ b/compiler/gdsMill/gdsMill/vlsiLayout.py
@@ -426,7 +426,8 @@ class VlsiLayout:
         self.structures[self.rootStructureName].texts.append(textToAdd)
 
     def padText(self, text):
-        if(len(text)%2 == 1):
+        debug.check(len(text) > 0, "Cannot have zero length text string.")
+        if(len(text) % 2 == 1):
             return text + '\x00'
         else:
             return text
@@ -700,7 +701,6 @@ class VlsiLayout:
 
         return max_pins
 
-
     def getAllPinShapes(self, pin_name):
         """
         Search for a pin label and return ALL the enclosing rectangles on the same layer
diff --git a/compiler/modules/bank.py b/compiler/modules/bank.py
index d1b5a90d..56cd854c 100644
--- a/compiler/modules/bank.py
+++ b/compiler/modules/bank.py
@@ -75,6 +75,11 @@ class bank(design.design):
         self.bank_array_ll = self.offset_all_coordinates().scale(-1, -1)
         self.bank_array_ur = self.bitcell_array_inst.ur()
         self.bank_array_ul = self.bitcell_array_inst.ul()
+
+        # These are used for other placements (e.g. address flops)
+        self.predecoder_top = self.port_address[0].predecoder_height + self.port_address_inst[0].by()
+        self.predecoder_bottom = self.port_address_inst[0].by()
+
         self.DRC_LVS()
 
     def add_pins(self):
@@ -227,7 +232,6 @@ class bank(design.design):
         x_offset = self.m2_gap + self.port_address[port].width
         self.port_address_offsets[port] = vector(-x_offset,
                                                  self.main_bitcell_array_bottom)
-        self.predecoder_height = self.port_address[port].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/control_logic.py b/compiler/modules/control_logic.py
index eb9a21ed..52c1dd87 100644
--- a/compiler/modules/control_logic.py
+++ b/compiler/modules/control_logic.py
@@ -346,9 +346,12 @@ class control_logic(design.design):
             row += 1
         self.place_wlen_row(row)
         row += 1
-        self.place_delay(row)
+
+        control_center_y = self.wl_en_inst.uy() + self.m3_pitch
+        
+        # Delay chain always gets placed at row 4
+        self.place_delay(4)
         height = self.delay_inst.uy()
-        control_center_y = self.delay_inst.by()
 
         # This offset is used for placement of the control logic in the SRAM level.
         self.control_logic_center = vector(self.ctrl_dff_inst.rx(), control_center_y)
@@ -387,19 +390,22 @@ class control_logic(design.design):
 
     def place_delay(self, row):
         """ Place the replica bitline """
-        y_off = row * self.and2.height + 2 * self.m1_pitch
+        debug.check(row % 2 == 0, "Must place delay chain at even row for supply alignment.")
+        
+        # It is flipped on X axis
+        y_off = row * self.and2.height + self.delay_chain.height
 
         # Add the RBL above the rows
         # Add to the right of the control rows and routing channel
-        offset = vector(self.delay_chain.width, y_off)
-        self.delay_inst.place(offset, mirror="MY")
+        offset = vector(0, y_off)
+        self.delay_inst.place(offset, mirror="MX")
 
     def route_delay(self):
 
-        out_pos = self.delay_inst.get_pin("out").bc()
+        out_pos = self.delay_inst.get_pin("out").center()
         # Connect to the rail level with the vdd rail
-        # Use pen since it is in every type of control logic
-        vdd_ypos = self.p_en_bar_nand_inst.get_pin("vdd").by()
+        # Use gated clock since it is in every type of control logic
+        vdd_ypos = self.gated_clk_buf_inst.get_pin("vdd").cy() + self.m1_pitch
         in_pos = vector(self.input_bus["rbl_bl_delay"].cx(), vdd_ypos)
         mid1 = vector(out_pos.x, in_pos.y)
         self.add_wire(self.m1_stack, [out_pos, mid1, in_pos])
@@ -676,7 +682,7 @@ class control_logic(design.design):
         # 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, self.and2.height)
+        mid_pos = vector(in_pos.x, self.gated_clk_buf_inst.get_pin("vdd").cy() - self.m1_pitch)
         rail_pos = vector(self.input_bus["clk_buf"].cx(), mid_pos.y)
         self.add_wire(self.m1_stack, [in_pos, mid_pos, rail_pos])
         self.add_via_center(layers=self.m1_stack,
@@ -794,3 +800,8 @@ class control_logic(design.design):
                                   to_layer="m2",
                                   offset=out_pos)
 
