nmigen.git
2 years agoback.cxxrtl: allow injecting black boxes.
whitequark [Sun, 14 Jun 2020 09:25:54 +0000 (09:25 +0000)]
back.cxxrtl: allow injecting black boxes.

2 years ago_yosys: add a way to retrieve Yosys data directory.
whitequark [Sun, 14 Jun 2020 00:04:18 +0000 (00:04 +0000)]
_yosys: add a way to retrieve Yosys data directory.

This is important for CXXRTL, since that's where its include files
are located.

2 years ago_yosys: fix typo in error message.
whitequark [Sun, 14 Jun 2020 00:03:36 +0000 (00:03 +0000)]
_yosys: fix typo in error message.

2 years agotest: fix example test after commit a7b8ced9.
whitequark [Thu, 11 Jun 2020 16:29:12 +0000 (16:29 +0000)]
test: fix example test after commit a7b8ced9.

2 years agoback.cxxrtl: new backend.
whitequark [Thu, 11 Jun 2020 16:19:40 +0000 (16:19 +0000)]
back.cxxrtl: new backend.

2 years ago_yosys: translate Yosys warnings to Python warnings.
whitequark [Thu, 11 Jun 2020 16:12:52 +0000 (16:12 +0000)]
_yosys: translate Yosys warnings to Python warnings.

This isn't used yet (the only Yosys warning we can get is useless),
but will be handy for CXXRTL.

2 years agofix yosys version import
Luke Kenneth Casson Leighton [Fri, 31 Dec 2021 14:07:39 +0000 (14:07 +0000)]
fix yosys version import

2 years agomanually patch in nmigen-yosys
Luke Kenneth Casson Leighton [Fri, 31 Dec 2021 14:02:09 +0000 (14:02 +0000)]
manually patch in nmigen-yosys

2 years agoback.verilog: fall back to nmigen_yosys package.
whitequark [Fri, 22 May 2020 16:50:45 +0000 (16:50 +0000)]
back.verilog: fall back to nmigen_yosys package.

The nmigen-yosys PyPI package provides a custom, minimal build of
Yosys that uses (at the moment) wasmtime-py to deliver a single
WASM binary that can run on many platforms, and eliminates the need
to build Yosys from source.

Not only does this lower barrier to entry for new nMigen developers,
but also decouples nMigen from Yosys' yearly release cycle, which
lets us use new features and drop workarounds for Yosys bugs earlier.

The source for the nmigen-yosys package is provided at:
  https://github.com/nmigen/nmigen-yosys
The package is built from upstream source and released automatically
with no manual steps.

Fixes #371.

2 years agonmigen.cli: fix file type autodetection code.
whitequark [Thu, 11 Jun 2020 15:36:43 +0000 (15:36 +0000)]
nmigen.cli: fix file type autodetection code.

2 years agoback.verilog: remove unused imports. NFC.
whitequark [Thu, 11 Jun 2020 15:17:49 +0000 (15:17 +0000)]
back.verilog: remove unused imports. NFC.

2 years agohdl.xfrm: preserve allow_reset_less when transforming ResetSignal.
Adam Greig [Sat, 6 Jun 2020 11:43:25 +0000 (12:43 +0100)]
hdl.xfrm: preserve allow_reset_less when transforming ResetSignal.

Fixes #400.

2 years agohdl.rec: preserve shapes when constructing a layout.
Shawn Anastasio [Fri, 5 Jun 2020 03:19:46 +0000 (22:19 -0500)]
hdl.rec: preserve shapes when constructing a layout.

Preserve the original user-provided shape, while still checking
its validity. This allows Enum decoders to work when specifying
record fields with Enums.

Fixes #393.

2 years agosetup: exclude tests.
whitequark [Fri, 29 May 2020 17:45:09 +0000 (17:45 +0000)]
setup: exclude tests.

The inclusion of tests in the package was completely unintentional
and a result of my lack of knowledge of setuptools. If you are using
nmigen.test.utils, please copy the parts you need to your own code
because the nmigen.test code can and will change without notice.

