whitequark [Thu, 27 Aug 2020 07:54:27 +0000 (07:54 +0000)]
sim._pyclock: new type of process.
The overhead of coroutine processes is fairly high. A clock driver
implemented through a coroutine process is mostly overhead. This was
partially addressed in commit
2398b792 by microoptimizing yielding.
This commit eliminates the coroutine process overhead completely by
introducing dedicated clock processes. It also simplifies the logic
to a simple toggle.
This change improves runtime by about 12% on Minerva SRAM SoC.
whitequark [Thu, 27 Aug 2020 07:11:14 +0000 (07:11 +0000)]
sim._pycoro: make src_loc() more robust.
* Guard for finished coroutines.
* Guard for coroutines yielding from iterators and not generators.
whitequark [Thu, 27 Aug 2020 06:53:14 +0000 (06:53 +0000)]
_toolchain.cxx: work around PyPy missing LDCXXSHARED sysconfig variable.
whitequark [Thu, 27 Aug 2020 06:24:18 +0000 (06:24 +0000)]
_toolchain.cxx: new toolchain.
whitequark [Thu, 27 Aug 2020 01:14:05 +0000 (01:14 +0000)]
hdl.ast: clarify exception message for out of bounds indexing.
Fixes #488.
whitequark [Thu, 27 Aug 2020 00:46:50 +0000 (00:46 +0000)]
nmigen.test.utils: restore FHDLTestCase to gracefully deprecate it.
Fixes #484.
whitequark [Thu, 27 Aug 2020 00:33:31 +0000 (00:33 +0000)]
tests: move out of the main package.
Compared to tests in the repository root, tests in the package have
many downsides:
* Unless explicitly excluded in find_packages(), tests and their
support code effectively become a part of public API.
This, unfortunately, happened with FHDLTestCase, which was never
intended for downstream use.
* Even if explicitly excluded from the setuptools package, using
an editable install, or setting PYTHONPATH still allows accessing
the tests.
* Having a sub-package that is present in the source tree but not
exported (or, worse, exported only sometimes) is confusing.
* The name `nmigen.test` cannot be used for anything else, such as
testing utilities that *are* intended for downstream use.
William D. Jones [Wed, 26 Aug 2020 22:49:49 +0000 (18:49 -0400)]
build.run: implement SSH remote builds using Paramiko.
whitequark [Wed, 26 Aug 2020 22:45:19 +0000 (22:45 +0000)]
back.rtlil: do not squash empty modules.
In commit
9faa1d37, the RTLIL backend was changed to ignore modules
without ports completely, since Yosys would recognize empty modules
as black boxes without explicit `write_verilog -noblackbox` and break
the design. That change had many flaws:
* It removed instances without ports, which are used in e.g. SoC
FPGAs to instantiate a dummy CPU.
* It removed fragments without ports, which can appear in e.g. SoC
FPGAs in case the fabric is not connected to any I/O ports.
* Finally, it was just conceptually unjustified.
This commit changes the logic to actually check for empty fragments,
and instead of removing them, it adds a dummy wire inside. It would
be possible to use the Yosys-specific (*noblackbox*) attribute.
However, it would be necessary to strip it for most targets right
away, and also the wire doubles as documentation.
Fixes #441.
whitequark [Wed, 26 Aug 2020 09:01:57 +0000 (09:01 +0000)]
back.verilog: omit Verilog initial trigger only if Yosys adds it.
Verilog has an edge case where an `always @*` process, which is used
to describe a combinatorial function procedurally, may not execute
at time zero because none of the signals in its implicit sensitivity
list change, i.e. when the process doesn't read any signals. This
causes the wires driven by the process to stay undefined.
The workaround to this problem (assuming SystemVerilog `always_comb`
is not available) is to introduce a dummy signal that changes only
at time zero and is optimized out during synthesis. nMigen has had
its own workaround, `$verilog_initial_trigger`, for a while. However,
`proc_prune`, while increasing readability, pulls references to this
signal out of the process. Because of this, a similar workaround was
implemented in Yosys' `write_verilog` itself.
This commit ensures we use our workaround on versions of Yosys
without the updated `write_verilog`, and Yosys' workaround on later
versions.
Fixes #418.
whitequark [Wed, 26 Aug 2020 10:18:02 +0000 (10:18 +0000)]
vendor.xilinx_{7series,ultrascale}: set BUFG* SIM_DEVICE as appropriate.
Fixes #438 (again).
whitequark [Wed, 26 Aug 2020 14:57:31 +0000 (14:57 +0000)]
vendor.xilinx_7series: unbreak.