+    def get_left_pins(self, name):
+        """
+        Return the left side supply pins to connect to a vertical stripe.
+        """
+        return(self.cntrl_dff_inst.get_pins(name) + self.delay_inst.get_pins(name))
diff --git a/compiler/modules/delay_chain.py b/compiler/modules/delay_chain.py
index d7e9c2d0..7e4830c9 100644
--- a/compiler/modules/delay_chain.py
+++ b/compiler/modules/delay_chain.py
@@ -31,6 +31,7 @@ class delay_chain(design.design):
 
         # number of inverters including any fanout loads.
         self.fanout_list = fanout_list
+        self.rows = len(self.fanout_list)
 
         self.create_netlist()
         if not OPTS.netlist_only:
@@ -43,10 +44,10 @@ class delay_chain(design.design):
 
     def create_layout(self):
         # Each stage is a a row
-        self.height = len(self.fanout_list) * self.inv.height
+        self.height = self.rows * self.inv.height
         # The width is determined by the largest fanout plus the driver
         self.width = (max(self.fanout_list) + 1) * self.inv.width
-
+        
         self.place_inverters()
         self.route_inverters()
         self.route_supplies()
@@ -62,14 +63,19 @@ class delay_chain(design.design):
         self.add_pin("gnd", "GROUND")
 
     def add_modules(self):
-        self.inv = factory.create(module_type="pinv")
+
+        self.dff = factory.create(module_type="dff_buf")
+        dff_height = self.dff.height
+        
+        self.inv = factory.create(module_type="pinv",
+                                  height=dff_height)
         self.add_mod(self.inv)
 
     def create_inverters(self):
         """ Create the inverters and connect them based on the stage list """
         self.driver_inst_list = []
         self.load_inst_map = {}
-        for stage_num, fanout_size in zip(range(len(self.fanout_list)), self.fanout_list):
+        for stage_num, fanout_size in zip(range(self.rows), self.fanout_list):
             # Add the inverter
             cur_driver=self.add_inst(name="dinv{}".format(stage_num),
                                      mod=self.inv)
@@ -77,7 +83,7 @@ class delay_chain(design.design):
             self.driver_inst_list.append(cur_driver)
 
             # Hook up the driver
-            if stage_num + 1 == len(self.fanout_list):
+            if stage_num + 1 == self.rows:
                 stageout_name = "out"
             else:
                 stageout_name = "dout_{}".format(stage_num + 1)
@@ -101,7 +107,7 @@ class delay_chain(design.design):
 
     def place_inverters(self):
         """ Place the inverters and connect them based on the stage list """
-        for stage_num, fanout_size in zip(range(len(self.fanout_list)), self.fanout_list):
+        for stage_num, fanout_size in zip(range(self.rows), self.fanout_list):
             if stage_num % 2:
                 inv_mirror = "MX"
                 inv_offset = vector(0, (stage_num + 1) * self.inv.height)
@@ -185,24 +191,26 @@ class delay_chain(design.design):
     def add_layout_pins(self):
 
         # input is A pin of first inverter
+        # It gets routed to the left a bit to prevent pin access errors
+        # due to the output pin when going up to M3
         a_pin = self.driver_inst_list[0].get_pin("A")
+        mid_loc = vector(a_pin.cx() - self.m3_pitch, a_pin.cy())
         self.add_via_stack_center(from_layer=a_pin.layer,
-                                  to_layer="m2",
-                                  offset=a_pin.center())
-        self.add_layout_pin(text="in",
-                            layer="m2",
-                            offset=a_pin.ll().scale(1, 0),
-                            height=a_pin.cy())
+                                        to_layer="m2",
+                                        offset=mid_loc)
+        self.add_path(a_pin.layer, [a_pin.center(), mid_loc])
+        
+        self.add_layout_pin_rect_center(text="in",
+                                        layer="m2",
+                                        offset=mid_loc)
 