2 years agovendor.lattice_ice40: reword confusing comment. NFC.
whitequark [Fri, 29 May 2020 17:45:09 +0000 (17:45 +0000)]
vendor.lattice_ice40: reword confusing comment. NFC.

2 years agohdl.ast: fix typo
Robin Ole Heinemann [Sun, 24 May 2020 15:54:17 +0000 (17:54 +0200)]
hdl.ast: fix typo

2 years agoUpdate .gitignore.
whitequark [Thu, 21 May 2020 10:41:23 +0000 (10:41 +0000)]
Update .gitignore.

2 years agovendor.intel: don't use `write_verilog -decimal`.
whitequark [Thu, 21 May 2020 09:49:42 +0000 (09:49 +0000)]
vendor.intel: don't use `write_verilog -decimal`.

See commit 702e41ba for details.

2 years agovendor.intel: double-quote Tcl values rather than brace-quoting.
whitequark [Thu, 21 May 2020 09:48:42 +0000 (09:48 +0000)]
vendor.intel: double-quote Tcl values rather than brace-quoting.

For unknown reasons, Quartus treats {foo} and "foo" in completely
different ways, which is not true for normal Tcl code; specifically,
it preserves the braces if they are used. Because of this, since
commit 6cee2804, the vendor.intel package was completely broken.

2 years agovendor.xilinx_{7series,ultrascale}: don't use `write_verilog -decimal`.
whitequark [Thu, 21 May 2020 08:57:30 +0000 (08:57 +0000)]
vendor.xilinx_{7series,ultrascale}: don't use `write_verilog -decimal`.

In commit 892cff05, `-decimal` was used when writing Verilog for
Vivado targets because it treats (* keep=32'd1 *) and (* keep=1 *)
differently in violation of Verilog LRM. However, it is possible
to avoid that workaround by using (* keep="TRUE" *). Do that,
and remove `-decimal` to avoid special-casing 32-bit constants.

Refs #373.

2 years agobuild.plat: skip clock constraints on unused signals.
whitequark [Wed, 20 May 2020 05:35:47 +0000 (05:35 +0000)]
build.plat: skip clock constraints on unused signals.

It's not very nice to add more internal mutable state to Platform
related classes, but our whole approach for Platform is inherently
stateful, and other solutions (like changing every individual vendor
platform to check for unused signals) are even worse.

Fixes #374.

2 years agovendor.xilinx_{7series,ultrascale}: add (*keep*) on constrained clocks.
whitequark [Wed, 20 May 2020 04:58:03 +0000 (04:58 +0000)]
vendor.xilinx_{7series,ultrascale}: add (*keep*) on constrained clocks.

If the clock signal is not a top-level port and has aliases, it can
be optimized out, and then the constraint will no longer apply.
To prevent this, make sure the constrained signal is preferred over
any aliases by using the `keep` attribute.

Vivado does not parse attributes like (* keep = 32'd1 *) as valid
even though, AFAICT, they are equivalent to (* keep = 1 *) or simply
(* keep *) per IEEE 1364. To work around this, use the solution we
currently use for Quartus, which is `write_verilog -decimal`.

Fixes #373.

2 years agohdl.ast: add const-shift operations.
whitequark [Wed, 20 May 2020 03:18:33 +0000 (03:18 +0000)]
hdl.ast: add const-shift operations.

Also, clean up the rotate code a bit.

Fixes #378.

2 years agohdl.ast: clarify docs for Value.rotate_{left,right}.
whitequark [Tue, 19 May 2020 23:43:25 +0000 (23:43 +0000)]
hdl.ast: clarify docs for Value.rotate_{left,right}.

"Rotate modulo 2**len(self)" is redundant because that's just how
rotates work.

2 years agohdl.dsl: check for unique domain name.
whitequark [Tue, 19 May 2020 23:40:49 +0000 (23:40 +0000)]
hdl.dsl: check for unique domain name.

Fixes #385.

