"similar no connect" pins so that it is not done on top-level
netlists. This is mainly to deal with the problem where connections
that would normally be pins (but may have, for example, been
connected to a voltage source in a schematic that was deleted because
it was not a netlistable device) are treated as no-connects.
The parallel combination of devices with similar no-connects can then
differ between netlists that differ in describing transistors as
fingered vs. individual devices. This is an obscure case, but the
output of netgen then becomes not only wrong but difficult to
understand what happened, so it is better to avoid.
parallel and series devices. This was a fairly major error
undermining the property sorting (the case where the same number
of devices are in parallel in both circuits and need to be sorted
by, e.g., width, prior to checking for matches).
sides of the output when presenting side-by-side results. This
matches up net or device names within a partition. Where net or
device names match, the contents are also aligned on either side.
I attempted to also do a "best match" of contents between sides,
but as this involves a lot of analyzing the contents, it is very
computationally expensive, and so the code has been disabled. It
could be added back in as an option. There are also various ways
to optimize it for speed.
pins for enabling parallel combinations, which could cause a crash.
Added a "-force" option to "equate pins" to allow pins to be matched
even on subcircuits that did not correctly match; this was done in
conjuction with an extra option to the "lvs" command "-noflatten="
to pass a list of cellname to not be flattened even if they do not
match. This is generally discouraged, as it prevents netgen from
resolving differences between layout and schematic hierarchy, but it
can be useful for checking that the hierarchy above a certain cell
is correct, given that if a subcell is really unmatched, then its
errors will keep propagating up the top level, making additional
errors hard to diagnose.
can be turned on or off from the setup using "property parallel open"
to allow parallelizing devices with no-connect pins vs. "property
parallel connected" to only allow parallelizing of devices with all
pins connected.
devices in parallel with unconnected pins would be confused with
N devices in parallel with those pins all tied together. This is
treated as a property error.
to generate incorrect results on occasion. The method to parallelize
cells with the same no-connect pins should avoid the worst-case
symmetry breaking that was previously plaguing the LVS of large
standard-cell layouts.
parallel if the same pins are no-connects. These were previously not
treated as parallel because each no-connect has a unique node number,
and cells were only considered as parallel if all pins connected to
the same node numbers. This avoids issues with long-running symmetry
breaking on standard cell designs due to cells like antenna taps or
any cell that is placed without connecting it up. To do: This
makes indistinguishable certain cases, e.g., N cells in circuit 1
with pin X open vs. N cells in circuit 2 with pin X all tied together.
This could be caught during property matching.
to go into an infinite loop and fill memory until it crashes, due
to a complete pin mismatch between devices causing one device to
have its pins removed and replaced with proxy pins.
was running through the entire object linked list to find the
predecessor of a record that it had already found. Solved by simply
keeping track of the predecessor record.
format files to the SPICE parser, so that netgen can be run
directly on a testbench file and not generate errors due to
statements in the control block.
I have a design that is taking quite a long time for netgen to complete LVS
checking. Profiles show a large chunk of runtime is in the hash functions.
Some of the hashtables are very sparsely populated, but others are
heavily used. One hashtable has chains of over 250. Longer term it would
be worth investigating resizing the hashtables (or perhaps using other
data strutures), but for now I looked at what changing the number of
hash buckets (OBJHASHSIZE) does for performance:
OBJHASHSIZE time (mm:ss)
997 24:18
10093 4:42
42073 3:12
104729 2:51
I somewhat arbitrarily chose 42073 which gives us a 7.6x improvement in
runtime.
file input so that pins occur first before nodes, as they do in a
SPICE netlist. Certain parts of the comparison code depend on pins
being first in the netlist, and reordering them when reading input
is easier than rewriting the rest of the code.
introduced in revision 150 can result in an incorrect result
reporting a bad match where the match is actually good (as proven
by running the full symmetry breaking on the same netlist).
Because the fast symmetry breaking is orders of magnitude faster
for large circuits, and because the false positive result appears
to be rare, I have introduced a command "symmetry" to switch
methods between fast and full. So fast symmetry breaking can be
run unless the result fails on symmetry breaking, in which case
the method can be switched to full to see if the problem is a
false positive or not. This is not an ideal solution, and some
investigation is needed to determine if there is a way to apply
fast symmetry breaking without encountering a false positive
error.
with the file number, so that it can get confused between libraries.