-        # output is A pin of last load inverter
+        # output is A pin of last load/fanout inverter
         last_driver_inst = self.driver_inst_list[-1]
         a_pin = self.load_inst_map[last_driver_inst][-1].get_pin("A")
         self.add_via_stack_center(from_layer=a_pin.layer,
-                                  to_layer="m2",
+                                  to_layer="m1",
                                   offset=a_pin.center())
-        mid_point = vector(a_pin.cx() + 3 * self.m2_width, a_pin.cy())
-        self.add_path("m2", [a_pin.center(), mid_point, mid_point.scale(1, 0)])
-        self.add_layout_pin_segment_center(text="out",
-                                           layer="m2",
-                                           start=mid_point,
-                                           end=mid_point.scale(1, 0))
+        self.add_layout_pin_rect_center(text="out",
+                                        layer="m1",
+                                        offset=a_pin.center())
+        
diff --git a/compiler/router/grid.py b/compiler/router/grid.py
index 793e5e89..ea59c80f 100644
--- a/compiler/router/grid.py
+++ b/compiler/router/grid.py
@@ -122,35 +122,45 @@ class grid:
             self.set_target(n)
             # self.set_blocked(n, False)
 
-    def add_perimeter_target(self, side="all"):
-        debug.info(3, "Adding perimeter target")
-
+    def get_perimeter_list(self, side="left", layers=[0, 1], width=1, margin=0, offset=0):
+        """
+        Side specifies which side.
+        Layer specifies horizontal (0) or vertical (1)
+        Width specifies how wide the perimter "stripe" should be.
+        """
         perimeter_list = []
         # Add the left/right columns
         if side=="all" or side=="left":
-            x = self.ll.x
-            for y in range(self.ll.y, self.ur.y, 1):
-                perimeter_list.append(vector3d(x, y, 0))
-                perimeter_list.append(vector3d(x, y, 1))
+            for x in range(self.ll.x + offset, self.ll.x + width + offset, 1):
+                for y in range(self.ll.y + margin, self.ur.y - margin, 1):
+                    for layer in layers:
+                        perimeter_list.append(vector3d(x, y, layer))
                 
         if side=="all" or side=="right":
-            x = self.ur.x
-            for y in range(self.ll.y, self.ur.y, 1):
-                perimeter_list.append(vector3d(x, y, 0))
-                perimeter_list.append(vector3d(x, y, 1))
+            for x in range(self.ur.x - width - offset, self.ur.x - offset, 1):
+                for y in range(self.ll.y + margin, self.ur.y - margin, 1):
+                    for layer in layers:
+                        perimeter_list.append(vector3d(x, y, layer))
 
         if side=="all" or side=="bottom":
-            y = self.ll.y
-            for x in range(self.ll.x, self.ur.x, 1):
-                perimeter_list.append(vector3d(x, y, 0))
-                perimeter_list.append(vector3d(x, y, 1))
+            for y in range(self.ll.y + offset, self.ll.y + width + offset, 1):
+                for x in range(self.ll.x + margin, self.ur.x - margin, 1):
+                    for layer in layers:
+                        perimeter_list.append(vector3d(x, y, layer))
 
         if side=="all" or side=="top":
-            y = self.ur.y
-            for x in range(self.ll.x, self.ur.x, 1):
-                perimeter_list.append(vector3d(x, y, 0))
-                perimeter_list.append(vector3d(x, y, 1))
+            for y in range(self.ur.y - width - offset, self.ur.y - offset, 1):
+                for x in range(self.ll.x + margin, self.ur.x - margin, 1):
+                    for layer in layers:
+                        perimeter_list.append(vector3d(x, y, layer))
 
+        return perimeter_list
+    
+    def add_perimeter_target(self, side="all", layers=[0, 1]):
+        debug.info(3, "Adding perimeter target")
+        
+        perimeter_list = self.get_perimeter_list(side, layers)
+        
         self.set_target(perimeter_list)
                 
     def is_target(self, point):
diff --git a/compiler/router/pin_group.py b/compiler/router/pin_group.py
index 7a5d8817..5e6d6f89 100644
--- a/compiler/router/pin_group.py
+++ b/compiler/router/pin_group.py
@@ -155,6 +155,10 @@ class pin_group:
         # Now simplify the enclosure list
         new_pin_list = self.remove_redundant_shapes(pin_list)
 
+        # Now add the right name
+        for pin in new_pin_list:
+            pin.name = self.name
+            
         debug.check(len(new_pin_list) > 0,
                     "Did not find any enclosures.")
 