2 years agoback.rtlil: handle signed and large Instance parameters correctly.
whitequark [Tue, 19 May 2020 23:33:14 +0000 (23:33 +0000)]
back.rtlil: handle signed and large Instance parameters correctly.

Fixes #388.

2 years agotracer: fix get_var_name() to work on toplevel attributes.
whitequark [Mon, 4 May 2020 05:51:08 +0000 (05:51 +0000)]
tracer: fix get_var_name() to work on toplevel attributes.

E.g. this now works:

    >>> class dummy: pass
    >>> self = dummy()
    >>> self.foo = Signal()
    >>> self.foo.name
    'foo'

2 years agovendor.lattice_machxo2: generate binary bitstreams.
Gwenhael Goavec-Merou [Fri, 8 May 2020 04:42:21 +0000 (06:42 +0200)]
vendor.lattice_machxo2: generate binary bitstreams.

2 years agoplat, vendor: systematically escape net and file names in Tcl.
whitequark [Sat, 2 May 2020 10:41:18 +0000 (10:41 +0000)]
plat, vendor: systematically escape net and file names in Tcl.

Before this commit, there was only occasional quoting of some names
used in any Tcl files. (I'm not sure what I was thinking.)

After this commit, any substs that may include Tcl special characters
are escaped. This does not include build names (which are explicitly
restricted to ASCII to avoid this problem), or attribute names (which
are chosen from a predefined set). Ideally we'd use a more principled
approach but Jinja2 does not support custom escaping mechanisms.

Note that Vivado restricts clock names to a more restrictive set that
forbids using Tcl special characters even when escaped.

Fixes #375.

2 years agoback.rtlil: fix incorrect escaping of signed parameters.
whitequark [Tue, 28 Apr 2020 02:16:12 +0000 (02:16 +0000)]
back.rtlil: fix incorrect escaping of signed parameters.

Also, improve escaping code in general.

2 years agohdl.ast: use SignalSet, not ValueSet, for _[lr]hs_signals().
whitequark [Mon, 27 Apr 2020 18:06:12 +0000 (18:06 +0000)]
hdl.ast: use SignalSet, not ValueSet, for _[lr]hs_signals().

The implementation of these functions likely predates support for
{Clock,Reset}Signal() in SignalKey.

2 years agolib.fifo: add r_rst output for AsyncFIFO{,Buffered}.
awygle [Fri, 24 Apr 2020 22:27:04 +0000 (15:27 -0700)]
lib.fifo: add r_rst output for AsyncFIFO{,Buffered}.

Fixes #181.

2 years agohdl.ir: typecheck `convert(ports=)` more carefully.
awygle [Fri, 24 Apr 2020 21:15:00 +0000 (14:15 -0700)]
hdl.ir: typecheck `convert(ports=)` more carefully.

The `ports` argument to the `convert` functions is a frequent hotspot of
beginner issues. Check to make sure it is either a list or a tuple, and
give an appropriately helpful error message if not.

Fixes #362.

2 years agoREADME: link directly to Yosys build instructions.
whitequark [Fri, 24 Apr 2020 21:09:58 +0000 (21:09 +0000)]
README: link directly to Yosys build instructions.

2 years agoback.verilog: add workaround for evaluation Verific behavior.
Teguh Hofstee [Thu, 23 Apr 2020 21:46:10 +0000 (14:46 -0700)]
back.verilog: add workaround for evaluation Verific behavior.

The evaluation version of Verific prints its license information to stdout,
and since it is against the EULA to change that in any way, this behavior
is not possible to fix in Yosys. Add a workaround in nMigen instead.

2 years agoback.verilog: make Yosys version check compatible with Verific.
Teguh Hofstee [Wed, 22 Apr 2020 12:23:14 +0000 (05:23 -0700)]
back.verilog: make Yosys version check compatible with Verific.

2 years agovendor: use nextpnr -12k for -12F devices; remove theoretical devices
Kate Temkin [Mon, 20 Apr 2020 22:25:50 +0000 (16:25 -0600)]
vendor: use nextpnr -12k for -12F devices; remove theoretical devices

nextpnr now supports -12k; which replaces the use of -25k and --idcode
together to build bitstreams compatible with -12F devices. Use this.

