previous behavior that had inadvertently been changed. In recent
versions, "load <absolute_path> -dereference" would incorrectly
apply the dereferencing to <absolute_path> rather than just its
subcells. Cleaned up the code around DBCellRead() in the process,
so everything is more straightforward (although probably more
could be done in that regard).
dereferencing, and making the behavior of "load" on the command
line (i.e., loading a cell from a file) the same as the
behavior of loading a cell as a result of expanding an unloaded
instance. In both cases, if "load -dereference" is used, and
a cell does not exist in any search path but does exist in the
original location, without dereferencing, then the cell will be
loaded from the original location. Also: Corrected an error
that has existed since adding the capability to read compressed
files, which causes magic to crash when attempting to run the
"crash recover" command (because that routine was mixing
compressed and regular file stream calls).
"cif *hier write disable" and "cif *array write disable" commands
for a specific cell def and its descendents. The revision ensures
that all descendents apply the override. The "cif write" command
uses a stack instead of recursion, which makes it difficult to
apply the same method. Currently the method only works for the
"gds write" command, and implementing the feature for "cif write"
is deemed not worth the effort.
This is diagnostic only and does not change the read-in
behavior.
(2) ext2spice: Corrected an error that had been introduced
into version 8.3.171 that accidentally marks all devices
as visited which causes all source/drain areas and
perimeters to be output as zero.
(3) extract: Sweeping changes to handling of fringe
capacitance. Removed the (recently added) "fringeshieldhalo"
parameter from the tech file. Reworked the fringe
capacitance models based on results from the "capiche"
project (github/RTimothyEdwards/capiche). Fringe shielding
is now done by clipping fringe at the boundary of a
shielding shape, rather than trying to calculate the
amount of shielding (as the "capiche" project proved this
to be equivalent). Values for partial fringing are modeled
by atan(x), which like the sidewall (1/x) curve, extends to
infinity and values are limited by the halo but do not
otherwise depend on the halo. Because of this, the halo can
be made variable and controlled by the user for deciding on
the tradeoff between accuracy and run time. A new command
option "extract halo" was added to allow this control over
the halo distance.
tech file to correct the underlying problem with the SkyWater
sky130 process in which a different layer/purpose pair is used
for TEXTTYPE and DATATYPE for the same layer. Previously, all
output from magic writes the same pair for both when writing a
port label. The new method preserves existing syntax, although
there are some differences based on what order the "port"
statement appears relative to other types for the same layer.
flattening, since the appearance of the layout will change even
though there are no physical changes. Finally got around to
debugging and correcting the input mask-hints, which can
preserve vendor GDS by marking areas where the vendor GDS differs
from magic's automatically generated output (the method was
almost correct and only needed an input scale factor change).
for the squares-grid operator and for GDS compression. But I
reverted the "calma contact" option to be false by default, because
the method does not exactly match the output when not using cell
instance arrays, and so may produce unexpected results. Will
need a different implementation that uses the same code to generate
the same (effective) layout.
default behavior of magic to make use of the "gds contacts true"
option to output contacts as arrays of subcells instead of
individual boundary entries, as the former is much more efficient
than the latter. Set the option to be true by default, and set
the "gds flatglob" option to have one entry "$$*$$" corresponding
to the contact subcells created by the "gds contacts" option, so
that GDS reads and writes as it did previously (but using a
different method). Expanded the method to include "squares-grid"
and "slots" operators (the latter should produce much more
efficient fill pattern arrays). Implemented for both compressed
and uncompressed GDS. Tested in all variations.
Alessandro De Laurenzis. That pull request cleaned up the vast
majority of compiler warnings. However, that cleanup exposed a
few additional warnings pointing to errors in the code that needed
fixing. The code now compiles cleanly except for one warning
about redefined CAD_DIR that I have not looked into.
This commit makes the code (mostly) C99-compatible, enabling to compile
it without the -Wno-error=implicit-function-declaration flag. This
way, Magic becomes usable on arm64 architectures, specifically on Apple
computers with M1/M2 SoC.
fringe capacitance halo where the default halo distance was set to
zero instead of one and caused divide-by-zero issues; (2) Found
extraction issues where labels picked up from cells flattened
during GDS reading cause the flattened/emptied cells to show up
in the extraction with extra pins that can mess up LVS. Solved
this by removing labels from flattened/emptied cells.
shrink routine that over-computes the diagonal position (the
equation failed to divide the intersecting angle in half).