diff --git a/compiler/router/router.py b/compiler/router/router.py
index 5213af4a..ca077572 100644
--- a/compiler/router/router.py
+++ b/compiler/router/router.py
@@ -28,7 +28,7 @@ class router(router_tech):
     route on a given layer. This is limited to two layer routes.
     It populates blockages on a grid class.
     """
-    def __init__(self, layers, design, gds_filename=None, bbox=None, margin=0, route_track_width=1):
+    def __init__(self, layers, design, bbox=None, margin=0, route_track_width=1):
         """
         This will instantiate a copy of the gds file or the module at (0,0) and
         route on top of this. The blockages from the gds/module will be
@@ -39,19 +39,7 @@ class router(router_tech):
 
         self.cell = design
 
-        # If didn't specify a gds blockage file, write it out to read the gds
-        # This isn't efficient, but easy for now
-        # start_time = datetime.now()
-        if not gds_filename:
-            gds_filename = OPTS.openram_temp+"temp.gds"
-            self.cell.gds_write(gds_filename)
-
-        # Load the gds file and read in all the shapes
-        self.layout = gdsMill.VlsiLayout(units=GDS["unit"])
-        self.reader = gdsMill.Gds2reader(self.layout)
-        self.reader.loadFromFile(gds_filename)
-        self.top_name = self.layout.rootStructureName
-        # print_time("GDS read",datetime.now(), start_time)
+        self.gds_filename = OPTS.openram_temp + "temp.gds"
 
         # The pin data structures
         # A map of pin names to a set of pin_layout structures
@@ -91,6 +79,16 @@ class router(router_tech):
         """
         Initialize the ll,ur values with the paramter or using the layout boundary.
         """
+
+        # If didn't specify a gds blockage file, write it out to read the gds
+        # This isn't efficient, but easy for now
+        # Load the gds file and read in all the shapes
+        self.cell.gds_write(self.gds_filename)
+        self.layout = gdsMill.VlsiLayout(units=GDS["unit"])
+        self.reader = gdsMill.Gds2reader(self.layout)
+        self.reader.loadFromFile(self.gds_filename)
+        self.top_name = self.layout.rootStructureName
+        
         if not bbox:
             # The boundary will determine the limits to the size
             # of the routing grid
@@ -178,6 +176,17 @@ class router(router_tech):
         """
         Find the pins and blockages in the design
         """
+
+        # If didn't specify a gds blockage file, write it out to read the gds
+        # This isn't efficient, but easy for now
+        # Load the gds file and read in all the shapes
+        self.cell.gds_write(self.gds_filename)
+        self.layout = gdsMill.VlsiLayout(units=GDS["unit"])
+        self.reader = gdsMill.Gds2reader(self.layout)
+        self.reader.loadFromFile(self.gds_filename)
+        self.top_name = self.layout.rootStructureName
+        # print_time("GDS read",datetime.now(), start_time)
+        
         # This finds the pin shapes and sorts them into "groups" that
         # are connected. This must come before the blockages, so we
         # can not count the pins themselves
@@ -881,12 +890,32 @@ class router(router_tech):
         # Clearing the blockage of this pin requires the inflated pins
         self.clear_blockages(pin_name)
 
+    def add_side_supply_pin(self, name, side="left", width=2):
+        """
+        Adds a supply pin to the perimeter and resizes the bounding box.
+        """
+        pg = pin_group(name, [], self)
+        if name == "vdd":
+            offset = width
+        else:
+            offset = 0
+        
+        pg.grids = set(self.rg.get_perimeter_list(side=side,
+                                                  width=width,
+                                                  margin=self.margin,
+                                                  offset=offset,
+                                                  layers=[1]))
+        pg.enclosures = pg.compute_enclosures()
+        pg.pins = set(pg.enclosures)
+        self.cell.pin_map[name].update(pg.pins)
+        self.pin_groups[name].append(pg)
+        
     def add_perimeter_target(self, side="all"):
         """
         This will mark all the cells on the perimeter of the original layout as a target.
         """
         self.rg.add_perimeter_target(side=side)
-    
+
     def num_pin_components(self, pin_name):
         """
         This returns how many disconnected pin components there are.
@@ -1214,12 +1243,12 @@ class router(router_tech):
                 return self.convert_track_to_pin(v)
             
         return None
-            
+
     def get_ll_pin(self, pin_name):
         """ Return the lowest, leftest pin group """
 
         keep_pin = None
-        for index,pg in enumerate(self.pin_groups[pin_name]):
+        for index, pg in enumerate(self.pin_groups[pin_name]):
             for pin in pg.enclosures:
                 if not keep_pin:
                     keep_pin = pin
@@ -1228,7 +1257,7 @@ class router(router_tech):
                         keep_pin = pin
                         
         return keep_pin
-
+    
     def check_all_routed(self, pin_name):
         """
         Check that all pin groups are routed.
diff --git a/compiler/router/router_tech.py b/compiler/router/router_tech.py
index b3e26f79..6cbbd422 100644
--- a/compiler/router/router_tech.py
+++ b/compiler/router/router_tech.py
@@ -123,7 +123,7 @@ class router_tech:
 
         min_wire_width = drc("minwidth_{0}".format(layer_name), 0, math.inf)
 