This also removes the LFEUM-12K and its 5G counterpart; as per Dave Shah
they're currently only theoretical FPGAs.

2 years agohdl.rec: make Record inherit from UserValue.
anuejn [Thu, 16 Apr 2020 16:46:55 +0000 (18:46 +0200)]
hdl.rec: make Record inherit from UserValue.

Closes #354.

2 years agoback.rtlil: translate enum decoders to Yosys enum attributes.
whitequark [Wed, 15 Apr 2020 14:42:46 +0000 (14:42 +0000)]
back.rtlil: translate enum decoders to Yosys enum attributes.

Fixes #254.

2 years agobuil.plat: enable strict undefined behavior in Jinja2.
whitequark [Tue, 14 Apr 2020 06:17:16 +0000 (06:17 +0000)]
buil.plat: enable strict undefined behavior in Jinja2.

By default, if an operation produces an undefined value (a Jinja2
concept that corresponds to Python's KeyError, AttributeError, etc)
then this value may be printed in a template, which is a nop. This
behavior can hide bugs.

This commit changes the Jinja2 behavior to raise an error instead of
producing an undefined value in all cases. (We produce undefined
values deliberately in a few places. Those are unaffected; it is OK
to use several kinds of undefined values in one Jinja2 environment.)

Fixes #337.

2 years agoback.rtlil: don't emit connections to zero width ports.
whitequark [Mon, 13 Apr 2020 17:04:13 +0000 (17:04 +0000)]
back.rtlil: don't emit connections to zero width ports.

Fixes #335.

2 years agoback.rtlil: refuse to create extremely large wires.
whitequark [Mon, 13 Apr 2020 16:38:36 +0000 (16:38 +0000)]
back.rtlil: refuse to create extremely large wires.

Such wires are likely to trigger pathological behavior in Yosys and,
if applicable, other toolchains that consume Verilog converted from
RTLIL.

Fixes #341.

2 years agoback.rtlil: fix expansion of Part() for partial dummy writes.
whitequark [Mon, 13 Apr 2020 15:56:39 +0000 (15:56 +0000)]
back.rtlil: fix expansion of Part() for partial dummy writes.

Before this commit, selecting a part that was fully out of bounds of
a value was correctly implemented as a write to a dummy wire, but
selecting a part that was only partially out of bounds resulted in
a crash.

Fixes #351.

2 years agoback.rtlil: fix legalization of Part() with stride.
whitequark [Mon, 13 Apr 2020 14:43:43 +0000 (14:43 +0000)]
back.rtlil: fix legalization of Part() with stride.

Also known as word_select().

2 years agoClarify a few comments. NFC.
whitequark [Mon, 13 Apr 2020 13:03:31 +0000 (13:03 +0000)]
Clarify a few comments. NFC.

2 years agohdl.ast: add Value.{rotate_left,rotate_right}.
Dan Ravensloft [Mon, 13 Apr 2020 13:40:39 +0000 (14:40 +0100)]
hdl.ast: add Value.{rotate_left,rotate_right}.

2 years agoTravis: require tests to pass on pypy3.
whitequark [Mon, 13 Apr 2020 11:04:25 +0000 (11:04 +0000)]
Travis: require tests to pass on pypy3.

Fixes #147.

2 years agoTravis: upgrade to bionic.
whitequark [Mon, 13 Apr 2020 09:40:00 +0000 (09:40 +0000)]
Travis: upgrade to bionic.

2 years agobuild.run: fix BuildProducts.extract to work with subdirectories.
whitequark [Sun, 12 Apr 2020 04:56:15 +0000 (04:56 +0000)]
build.run: fix BuildProducts.extract to work with subdirectories.

Fixes #353.

2 years agohdl.rec: improve repr() for Layout.
whitequark [Sun, 12 Apr 2020 04:47:40 +0000 (04:47 +0000)]
hdl.rec: improve repr() for Layout.

Fixes #326.

