- Add OpenChams::SlicingNode class to drive/parse XML files.
OpenChams::SlicingNode contains information to create a real slicing tree (SlicingNode from hurricaneAMS)
- Minor modification in Circuit.cpp for case when HBTree are not described in XML file.
* New: In boostrap, in ccb.py, added support for the "support" project
which contains external dependencies. Currently contains only
RapidJSON. Those dependencies must be cloned here manually.
* New: In Hurricane, Initializer class for an ordered static initialization
across different compilation modules.
* Change: In Hurricane, at DataBase level, no longer save the Cells in
the libraries, but create and ordered list of Cell. Deepest cells
(leaf cells) first. To avoid dependency problems.
* New: In Hurricane, JSON support for Property.
* New; In CRL, added JSON support for AllianceFramework, Environement &
Catalog and their associated properties. Adjustment of the
AllianceLibrary creation policy.
* New: In Hurricane, added first support for DataBase native import/export
using JSON.
We choose RapidJSON, in SAX mode, to manage the JSON format low level
Read/Write. Thus, it's Git repository http://github.com/miloyip/rapidjson
must be added under ~/coriolis-2.x/src and manually build and installed
in the Coriolis installation tree (to be integrated in ccb later).
Two mode are being supported:
1. Cell mode: one Cell only is saved. In that mode, Entities
referred by Occurrences are "outside" the file. They are coded
through their "signature" (mostly, all the values of their
attributes). The ids saved in the file cannot be restored
identically as we cannot predict when and in which context the
Cell will be reloaded.
2. Design Blob mode: the whole design hierarchy, down and including
the standard cells is saved. This way the design is completly
self contained and Entities ared referred through their ids.
A design blob can only be loaded immediatly after starting cgt
as the DataBase *must* be empty. This way we restore the whole
design hierarchy with *exactly* the same ids.
Now, Hurricane object should provide a "toJson()" method for driving
JSON, and be associated with a JsonObject derived class for parsing.
* New: In Hurricane, ability to force the next id that will be used for a
DBo (used by Design Blob Mode).
* New: In Hurricane, in DataBase, added getCell() and getLibrary() functions
to allow the hierarchical access of a Cell/Library in native mode
(i.e. whithout the requirement of AllianceFramework).
* New: In Hurricane, In CellViewer, added menu entry for Save/Load of
JSON Design Blobs. Added at this level because we consider it as the
"native" format of Hurricane.
* New: In Unicorn, added support of import/export of JSON Cell.
* Bug: In Hurricane, in Instance, when cloning an Instance or uniquifying
it's master Cell, we forgot about the Occurrences (through shared pathes).
When an instance is cloned the Shared pathes still points toward the
original Instance. And when it's the master Cell that is uniquifyed
it's the Entities pointed to that remains in the original Cell.
This is a software design problem. It is difficult to define what
policy to adopt when uniquifying: basically that means that one
Occurence is either moved onto the clone or duplicated. Furthermore,
it is not trivial to known what Occurrence is pointing on the
uniquifyed/cloned item. Have to think about it a little more.
* Bug: In Etesian, in EtesianEngine, build the flattened nets and their
RoutingPads *after* uniquifying (through slaving bounding boxes).
This way we avoid the Occurrences problem described above.
* Bug: In Etesian, in EtesianEngine, invalidate the RoutingPad after
processing the placement so they are put into the right quadtree.
This problem is due to the fact that the RoutingPads do not belong
to the Instance that they refer. And when this instance is moved
around, she doesn't inform the RoutingPad that is has moved.
More software architecture design to review...
* Bug: In CRL Core, in Vst driver, remove VhdlEntity (from Cell) and
BitProperty/Bit (from Net) with the property remove and not the
destroy() method. The BitProperty removal was completly forgotten
leading to the use of removed Signals when doing multiple saves
(hence core-dump).
* Change: In CRL Core, in Vst driver, never save as Signals the DeepNets
as they are created by a virtual flatten and do not connect any
instances at top level. Note that they will exists in the physical
file if routing layout has been created.