-        min_width = drc("minwidth_{0}".format(layer_name), self.route_track_width * min_wire_width, math.inf)
+        min_width = self.route_track_width * drc("minwidth_{0}".format(layer_name), self.route_track_width * min_wire_width, math.inf)
         min_spacing = drc(str(layer_name)+"_to_"+str(layer_name), self.route_track_width * min_wire_width, math.inf)
 
         return (min_width, min_spacing)
diff --git a/compiler/router/signal_escape_router.py b/compiler/router/signal_escape_router.py
index a9531290..7cc41d97 100644
--- a/compiler/router/signal_escape_router.py
+++ b/compiler/router/signal_escape_router.py
@@ -17,7 +17,7 @@ class signal_escape_router(router):
     A router that routes signals to perimeter and makes pins.
     """
 
-    def __init__(self, layers, design, bbox=None, margin=0, gds_filename=None):
+    def __init__(self, layers, design, bbox=None, margin=0):
         """
         This will route on layers in design. It will get the blockages from
         either the gds file name or the design itself (by saving to a gds file).
@@ -25,7 +25,6 @@ class signal_escape_router(router):
         router.__init__(self,
                         layers=layers,
                         design=design,
-                        gds_filename=gds_filename,
                         bbox=bbox,
                         margin=margin)
 
diff --git a/compiler/router/signal_router.py b/compiler/router/signal_router.py
index 89edcb01..9fbf871f 100644
--- a/compiler/router/signal_router.py
+++ b/compiler/router/signal_router.py
@@ -15,12 +15,12 @@ class signal_router(router):
     route on a given layer. This is limited to two layer routes.
     """
 
-    def __init__(self, layers, design, gds_filename=None, bbox=None):
+    def __init__(self, layers, design, bbox=None):
         """
         This will route on layers in design. It will get the blockages from
         either the gds file name or the design itself (by saving to a gds file).
         """
-        router.__init__(self, layers, design, gds_filename, bbox)
+        router.__init__(self, layers, design, bbox)
 
     def route(self, src, dest, detour_scale=5):
         """
diff --git a/compiler/router/supply_grid_router.py b/compiler/router/supply_grid_router.py
index 8a201474..f24498ab 100644
--- a/compiler/router/supply_grid_router.py
+++ b/compiler/router/supply_grid_router.py
@@ -21,7 +21,7 @@ class supply_grid_router(router):
     routes a grid to connect the supply on the two layers.
     """
 
-    def __init__(self, layers, design, gds_filename=None, bbox=None):
+    def __init__(self, layers, design, margin=0, bbox=None):
         """
         This will route on layers in design. It will get the blockages from
         either the gds file name or the design itself (by saving to a gds file).
@@ -29,9 +29,9 @@ class supply_grid_router(router):
         start_time = datetime.now()
 
         # Power rail width in minimum wire widths
-        self.route_track_width = 2
+        self.route_track_width = 1
 
-        router.__init__(self, layers, design, gds_filename, bbox, self.route_track_width)
+        router.__init__(self, layers, design, bbox=bbox, margin=margin, route_track_width=self.route_track_width)
 
         # The list of supply rails (grid sets) that may be routed
         self.supply_rails = {}
@@ -357,8 +357,9 @@ class supply_grid_router(router):
 
             # This is inefficient since it is non-incremental, but it was
             # easier to debug.
-            self.prepare_blockages(pin_name)
-
+            self.prepare_blockages()
+            self.clear_blockages(self.vdd_name)
+            
             # Add the single component of the pin as the source
             # which unmarks it as a blockage too
             self.add_pin_component_source(pin_name, index)
@@ -369,7 +370,7 @@ 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")
 
             # if index==3 and pin_name=="vdd":
             #     self.write_debug_gds("route.gds",False)
diff --git a/compiler/router/supply_tree_router.py b/compiler/router/supply_tree_router.py
index ba54ee39..97ba87d5 100644
--- a/compiler/router/supply_tree_router.py
+++ b/compiler/router/supply_tree_router.py
@@ -21,7 +21,7 @@ class supply_tree_router(router):
     routes a grid to connect the supply on the two layers.
     """
 
