mirror of https://github.com/VLSIDA/OpenRAM.git
278 lines
11 KiB
Python
278 lines
11 KiB
Python
# See LICENSE for licensing information.
|
|
#
|
|
# Copyright (c) 2016-2019 Regents of the University of California and The Board
|
|
# of Regents for the Oklahoma Agricultural and Mechanical College
|
|
# (acting for and on behalf of Oklahoma State University)
|
|
# All rights reserved.
|
|
#
|
|
import contact
|
|
import pgate
|
|
import debug
|
|
from tech import drc, parameter, spice
|
|
from vector import vector
|
|
from globals import OPTS
|
|
import logical_effort
|
|
from sram_factory import factory
|
|
|
|
class pnand2(pgate.pgate):
|
|
"""
|
|
This module generates gds of a parametrically sized 2-input nand.
|
|
This model use ptx to generate a 2-input nand within a cetrain height.
|
|
"""
|
|
def __init__(self, name, size=1, height=None):
|
|
""" Creates a cell for a simple 2 input nand """
|
|
|
|
debug.info(2, "creating pnand2 structure {0} with size of {1}".format(name, size))
|
|
self.add_comment("size: {}".format(size))
|
|
|
|
self.size = size
|
|
self.nmos_size = 2*size
|
|
self.pmos_size = parameter["beta"]*size
|
|
self.nmos_width = self.nmos_size*drc("minwidth_tx")
|
|
self.pmos_width = self.pmos_size*drc("minwidth_tx")
|
|
|
|
# FIXME: Allow these to be sized
|
|
debug.check(size==1,"Size 1 pnand2 is only supported now.")
|
|
self.tx_mults = 1
|
|
|
|
# Creates the netlist and layout
|
|
pgate.pgate.__init__(self, name, height)
|
|
|
|
#For characterization purposes only
|
|
#self.exclude_nmos_from_graph()
|
|
|
|
def create_netlist(self):
|
|
self.add_pins()
|
|
self.add_ptx()
|
|
self.create_ptx()
|
|
|
|
def create_layout(self):
|
|
""" Calls all functions related to the generation of the layout """
|
|
|
|
self.setup_layout_constants()
|
|
self.route_supply_rails()
|
|
self.place_ptx()
|
|
self.connect_rails()
|
|
self.add_well_contacts()
|
|
self.extend_wells(self.well_pos)
|
|
self.route_inputs()
|
|
self.route_output()
|
|
|
|
def add_pins(self):
|
|
""" Adds pins for spice netlist """
|
|
pin_list = ["A", "B", "Z", "vdd", "gnd"]
|
|
dir_list = ["INPUT", "INPUT", "OUTPUT", "POWER", "GROUND"]
|
|
self.add_pin_list(pin_list, dir_list)
|
|
|
|
|
|
def add_ptx(self):
|
|
""" Create the PMOS and NMOS transistors. """
|
|
self.nmos = factory.create(module_type="ptx",
|
|
width=self.nmos_width,
|
|
mults=self.tx_mults,
|
|
tx_type="nmos",
|
|
connect_poly=True,
|
|
connect_active=True)
|
|
self.add_mod(self.nmos)
|
|
|
|
self.pmos = factory.create(module_type="ptx",
|
|
width=self.pmos_width,
|
|
mults=self.tx_mults,
|
|
tx_type="pmos",
|
|
connect_poly=True,
|
|
connect_active=True)
|
|
self.add_mod(self.pmos)
|
|
|
|
def setup_layout_constants(self):
|
|
""" Pre-compute some handy layout parameters. """
|
|
|
|
# metal spacing to allow contacts on any layer
|
|
self.input_spacing = max(self.poly_space + contact.poly.first_layer_width,
|
|
self.m1_space + contact.m1m2.first_layer_width,
|
|
self.m2_space + contact.m2m3.first_layer_width,
|
|
self.m3_space + contact.m2m3.second_layer_width)
|
|
|
|
|
|
# Compute the other pmos2 location, but determining offset to overlap the
|
|
# source and drain pins
|
|
self.overlap_offset = self.pmos.get_pin("D").ll() - self.pmos.get_pin("S").ll()