2 years agohdl.ast: improve repr() for Shape.
whitequark [Sun, 12 Apr 2020 03:59:56 +0000 (03:59 +0000)]
hdl.ast: improve repr() for Shape.

The default __repr__() from typing.NamedTuple does not include
the module name, so the replacement (which uses the preferred syntax
for specifying these shapes) doesn't either.

2 years agobuild.plat: don't check for toolchain presence if do_build=False.
whitequark [Sun, 12 Apr 2020 03:28:29 +0000 (03:28 +0000)]
build.plat: don't check for toolchain presence if do_build=False.

2 years agoback.pysim: Clear pending updates after they are effected
Stuart Olsen [Tue, 7 Apr 2020 11:17:14 +0000 (04:17 -0700)]
back.pysim: Clear pending updates after they are effected

2 years agoback.pysim: Eliminate duplicate dict lookup in VCD update
Stuart Olsen [Tue, 7 Apr 2020 05:22:45 +0000 (22:22 -0700)]
back.pysim: Eliminate duplicate dict lookup in VCD update

2 years agoback.pysim: Reuse clock simulation commands
Stuart Olsen [Tue, 7 Apr 2020 05:22:09 +0000 (22:22 -0700)]
back.pysim: Reuse clock simulation commands

2 years agohdl.mem: fix source location of ReadPort.en.
whitequark [Sun, 5 Apr 2020 02:00:06 +0000 (02:00 +0000)]
hdl.mem: fix source location of ReadPort.en.

2 years agoback.pysim: fix emission of undriven traces to VCD files.
whitequark [Fri, 3 Apr 2020 05:20:42 +0000 (05:20 +0000)]
back.pysim: fix emission of undriven traces to VCD files.

This has been originally implemented in commit d3775eed (which fixed
`write_vcd(traces=)` to do something at all), but had a flaw where
undriven traces would not be correctly placed in hierarchy. This
used to produce incorrect results on pyvcd 0.1, but started causing
assertion failures on pyvcd 0.2.

Fixes #345.

2 years agosetup: bump pyvcd to ~=0.2.
whitequark [Thu, 2 Apr 2020 11:17:41 +0000 (11:17 +0000)]
setup: bump pyvcd to ~=0.2.

We don't use any of the deprecated functionality, and the added
requirement of Python 3.6+ matches ours.

Having a requirement for pyvcd ~=0.1.4 was actually the cause of
the error I tried to fix in commit 6e1145e2. It had nothing to do
with Jinja2 (though it is definitely still good to have the tighter
requirement on Jinja2); the cause of the error was that pip would
install Jinja2, schedule installing markupsafe, install pyvcd 0.2.0
and then choke on the pyvcd 0.1.4 requirement, which would prevent
it from installing markupsafe &c. Why it does that is beyond me.

2 years agoAdd support for using non-compat Elaboratable instances with compat.fhdl.verilog...
Jacob Lifshay [Thu, 2 Apr 2020 02:38:14 +0000 (19:38 -0700)]
Add support for using non-compat Elaboratable instances with compat.fhdl.verilog.convert and compat.run_simulation

Fixes #344

2 years agosetup: tighten version constraint on Jinja2.
whitequark [Thu, 2 Apr 2020 01:35:39 +0000 (01:35 +0000)]
setup: tighten version constraint on Jinja2.

I remember thinking that not constraining it properly might bite us
someday, but assumed that Jinja2 will always stay version 2. Now it
looks like Jinja2 3.0.0a1 got released, pip picks it by default, and
it's currently broken (something about markupsafe missing).

Do what needed to be done in the first place.

2 years agohdl.ast: implement abs() on values.
whitequark [Sun, 22 Mar 2020 20:50:07 +0000 (20:50 +0000)]
hdl.ast: implement abs() on values.

2 years agovendor.lattice_ice40: add support for SB_[LH]FOSC as default_clk.
WRansohoff [Fri, 20 Mar 2020 08:10:48 +0000 (04:10 -0400)]
vendor.lattice_ice40: add support for SB_[LH]FOSC as default_clk.

