lots more work on klayout approach

This commit is contained in:
Jan Petykiewicz 2023-06-09 00:38:19 -07:00
commit 0ffe18c9f1
9 changed files with 303 additions and 170 deletions

View file

@ -1,7 +1,6 @@
from typing import Sequence, Callable
from typing import Sequence, Iterable
import logging
from collections import Counter
from dataclasses import dataclass
from itertools import chain
from klayout import db
@ -11,83 +10,151 @@ from .types import lnum_t, layer_t
logger = logging.getLogger(__name__)
def get_topcell(
layout: db.Layout,
name: str | None = None,
) -> db.Cell:
if name is None:
return layout.top_cell()
else:
ind = layout.cell_by_name(name)
return layout.cell(ind)
class TraceAnalysis:
"""
Short/Open analysis for a list of nets
"""
def write_net_layout(
l2n: db.LayoutToNetlist,
filepath: str,
layers: Sequence[lnum_t],
) -> None:
layout = db.Layout()
top = layout.create_cell('top')
lmap = {layout.layer(*layer) for layer in layers}
l2n.build_all_nets(l2n.cell_mapping_into(ly, top), ly, lmap, 'net_', 'prop_', l2n.BNH_Flatten, 'circuit_')
layout.write(filepath)
def merge_labels_from(
filepath: str,
into_layout: db.Layout,
lnum_map: dict[lnum_t, lnum_t],
topcell: str | None = None,
) -> None:
layout = db.Layout()
lm = layout.read(filepath)
topcell_obj = get_topcell(layout, topcell)
for labels_layer, conductor_layer in lnum_map:
layer_ind_src = layout.layer(*labels_layer)
layer_ind_dst = into_layout.layer(*conductor_layer)
shapes_dst = topcell_obj.shapes(layer_ind_dst)
shapes_src = topcell_obj.shapes(layer_ind_src)
for shape in shapes_dst.each():
new_shape = shapes_dst.insert(shape)
shapes_dst.replace_prop_id(new_shape, 0) # clear shape properties
@dataclass
class TraceResult:
shorts: list[str]
opens: list[str]
nets: list[set[str]]
""" List of nets (connected sets of labels) """
opens: dict[str, int]
""" Labels which appear on 2+ disconnected nets, and the number of nets they touch """
shorts: list[set[str]]
""" Nets containing more than one unique label """
def __init__(
self,
nets: Sequence[Iterable[str]],
) -> None:
"""
Args:
nets: Sequence of nets. Each net is a sequence of labels
which were found to be electrically connected.
"""
setnets = [set(net) for net in nets]
# Shorts contain more than one label
shorts = [net for net in setnets if len(net) > 1]
# Check number of times each label appears
net_occurences = Counter(chain.from_iterable(setnets))
# Opens are where the same label appears on more than one net
opens = {
nn: count
for nn, count in net_occurences.items()
if count > 1
}
self.nets = setnets
self.shorts = shorts
self.opens = opens
def __repr__(self) -> str:
def format_net(net: Iterable[str]) -> str:
names = [f"'{name}'" if any(cc in name for cc in ' \t\n') else name for name in sorted(net)]
return ','.join(names)
def sort_nets(nets: Sequence[Iterable[str]]) -> list[Iterable[str]]:
return sorted(nets, key=lambda net: ','.join(sorted(net)))
ss = 'Trace analysis'
ss += '\n============='
ss += '\nNets'
ss += '\n(groups of electrically connected labels)\n'
for net in sort_nets(self.nets):
ss += '\t' + format_net(net) + '\n'
ss += '\nOpens'
ss += '\n(2+ nets containing the same name)\n'
for label, count in sorted(self.opens.items()):
ss += f'\t{label} : {count} nets\n'
ss += '\nShorts'
ss += '\n(2+ unique names for the same net)\n'
for net in sort_nets(self.shorts):
ss += '\t' + format_net(net) + '\n'
ss += '=============\n'
return ss
def trace_layout(
filepath: str,
connectivity: list[layer_t, layer_t | None, layer_t],
connectivity: Sequence[tuple[layer_t, layer_t | None, layer_t]],
layer_map: dict[str, lnum_t] | None = None,
topcell: str | None = None,
*,
labels_map: dict[layer_t, layer_t] = {},
labels_map: dict[layer_t, layer_t] | None = None,
lfile_path: str | None = None,
lfile_map: dict[layer_t, layer_t] | None = None,
lfile_layer_map: dict[str, lnum_t] | None = None,
lfile_topcell: str | None = None,
output_path: str | None = None,
parse_label: Callable[[str], str] | None = None,
) -> TraceResult:
) -> list[set[str]]:
"""
Trace a layout to identify labeled nets.
To label a net, place a text label anywhere touching the net.
Labels may be mapped from a different layer, or even a different
layout file altogether.
Note: Labels must not contain commas (,)!!
Args:
filepath: Path to the primary layout, containing the conductor geometry
(and optionally also the labels)
connectivity: List of (conductor1, via12, conductor2) tuples,
which indicate that the specified layers are electrically connected
(conductor1 to via12 and via12 to conductor2). The middle (via) layer
may be `None`, in which case the outer layers are directly connected
at any overlap (conductor1 to conductor2).
layer_map: {layer_name: (layer_num, dtype_num)} translation table.
Should contain any strings present in `connectivity` and `labels_map`.
Default is an empty dict.