|
|
|
|
# Two PMOS devices and a well contact. Separation between each.
|
|
# Enclosure space on the sides.
|
|
self.well_width = 2*self.pmos.active_width + contact.active.width \
|
|
+ 2*drc("active_to_body_active") + 2*drc("well_enclosure_active")
|
|
|
|
self.width = self.well_width
|
|
# Height is an input parameter, so it is not recomputed.
|
|
|
|
# This is the extra space needed to ensure DRC rules to the active contacts
|
|
extra_contact_space = max(-self.nmos.get_pin("D").by(),0)
|
|
# This is a poly-to-poly of a flipped cell
|
|
self.top_bottom_space = max(0.5*self.m1_width + self.m1_space + extra_contact_space,
|
|
drc("poly_extend_active"), self.poly_space)
|
|
|
|
def route_supply_rails(self):
|
|
""" Add vdd/gnd rails to the top and bottom. """
|
|
self.add_layout_pin_rect_center(text="gnd",
|
|
layer="metal1",
|
|
offset=vector(0.5*self.width,0),
|
|
width=self.width)
|
|
|
|
self.add_layout_pin_rect_center(text="vdd",
|
|
layer="metal1",
|
|
offset=vector(0.5*self.width,self.height),
|
|
width=self.width)
|
|
|
|
def create_ptx(self):
|
|
"""
|
|
Add PMOS and NMOS to the netlist.
|
|
"""
|
|
|
|
self.pmos1_inst=self.add_inst(name="pnand2_pmos1",
|
|
mod=self.pmos)
|
|
self.connect_inst(["vdd", "A", "Z", "vdd"])
|
|
|
|
self.pmos2_inst = self.add_inst(name="pnand2_pmos2",
|
|
mod=self.pmos)
|
|
self.connect_inst(["Z", "B", "vdd", "vdd"])
|
|
|
|
self.nmos1_inst=self.add_inst(name="pnand2_nmos1",
|
|
mod=self.nmos)
|
|
self.connect_inst(["Z", "B", "net1", "gnd"])
|
|
|
|
self.nmos2_inst=self.add_inst(name="pnand2_nmos2",
|
|
mod=self.nmos)
|
|
self.connect_inst(["net1", "A", "gnd", "gnd"])
|
|
|
|
|
|
def place_ptx(self):
|
|
"""
|
|
Place PMOS and NMOS to the layout at the upper-most and lowest position
|
|
to provide maximum routing in channel
|
|
"""
|
|
|
|
pmos1_pos = vector(self.pmos.active_offset.x,
|
|
self.height - self.pmos.active_height - self.top_bottom_space)
|
|
self.pmos1_inst.place(pmos1_pos)
|
|
|
|
self.pmos2_pos = pmos1_pos + self.overlap_offset
|
|
self.pmos2_inst.place(self.pmos2_pos)
|
|
|
|
|
|
nmos1_pos = vector(self.pmos.active_offset.x, self.top_bottom_space)
|
|
self.nmos1_inst.place(nmos1_pos)
|
|
|
|
self.nmos2_pos = nmos1_pos + self.overlap_offset
|
|
self.nmos2_inst.place(self.nmos2_pos)
|
|
|
|
# Output position will be in between the PMOS and NMOS
|
|
self.output_pos = vector(0,0.5*(pmos1_pos.y+nmos1_pos.y+self.nmos.active_height))
|
|
|
|
# This will help with the wells
|
|
self.well_pos = vector(0,self.nmos1_inst.uy())
|
|
|
|
def add_well_contacts(self):
|
|
""" Add n/p well taps to the layout and connect to supplies AFTER the wells are created """
|
|
|
|
self.add_nwell_contact(self.pmos, self.pmos2_pos)
|
|
self.add_pwell_contact(self.nmos, self.nmos2_pos)
|
|
|
|
|
|
def connect_rails(self):
|
|
""" Connect the nmos and pmos to its respective power rails """
|
|
|
|
self.connect_pin_to_rail(self.nmos1_inst,"S","gnd")
|
|
|
|
self.connect_pin_to_rail(self.pmos1_inst,"S","vdd")
|
|
|
|
self.connect_pin_to_rail(self.pmos2_inst,"D","vdd")
|
|
|
|
def route_inputs(self):
|
|
""" Route the A and B inputs """
|
|
inputB_yoffset = self.nmos2_pos.y + self.nmos.active_height + self.m2_space + 0.5*self.m2_width
|
|
self.route_input_gate(self.pmos2_inst, self.nmos2_inst, inputB_yoffset, "B", position="center")
|
|
|
|
