Presently, every IP block exports Rust/C headers using rules in
`autogen.bzl`. However, this is suboptimal because every time we
want to make a change, it needs to be replicated in all IP blocks.
Furthermore, IP blocks should not be concerned about how the hjson
is used to create register definitions.

This commit create a new macro `autogen_hjson_sw_headers` that every
IP block must call to create all necessary SW artifacts.

This commit introduces no functional change, all bazel targets are
still exactly the same as before.

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

The source tree location is computed once at the top, make sure that
the rest of code uses this location instead of recomputing some path.
This makes the code more consistent and understandable.

Signed-off-by: Amaury Pouly <[email protected]>

The top-level C library files should include a warning about being
generated and how they were generated.

Signed-off-by: Amaury Pouly <[email protected]>

This list will be useful for the multi-top support.

Signed-off-by: Amaury Pouly <[email protected]>

Currently, this BUILD file is created manually which is not consistent.
With this commit, it is now generated from a template. The generated content
is exactly identical to the current one except for the warning message.

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

All IP blocks now generate a new bazel target for the IP-specific
part of the devicetables.

Signed-off-by: Amaury Pouly <[email protected]>

We lack a unified scheme where we put the generated headers (C, rust, DT).
Therefore we need a manual target for each tops that lists the DT library
for all IP blocks. Maybe this can be autogenerated in the future.

Signed-off-by: Amaury Pouly <[email protected]>

This commit introduces a new rule `autogen_dttool` to run the
`dttool` and generate the top-specific DT code (dt_api.h,
devicetables.c/h). It also introduces a convenience macro
`autogen_top_dt` that will run `dttool` and create a necessary
libraries to write the devicetables library.

There is a small annoyance here due to two details of our
current flow:
- dttool needs access to all hjson files, not just the generate
  top configuration, but which hjson files were used is not
  recorded anywhere so it must be manually provided to the rule
- we don't have a standard naming scheme for IP headers so the
  generate DT definitions are all over the place in hw/ip/<ip>/data,
  sometimes in /data/autogen, and so on. Therefore the devicetables
  depends on some manually specified (`dt_ip_deps`) deps.

Signed-off-by: Amaury Pouly <[email protected]>

In order to run the dttool using bazel, we need not only to provide
information about the top but also point to the topgen-generated
cfg (top_<name>.gen.hjson) AND the IP block hjson. Unfortunately,
this latter information is currently not recorded anywhere but topgen
knows it so it can render it in the bazel file.

Note that even topgen cannot reliable generate the list of IP DT
dependencies without some hacks because of our very inconsistent
naming scheme. Therefore we for now point to `//hw/<top>:dt_headers`
which is a manually created list.

Signed-off-by: Amaury Pouly <[email protected]>

All changes in this commit are autogenerated by running:
```
util/topgen.py -t hw/top_earlgrey/data/top_earlgrey.hjson
```

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>
Co-authored-by: Alexander Williams <[email protected]>

This commit changes the generator as follows:
- a new `dif_<ip>_init_from_dt` function to initialize from DT
  instead of from an IO region
- the dif IRQ names and types are now defines/typedefs to the DT
  types: this ensures compatibility with existing code and an
  incremental transition

Signed-off-by: Amaury Pouly <[email protected]>

All changes in this commit are autogenerated by running:
```
util/make_new_dif.py --mode=regen --only=autogen
```

Signed-off-by: Amaury Pouly <[email protected]>

This is necessary since the autogen part of DIFs now depends on DT.

Signed-off-by: Amaury Pouly <[email protected]>
Co-authored-by: Alexander Williams <[email protected]>

The DT commit changed the type `dif_rv_plic_irq_id_t` from `uint32_t`
to an enumeration. This makes certain implicit casts in the C++ code
invalid now.

Signed-off-by: Amaury Pouly <[email protected]>

This is for documentation purposes, to make it clear that DT IRQ
IDs are PLIC IRQ IDs.

Signed-off-by: Amaury Pouly <[email protected]>

The compiler is not happy about a missing case for kDtI2cIrqCount
which is obviously invalid. Add a default to please the compiler.

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>
Co-authored-by: Alexander Williams <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

In the future, we want SW to depend on top elements without knowing
which top is it compiled for. The way to achieve this is to introduce
a layer of indirection: SW would depend on e.g. //hw/top:something
instead of //hw/top_earlgrey:something. Currently, this commit
only points everything to earlgrey but more tops can be added in
the future and the top selection becomes part of the bazel
configuration using a config_setting.

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

Currently, topgen generates plic_all_irqs_test and alert_test
in sw/device/tests/autogen. But this is incompatible with multi-top
so change this behaviour to create a one directory per top in
sw/device/test/autogen.

Signed-off-by: Amaury Pouly <[email protected]>

This commit contains two changes:
- regenerate the top using
  util/topgen.py -t hw/top_earlgrey/data/top_earlgrey.hjson
- manually tell git about the moved files

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

We now treat english_breakfast as a proper top and we check-in the autogenerated
files just like earlgrey.

All files were generated using
  util/topgen.py -t hw/top_englishbreakfast/data/top_englishbreakfast.hjson
and the new files were checked-in.

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

This commit converts just the necessary files to DT to be able
to build the test_rom for english breakfast.

Signed-off-by: Amaury Pouly <[email protected]>

The script previously outputed to a dedicated temporary directory
and deletes some files before generating them. Now that we are checking
files in, we need to change this behaviour. The script now doesn't
remote anything and outputs to hw/<top_name>.

Signed-off-by: Amaury Pouly <[email protected]>

All files were generated by running:
```
./util/topgen-fusesoc.py --files-root=. --topname=top_englishbreakfast
```
and checking-in new files.

Signed-off-by: Amaury Pouly <[email protected]>

The two tops do not quite use the same flow, in particular they use
different flags to specify the fileset. This requires to tweak the
fusesoc rule to make it more general.

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>

The script assumes that scrambling is always enabled but in fact there
is a setting for that and it is used on english breakfast.

Signed-off-by: Amaury Pouly <[email protected]>

This setting is not yet wired with the rest of the build system.

Signed-off-by: Amaury Pouly <[email protected]>

With the upcoming change to use //hw/top:top to select the top,
the per-top copts and feature will not be set on the command-line
but instead added when transitioning to build RV targets.

Signed-off-by: Amaury Pouly <[email protected]>

All select are now changed to use //hw/top:top

Signed-off-by: Amaury Pouly <[email protected]>

To avoid a massive duplication, and as a first step toward a new
exec_env-less system, we will move all exec_env to //hw/top so that
they work for all tops. The easiest to change is verilator.
For now, only convert the test_rom_test to point to this new env.

Signed-off-by: Amaury Pouly <[email protected]>

This is enough to be able to run the test_rom_test in verilator

Signed-off-by: Amaury Pouly <[email protected]>

Some IP instances have a name, e.g. sram_ctrl_ret_aon
and sram_ctrl_main. In some cases, it is important to refer to
a specific instance, e.g. `kDtSramCtrl[<index of the ret_aon>]`.
This commit generates names so that one can now do
`kDtSramCtrl[kDtIndexSramCtrlRetAon]`.

Signed-off-by: Amaury Pouly <[email protected]>

Signed-off-by: Amaury Pouly <[email protected]>