These oscillators are only available on iCE40 UltraPlus devices.

2 years agovendor: fix typo `async_ff_sync`
Nicolas Robin [Sun, 15 Mar 2020 09:33:22 +0000 (10:33 +0100)]
vendor: fix typo `async_ff_sync`

2 years agoback.pysim: implement modulus operator.
Stuart Olsen [Sun, 15 Mar 2020 05:22:03 +0000 (22:22 -0700)]
back.pysim: implement modulus operator.

2 years agoCorrectly handle resets in AsyncFIFO.
awygle [Sat, 14 Mar 2020 23:26:07 +0000 (16:26 -0700)]
Correctly handle resets in AsyncFIFO.

This commit improves handling of resets in AsyncFIFO in two ways:
  * First, resets no longer violate Gray counter CDC invariants.
  * Second, write domain reset now empties the entire FIFO.

4 years agovendor: fix a few issues in commit 2f8669ca.
whitequark [Thu, 12 Mar 2020 20:28:41 +0000 (20:28 +0000)]
vendor: fix a few issues in commit 2f8669ca.

4 years agolib.cdc: extract AsyncFFSynchronizer.
awygle [Sun, 8 Mar 2020 21:37:40 +0000 (14:37 -0700)]
lib.cdc: extract AsyncFFSynchronizer.

In some cases, it is necessary to synchronize a reset-like signal but
a new clock domain is not desirable. To address these cases, extract
the implementation of ResetSynchronizer into AsyncFFSynchronizer,
and replace ResetSynchronizer with a thin wrapper around it.

4 years agohdl.ast: fix off-by-1 in Initial.__init__().
whitequark [Wed, 19 Feb 2020 01:28:14 +0000 (01:28 +0000)]
hdl.ast: fix off-by-1 in Initial.__init__().

4 years agoback.pysim: fix RHS codegen for Cat() and Repl(..., 0).
whitequark [Wed, 19 Feb 2020 01:21:00 +0000 (01:21 +0000)]
back.pysim: fix RHS codegen for Cat() and Repl(..., 0).

Fixes #325.

4 years agoback.pysim: optionally allow introspecting generated code.
whitequark [Wed, 19 Feb 2020 00:56:24 +0000 (00:56 +0000)]
back.pysim: optionally allow introspecting generated code.

4 years agonmigen.compat.genlib.cdc: add PulseSynchronizer.
awygle [Sun, 16 Feb 2020 07:01:44 +0000 (23:01 -0800)]
nmigen.compat.genlib.cdc: add PulseSynchronizer.

4 years agonmigen.lib.cdc: port PulseSynchronizer.
awygle [Sun, 16 Feb 2020 06:51:53 +0000 (22:51 -0800)]
nmigen.lib.cdc: port PulseSynchronizer.

Co-authored-by: Luke Wren <wren6991@gmail.com>
4 years agoTravis: prune dependencies.
whitequark [Fri, 14 Feb 2020 06:34:28 +0000 (06:34 +0000)]
Travis: prune dependencies.

bitarray is not required since commit 7df70059.

4 years agoTravis: test on Python 3.8.
whitequark [Fri, 14 Feb 2020 06:33:08 +0000 (06:33 +0000)]
Travis: test on Python 3.8.

4 years agocli: update use of deprecated code.
whitequark [Wed, 12 Feb 2020 14:42:24 +0000 (14:42 +0000)]
cli: update use of deprecated code.

4 years agoback.pysim: accept write_vcd(vcd_file=None).
whitequark [Wed, 12 Feb 2020 14:42:06 +0000 (14:42 +0000)]
back.pysim: accept write_vcd(vcd_file=None).

Because write_vcd() is a context manager, this is useful if the VCD
file should be sometimes not written, since it avoids awkward
conditionals with duplicated code. It's not very elegant though.

Fixes #319.

4 years agosetup: update project URLs.
whitequark [Sun, 9 Feb 2020 17:03:06 +0000 (17:03 +0000)]
setup: update project URLs.