This commit fixes a series of typos introduced in commit
4e208b0a.
whitequark [Wed, 26 Aug 2020 13:26:38 +0000 (13:26 +0000)]
sim._pyrtl: optimize uses of reflexive operators.
When a literal is used on the left-hand side of a numeric operator,
Python is able to constant-fold some expressions:
>>> dis.dis(lambda x: 0 + 0 + x)
1 0 LOAD_CONST 1 (0)
2 LOAD_FAST 0 (x)
4 BINARY_ADD
6 RETURN_VALUE
If a literal is used on the right-hand side such that the left-hand
side is variable, this doesn't happen:
>>> dis.dis(lambda x: x + 0 + 0)
1 0 LOAD_FAST 0 (x)
2 LOAD_CONST 1 (0)
4 BINARY_ADD
6 LOAD_CONST 1 (0)
8 BINARY_ADD
10 RETURN_VALUE
PyRTL generates fairly redundant code due to the pervasive masking,
and because of that, transforming expressions into the former form,
where possible, improves runtime by about 10% on Minerva SRAM SoC.
whitequark [Wed, 26 Aug 2020 09:16:46 +0000 (09:16 +0000)]
back.cxxrtl: actualize Yosys version requirement.
whitequark [Wed, 26 Aug 2020 06:58:22 +0000 (06:58 +0000)]
hdl.ast: avoid unnecessary sign padding in ArrayProxy.
Before this commit, ArrayProxy would add sign padding (an extra bit)
a homogeneous array of signed values, or an array where all unsigned
values are smaller than the largest signed one. After this commit,
ArrayProxy would only add padding in arrays with mixed signedness
where all signed values are smaller or equal in size to the largest
unsigned value.
Fixes #476.
Co-authored-by: Pepijn de Vos <pepijndevos@gmail.com>
whitequark [Wed, 26 Aug 2020 04:15:26 +0000 (04:15 +0000)]
sim._pyrtl: fix miscompilation of -(Const(0b11, 2).as_signed()).
Fixes #473.
whitequark [Wed, 26 Aug 2020 03:19:13 +0000 (03:19 +0000)]
lib.cdc: in AsyncFFSynchronizer(), rename domain= to o_domain=.
This is for consistency with other synchronizers.
Fixes #467.
Robin Ole Heinemann [Tue, 4 Aug 2020 15:30:18 +0000 (17:30 +0200)]
vendor.lattice_machxo_2_3l: add SRAM svf generation
Mariusz Glebocki [Sun, 2 Aug 2020 16:48:26 +0000 (18:48 +0200)]
vendor: Add initial support for Symbiflow for Xilinx 7-series
Mariusz Glebocki [Mon, 24 Aug 2020 11:03:59 +0000 (13:03 +0200)]
vendor.xilinx_7series: add `_part` property getter
Xiretza [Sat, 22 Aug 2020 13:46:58 +0000 (15:46 +0200)]
cli: Improve help texts
545e49c2 added the option to export as CXXRTL, but the help texts for
the CLI options don't reflect this yet.
whitequark [Sat, 15 Aug 2020 13:00:50 +0000 (13:00 +0000)]
docs/lang: use less confusing placeholder variable names.
Fixes #474.
awygle [Sat, 15 Aug 2020 08:40:56 +0000 (01:40 -0700)]
lib.fifo: add `r_level` and `w_level` to all FIFOs
whitequark [Thu, 13 Aug 2020 03:10:17 +0000 (03:10 +0000)]
Add Linguist tags to .gitattributes.
This should make it possible to navigate to nmigen/vendor/*.py using
GitHub's file finder.
Robin Ole Heinemann [Mon, 10 Aug 2020 15:23:29 +0000 (17:23 +0200)]
vendor.lattice_{ecp5,machxo_2_3l}: specify impl-dir correctly
whitequark [Fri, 31 Jul 2020 13:17:39 +0000 (13:17 +0000)]
build,vendor: never carry around parts of differential signals.
When a port component is skipped, it should appear neither in the RTL
nor in the constraint file. However, passing around components of
differential ports explicitly makes that harder.
Fixes #456.
Supersedes #457.
Co-authored-by: Jean THOMAS <git0@pub.jeanthomas.me>
whitequark [Fri, 31 Jul 2020 14:45:38 +0000 (14:45 +0000)]
vendor.xilinx_{7series,ultrascale}: use BUFGCTRL rather than BUFGCE.
Fixes #438 (again).
Adam Greig [Thu, 30 Jul 2020 07:05:18 +0000 (08:05 +0100)]
hdl.mem: cast reset value for transparent read ports to integer.