# This will help with the wells and the input/output placement
|
|
self.inputA_yoffset = inputB_yoffset + self.input_spacing
|
|
self.route_input_gate(self.pmos1_inst, self.nmos1_inst, self.inputA_yoffset, "A")
|
|
|
|
|
|
def route_output(self):
|
|
""" Route the Z output """
|
|
# PMOS1 drain
|
|
pmos_pin = self.pmos1_inst.get_pin("D")
|
|
top_pin_offset = pmos_pin.center()
|
|
# NMOS2 drain
|
|
nmos_pin = self.nmos2_inst.get_pin("D")
|
|
bottom_pin_offset = nmos_pin.center()
|
|
|
|
# Output pin
|
|
out_offset = vector(nmos_pin.center().x + self.m1_pitch,self.inputA_yoffset)
|
|
|
|
# Midpoints of the L routes go horizontal first then vertical
|
|
mid1_offset = vector(out_offset.x, top_pin_offset.y)
|
|
mid2_offset = vector(out_offset.x, bottom_pin_offset.y)
|
|
|
|
self.add_via_center(layers=("metal1", "via1", "metal2"),
|
|
offset=pmos_pin.center(),
|
|
directions=("V","H"))
|
|
self.add_via_center(layers=("metal1", "via1", "metal2"),
|
|
offset=nmos_pin.center(),
|
|
directions=("V","H"))
|
|
self.add_via_center(layers=("metal1", "via1", "metal2"),
|
|
offset=out_offset)
|
|
|
|
|
|
# PMOS1 to mid-drain to NMOS2 drain
|
|
self.add_path("metal2",[top_pin_offset, mid1_offset, out_offset, mid2_offset, bottom_pin_offset])
|
|
|
|
# This extends the output to the edge of the cell
|
|
self.add_layout_pin_rect_center(text="Z",
|
|
layer="metal1",
|
|
offset=out_offset,
|
|
width=contact.m1m2.first_layer_height,
|
|
height=contact.m1m2.first_layer_width)
|
|
|
|
def analytical_power(self, corner, load):
|
|
"""Returns dynamic and leakage power. Results in nW"""
|
|
c_eff = self.calculate_effective_capacitance(load)
|
|
freq = spice["default_event_rate"]
|
|
power_dyn = self.calc_dynamic_power(corner, c_eff, freq)
|
|
power_leak = spice["nand2_leakage"]
|
|
|
|
total_power = self.return_power(power_dyn, power_leak)
|
|
return total_power
|
|
|
|
def calculate_effective_capacitance(self, load):
|
|
"""Computes effective capacitance. Results in fF"""
|
|
c_load = load
|
|
c_para = spice["min_tx_drain_c"]*(self.nmos_size/parameter["min_tx_size"])#ff
|
|
transition_prob = spice["nand2_transition_prob"]
|
|
return transition_prob*(c_load + c_para)
|
|
|
|
def get_cin(self):
|
|
"""Return the relative input capacitance of a single input"""
|
|
return self.nmos_size+self.pmos_size
|
|
|
|
def input_load(self):
|
|
"""Return the relative input capacitance of a single input"""
|
|
return self.nmos_size+self.pmos_size
|
|
|
|
def get_stage_effort(self, cout, inp_is_rise=True):
|
|
"""Returns an object representing the parameters for delay in tau units.
|
|
Optional is_rise refers to the input direction rise/fall. Input inverted by this stage.
|
|
"""
|
|
parasitic_delay = 2
|
|
return logical_effort.logical_effort(self.name, self.size, self.get_cin(), cout, parasitic_delay, not inp_is_rise)
|
|
|
|
def exclude_nmos_from_graph(self):
|
|
"""Exclude the nmos TXs from the graph to reduce complexity"""
|
|
#The pull-down network has an internal net which causes 2 different in->out paths
|
|
#Removing them simplifies generic path searching.
|
|
self.graph_inst_exclude.add(self.nmos1_inst)
|
|
self.graph_inst_exclude.add(self.nmos2_inst)
|
|
|
|
def build_graph(self, graph, inst_name, port_nets):
|
|
"""Adds edges based on inputs/outputs. Overrides base class function."""
|
|
self.add_graph_edges(graph, port_nets)
|