-    def __init__(self, layers, design, gds_filename=None, bbox=None):
+    def __init__(self, layers, design, bbox=None, side_pin=None):
         """
         This will route on layers in design. It will get the blockages from
         either the gds file name or the design itself (by saving to a gds file).
@@ -31,11 +31,19 @@ class supply_tree_router(router):
         # for prettier routes.
         self.route_track_width = 1
 
-        router.__init__(self, layers, design, gds_filename, bbox, self.route_track_width)
+        # The pin escape router already made the bounding box big enough,
+        # so we can use the regular bbox here.
+        self.side_pin = side_pin
+        router.__init__(self,
+                        layers,
+                        design,
+                        bbox=bbox,
+                        route_track_width=self.route_track_width)
 
     def route(self, vdd_name="vdd", gnd_name="gnd"):
         """
-        Route the two nets in a single layer)
+        Route the two nets in a single layer. 
+        Setting pin stripe will make a power rail on the left side.
         """
         debug.info(1, "Running supply router on {0} and {1}...".format(vdd_name, gnd_name))
         self.vdd_name = vdd_name
@@ -50,11 +58,17 @@ class supply_tree_router(router):
             # but this is simplest for now.
             self.create_routing_grid(signal_grid)
 
-        # Get the pin shapes
         start_time = datetime.now()
+
+        # Get the pin shapes
         self.find_pins_and_blockages([self.vdd_name, self.gnd_name])
         print_time("Finding pins and blockages", datetime.now(), start_time, 3)
 
+        # Add side pins if enabled
+        if self.side_pin:
+            self.add_side_supply_pin(self.vdd_name)
+            self.add_side_supply_pin(self.gnd_name)
+        
         # Route the supply pins to the supply rails
         # Route vdd first since we want it to be shorter
         start_time = datetime.now()
@@ -87,16 +101,21 @@ class supply_tree_router(router):
         pin_size = len(self.pin_groups[pin_name])
         adj_matrix = [[0] * pin_size for i in range(pin_size)]
 
-        for index1,pg1 in enumerate(self.pin_groups[pin_name]):
-            for index2,pg2 in enumerate(self.pin_groups[pin_name]):
+        for index1, pg1 in enumerate(self.pin_groups[pin_name]):
+            for index2, pg2 in enumerate(self.pin_groups[pin_name]):
                 if index1>=index2:
                     continue
                 dist = int(grid_utils.distance_set(list(pg1.grids)[0], pg2.grids))
                 adj_matrix[index1][index2] = dist
 
         # Find MST
-        debug.info(2, "Finding MinimumSpanning Tree")
+        debug.info(2, "Finding Minimum Spanning Tree")
         X = csr_matrix(adj_matrix)
+        from scipy.sparse import save_npz
+        #print("Saving {}.npz".format(self.cell.name))
+        #save_npz("{}.npz".format(self.cell.name), X)
+        #exit(1)
+
         Tcsr = minimum_spanning_tree(X)
         mst = Tcsr.toarray().astype(int)
         connections = []
diff --git a/compiler/sram/sram_1bank.py b/compiler/sram/sram_1bank.py
index 6bc2cd43..f9ea1d43 100644
--- a/compiler/sram/sram_1bank.py
+++ b/compiler/sram/sram_1bank.py
@@ -120,8 +120,9 @@ class sram_1bank(sram_base):
         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.predecoder_height)
+        # It is above the control logic and the predecoder array
+        y_offset = max(self.control_logic_insts[port].uy(), self.bank.predecoder_top)
+
         self.row_addr_pos[port] = vector(x_offset, y_offset)
         self.row_addr_dff_insts[port].place(self.row_addr_pos[port])
 
@@ -130,7 +131,7 @@ class sram_1bank(sram_base):
             # 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
+            y_offset = self.bank.predecoder_bottom
             self.row_addr_pos[port] = vector(x_offset, y_offset)
             self.row_addr_dff_insts[port].place(self.row_addr_pos[port], mirror="XY")
 
@@ -419,11 +420,11 @@ class sram_1bank(sram_base):
         if len(route_map) > 0:
 
