This is an unfortunate necessity needed to fix memory inference regressions
introduced when we switched to using v2 cells. A better approach, compatible
with RFC 54, will need to be figured out for Amaranth 0.6.
Fixes#1011.
This feature does not exactly follow the RFC because the RFC as written
is not implementable; the treatment of async resets in `tick()` triggers
had to be changed. In addition, iterating a trigger was made to watch
for missed events, in case the body of the `async for` awaited for too
long.
Co-authored-by: Wanda <wanda-phi@users.noreply.github.com>
This also changes `decoder` a bit: when an enum is used as a decoder,
it is converted to a `Format.Enum` instead. The original enum is still
stored on the `decoder` attribute, so that it can be propagated
on `Signal.like`.
This includes a few minor code changes:
- Removing redundant `lib.memory.Memory.Init = hdl.MemoryData.Init`
re-export;
- Renaming `FrozenError` to `FrozenMemory` and moving it to `.hdl`;
- Marking `ReadPort` and `WritePort` as `@final`.
This check was originally added out of abundance of caution, but since
then it was observed that reset-less-ness is purely an implementation
detail (see #1220), and furthermore it interferes with adaptation of
`FIFOInterface`` signals (where `[rw]_data` are reset-less) for RFC 61.
These accessors used to be necessary (in addition to `.shape()`) while
the AST nodes were mutable. However, after commit 2bf1b4da that made
AST nodes immutable, there is no technical requirement to keep them
around. Additionally:
- `len(value)` is shorter than `value.width` and works with any `value`
- `value.shape().signed` is longer than `value.signed` but works with
any `value`
* Given a private name `$\d+` in RTLIL (as they are not named in the IR)
* Not automatically added to VCD files (as they are not named in the IR)
* Cannot be traced to a VCD (as they have no name to put in the file)
* Cannot be used with an unnamed top-level port (as there is no name)