Rewrote the equation for the correct grow distance, still
accounting for the grid limit (if set).
The extSubtree() routine cuts a layout into squares and extracts
each separately, checking for subcell interactions. In each
square it parses all labels looking for unconnected ones. This
section of code not only parses all labels M x N times, but it
then marks interaction areas where there may be none, forcing
additional unnecessary processing. This commit makes the first
quick optimization, which is to change the return value of
DRCFindInteractions() from boolean to integer, allowing it to
return a value indicating that there are no subcells in the
area. This prevents the loop through labels from happening in
cases where there can never be interactions. More to come.
the timestamp is updated after reading in CIF or GDS, and managed to
get the timestamp dirty flag to remain clear after reading when
"gds datestamp" is used. This includes a modification of the timestamp
update routine that only updates timestamps on a single file if only a
single file is being written.
yesterday's commit to allow the syntax "gds maskhints <types>", in
which mask hints can be restricted to a specific list of layers
rather than all layers which define mask hints in the cifoutput
rule.
in a cell to account for the difference between what's in an input
GDS file and what magic would write out itself from the processed
data. This potentially allows library cells to be read in that
will generate the equivalent mask data as output without resorting
to using GDS file references as properties. The method is activated
with the new command option "gds maskhints on" and the default is
off.
this limited ports to 16384, which seemed reasonable at the time.
However, the sky130_sram_macro layouts connect power and ground in a
way that when coupled with "extract unique" can generate tens of
thousands of ports and overrun the bit field, showing that automation
can do the unexpected. The solution was to split out the port number
from the label record as its own 32-bit value.
that it correctly lands on a grid limit boundary (which the
previous commit did not do). Note that work is still ongoing to
detect some pathological cases where the shapes end up off-grid
where two non-manhattan shapes intersect at different angles
(such as an inside corner).
calculation for "limit" in the CIFgen routines was wrong, not the
interpretation of the "gridlimit" value in the tech file. The
parsing of "gridlimit" has been put back the way it was before the
last commit, and the "limit" value calculations have been corrected.
reading a bad "bloat-all" statement in a techfile. This prevents
magic from eventually segfaulting when exercising the recipe in which
the bad statement occurs.
tile corners in the check area to find areas needing bridges.
This prevents generation of unnecessary bridging geometry; and
since the error made the check rotation-dependent, this may
resolve some "parent and child disagree on CIF" errors.
rules do not get hierarchical processing. I am not sure why I added
that exception, which clearly is not like "bound" or "net" in causing
serious issues when used hierarchically. The current counterexample
is the use in sky130A.tech for the NPC layer. Based on that usage,
the exception has been relaxed to consider any "squares" and similar
rule during hierarchical processing if followed by a "grow" operator.
Possibly this is still not relaxed enough to capture all meaningful
use cases, but should suffice for now.
were incorrect assumptions made in the code from 35 years back or so,
for the case where the CIF layer halo is larger than the size/separation
of the cells in the array. The new code will prevent the array routine
from copying hierarchical additions to the mask layers outside of the
array area. Whether or not the new code has its own faulty assumptions
remains to be seen through thorough vetting.
The previous behavior was to generate hierarchical names for all
labels when copying contents of subcells. This is "safe" for
copying selections without accidentally shorting things through
labeling, but it can make a mess of the selection. Options are
now "select do labels" for the existing behavior, "select no labels"
to not show any labels, and "select simple labels" to show only the
root name of labels in subcells.
"bridge" GDS output operator; the previous equations were, under
some situations, failing to meet the width requirement. Also:
corrected the "cif style" check so that it does not claim that a
style name with an exact match is ambiguous. Also: Corrected the
use of "grid limit", so that the limit correctly scales with the
output expander value.
to stop the search whenever a cell is not found. Used this to implement
a new option for GDS writes, "gds undefined allow|disallow" (default
"disallow") controls whether or not GDS with undefined references will
be allowed to be written. Similarly affects CIF and LEF writes, extraction,
and DRC (when running "drc check" from the top).
values in a mask hints property, and added back the handling of
mask hints in the top level cell, since the hierarchical function
on cells does not apply the function to the top level.
that have been removed by flattening into the parent cell due to lack
of devices. Previously the checks on writing the subcircuit and writing
the call were slightly different, leading to instances in which the
subcircuit call would be written to the netlist output without the
subcircuit being defined. (2) Corrected an error in the "bridge" CIF/GDS
output operator. In certain (somewhat rare) geometries, the tile behind
(instead of in front of) the corner being checked may be incorrectly
flagged as a DRC spacing error. The fix is to ignore tiles that are
behind the corner being checked.