             # The write masks will have blockages on M1
-            if self.num_wmasks > 0 and port in self.write_ports:
-                layer_stack = self.m3_stack
-            else:
-                layer_stack = self.m1_stack
-
+            # if self.num_wmasks > 0 and port in self.write_ports:
+            #     layer_stack = self.m3_stack
+            # else:
+            #     layer_stack = self.m1_stack
+            layer_stack = self.m3_stack
             if port == 0:
                 offset = vector(self.control_logic_insts[port].rx() + self.dff.width,
                                 - self.data_bus_size[port] + 2 * self.m3_pitch)
@@ -527,13 +528,13 @@ class sram_1bank(sram_base):
             # 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_{0}_{0}".format(port))
-            self.add_wire(self.m2_stack[::-1],
+            self.add_wire(self.m3_stack,
                           [src_pin.center(), vector(src_pin.cx(), dest_pin.cy()), dest_pin.rc()])
             self.add_via_stack_center(from_layer=src_pin.layer,
-                                      to_layer="m2",
+                                      to_layer="m4",
                                       offset=src_pin.center())
             self.add_via_stack_center(from_layer=dest_pin.layer,
-                                      to_layer="m2",
+                                      to_layer="m3",
                                       offset=dest_pin.center())
 
     def route_row_addr_dff(self):
diff --git a/compiler/sram/sram_base.py b/compiler/sram/sram_base.py
index ca9223e1..28234bf7 100644
--- a/compiler/sram/sram_base.py
+++ b/compiler/sram/sram_base.py
@@ -225,11 +225,7 @@ class sram_base(design, verilog, lef):
         for pin_name in ["vdd", "gnd"]:
             for inst in self.insts:
                 self.copy_power_pins(inst, pin_name)
-
-        if not OPTS.route_supplies:
-            # Do not route the power supply (leave as must-connect pins)
-            return
-
+        
         try:
             from tech import power_grid
             grid_stack = power_grid
@@ -238,43 +234,99 @@ class sram_base(design, verilog, lef):
             # Route a M3/M4 grid
             grid_stack = self.m3_stack
 
-        if OPTS.route_supplies == "grid":
-            from supply_grid_router import supply_grid_router as router
-        elif OPTS.route_supplies:
-            from supply_tree_router import supply_tree_router as router
+        # lowest_coord = self.find_lowest_coords()
+        # highest_coord = self.find_highest_coords()
             
-        rtr=router(grid_stack, self)
+        # # Add two rails to the side
+        # if OPTS.route_supplies == "side":
+        #     supply_pins = {}
+        #     # Find the lowest leftest pin for vdd and gnd
+        #     for (pin_name, pin_index) in [("vdd", 0), ("gnd", 1)]:
+        #         pin_width = 8 * getattr(self, "{}_width".format(grid_stack[2]))
+        #         pin_space = 2 * getattr(self, "{}_space".format(grid_stack[2]))
+        #         supply_pitch = pin_width + pin_space
+
+        #         # Add side power rails on left from bottom to top
+        #         # These have a temporary name and will be connected later.
+        #         # They are here to reserve space now and ensure other pins go beyond
+        #         # their perimeter.
+        #         supply_height = highest_coord.y - lowest_coord.y
+                
+        #         supply_pins[pin_name] = self.add_layout_pin(text=pin_name,
+        #                                                     layer=grid_stack[2],
+        #                                                     offset=lowest_coord + vector(pin_index * supply_pitch, 0),
+        #                                                     width=pin_width,
+        #                                                     height=supply_height)
+
+        if not OPTS.route_supplies:
+            # Do not route the power supply (leave as must-connect pins)
+            return
+        elif OPTS.route_supplies == "grid":
+            from supply_grid_router import supply_grid_router as router
+        else:
+            from supply_tree_router import supply_tree_router as router
+
+        rtr=router(grid_stack, self, side_pin=(OPTS.route_supplies == "side"))
         rtr.route()
 
-        # Find the lowest leftest pin for vdd and gnd
-        for pin_name in ["vdd", "gnd"]:
-            # Copy the pin shape(s) to rectangles
-            for pin in self.get_pins(pin_name):
+        if OPTS.route_supplies == "side":
+            # Find the lowest leftest pin for vdd and gnd
+            for pin_name in ["vdd", "gnd"]:
+                # Copy the pin shape(s) to rectangles
+                for pin in self.get_pins(pin_name):
+                    self.add_rect(pin.layer,
+                                  pin.ll(),
+                                  pin.width(),
+                                  pin.height())
+
+                # Remove the pin shape(s)
+                self.remove_layout_pin(pin_name)
+
+                # Get the lowest, leftest pin
+                pin = rtr.get_ll_pin(pin_name)
+                self.add_layout_pin(pin_name,
+                                    pin.layer,
+                                    pin.ll(),
+                                    pin.width(),
+                                    pin.height())
+            
+        elif OPTS.route_supplies:
+            # Update these as we may have routed outside the region (perimeter pins)
+            lowest_coord = self.find_lowest_coords()
+        
+            # Find the lowest leftest pin for vdd and gnd
+            for pin_name in ["vdd", "gnd"]:
+                # Copy the pin shape(s) to rectangles
+                for pin in self.get_pins(pin_name):
+                    self.add_rect(pin.layer,
+                                  pin.ll(),
+                                  pin.width(),
+                                  pin.height())
+
+                # Remove the pin shape(s)
+                self.remove_layout_pin(pin_name)
+
+                # Get the lowest, leftest pin
+                pin = rtr.get_ll_pin(pin_name)
+
+                pin_width = 2 * getattr(self, "{}_width".format(pin.layer))
+
+                # Add it as an IO pin to the perimeter
+                route_width = pin.rx() - lowest_coord.x
+                pin_offset = vector(lowest_coord.x, pin.by())
                 self.add_rect(pin.layer,
-                              pin.ll(),
-                              pin.width(),
+                              pin_offset,
+                              route_width,
                               pin.height())
 