Also made a fix to coerce one cell class to be forced to be the
same in both circuits under some circumstances.
the top level circuits to be declared matching with no errors even
though the pins do not match. "proxy pins" are fine for subcells
to detect cases where one subcell has an unused pin and the matching
subcell does not declare it, but that should not be allowed on the
top level, as it cannot be known whether the pin is unused or not.
corrected the error statement so that it refers both to the (corrected)
left-hand side and also the portion of the right-hand side that cannot
be parsed as structural verilog.
breaking by property was only matching properties between circuits but
not within the same circuit, which is needed for correct symmetry
breaking. But the PropertyMatch() routine assumed that it is passed
one item from each circuit, leading to a segfault when running the
symmetry breaking within a single circuit. This has been fixed.
that are not declared in the verilog netlist because they don't
connect to anything, and their presence is not required by verilog
syntax) and the printing of proxy pins created to act as placeholders
for those implicit pins. Also removed the pinting of the "disconnected
pin" messages for black-box modules (since by definition they have
disconnected pins, because black-box modules have no contents).
black-box circuits, especially those coming from verilog netlists
where a pin does not need to be declared and is implicitly floating.
This prevents the need to have an explicit black-box entry for any
verilog module that may have an instance that does not declare all
the pin connections. Also corrected an error which causes mysterious
failures if a verilog netlist is read before a SPICE netlist,
because the former gets hashed case-sensitive and the latter changes
the hashing to case-insensitive. Modified to force the SPICE
netlist to be treated case-sensitive, which may cause errors, but
is consistent with the reverse order handling, and doesn't cause
unexplained errors.
symmetry of all elements in all symmetric partitions, rather than
(as previously done) all elements in each partition, before re-
running iterations to convergence. This solves the problem of
having a very large number of partitions with a few elements each
taking a long time to run.
automorphisms so that it arbitrarily assigns all pairs from
circuit1 and circuit2 at once rather than assigning one pair at
a time and rerunning to convergence. I'm not sure of the validity
of this, other than that I have never seen a circuit fail to match
after resolving automorphisms, leading me to believe that the way
the symmetry breaking is done is irrelevant.
long run-times even when there are no properties to check. Corrected
a problem with SPICE and verilog netlist reading which arbitrarily
replaces file extensions even when a file extension is given,
resulting in reading the wrong file.
to include type CLASS_MODULE in the list of types to descend into,
since "module" (black-box) types need to be checked for pin
matching even if they have no contents. This allows two verilog
netlists to be compared against each other.
definitions: Now correctly parses everything from the definition
name to the end of line as the definition value. Also: The
search for definitions in the body of the text does not reject
non-alphanumerics "_" and "$" in the definition name, without
which definition names containing those characters will go
unrecongized. Have not yet extended this to multi-line definition
values.
of the root name of the LHS net, and so would use the last root
name copied, which might have belonged to something entirely
different, or nothing at all.
scan structure because it is used in two different places and
would have to run a cost-prohibitive search of the cell's
object list. Also, was missing recording a bus input/output
signal from an "input" or "output" statement (as opposed to
in-line signals in the I/O list).
to bus pins over an array of instances. Takes care of the three
situations where the length of the signal bus equals the number of
instances; where the length of the signal bus is a multiple of
the number of instances; and where the number of instances is a
multiple of the length of the signal bus.
from 200 to 1024. Probably this should be dynamically allocated
and expanded as needed, as it is holding names that are of
increasing length as a hierarchy is descended and the instance
prefixes appended to the name.
However, I have identified an issue that has not yet been resolved,
which is that there can be automorphisms that are potentially
broken by property matching. Currently, the automorphisms are
arbitrarily resolved, then properties are matched---and may fail
accordingly.
in an "include" statement in either SPICE or verilog. Modified pin
matching behavior to force cells in both netlists to be marked as
black-box entries if either one is marked as a black-box entry (this
may not be needed, but shouldn't do any harm, either).
be removed because they do not match pins on the other cell being
compared, are marked with a different number (-2) than the usual
(-1). CleanupPins then only removes those pins that are marked,
rather than all disconnected pins.