4 years agodoc: remove outdated files and references to them.
whitequark [Sun, 9 Feb 2020 14:15:51 +0000 (14:15 +0000)]
doc: remove outdated files and references to them.

4 years agoREADME: link to IRC channel.
whitequark [Sat, 8 Feb 2020 11:00:08 +0000 (11:00 +0000)]
README: link to IRC channel.

4 years agoREADME: consolidate requirements in the Installation section.
whitequark [Sat, 8 Feb 2020 10:54:01 +0000 (10:54 +0000)]
README: consolidate requirements in the Installation section.

4 years agotest_build_res: fix after commit 3e2ecdf2.
whitequark [Fri, 7 Feb 2020 00:07:19 +0000 (00:07 +0000)]
test_build_res: fix after commit 3e2ecdf2.

4 years agobuild.res,vendor: place clock constraint on port, not net, if possible.
whitequark [Thu, 6 Feb 2020 23:37:15 +0000 (23:37 +0000)]
build.res,vendor: place clock constraint on port, not net, if possible.

For most toolchains, these are functionally identical, although ports
tend to work a bit better, being the common case. For Vivado, though,
it is necessary to place them on the port because its timing analyzer
considers input buffer delay.

Fixes #301.

4 years agoxilinx_{7series,ultrascale}: run `report_methodology`.
whitequark [Thu, 6 Feb 2020 19:38:21 +0000 (19:38 +0000)]
xilinx_{7series,ultrascale}: run `report_methodology`.

This can expose important timing issues, such as #301.

4 years agohdl.ast: add Value.{as_signed,as_unsigned}.
whitequark [Thu, 6 Feb 2020 18:27:55 +0000 (18:27 +0000)]
hdl.ast: add Value.{as_signed,as_unsigned}.

Before this commit, there was no way to do so besides creating and
assigning an intermediate signal, which could not be extracted into
a helper function due to Module statefulness.

Fixes #292.

4 years agotest_lib_fifo: define all referenced FSM states.
whitequark [Thu, 6 Feb 2020 18:10:15 +0000 (18:10 +0000)]
test_lib_fifo: define all referenced FSM states.

Broken in commit a1c58633.

4 years agohdl.dsl: make referencing undefined FSM states an error.
whitequark [Thu, 6 Feb 2020 17:47:46 +0000 (17:47 +0000)]
hdl.dsl: make referencing undefined FSM states an error.

Before this commit, doing something like:

    with m.FSM():
        with m.State("FOO"):
            m.next = "bAR"
        with m.State("BAR"):
            m.next = "FOO"

would silently create an empty state `bAR` and get stuck in it until
the module is reset. This was done intentionally (in Migen, this code
would in fact miscompile), but in retrospect was clearly a bad idea;
it turns typos into bugs, while in the rare case that branching to
a completely empty state is desired, it is trivial to define one.

Fixes #315.

4 years agohdl.ir: type check ports.
whitequark [Thu, 6 Feb 2020 17:33:41 +0000 (17:33 +0000)]
hdl.ir: type check ports.

Fixes #290.

4 years agoback.pysim: emit toplevel inputs in VCD files as well.
whitequark [Thu, 6 Feb 2020 17:19:47 +0000 (17:19 +0000)]
back.pysim: emit toplevel inputs in VCD files as well.

Before this commit, only signals driven from fragments (in practice,
everything except toplevel inputs) would get written to a VCD file.
Not having toplevel inputs in the dump made debugging ~impossible.

After this commit, all signals the fragment refers to get written to
a VCD file. (More specifically, all signals the compiler assigns
an index to, i.e. signals the generated code reads or writes.)

Fixes #280.

4 years agoback.pysim: make `write_vcd(traces=)` actually use those traces.
whitequark [Thu, 6 Feb 2020 17:07:48 +0000 (17:07 +0000)]
back.pysim: make `write_vcd(traces=)` actually use those traces.

Reported in #280.

4 years agohdl.dsl: reject name mismatch in `m.domains.<name> +=`.
whitequark [Thu, 6 Feb 2020 16:13:59 +0000 (16:13 +0000)]
hdl.dsl: reject name mismatch in `m.domains.<name> +=`.