* Change: In vlsispad, in Dots, add an enable/disable flag because when printing
into a cmess, it is only the base class ostream which is took into account
as none of it's methods are virtuals (silly me).
* Bug: In Etesian, print into cmess instead of cout and make use of the Dots
enabling feature.
* New: In Katabatic, added AutoContact::migrateConstraintBox() to transfert
constraint from one contact to another.
New flag for AutoContact, CntWeakTerminal for AutoContact which are at
the other of a segment directly connected to a terminal. They may hold
special constraints in case of caged terminals (computed in Kite).
In AutoHorizontal & AutoVertical, in ::_makeDogleg(), transfert flags
and constraints when breaking a segment directly connected to a terminal.
* New: In Kite, in protectCagedTerminals(), uses cross constraint on the
AutoContact opposite to the ContactTerminal (CntWeakTerminal) instead
of moving up one terminal over two consecutives. This is simpler without
degrading the routing quality.
* Change: In Knik, in Vertex, add a "blocked" flag to signal disabled
vertexes in the grid (must not be used by the global router).
Modificate the Graph::getVertex() method so that when a vertex
is geometrically queried, if is a blocked one, return a non-blocked
neighbor. This mechanism is introduced to, at last, prevent the
global router to go *under* the pad in case of a commplete chip.
* New: In Katabatic, in AutoSegment, a new state has been added: "reduced".
A reduced segment is in the same layer as it's perpandiculars.
To be reduced, a segments has to be connected on source & target to
AutoContactTurn, both of the perpandiculars must be of the same layer
(below or above) and it's length must not exceed one pitch in the
perpandicular direction.
To reduce an AutoSegment, call ::reduce() and to revert the state,
call ::raise(). Two associated predicates are associated:
::canReduce() and ::mustRaise().
Note: No two adjacent segments can be reduced at the same time.
* Bug: In Katabatic, in GCellTopology, add a new method ::doRp_AccessPad()
to connect to the pads. Create wiring, fixed and non managed by
Katabatic, to connect the pad connector layer to the lowest routing
layers (depth 1 & 2). The former implementation was sometimes leading
to gaps (sheared contact) that *must not* occurs during the building
stage.
Remark: This bug did put under the light the fact that the initial
wiring must be created without gaps. Gaps are closed by making doglegs
on contacts. But this mechanism could only work when the database if
fully initialised (the cache is up to date). Otherwise various problems
arise, in the canonization process for example.
* New: In Katabatic, in AutoContactTerminal::getNativeConstraintBox(),
when anchored on a RoutingPad, now take account the potential rotation
of the Path's transformation. Here again, for the chip's pads.
* New: In Kite, support for reduced AutoSegment. TrackSegment associateds
to reduced AutoSegment are *not* inserted into track to become
effectively invisibles. When a segment becomes reduced, a TrackEvent
is generated to remove it. Conversely when it is raised a RoutingEvent
is created/rescheduled to insert it. All this is mostly managed inside
the Session::revalidate() method.
* New: In Kite, in KiteEngine::createGlobalGraph(), in case of a chip,
mark all global routing vertexes (Knik) that are under a pad, as blockeds.
* Bug: In Cumulus, in PadsCorona.Side.getAxis(), inversion between X and
Y coordinate of the chip size. Did not show until a non-square chip
was routed (i.e. our MIPS R3000).
* Change: In Stratus1, in st_placement.py add the ClockBuffer class for
backward compatibility with the MIPS32 bench. Have to review this
functionnality coming from the deprecated placeAndroute.py.
In st_instance.py, no longer creates the Plug ring of a Net.
In my opinion it just clutter the display until the P&R is called.
Can re-enable later as an option (in Unicorn).
* Change: In Unicorn, in cgt.py, more reliable way of loading then running
user supplied scripts. Borrowed from alliance-checker-toolkit doChip.py .