<col2_width>" to set the output format width, to avoid the fixed column
widths of 41 characters which can truncate long strings in the output
such as deep cell hierarchies. This command can be placed in the setup
script to widen the output columns to accomodate the result (to-do:
provide an "auto" mode to automatically determine the best width).
with increment/decrement syntax at the end. Also the parser now
handles additional keywords associated with behavioral verilog
(initial, specify) and flags modules with them as black-box entries.
that affects vectors sliced across instance arrays. Can cause vector
numbers to be out-of-bounds if an instance is arrayed but each
instance is listed separately.
inline-I/O syntax with "wire" (e.g., "input wire [3:0] test")
and addressed the failure to add buses declared in inline I/O
to the list of known buses.
verilog parser. The parser should now be able to handle any
conditionals anywhere in the verilog code. Also a bug was found
in the code that handles "a = b" assignments, and corrected.
code using the backtick expression. Also expanded the parsing of
"ifdef", "ifndef", and "endif" to include "elsif" and "else". All
forms of "if" statements should now be handled, since verilog does
not define boolean expressions in ifdef operators like most languages
do.
SPICE and verilog, in spite of the syntactical differences, to
account for the trick that qflow uses to replace the trailing space
in a verilog backslash-escaped name with a second backslash to get
a SPICE-compatible name that can be easily converted back to its
original verilog name without loss of information. What this means
is that verilog can read SPICE files containing verilog names (which
is illegal SPICE) and verilog files containing hacked-backslash
names (which is illegal verilog). This should be mostly harmless
although the wisdom of it is surely questionable.
assignment statements in verilog netlists, including assignment of
signal bundles. Also corrected handling of signal bundles in pin
connections, which had been corrected in qflow's verilog parser but
not copied back to netgen. Note that the syntax for signals multiplied
N times is still not handled.
a "critical property". This lets a device define multiple critical
properties, all of which must match before additive properties can
be combined in the same device.
re-worked property command, "critical" properties must match
across devices so that all other properties can be combined as
specified (addition, parallel combination, or none). The code was
still based on the older notion of the "critical" property being
the one to add (and thus preventing multiple properties from being
added in different ways). This has now been fixed.
handle verilog syntax. Also: Added SPICE voltage and current
sources as separate classes (as opposed to being converted to
subcircuits, which was how they were previously handled). That
allowed voltage sources to be checked for zero value and removed
by shorting the ends together, as was being done for zero value
resistors (note that like zero-value resistors, removal is only
done if removing the component makes a better match than leaving
it in). In particular, yosys has SPICE netlist output that
converts equality assignments ("assign a = b") into zero-value
voltage sources, so these components need to be treated as
non-physical elements.
the possibility that a device (e.g., resistor or capacitor) may
not be a semiconductor device (in other words, a parasitic or
ideal device), and therefore uses "value" but not width and
length, and therefore "value" is a critical property to merge
both in series and parallel. Corrected the series/parallel
network optimization to prevent it from setting both M and S
records > 1 on the same device (which is ambiguous). To try
to get number of devices to match, where there are both series
and parallel devices, they will be merged across the critical
property early (before property matching).
accidentally erasing array delimiters from node names, a move that
surprisingly has no effect at all on LVS until the cell containing
the truncated nodes is flattened, at which point it causes odd and
confusing behavior that seems to have nothing to do with node names
at all.
type double before sorting (previously, property promotion was
being done after parallel sorting). Also, modified the verilog
file reading so that parameters and definitions that are numeric
are stored as numeric (not string) properties.
perfect). Given the complexities of the verilog language, the simple
strtok() tokenizer used by the SPICE parser is not sufficient. Wrote
a better tokenizer that can distinguish between whitespace and
functional tokens like parentheses, semicolons, etc., which are tokens
themselves but also token separators.
determine if a parameter is a floating-point number, integer, or
string, and set the parameter accordingly. Found an error in the
parameter comparison if the subcircuit definitions don't agree
on the type of parameter. Now all values are promoted to a single
type based on preference order (double, integer, string). Tested
on a verilog file with a primitive device type defined as a module
with its properties encoded as parameters. This successfully
matched against the SPICE primitive device.