- Don't print error if there is no more data to process
- Fix error message: `propname` already has the
MASKHINTS_ prefix
Signed-off-by: Sylvain Munaut <tnt@246tNt.com>
and made the default behavior equivalent to "see-vendor". While it is
true that running "cif see" on a readonly cell shows layers that are
not necessarily representative of what is in the file, the opposite
behavior manifests itself in ways that are confusing.
layers that are normally automatically generated to be supplemented
with additional geometry in the form of properties. The first
commit implements a CIF operator "mask-hints" that tells CIFGenLayer
that additional geometry may be specified with a property named
"MASKHINTS_" plus the name passed to the operator as its only
argument. A more extensive commit to be done will allow this
operator to be used on cifinput to use mask hints to retain the
exact geometry of mask layers used in the input file.
for infinities in space tiles by checking tile dimensions against
TiPlaneRect, where in fact TiPlaneRect is slightly smaller than
the plane boundaries, so this check would always fail, causing
unpredictable behavior due to integer overflow.
used both for counting cells during GDS write and for saving
geometry data from the "copyup" operator during GDS read. The write
routine does not clear the client record, and the read routine was
checking if the cd_client value was default. Corrected the resulting
crash condition by resetting cd_client before GDS reads. However, the
underlying problem is that the GDS read is reading data into a cell
that already exists in the database, and is not handling it robustly
by renaming the existing cell. So this should be revisited.
disagreements between parent and child cells on GDS generation for
templayers, as these layers are not output; any resulting differences
showing up eventually on an output layer will be reported.
bloat-all, that was resetting flags in the entire plane within the
callback for each tile processed. This would push DRC run-times from
minutes to hours. Also corrected another, much more minor, efficiency,
in which the connection mask was generated in the callback routine
for each tile, instead of calculating before the plane search and
passing the mask to the callback function in the client data.
hierarchical processing from CIFGenSubcells() and CIFGenArrays(), and to
avoid certain operators that are useless and harmful when applied
hierarchically; namely squares, slots, bbox, boundary, and net.
which was failing to clear tiles in the layout of the "processed" state,
leading to unpredicatable results if the same layout layer is used in
a subsequent CIF operation.
The new CIF operator BRIDGE-LIM is similar to the BRIDGE operator with the difference that the material created to meet the minimum spacing/width rules do not overlap the limiting layers. Syntax of the new operator is:
bridge-lim spacing width layers
where "layers" are the limiting layers.
the rule is a normal database rule or a CIF-DRC rule. For the latter,
the flag is used when substituting for escape strings in the "why"
rule explanation to produce the correct value in microns.
read". While "lef read" normally annotates existing layout, this
option ensures that no additional cells are created from macros in
the input LEF file. (2) Added a check on the "Input off lambda grid"
warning during CIF/GDS input such that it is not repeated once issued,
as it tends to be output many times when it occurs.
to accomodate a method used for processes that require additional
spacing between contacts for large via arrays; this requires
distinguishing between large and small areas to output vias, and
so requires use of "and" and "and-not" before "squares". This
highlights the arbitrary nature of this routine, which probably
works better getting data from cifinput, or not at all (i.e., it is
used when reading LEF, but it is known that the LEF read routines
would be better implemented by running input through CIFGen(). If
that were done, then this problem would not come up).
exclude layers created using the "boundary" or "bbox" operators,
since they only exist in specific levels of the hierarchy. Pretty
much by definition they should not be used in a way that requires
additional geometry to be added to the parent cell. This greatly
reduces parent-vs.-child disagreements (and the corresponding error
messages), but does not entirely eliminate them.
caused by the new bloat-all with templayers, which used the distance
record in the bloat structure differently than the other bloat
functions, and therefore was messing up the enumeration of layers
needing to be handled by the hierarchical output. This was probably
also wrong for the existing bloat-all function, which might explain
some problems recently with the GDS output.
generation in the "cifoutput" section of the techfile. This operator
solves the problems with the catecorner position of geometry when
attempting to automatically close up gaps between islands of a layer.