Jean THOMAS [Tue, 28 Jul 2020 21:02:01 +0000 (23:02 +0200)]
nmigen.lib.scheduler: add RoundRobin.
Jacob Graves [Tue, 28 Jul 2020 19:35:25 +0000 (13:35 -0600)]
tests: fix remove unnecessary workaround for some unittest assertions.
whitequark [Wed, 22 Jul 2020 02:13:10 +0000 (02:13 +0000)]
vendor.xilinx_{7series,ultrascale}: add SIM_DEVICE parameter.
The parameter defaults to "ULTRASCALE", even when synthesizing for
7-series devices. This could lead to a simulation/synthesis mismatch,
and causes a warning.
Fixes #438.
Jean THOMAS [Thu, 23 Jul 2020 12:24:31 +0000 (14:24 +0200)]
vendor.lattice_ecp5: add missing differential IO types.
whitequark [Wed, 22 Jul 2020 14:43:44 +0000 (14:43 +0000)]
back.rtlil: lower maximum accepted wire size.
In practice wires of just 100000 bits sometimes have unacceptable
performance with Yosys, so stick to Verilog's minimum limit of 65536
bits.
whitequark [Wed, 22 Jul 2020 14:32:45 +0000 (14:32 +0000)]
sim._pycoro: avoid spurious wakeups.
This bug was introduced in commit
e435a217.
whitequark [Wed, 22 Jul 2020 08:11:59 +0000 (08:11 +0000)]
CI: replace Travis with GitHub Actions.
Fixes #445.
whitequark [Tue, 21 Jul 2020 02:53:29 +0000 (02:53 +0000)]
compat.fhdl.bitcontainer: fix value_bits_sign().
This function was broken in commit
659b0e81; some downstream code
expects bits_sign to be e.g. indexable.
whitequark [Thu, 16 Jul 2020 08:00:10 +0000 (08:00 +0000)]
CI: use WASM yosys instead of building our own.
Fixes #434.
whitequark [Wed, 15 Jul 2020 04:09:58 +0000 (04:09 +0000)]
back.rtlil: fix guard for division by zero.
Oops... that should be checking the divisor, not the dividend. This
was discovered by running the test suite on cxxsim.
Filipe Laíns [Mon, 13 Jul 2020 23:42:02 +0000 (00:42 +0100)]
docs: add install instructions for arch
Signed-off-by: Filipe Laíns <lains@archlinux.org>
whitequark [Tue, 14 Jul 2020 00:25:11 +0000 (00:25 +0000)]
CI: run on pull requests as well, not just pushes.
whitequark [Mon, 13 Jul 2020 23:16:27 +0000 (23:16 +0000)]
lib.cdc: fix typo.
Co-authored-by: @ECP5-PCIe
Jacob Lifshay [Mon, 13 Jul 2020 02:10:01 +0000 (19:10 -0700)]
sim.pysim: write the next, not curr signal value to the VCD file
This is a temporary fix for #429.
whitequark [Sat, 11 Jul 2020 12:25:31 +0000 (12:25 +0000)]
sim.pysim: use VCD aliases to reduce space and time overhead.
On Minerva SoC, this reduces VCD file size by about 35%, and reduces
runtime overhead of writing VCDs by 10% or less.
whitequark [Wed, 8 Jul 2020 17:30:06 +0000 (17:30 +0000)]
sim: simplify. NFC.
whitequark [Wed, 8 Jul 2020 12:49:38 +0000 (12:49 +0000)]
back.pysim→sim.pysim; split into more manageable parts.
This is necessary to add cxxrtl as an alternate simulation engine.
whitequark [Wed, 8 Jul 2020 09:08:00 +0000 (09:08 +0000)]
vendor.xilinx_{7series,ultrascale}: remove `grade` property.
This was added in commit
bfd4538d based on a misunderstanding of how
Xilinx part numbers work.
* non-ultrascale 7-series parts don't have temperature grades;
* ultrascale parts have temperature grade as a part of speed grade.
whitequark [Wed, 8 Jul 2020 08:29:20 +0000 (08:29 +0000)]
back.pysim: only extract signal names if VCD is requested.
This commit also fixes an issue introduced in
2606ee33 that regressed
simulator startup time and bloated VCD files. (It's actually about
10% faster now than *before* the regression was introduced.)
whitequark [Wed, 8 Jul 2020 07:12:00 +0000 (07:12 +0000)]
back.pysim: reset timeline as well.