* New: In Hurricane, in NetlistWidget & NetlistModel, adds a column for
the RoutingPads count. Special display to mark global nets.
* Change: In Hurricane, in all MVC's models, some improvements on the
way the columns are sized and the titles displayeds.
* Bug: In Knik, in Graph::Dijsktra(), when all the RoutingPads of a net
are aligned, the seaching area is only one vertex wide (in width or
height). That is, it contains only one edge in it's main direction.
So when that sub-graph contains a fully blocked edge, Dijkstra cannot
find a path... Now enlarge the one edge graph to at least three by
inflating the search area. Curiously a code for similar purpose seems
to be present, but commented.
* New: In Hurricane, added hierarchy browser in the controller.
The hierarchy is presented as a tree. New MVC, HierarchyInformations,
HierarchyModel and HierarchyBrowser.
When using a data model which is not directly connect inside the
model, we must use the internalPointer field of the QModelIndex.
But this field is also used by the QSortFilterproxymodel, this
prevent using it with a Tree model.
The filter is directly implemented and is specific to each
Instance entry.
* Change: In CellWidget & CellWidget::State, now store the Cell and
a Path, this way we can keep track of the hierarchy when browsing.
The Path is in the State and so propagated back to the Controller.
* Change: In HierarchyCommand, use the Path with setCell().
* New: In Hurricane, in Cell & Instance, add the ability to merge the
QuadTree when *second level* instances of a Cell are placed in
the same space as the top Cell. This is the case of a deeply
hierarchical design made of only standard cells that are to
be placed in a "flat" manner.
The design is uniquified then the intermediate instances models,
which should be unique at that point have their QuadTree merged
through a call to Instance::slaveAbutmentBox(). That method will
make the model of the instance use the QuadTree of the Cell to
which the instance belong. The instance model no longer posseses
a dedicated QuadTree. As a corollary the abutment box of both
Cell are kept identical and the Instance has it's transformation
set to (0,0,ID).
Remark: when we talk about "QuadTree", we mean in fact the
QuadTree for the instances *and* the SliceMap (Layer+QuadTree).
Consequence in Query: when going through the resulting
"flattened" QuadTree we will find objects with an incomplete
Path du to the fact that we didn't have to explore their
Instance/Cell level to reach them. The shunted part of the
Path is stored in the Go master Cell in the _shuntedPath
attribute. This also affect the displayed depth of hierarchy,
but not too badly.
* New: In Hurricane, in Cell, new methods:
- Cell::updatePlacedFlag() : set the placement flags.
- Cell::isUnique() : one or less instance.
- Cell::isUniquified() : is the result of an uniquification.
- Cell::isUniquifyMaster() : is the reference cell of the
uniquification.
* Change: In Hurricane, in Cell::Uniquify(), uniquify a Cell only
if it is unplaced. We do not need to duplicate placed Cells
(see datapathes).
* Change: In KiteEngine::wipeoutRouting(), remove only the Contacts
that are *not* anchored. Any other will be indirectly destroyed.
* New: In GraphicKiteEngine, add an encapsulation and a menu for
wipeoutRouting().
* New: In Hurricane, in Cell, no longer create rings of RoutingPads when
flattening the nets. Put that functionnality into a separate method
::createRoutingPadRings().
This allow to perform the Net flattening in Etesian *without*
the rings, which slow it down. Then the rings are created by
Knik/Kite. This also solves the double ring creation when doing
P&R of a complete chip (rings where created twice: in the core
block for Etesian and at chip level for Kite).
* Change: In Etesian, slight beautification of the printed informations.
(psychorigid me)
* Bug: In Hurricane, In NetAlias::removeAlias(), perform the isMaster()
check on the *current* hook, not the *this* one... Was preventing the
removal of most aliases.
* New: In CRL Core, in DefExport, perform a virtual flatten before exporting
a design. Replace the hard-wired "sxlib" gauge by the current default
one. More secure management of flattened instance and net names (removal
of all potentially special characters). In net names, correct management
of vectorized names (do not systematically add "_net" at the end).