SPICE subcircuit definitions from a netlist file and apply the port
order in the netlist to the port labels in the corresponding cell
or cells in the magic database. Also: Corrected an error in the
bloat-all code introduced in a recent commit that can cause a
segfault.
templayers. This permits some useful interactions like growing to
the size of a bounding box, or abutment box, as well as many other
possibilities. Also: Corrected the use of "cif see" for the boundary
(abutment box) layer, which was not working because the "cif see"
command uses a flattened CellDef that does not have the boundary
property of the cell it was flattened from.
not known, though, why the flag is set indicating valid planes
in the clientdata record when there in fact is none. That seems
to be the underlying bug.
not copy labels; not copying labels speeds up the antenna checks
(which don't need labels) greatly. Also fixed several numerical
overflow problems in the antenna checks, which resulted in false
positive errors, as well as nonsensical results.
uninitialized variable, with the result that writing GDS would
claim that it cannot scale down enough and that the output units
are either wrong or the output style must have "units angstroms".
Using angstrom units would solve the problem but did not treat
the root of the problem.
only at the time of running the command "load". But cells are generally
loaded only on an as-needed basis, so the dereferencing option must be
saved as a flag in the cell and honored whenever its subcells are expanded
or otherwise read at a later time.
limit of the Calma definition, and probably has not done so for ages.
Nobody informed me of this. The restriction has been lifted from
GDS input and output in Magic. It can be reinstated if necessary by
setting a flag in the cifoutput section of the techfile, but it is
likely that this will not be necessary unless there are other tools
that enforce the limit and will not read a GDS file that exceeds it.
a long-standing error (introduced with the "extresist geometry"
option) that can cause nets not to be extracted (due to the first
record not having extraction data, which was itself a long-standing
error in the code but which was not fixed correctly); (2) handle
"device mosfet" type transistors (previously only handled the old
"fet" type extraction devices); and (3) correct for the res.ext
file having a different scalefactor relative to the .ext file. The
latter item was solved by forcing all input to scale like
ExtCurStyle->exts_unitsPerLambda, locally correcting all input as
needed. Note that extresist still needs to handle other extraction
devices (e.g., resistors and capacitors) but those will require
additional handling in the routines which analyze the current path
to determine how to break up wires into paths.
which takes the existing generated CIF plane, finds all enclosed
areas that have an area less than <area>, and fills them in. This
satisfies a minimum hole area rule in a way that is not possible
with any of the existing CIF operators.
which adds an offset value of "start" to both X and Y from the
lower left corner of the fill area. This allows the use of the
"offset" (from the previous git commit) to be declared on different
layers without creating an exact overlap, as is often required by
foundries for fill patterns.
from a selection, which can then be used to index into other lists.
This lets one selection be made on a list of arbitrary names, and
then additional parameters can be linked together with the same
index). Also, implemented (finally!) the "offset" parameters of
the "slots" function (as advertised in the documentation).
like resistors where a tile other than space may border the resistor
device on its non-terminal sides (which is handled correctly, and
should not be considered an error).
rid of redundant port entries in subcircuits. There is still an outstanding
issue as to whether nodes and connections need to be recursively iterated
to the hierarchy bottom. The current fix corrected the test case. Also,
added a "-dereference" option to the "load" command to revert to the
original behavior of using only search paths from "addpath" when searching
for files to load.
used with "ext2spice hierarchy on" because the device index is not
reset between calls to output cells in the hierarchy, leading to
a mismatch of the index for all cells after the first one output.
simple FET device in extresist. Also: Extended the bloat-all CIF operator
again, allowing the trigger layer for the bloat operation to include both
CIF layers and magic layers (previously only magic layers were supported).
This extension is possible due to the previous extension allowing the
trigger layer and bloating layers to be on separate planes. This operator
extension is useful for tagging geometry that is in the proximity of, but
not overlapping, geometry on another plane.
reduction of memory and startup time, which was to maintain only
one CIF style in memory. The new method is just to read in and
keep the DRC CIF style separately from the output CIF style.
Because the CIF sections of the techfile are read before the DRC
sections, and the CIF DRC style is declared in the DRC section,
the CIF DRC style is read in on the fly during the first DRC
checking.
option "labels ... cellid" to handle some vendor files where
apparently to get around the 30-character cell name limit, the
actual cellname is encoded on a text layer. Added new cifop
"boundary" (no arguments) for cases where a cell abutment box
is encoded on a GDS layer; this now translates the bounding
box to the FIXED_BBOX property, as is done with the LEF bounding
box. Also corrected the property set function to free existing
property value allocated memory when overwriting a property with
a new value.
Tim Edwards
Alessandro De Laurenzis
Anton Blanchard
Sylvain Munaut
fabian-l-c