topcell: Cell name of the topcell. If `None`, it is automatically chosen.
labels_map: {label_layer: metal_layer} mapping, which allows labels to
reside on a different layer from their corresponding metals.
Only labels on the provided label layers are used, so
{metal_layer: metal_layer} entries must be explicitly specified if
they are desired.
If `None`, labels on each layer in `connectivity` are used alongside
that same layer's geometry ({layer: layer} for all participating
geometry layers)
Default `None`.
lfile_path: Path to a separate file from which labels should be merged.
lfile_map: {lfile_layer: primary_layer} mapping, used when merging the
labels into the primary layout.
lfile_layer_map: {layer_name: (layer_num, dtype_num)} mapping for the
secondary (label) file. Should contain all string keys in
`lfile_map`.
`None` reuses `layer_map` (default).
lfile_topcell: Cell name for the topcell in the secondary (label) file.
`None` automatically chooses the topcell (default).
output_path: If provided, outputs the final net geometry to a layout
at the given path. Default `None`.
Returns:
List of labeled nets, where each entry is a set of label strings which
were found on the given net.
"""
if layer_map is None:
layer_map = {}
if parse_label is None:
def parse_label(label: str) -> str:
return label
if labels_map is None:
labels_map = {
layer: layer
for layer in chain(*connectivity)
if layer is not None
}
layout = db.Layout()
lm = layout.read(filepath)
layout.read(filepath)
topcell_obj = get_topcell(layout, topcell)
topcell_obj = _get_topcell(layout, topcell)
# Merge labels from a separate layout if asked
if lfile_path:
@ -106,7 +173,7 @@ def trace_layout(
lshape = layer_map[lshape]
lnum_map[ltext] = lshape
merge_labels_from(lfile_path, layout, lnum_map, lfile_topcell)
_merge_labels_from(lfile_path, layout, lnum_map, lfile_topcell)
#
# Build a netlist from the layout
@ -117,6 +184,8 @@ def trace_layout(
# Create l2n polygon layers
layer2polys = {}
for layer in set(chain(*connectivity)):
if layer is None:
continue
if isinstance(layer, str):
layer = layer_map[layer]
klayer = layout.layer(*layer)
@ -143,7 +212,7 @@ def trace_layout(
top = layer_map[top]
if isinstance(via, str):
via = layer_map[via]
if isinstance(top, str):
if isinstance(bot, str):
bot = layer_map[bot]
if via is None:
@ -162,43 +231,78 @@ def trace_layout(
l2n.connect(layer2polys[metal_layer], layer2texts[label_layer])
# Get netlist
nle = l2n.extract_netlist()
l2n.extract_netlist()
nl = l2n.netlist()
nl.make_top_level_pins()
if output_path:
write_net_layout(l2n, output_path, layer2polys.keys())
_write_net_layout(l2n, output_path, layer2polys)
#
# Analyze traced nets
# Return merged nets
#
top_circuits = [cc for cc, _ in zip(nl.each_circuit_top_down(), range(nl.top_circuit_count()))]
# Nets with more than one label get their labels joined with a comma
nets = [
{parse_label(ll) for ll in nn.name.split(',')}
nets = [
set(nn.name.split(','))
for cc in top_circuits
for nn in cc.each_net()
if nn.name
]
nets2 = [
nn.name
for cc in top_circuits
for nn in cc.each_net()
]
print(nets2)
return nets
# Shorts contain more than one label
shorts = [net for net in nets if len(net) > 1]
# Check number of times each label appears
net_occurences = Counter(chain.from_iterable(nets))
def _get_topcell(
layout: db.Layout,
name: str | None = None,
) -> db.Cell:
"""
Get the topcell by name or hierarchy.
# If the same label appears on more than one net, warn about an open
opens = [
(nn, count)
for nn, count in net_occurences.items()
if count > 1
]
Args:
layout: Layout to get the cell from
name: If given, use the name to find the topcell; otherwise use hierarchy.
return TraceResult(shorts=shorts, opens=opens, nets=nets)
Returns:
Cell object
"""
if name is None:
return layout.top_cell()
else:
ind = layout.cell_by_name(name)
return layout.cell(ind)
def _write_net_layout(
l2n: db.LayoutToNetlist,
filepath: str,
layer2polys: dict[lnum_t, db.Region],
) -> None:
layout = db.Layout()
top = layout.create_cell('top')
lmap = {layout.layer(*layer): polys for layer, polys in layer2polys.items()}
l2n.build_all_nets(l2n.cell_mapping_into(layout, top), layout, lmap, 'net_', 'prop_', l2n.BNH_Flatten, 'circuit_')
layout.write(filepath)
def _merge_labels_from(
filepath: str,
into_layout: db.Layout,
lnum_map: dict[lnum_t, lnum_t],
topcell: str | None = None,
) -> None:
layout = db.Layout()
layout.read(filepath)
topcell_obj = _get_topcell(layout, topcell)
for labels_layer, conductor_layer in lnum_map:
layer_ind_src = layout.layer(*labels_layer)
layer_ind_dst = into_layout.layer(*conductor_layer)
shapes_dst = topcell_obj.shapes(layer_ind_dst)
shapes_src = topcell_obj.shapes(layer_ind_src)
for shape in shapes_src.each():
new_shape = shapes_dst.insert(shape)
shapes_dst.replace_prop_id(new_shape, 0) # clear shape properties