When implementing concepts such as partition pins or deliberately split
nets, there's a need for something that looks like a cell (starts/ends
routing with pins on nets, has timing data) but isn't mapped to a fixed
bel in the architecture, but instead can have pin mappings defined at
runtime.
The PseudoCell allows this by providing an alternate, virtual-function
based API for such cells. When a cell has `pseudo_cell` used, instead of
calling functions such as getBelPinWire, getBelLocation or getCellDelay
in the Arch API; such data is provided by the cell itself, fully
flexible at runtime regardless of arch, via methods on the PseudoCell
implementation.
This won't affect Python-built arches significantly; but will be useful
for the future 'viaduct' functionality where generic routing graphs can
be built on the C++ side; too.
Signed-off-by: gatecat <gatecat@ds0.me>
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>
This is important for distiguishing valid pseudo pips in the FPGA
interchange arch. This also avoids a double or triple lookup of
pip->net map.
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>
Single argument constructors will silently convert to that type. This
is typically not the right thing to do. For example, the nexus and
ice40 arch_pybindings.h files were incorrectly parsing bel name strings,
etc.
Signed-off-by: Keith Rothman <537074+litghost@users.noreply.github.com>
It has not actually been implemented in any router for over 2.5 years and causes nothing more than confusion. It can always be added back if it forms part of a future solution; possibly as part of a more general database structure rethink.
Signed-off-by: David Shah <dave@ds0.me>
