Includes information on additional pin functions such as RPLL_C_IN, GCLKC_3, SCLK and others.
This allows a decision to be made about special network routing of such pins
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
The rPLL primitive for the simplest chip (GW1N-1) in the family is
processed. All parameters of the primitive are passed on to gowin_pack,
and general-purpose wires are used for routing outputs of the primitive.
Compatible with older versions of apicula, but in this case will refuse
to place the new primitive.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
There is no need to multiply item names, it is a rudiment of my very
first addition to nextpnr.
Fully compatible with older versions of Apicula.
Note: the cosmetic changes in lines with RAM are not my initiative, but
the result of applying clang-format.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
Refactor in order to detect networks that will be routed in a special
mode earlier. This makes it possible to mark the source of such networks
as a global buffer, thereby removing their influence on element
placement.
In addition, timing classes are set for some cells.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
A simple router that takes advantage of the fact that in each cell with
DFFs their CLK inputs can directly connect to the global clock network.
Networks with a large number of such sinks are sought and then each
network is assigned to the available independent global clock networks.
There are limited possibilities for routing mixed networks, that is,
when the sinks are not only CLKs: in this case an attempt is made to use
wires such as SN10/20 and EW10/20, that is, one short transition can be
added between the global clock network and the sink.
* At this time, networks with a source other than the I/O pin are not
supported. This is typical for Tangnano4k and runber boards.
* Router is disabled by default, you need to specify option
--enable-globals to activate
* No new chip bases are required. This may change in the distant future.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
Gowin chips have a highly sophisticated system of long wires that are
wired to each cell and allow the clock or logic to spread quickly.
This commit implements some of the capabilities of the long wire system
for quadrants, leaving out the fine-tuning of them for each column.
To make use of the long wire system, the specified wire is cut at the
driver and a special cell is placed between the driver and the rest of
the wire.
* VCC and GND can not use long wires because they are in every cell and
there is no point in using a net
* Long wire numbers can be specified manually or assigned automatically.
* The route from the driver to the port of the new cell can be quite
long, this will have to be solved somehow.
* It might make sense to add a mechanism for automatically finding
candidates for long wires.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
This chip has a different default state for one type of I/O buffer ---
you have to explicitly switch it to the normal state by feeding VCC/VSS
to certain inputs.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
Compatible with older versions of apicula bases.
Also small fixes and as the number of virtual Bels grows it is necessary
to assign them Z coordinate in a centralized way to avoid conflicts and
for this purpose introduced the BelZ enum.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
* Items such as LUT, DFF, MUX, ALU, IOB are displayed;
* Local wires, 1-2-4-8 wires are displayed;
* The clock spines, taps and branches are displayed with some caveats.
For now, you can not create a project in the GUI because of possible
conflict with another PR (about GW1NR-9C support), but you can specify
the board in the command line and load .JSON and .CST in the GUI.
Although ALUs are displayed, but the CIN and COUT wires are not. This is
still an unsolved problem.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
gridDimX and gridDimY are not initialized explicitly, which leads to
effects when the design is reloaded, say, from the GUI.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
This makes predictDelay be based on an arbitrary belpin pair rather
than a arc of a net based on cell placement. This way 'what-if'
decisions can be evaluated without actually changing placement;
potentially useful for parallel placement.
A new helper predictArcDelay behaves like the old predictDelay to
minimise the impact on existing passes; only arches need be updated.
Signed-off-by: gatecat <gatecat@ds0.me>
Instead of parsing the partnumber with a regular expression,
a simple table is used. This is done because the structure
of the partnumber changes as new features appear (e.g., ES instead of C6/I5)
This commit does not yet disable the very first regular expression check.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
* A hardwired MUX within each logical cell is used.
* The delay is equal 0.
* No user placement constraints.
* The output route contains dummy PIPs. They are ignored by gowin_pack, but it may be worth removing them.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
Added support for the INS_LOC instruction in the constraints file
(.CST), which is used to specify object placement.
Expanded treatment of IO_LOC/IO_PORT constraints, which now can
be applied to both ports and IO buffers.
Port constraints have priority.
Signed-off-by: YRabbit <rabbit@yrabbit.cyou>
"nextpnr.h" is no longer the god header. Important improvements:
- Functions in log.h can be used without including
BaseCtx/Arch/Context. This means that log_X functions can be called
without included "nextpnr.h"
- NPNR_ASSERT can be used without including "nextpnr.h" by including
"nextpnr_assertions.h". This allows NPNR_ASSERT to be used safely in
any header file.
- Types defined in "archdefs.h" are now available without including
BaseCtx/Arch/Context. This means that utility classes that will be
used inside of BaseCtx/Arch/Context can be defined safely in a
self-contained header.
Signed-off-by: Keith Rothman <537074+litghost@users.noreply.github.com>
This replaces the arch-specific DelayInfo structure with new DelayPair
(min/max only) and DelayQuad (min/max for both rise and fall) structures
that form part of common code.
This further reduces the amount of arch-specific code; and also provides
useful data structures for timing analysis which will need to delay
with pairs/quads of delays as it is improved.
While there may be a small performance cost to arches that didn't
separate the rise/fall cases (arches that aren't currently separating
the min/max cases just need to be fixed...) in DelayInfo, my expectation
is that inlining will mean this doesn't make much difference.
Signed-off-by: gatecat <gatecat@ds0.me>
This Arch API dates from when we were first working out how to
implement placement validity checking, and in practice is little used by
the core parts of placer1/HeAP and the Arch implementation involves a
lot of duplication with isBelLocationValid.
In the short term; placement validity checking is better served by the
combination of checkBelAvail and isValidBelForCellType before placement;
followed by isBelLocationValid after placement (potentially after
moving/swapping multiple cells).
Longer term, removing this API makes things a bit cleaner for a new
validity checking API.
Signed-off-by: gatecat <gatecat@ds0.me>
This is a basic implementation, without considering "M of N"
arrangements (e.g. for LUT permuation where you only want to route to 1
out of 4/6 sinks) or using a type other than IdString to identify bel
pins.
But this is also enough to start working out where in nextpnr will break
due to removing the 1:1 cell:bel pin cardinality, as a next step.
Signed-off-by: gatecat <gatecat@ds0.me>