This is a bug that was introduced in
94faf497b.
whitequark [Wed, 8 Jul 2020 06:29:34 +0000 (06:29 +0000)]
back.pysim: simplify. NFC.
whitequark [Wed, 8 Jul 2020 06:04:50 +0000 (06:04 +0000)]
back.pysim: extract timeline handling to class _Timeline. NFC.
whitequark [Wed, 8 Jul 2020 05:42:33 +0000 (05:42 +0000)]
back.pysim: extract simulator commands to sim._cmds. NFC.
whitequark [Wed, 8 Jul 2020 03:55:09 +0000 (03:55 +0000)]
back.pysim: simplify. NFC.
awygle [Tue, 7 Jul 2020 05:17:03 +0000 (22:17 -0700)]
hdl.ast: don't inherit Shape from NamedTuple.
Fixes #421.
whitequark [Tue, 7 Jul 2020 04:29:13 +0000 (04:29 +0000)]
back.pysim: simplify.
Compiled process names were never particularly useful (comments in
the source would make more sense for debugging), and coroutine
process names were actually source locations.
whitequark [Tue, 7 Jul 2020 04:19:05 +0000 (04:19 +0000)]
back.pysim: simplify. NFC.
whitequark [Tue, 7 Jul 2020 04:09:10 +0000 (04:09 +0000)]
back.pysim: simplify. NFC.
whitequark [Tue, 7 Jul 2020 04:06:06 +0000 (04:06 +0000)]
back.pysim: synchronize waveform writing with cxxrtl.
whitequark [Tue, 7 Jul 2020 02:35:04 +0000 (02:35 +0000)]
back.pysim: synchronize terms with cxxrtl. NFC.
whitequark [Tue, 7 Jul 2020 02:14:06 +0000 (02:14 +0000)]
back.pysim: simplify. NFC.
whitequark [Tue, 7 Jul 2020 01:59:25 +0000 (01:59 +0000)]
back.pysim: simplify. NFC.
whitequark [Mon, 6 Jul 2020 16:01:49 +0000 (16:01 +0000)]
Remove everything deprecated in nmigen 0.2.
Alan Green [Wed, 1 Jul 2020 21:29:30 +0000 (17:29 -0400)]
Update license and copyright info
Remove non-license explanatory text from LICENSE.txt.
Create CONTRIBUTING file with instructions and notes for contributors.
This change relates to issue #412
Konrad Beckmann [Mon, 6 Jul 2020 14:04:24 +0000 (16:04 +0200)]
vendor.lattice_ecp5: Add support for io with xdr=7
This adds support for IOs with xdr=7 using the
IODDR71B and ODDR71B primitives.
Konrad Beckmann [Mon, 6 Jul 2020 14:01:19 +0000 (16:01 +0200)]
vendor.lattice_ecp5: Add support for io with xdr=4
This adds support for IOs with xdr=4 using the
IDDRX2F and ODDRX2F primitives.
whitequark [Sun, 5 Jul 2020 23:51:14 +0000 (23:51 +0000)]
docs: use working sphinxcontrib-platformpicker.
whitequark [Sun, 5 Jul 2020 23:39:47 +0000 (23:39 +0000)]
docs: use sphinxcontrib-platformpicker.
Fixes #416.
whitequark [Sat, 4 Jul 2020 02:09:35 +0000 (02:09 +0000)]
docs: link to community tutorials until we have an official one.
whitequark [Thu, 2 Jul 2020 23:11:35 +0000 (23:11 +0000)]
docs/lang: document constshifts.
whitequark [Thu, 2 Jul 2020 23:08:10 +0000 (23:08 +0000)]
docs/index: rename to "Language & toolchain".
whitequark [Thu, 2 Jul 2020 22:49:04 +0000 (22:49 +0000)]
test: remove FHDLTestCase.assertRaisesRegex.
This method is only there because I misunderstood the documentation
of unittest.
whitequark [Thu, 2 Jul 2020 22:22:44 +0000 (22:22 +0000)]
compat.fhdl.specials: fix handling of tristate (i=None) pins.
Fixes #406.
whitequark [Thu, 2 Jul 2020 18:26:08 +0000 (18:26 +0000)]
_yosys→_toolchain.yosys
whitequark [Thu, 2 Jul 2020 18:13:54 +0000 (18:13 +0000)]
vendor: `yosys` is not a required tool for proprietary toolchains.
Since commit
b9799b4c, the discovery mechanism for the Yosys required
to produce Verilog is different from the usual require_tool(); namely
it is possible to produce Verilog without a `yosys` binary on PATH.
