This code now takes the AST nodes of type AST_BIND and generates a
representation in the RTLIL for them.
This is a little tricky, because a binding of the form:
bind baz foo_t foo_i (.arg (1 + bar));
means "make an instance of foo_t called foo_i, instantiate it inside
baz and connect the port arg to the result of the expression 1+bar".
Of course, 1+bar needs a cell for the addition. Where should that cell
live?
With this patch, the Binding structure that represents the construct
is itself an AST::AstModule module. This lets us put the adder cell
inside it. We'll pull the contents out and plonk them into 'baz' when
we actually do the binding operation as part of the hierarchy pass.
Of course, we don't want RTLIL::Binding to contain an
AST::AstModule (since kernel code shouldn't depend on a frontend), so
we define RTLIL::Binding as an abstract base class and put the
AST-specific code into an AST::Binding subclass. This is analogous to
the AST::AstModule class.
This also aligns the functionality:
- in all cases, the onehot attribute is used to create appropriate
constraints (previously, opt_dff didn't do it at all, and share
created one-hot constraints based on $pmux presence alone, which
is unsound)
- in all cases, shift and mul/div/pow cells are now skipped when
importing the SAT problem (previously only memory_share did this)
— this avoids creating clauses for hard cells that are unlikely
to help with proving the UNSATness needed for optimization
While this helper is already useful to squash sequential initializations
into one in cxxrtl, its main purpose is to squash overlapping masked memory
initializations (when they land) and avoid having to deal with them in
cxxrtl runtime.
Transparency is meaningless for asynchronous ports, so we assume
transparent == false to simplify the code in this case. Likewise,
enable is meaningless, and we assume it is const-1. However,
turns out that nMigen emits the former, and Verilog frontend emits
the latter, so squash these issues when ingesting a $memrd cell.
Fixes#2811.
This essentially adds wide port support for free in passes that don't
have a usefully better way of handling wide ports than just breaking
them up to narrow ports, avoiding "please run memory_narrow" annoyance.
When converting a sync transparent read port with const address to async
read port, nothing at all needs to be done other than clk_enable change,
and thus we have no FF cell to return. Handle this case correctly in
the helper and in its users.
When extracting read register from a transparent port that has an
enable, reset, or initial value, the usual trick of putting a register
on the address instead of data doesn't work. In this case, create soft
transparency logic instead.
When transparency masks land, this will also be used to handle ports
that are transparent to only a subset of write ports.
Like wide port support, this is still completely unusable, and support
in various passes will be gradually added later. It also has no support
at all in the cell library, so attempting to create a read port with
a reset or initial value will cause an assert failure for now.
Since the packed cell doesn't actually support wide ports yet, we just
auto-narrow them on emit. The future packed cell will add
RD_WIDE_CONTINUATION and WR_WIDE_CONTINUATION parameters so the
transform will be trivially reversible for proper serialization.
Such ports cannot actually be created or used yet, this just adds the
necessary plumbing in the helper. Subsequent commits will gradually
add wide port support to various yosys passes.
This is going to be used to store arbitrary priority masks in the
future. Right now, it is not supported by our cell library, so the
priority_mask is computed from port order on helper construction,
and discarded when emitted. However, this allows us to already convert
helper-using passes to the new model.
There will soon be more (versioned) memory cells, so handle passes that
only care if a cell is memory-related by a simple helper call instead of
a hardcoded list.
cell_inputs and cell_outputs retain cell pointers as their keys across
invocations of setup(), which may however be invalidated in the meantime
(as happens in e.g. passes/opt/share.cc:1432). A later rehash of the
dicts (caused by inserting in ModWalker::add_wire()) will cause them to
be dereferenced.
Among other problems, this also fixes equality comparisons between
SigSpec by enforcing a canonical form.
Also fix another minor issue with possible non-canonical SigSpec.
Fixes#2623.
Marcelina Kościelnicka
Rupert Swarbrick
whitequark
gatecat
Zachary Snow
Claire Xenia Wolf
N. Engelhardt
Miodrag Milanović
Xiretza
Michael Singer
Lofty