This would violate invariants later in the elaboration process.

Fixes #282.

4 years agohdl.dsl: type check when adding to m.domains.
whitequark [Thu, 6 Feb 2020 15:19:16 +0000 (15:19 +0000)]
hdl.dsl: type check when adding to m.domains.

4 years agohdl.mem: add synthesis attribute support.
whitequark [Thu, 6 Feb 2020 14:48:48 +0000 (14:48 +0000)]
hdl.mem: add synthesis attribute support.

Fixes #291.

4 years agohdl.mem: document Memory.
whitequark [Thu, 6 Feb 2020 13:47:13 +0000 (13:47 +0000)]
hdl.mem: document Memory.

4 years agohdl.{ast,dsl}: allow whitespace in bit patterns.
whitequark [Tue, 4 Feb 2020 07:54:54 +0000 (07:54 +0000)]
hdl.{ast,dsl}: allow whitespace in bit patterns.

Fixes #316.

4 years agohdl.ast: update documentation for Signal.
whitequark [Sat, 1 Feb 2020 23:15:18 +0000 (23:15 +0000)]
hdl.ast: update documentation for Signal.

Fixes #288.

4 years agohdl.ast: prohibit shifts by signed value.
whitequark [Sat, 1 Feb 2020 23:04:25 +0000 (23:04 +0000)]
hdl.ast: prohibit shifts by signed value.

These are not desirable in a HDL, and currently elaborate to broken
RTLIL (after YosysHQ/yosys#1551); prohibit them completely, like
we already do for division and modulo.

Fixes #302.

4 years agobuild.plat: align pipeline with Fragment.prepare().
whitequark [Sat, 1 Feb 2020 03:24:26 +0000 (03:24 +0000)]
build.plat: align pipeline with Fragment.prepare().

Since commit 7257c20a, platform code calls create_missing_domains()
before _propagate_domains_up() (as a part of prepare() call). Since
commit a7be3b48, without a platform, create_missing_domains() is
calle after _propagate_domains_up(); because of that, it adds
the missing domain to the fragment. When platform code then calls
prepare() again, this causes an assertion failure.

The true intent behind the platform code being written this way is
that it *overrides* a part of prepare()'s mechanism. Because it was
not changed when prepare() was modified in 7257c20a, the override,
which happened to work by coincidence, stopped working. This is
now fixed by inlining the relevant parts of Fragment.prepare() into
Platform.prepare().

This is not a great solution, but given the amount of breakage this
causes (no platform-using code works), it is acceptable for now.

Fixes #307.

4 years agohdl.dsl: don't allow inheriting from Module.
whitequark [Sat, 1 Feb 2020 02:15:45 +0000 (02:15 +0000)]
hdl.dsl: don't allow inheriting from Module.

`Module` is an object with a lot of complex and sometimes fragile
behavior that overrides Python attribute accessors and so on.
To prevent user designs from breaking when it is changed, it is not
supposed to be inherited from (unlike in Migen), but rather returned
from the elaborate() method. This commit makes sure it will not be
inherited from by accident (most likely by users familiar with
Migen).

Fixes #286.

4 years agohdl.ast: warn on unused property statements (Assert, Assume, etc).
whitequark [Sat, 1 Feb 2020 01:55:23 +0000 (01:55 +0000)]
hdl.ast: warn on unused property statements (Assert, Assume, etc).

A property statement that is created but not added to a module is
virtually always a serious bug, since it can make formal verification
pass when it should not. Therefore, add a warning to it, similar to
UnusedElaboratable.

Doing this to all statements is possible, but many temporary ones are
created internally by nMigen, and the extensive changes required to
remove false positives are likely not worth the true positives.
We can revisit this in the future.

Fixes #303.

4 years ago_unused: extract must-use logic from hdl.ir.
whitequark [Sat, 1 Feb 2020 01:35:05 +0000 (01:35 +0000)]
_unused: extract must-use logic from hdl.ir.