CDL syntax) parameters that have values which are multi-word and
space separated (who the hell comes up with these things?). One
can presumably safely assume that additional pin names are not
intermixed with parameters, so now if a multi-token parameter
value is encountered, the first word is taken as the value and
the remaining words are ignored, generating a warning message.
Since I have no idea what these values are used for, I cannot say
with certainty whether or not this would break LVS, but in the
test case given to me, the parameter had no impact on LVS, but
was apparently an annotation for the layout editor.
ifndef, endif conditional statements. Pre-define the key "LVS"
for use with netgen. Also corrected some problems stemming from
the way delimiters are handled and the flexible use of whitespace
in verilog.
be in any specific order since all ports are named. Also corrected
problem with pin names not using the same string matching function
as used for nets in general (which affects the ability to match
against different bus delimiters).
some problems stemming from comparing a case-sensitive netlist
against a case-insensitive one. Verilog netlist reading does
not yet have support for macros other than "`include", and it
does not yet have support for bit vectors constructed with
braces ({}).
calls "equate pins". This could fail because the routine that forces
uniqueness of pins was being called by the "compare" command but
outside of PinMatch. Fixed by duplicating the call to force uniqueness
of pins inside the "equate" function. Redundant calls should not
matter as uniqueness is resolved on the first call and subsequent calls
will need no further action.
time warnings and errors. Removed the "-lazy" option from the
Tcl load command, which is not needed when the stubs libraries
are compiled in correctly, and which causes issues on some
systems (e.g., Mac OS). Thanks to Matt Guthaus for the patch.
from a ".include" line by trying alternate extensions. This
should be discouraged, as it happened that a file included
"name.defparams", a file that didn't exist, and instead of calling
out the missing file, it recast it to "name.spice" and caused it
to drop into an infinite loop. Oops.
the default before comparing instances against each other for
serial/parallel combination. In particular, this avoids a
failure to serially combine a device with M = 1 vs. a device with
no M declared.
Netgen was incorrectly treating mismatched networks as a missing
set of parameters on whichever device had more property records,
resulting in misleading output.
attempt to resolve values by combining over serial chains was
attempting to access a property "S" in the component's master
record, which generally won't exist unless it has been explicitly
set in the netlist (which is unlikely since "S" is not a standard
SPICE/CDL parameter like "M").
misses a device if it has been already moved due to earlier
merging in the serial combine routine, and runs off the end of
the list; the other if the pin check routine falls on the last
device in the list, leading to an incorrect check for a record
where there is only a NULL.
unordered at the time of reaching reorderpins(). Pins will be
ordered arbitrarily (in the order of appearance in the linked
list), but netgen will not crash.
for subcircuits (.subckt ... .ends pair with cellname and pin names
and pin order, but no contents) are automatically treated as black-
box circuits if found and if the "-blackbox" option is passed to the
"lvs" (scripted) command. The "equate pins" command can be used
outside of a comparison to force two circuits (black-box or
otherwise) to be matched by pin name (if not a black-box circuit,
then this is a provisional name match, as a circuit comparison will
order based on connectivity first, not pin names). So two sets of
black-box circuit libraries can be used as long as their pin names
match. One hack added to ignore the "!" at the end of global names
when comparing pin names for matching. Otherwise, pin names must
compare by case-insensitive string match.
treat empty subcircuits as blackbox cells automatically without
requiring specific callse to "model <cell> blackbox" for each.
Enabled in LVS script by giving option "-blackbox" at the end
of the LVS command.
about property errors to show up, not due to property errors, but
due to proxy pins being inserted in the middle of a device record.
However, the first one was fixed for the case of proxy pins being
added to circuit 1, but the same fix was not made for the opposite
case of proxy pins being added to circuit 2. This commit corrects
that omission.
to stop at the end of an instance record without properties, leading
to strange errors where netgen declares "There were property errors"
but does not print any errors (because there aren't any).
the serial combination routine was disabled so as not to post a
non-working version, since the parallel/serial property networks
are not analyzed. This should be completed soon.
making any subcircuit serializable by using the new command option
"property (device) serial|parallel enable|disable". Note that as
of this commit, serial device detection is enabled but serial
networks are not collapsed for matching, which will tend to lead
to property errors in serial devices until this code is added,
which should be in a day or two.
an instance property did not match the cell property type, as
long as the cell property types of the two compared cells matched.