In DefImport, manage the BUSBITS property and revectorize accordingly
the net names. When loading nets, systematically remove the alias of
the temporary net "__prebuild__" (a bug in NetAlias was preventing it).
In LefExport, the right value of MANUFACTURINGGRID in pseudo symbolic
mode is 0.5 (expressed in lamba). That way all length are even multiples
of it. Note that there is still the order bug of the 5.7 drivers that
puts that entry *before* UNITS in the LEF file (bad bad Cadence).
* Bug: In Hurricane, in Instance::getClone(), the Instance::create()
was called with a wrong number of parameters, causing the the
wrong overload to be called. The position was thus not copied
generating an unplaced copy. This was causing error in the
AM2901/datapath bench, because fixed blocks where becoming
unfixed.
* Bug: In <bootstrap>, in coriolisEnv (again) an error in the
detection of the LD_LIBRARY_PATH.
* New: In <bootstrap>, socInstaller.py, a script to perform standard
SoC network installations and nighltly builds. Clone/update the
git repositories (Coriolis & Chams), build binaries and docs
in two passes, then send a mail report.
* New: In Hurricane, dedicated class BaseFlags to wrap a set of flags.
Similar to the Mask class, only with a slightly different semantic.
Encapsulation of <unsigned int>.
Also provide support for the Inspector, to have a human-readable
display of the flags.
* Change: In Hurricane, in Cell, regroup all the flags under a Flags
sub-class of Cell. No more mixing between booleans and bit flags.
(first use of BaseClass).
* In Hurricane, CRL Core & Knik, many small corrections to suppress
annoying warning messages at compile time.
Most noticeably, Bison errors in VstParserGrammar generated by
rules and token with mismatched return value type. This was not
problematic because the badly defined return value where not used.
But still...
* Bug: In CRL Core, in VectorPortMap::VhdlPortMap(), if the connection
was made to *non-contiguous* bits of an otherwise *contiguous* vector,
it was using a span instead of the separate bits. Now check that
bits are contiguous (delta: +1/-1) and the delta do not change of
sign.
* Change: In Etesian & Kite, the Python interface function ::setViewer()
was checking that the argument was indeed a CellViewer, but in text
mode it is None. Now, silently ignore the argument if it cannot be
converted into CellViewer.
* New: In Hurricane, in Collection add simple STL iterator support.
So now the C++11 "for" instruction can be used instead of the
in-house "forEach".
Example:
forEach ( Component*, icomponent, net->getComponents() ) {
cout << (*icomponent) << endl;
}
Become:
for ( Component icomponent : net->getComponents() ) {
cout << icomponent << endl;
}
* Change: In bootstrap, in FindLEFDEF.cmake, adds my personal path where
the LEF/DEF libraries are to be found (~jpc/oa/lefdef/5.7-s038/).
* Change: In CRL Core, in LEF/DEF exporters, use the default routing
gauge and cell gauge, instead of systematically requiring "sxlib".
IMPORTANT: In order to build a *shared* library including the
*static* LEF/DEF libraries, those libraries, even if static, must
be build with the "-fPIC" flag (otherwise the compiler complains).
I have done so on my 64 bit build of LEF/DEF, but it's a completly
*local* patch. You may have to do it manually on your installations.
* Change: In CRL Core, replace duplicated libraries, even when performing
a Append/Prepend operation. Also correct a bug in library name
uniquification, event if it's no used. The policy in cas of duplicate
libraries is not completly clear for me, still.
* New: In Hurricane, in Graphics, add a new static method ::isHighDpi()
which returns true if the DPI is greater than 192. A "normal" screen
is typically 96dpi, we sets the trigger at twice that value.
Then in all the widgets with size settings, we uses that method to
select the more appropriate size.
* New: In Kite, in BuildPowerRails more accurate GCell density computation
under the pad area, to limit the overloaded messages.