-            # Remove the pin shape(s)
-            self.remove_layout_pin(pin_name)
-
-            # Get the lowest, leftest pin
-            pin = rtr.get_ll_pin(pin_name)
-
-            # Add it as an IO pin to the perimeter
-            lowest_coord = self.find_lowest_coords()
-            route_width = pin.rx() - lowest_coord.x
-            pin_width = 2 * getattr(self, "{}_width".format(pin.layer))
-            pin_offset = vector(lowest_coord.x, pin.by())
-            self.add_layout_pin(pin_name,
-                                pin.layer,
-                                pin_offset,
-                                pin_width,
-                                pin.height())
-            self.add_rect(pin.layer,
-                          pin_offset,
-                          route_width,
-                          pin.height())
+                self.add_layout_pin(pin_name,
+                                    pin.layer,
+                                    pin_offset,
+                                    pin_width,
+                                    pin.height())
+        else:
+            # Grid is left with many top level pins
+            pass
 
     def route_escape_pins(self):
         """
@@ -319,7 +371,7 @@ class sram_base(design, verilog, lef):
         from signal_escape_router import signal_escape_router as router
         rtr=router(layers=self.m3_stack,
                    design=self,
-                   margin=4 * self.m3_pitch)
+                   margin=8 * self.m3_pitch)
         rtr.escape_route(pins_to_route)
 
     def compute_bus_sizes(self):
diff --git a/compiler/tests/golden/sram_2_16_1_freepdk45.v b/compiler/tests/golden/sram_2_16_1_freepdk45.v
index 46f89fa5..4441b717 100644
--- a/compiler/tests/golden/sram_2_16_1_freepdk45.v
+++ b/compiler/tests/golden/sram_2_16_1_freepdk45.v
@@ -3,6 +3,10 @@
 // Word size: 2
 
 module sram_2_16_1_freepdk45(
+`ifdef USE_POWER_PINS
+    vdd,
+    gnd,
+`endif
 // Port 0: RW
     clk0,csb0,web0,addr0,din0,dout0
   );
@@ -15,6 +19,10 @@ module sram_2_16_1_freepdk45(
   parameter VERBOSE = 1 ; //Set to 0 to only display warnings
   parameter T_HOLD = 1 ; //Delay to hold dout value after posedge. Value is arbitrary
 
+`ifdef USE_POWER_PINS
+    inout vdd;
+    inout gnd;
+`endif
   input  clk0; // clock
   input   csb0; // active low chip select
   input  web0; // active low write control
diff --git a/compiler/tests/golden/sram_2_16_1_scn4m_subm.v b/compiler/tests/golden/sram_2_16_1_scn4m_subm.v
index cb95036b..fd77a66e 100644
--- a/compiler/tests/golden/sram_2_16_1_scn4m_subm.v
+++ b/compiler/tests/golden/sram_2_16_1_scn4m_subm.v
@@ -3,6 +3,10 @@
 // Word size: 2
 
 module sram_2_16_1_scn4m_subm(
+`ifdef USE_POWER_PINS
+    vdd,
+    gnd,
+`endif
 // Port 0: RW
     clk0,csb0,web0,addr0,din0,dout0
   );
@@ -15,6 +19,10 @@ module sram_2_16_1_scn4m_subm(
   parameter VERBOSE = 1 ; //Set to 0 to only display warnings
   parameter T_HOLD = 1 ; //Delay to hold dout value after posedge. Value is arbitrary
 
+`ifdef USE_POWER_PINS
+    inout vdd;
+    inout gnd;
+`endif
   input  clk0; // clock
   input   csb0; // active low chip select
   input  web0; // active low write control