Fixes #419.
whitequark [Wed, 1 Jul 2020 21:04:25 +0000 (21:04 +0000)]
whitequark [Wed, 1 Jul 2020 21:04:04 +0000 (21:04 +0000)]
setup: gracefully recover from missing setuptools_scm.
whitequark [Wed, 1 Jul 2020 20:49:41 +0000 (20:49 +0000)]
setup: link to the right documentation version from pip metadata.
whitequark [Wed, 1 Jul 2020 20:17:10 +0000 (20:17 +0000)]
Update README.
whitequark [Wed, 1 Jul 2020 19:16:01 +0000 (19:16 +0000)]
Update README.
whitequark [Wed, 1 Jul 2020 08:58:36 +0000 (08:58 +0000)]
docs/install: use pip/pip3 more consistently.
whitequark [Wed, 1 Jul 2020 08:53:52 +0000 (08:53 +0000)]
docs: clarify naming.
whitequark [Wed, 1 Jul 2020 08:16:28 +0000 (08:16 +0000)]
Revert "Add PEP 518 `pyproject.toml`."
This reverts commit
7fca037f9c14974f499195171bb84c38ebddc1e1.
This broke editable installs and has to be reverted due to a number
of pip issues:
* pypa/pip#6375
* pypa/pip#6434
* pypa/pip#6438
We can put this back once PEP 517/518 support editable installs.
Until then the legacy behavior will suffice, and we should just teach
people to install the dependencies in virtualenvs or something...
whitequark [Wed, 1 Jul 2020 08:03:57 +0000 (08:03 +0000)]
docs: add a crude approximation of intersphinx toctrees.
This is a pretty awful hack. We could do this properly after one of
the following issues is fixed:
* sphinx-doc/sphinx#701
* sphinx-doc/sphinx#1836
whitequark [Wed, 1 Jul 2020 07:00:02 +0000 (07:00 +0000)]
Gracefully handle missing dependencies.
Some people's workflows involve not using `pip`. This is not
a recommended way to use nMigen, but is prevalent enough for good
enough reason that we try to keep them working anyway.
whitequark [Wed, 1 Jul 2020 06:33:03 +0000 (06:33 +0000)]
docs/install: fix dead link.
whitequark [Wed, 1 Jul 2020 06:32:15 +0000 (06:32 +0000)]
docs/install: quote special chars to avoid issues with extended glob.
whitequark [Wed, 1 Jul 2020 06:15:16 +0000 (06:15 +0000)]
docs/install: explain how to install non-editable snapshot from git.
whitequark [Wed, 1 Jul 2020 05:58:05 +0000 (05:58 +0000)]
Add PEP 518 `pyproject.toml`.
This is necessary to be able to install nMigen into a virtualenv that
does not have `wheel` installed in certain cases.
See #349.
whitequark [Wed, 1 Jul 2020 05:06:20 +0000 (05:06 +0000)]
docs/install: use `pip install --user` on *nix.
whitequark [Wed, 1 Jul 2020 04:11:51 +0000 (04:11 +0000)]
docs/install: reduce repetition.
whitequark [Wed, 1 Jul 2020 04:08:02 +0000 (04:08 +0000)]
README: update to refer to documentation where applicable.
whitequark [Wed, 1 Jul 2020 03:59:16 +0000 (03:59 +0000)]
docs: reword slightly.
whitequark [Wed, 1 Jul 2020 03:46:59 +0000 (03:46 +0000)]
docs: explain how to install GTKWave on Windows.
whitequark [Wed, 1 Jul 2020 03:37:29 +0000 (03:37 +0000)]
CI: build docs on all pushes, update only on
whitequark [Wed, 1 Jul 2020 03:32:17 +0000 (03:32 +0000)]
docs: rewrite install instructions to be easier to understand.
whitequark [Tue, 30 Jun 2020 23:13:44 +0000 (23:13 +0000)]
setup: link to proper location for docs.
whitequark [Tue, 30 Jun 2020 22:29:43 +0000 (22:29 +0000)]
docs: fix syntax.
whitequark [Tue, 30 Jun 2020 22:28:12 +0000 (22:28 +0000)]
docs: fix CI workflow.
whitequark [Tue, 30 Jun 2020 22:26:54 +0000 (22:26 +0000)]
docs: explain that `pip3 install -e` should be run after pulling.
whitequark [Mon, 27 Apr 2020 07:21:31 +0000 (07:21 +0000)]
Add (heavily work in progress) documentation.
To render correctly, the docs require:
* pygments/pygments#1441
whitequark [Tue, 30 Jun 2020 22:08:59 +0000 (22:08 +0000)]
Don't use pkg_resources.
This package is deprecated and introduces a massive amount of startup
latency. On my machine with 264 installed Python packages, it reduces
the time required to `import nmigen` from ~100ms to ~200ms.