* New: In Cumulus, In BlockCorona, remove the vertical METAL3 blockages
and replace them by horizontal blockages completly enclosing the
corona big vias in all layers. This way the I/O wires that goes
through the corona are guided to be straight and do not creates
unsolvable doglegs due to the presence of big vias.
* Change: In Kite & Etesian, small typo: refactor the method
"::wipeOutRouting()" into "::wipeoutRouting()" (indulge me, I'm a
code maniac).
* Bug: In Hurricane, in PyHurricane.h the macro HCATCH was not catching
standard STL exceptions. This was the source of the cryptic message:
"Fatal Python error: Py_EndInterpreter: thread still has a frame"
The Python interpreter was interrupted uncleanly bypassing it's own
exceptions mechanism.
In PyViewer, the Viewer *do not* inherit from a base class (in the
Python export).
* New: In Hurricane, in DbU, compute maximum values (in double) for
grid, lambda & physical (in meter) so now the DbU::toGrid(),
DbU::toLambda() & DbU::toPhysical() methods can check for out of
bound values, and throw an exception.
* Change: In Hurricane, ExceptionWidget::catchAllWrapper() now returns
a boolean, set to <true> if an exception has been catched. Allow
callers to interrupt themselves if a problem has occured.
* Bug: In Kite & Etesian, in the Python wrapper, send a Python exception
if catchAllwrapper() did return true, instead of continuing...
* Change: In Kite & Etesian, adds a setViewer() method (exported in Python)
to use the graphical ExceptionWidget when in graphic mode.
* Bug: In Cumulus, in PadsCorona.py the check for the core vs. chip size
was not returning False when invalid.
* New: In CRL Core, in Vst driver, add a support IEEE VHDL. Inactive for
now as I don't see clearly the policy for selecting it or not.
Remove the code of the old Vst driver.
In Blif parser, check for non-existent models (incomplete or
corrupted Blif file). Found by G. Gouvine.
* New: Added extras file for IDE-like support under Emacs.
* New: In CRL Core, in VstParser, support IEEE VHDL, with tokens
<library> and <use>. If "use IEEE.std_logic_1164.ALL" is present
the file will be considered to be IEEE compliant.
To be precise, the parser now support any mix between Alliance
and IEEE VHDL. So you can have both <std_logic> and <wor_bit>
in the same file, but it is unclean to do that.
The two extensions ".vhd" & ".vhdl" are supported.
The drivers still always creates Alliance VHDL.
* New: In CRL Core, created a LibraryManager widget. It provides a
composite information based on what is present, for each
Alliance library:
1. - A Cell in memory, without Catalog::State.
2. - A Catalog::State, with or whithout the Cell in memory.
3. - The files of the Cell in the librariy's directory.
4. - A file with a format referenced for one of the importers.
File type recognition is based *only* on the file extension,
so it may easily confused. Be careful about what you put in
the library's directory.
One of the big limitation is that it will not display Hurricane
libraries that do not have the AllianceLibrary extension.
This widget is put in a separate library <libmanager>, included
in the default CRLCORE_LIBRARIES.
* Change: In CRL Core, in State (through the loader), now sets the
InMemory flag (event if nobody uses it yet...). Display it in
the state _getString().
In AllianceFramework, new getAllianceLibraries() method.
In CatalogExtension, make the static method "get()" publicly
accessible, for sometimes we want the whole State.
* Bug: In vlsisapd, in Path, the pathcache was not rebuild when it
should, leading to incorrect results.
* New: In vlsisapd, in Path, added a listdir() method to access the
contents of a directory and a stat() method to poll the status
of a file/directory.
Rename the ".string()" method in ".toString()" to avoid
tricky name resolution with std::string, refactor in all the
other tools.
* Change: In Hurricane, in Controller, no longer oversize the fonts
of the table's headers.
* New: In Unicorn, in UnicornGui, integrate LibraryManager.
Jean-Paul Chaput
EricLaoGitHub
Gabriel Gouvine