Along with a recent change that left "M" as a type double during
SPICE netlist read-in, this caused "M" mismatches to be ignored,
because the double value was ignored and the integer value was
always zero.
avoid the problem where non-critical properties cause devices not
to match, resulting in apparent mismatches of matched devices.
The current behavior now prints a statement about each device.
However, the result is still somewhat ambiguous.
as failing on certain compilers. This undoubtedly reflects some
change in gcc or the OS setup, but since modern compilers should
be able to figure out for themselves when to inline a subroutine
(or not), the inline hint is somewhat arcane and unnecessary.
'ignore class', but removes instances of the specified class whose
pins are shorted together. Currently requires that all pins must
be shorted together.
leading '/' of pin names and therefore failing to print anything;
(2) Corrected 'addproxies', which was ending abruptly at the end of
a circuit's object list, such that if an instance needing proxy pins
added was the last object in the circuit, it would not get the proxy
pins added, and therefore would fail LVS.
include calculations of effective width due to the addition of
width of multiple device instances in parallel. The original
behavior of splitting all "M=" devices into individual instances
has been effectively inverted, instead combining all parallel
devices of the same class into one, with multiple property
records for devices with non-matching properties (e.g., width,
length, etc.). Property matching combines devices with different
"critical properties" (e.g., FET gate width) if these are defined
in the setup using the "property merge" command.
Not yet done: Matching of multiple property records when
critial properties are not defined, handling of critical properties
that combine in parallel instead of simple addition, handling of
devices that combine in series, and the combination of non-critical
properties (e.g., source and drain area, although these are usually
removed from matching).
based on properties that can be traded with number of devices,
such as MOSFET width, by merging. This initial implementation
is somewhat limited, only dealing with properties that merge
by summing. Only devices that do not match at all in the other
circuit will be considered for merging. The feature includes
a command option "property ... merge ..." that allows control
over which devices can and cannot be merged.
variables to iterate over hash table contents. This led to the
inability to nest hash table iterators. Fixed by defining a
wrapper structure that holds the actual hash table plus the size
and iterator indexes. Not only does this solve the nesting
problem, but it also avoids the need to pass the hash table size
on every call, and that reduces the number of ways a hash table
subroutine can go wrong (e.g., cannot access the table out of
bounds simply by passing a size that is larger than was used to
initialize the table).
comparisons between, duplicate cells (cells with the same netlist
that may have more than one name in a circuit, or which for some
reason appear with the same name more than once in a netlist).
Added more checks to the list prematching, which prevents various
troubles with cells having a mismatched hierarchy. Added a
command option to "flatten class" to flatten instances only within
a specific cell. Corrected one error in the pin matching routine.
Added a check in the pin matching routine to look for pins that
have been found to be no-connects after cleaning up the pin lists
of the children of that cell.
equates class "a" in circuit 1 with class "b" in circuit2, and
if circuit 1 has a class called "b" and/or circuit 2 has a class
called "a", then both classes are given a new hash to avoid
conflicts with the (presumably) unrelated cells of the same name
in the other netlist.
been modified in the Tcl/Tk version to take an additional argument
for the file number. In the non-Tcl/Tk-compatibility mode, netgen
should operate in a backwards-compatibility mode with filenum = -1.
comparison would look for matching names from the "equate
classes" command. However, it failed to do the reverse
check, which is to make sure that if no forced match was
found for a cell, but there was a name match, that the
name-matched cell found is not being forcibly matched to
something else.
that matches explicitly called out in the setup file using
"equate classes" will take precedence over same-name matching,
instead of the other way around.
after ".ends" (seems to be unique to that situation). Also,
corrected a segfault caused by a cell definition containing
only components that are uninstantiated subcircuits (recently
added code makes references to CurrentCell without checking if
it exists).
Tim Edwards
Alessandro De Laurenzis
Anton Blanchard