Commit graph

165 commits

Author SHA1 Message Date
whitequark ea25806971 back.rtlil: emit \sig$next wires instead of \$next\sig. NFC.
Just a bit more readable.
2019-07-02 18:06:50 +00:00
whitequark dd5e513e42 back.rtlil: do not emit $next wires for comb signals.
According to RTLIL semantics (that was undocumented before today),
the only purpose of `sync always` is to enable inference of latches,
because there is no other way to express them in terms of RTLIL
processes without ending up with a combinatorial loop. But, nMigen
specifically avoids latches, so this is not necessary.

This change results in major improvements in Verilog readability.

See also #98.
2019-07-02 18:05:34 +00:00
whitequark f75a0163f9 back.rtlil: fix Array regression in 32446831.
Fixes #117.
2019-07-01 01:53:56 +00:00
whitequark 9c54d0c061 back.pysim: create unique ResetSynchronizer internal domains.
Commit 300d47ca introduced the same bug commit 779f3ee9 was trying to
avoid, but now only in the simulator. Since the names in simulator
don't have to make any sense, just use DUID to generate them.
2019-06-28 08:34:43 +00:00
whitequark 300d47ca2e back.pysim: override ResetSynchronizer implementation.
This was rewritten to use Yosys cells in 779f3ee9 to avoid leaking
the interior clock domain, but the simulator doesn't understand Yosys
cells. So, use the old implementation in the simulator.
2019-06-28 07:49:14 +00:00
whitequark 32446831b4 hdl.{ast,dsl}, back.{pysim,rtlil}: allow multiple case values.
This means that instead of:

    with m.Case(0b00):
        <body>
    with m.Case(0b01):
        <body>

it is legal to write:

    with m.Case(0b00, 0b01):
        <body>

with no change in semantics, and slightly nicer RTLIL or Verilog
output.

Fixes #103.
2019-06-28 04:37:08 +00:00
whitequark 48d4ee4031 hdl.ir, back.rtlil: allow specifying attributes on instances.
Fixes #107.
2019-06-28 04:14:38 +00:00
whitequark 6f4e3156d8 back.pysim: fix scope screwup. 2019-06-26 05:22:09 +00:00
whitequark e5e23644a4 hdl.{ast,dst}: directly represent RTLIL default case.
This makes RTLIL mildly nicer:

 casez ({ \$5 , \$3 , \$1  })
   3'bzz1:
       \$next\o  = \$7 ;
   3'bz1z:
       \$next\o  = \$9 ;
   3'b1zz:
       \$next\o  = \$11 ;
-  3'bz:
+  default:
       { \$next\co , \$next\o  } = \$13 ;
 endcase
2019-06-25 22:01:14 +00:00
whitequark 066dd799e8 back.pysim: check for a clock being added twice.
This commit adds a best-effort error for a common mistake of adding
a clock driving the same domain twice, such as a result of
a copy-paste error.

Fixes #27.
2019-06-11 03:54:22 +00:00
whitequark d2d8c2b8bf back.rtlil: mask memory init values.
This handles both init values that are too wide, which happens if
their magnitude is too high, or if they're negative.

Fixes #96.
2019-06-11 03:43:09 +00:00
whitequark f17375a60b back.rtlil: allow specifying platform for convert(). 2019-05-26 17:10:56 +00:00
whitequark 921f506e69 back.rtlil: assign undriven signals to their reset value.
Fixes #35.
2019-05-13 08:33:55 +00:00
whitequark 97af266645 back.verilog: allow stripping the src attribute, for cleaner output. 2019-04-22 14:59:53 +00:00
whitequark 585514e6ed hdl.ir: rework named port handling for Instances.
The main purpose of this rework is cleanup, to avoid specifying
the direction of input ports in an implicit, ad-hoc way using
the named ports and ports dictionaries.

While working on this I realized that output ports can be connected
to anything that is valid on LHS, so this is now supported too.
2019-04-22 07:46:47 +00:00
whitequark 85ae99c1b4 back.rtlil: emit nmigen.hierarchy attribute.
Fixes #54.
2019-04-21 07:55:08 +00:00
whitequark 083016d747 back.rtlil: only expand legalized values in Array/Part context on RHS.
Otherwise the following code fails to compile:

    index = Signal(1)
    array = Array(range(2))
    with m.If(0 == array[index]):
        m.d.sync += index.eq(0)

Fixes #51.
2019-04-21 06:43:31 +00:00
whitequark f22106e5ef back.rtlil: allow record slices on LHS. 2019-04-20 08:12:29 +00:00
whitequark a57c72d606 back.rtlil: fix off-by-one in Part legalization.
Fixes #52.
2019-03-28 05:12:12 +00:00
whitequark 43e4833ddb back.rtlil: accept ast.Const as cell parameter. 2019-01-26 23:25:54 +00:00
whitequark e74dbc3377 back.pysim: support async reset. 2019-01-26 18:07:43 +00:00
whitequark 8686e9aa06 back.pysim: give better names to unnamed fragments and their signals.
Was: top.#0, top.None_clk
Now: top.U0, top.U0_clk

(U for Unnamed, or similarly, an unit refdes.)
2019-01-26 18:07:16 +00:00
whitequark b133eb735f back.rtlil: accept any elaboratable, not just fragments. 2019-01-26 16:11:29 +00:00
whitequark 4948162f33 hdl.ir: rename .get_fragment() to .elaborate().
Closes #9.
2019-01-26 02:31:12 +00:00
whitequark 7b25665fde back.pysim: fix behavior of initial cycle for sync processes.
The current behavior was introduced in 65702719, which was a wrong
fix for an issue that was actually fixed in 12e04e4e. This commit
effectively reverts 65702719 and 1782b841.
2019-01-25 20:37:56 +00:00
whitequark 12e04e4ee5 back.pysim: wake up processes before ever committing any values.
Otherwise, the contract of the simulator to sync processes is not
always fulfilled.
2019-01-21 16:00:25 +00:00
whitequark b50b47d984 hdl.ast: give Assert and Assume their own src_loc.
This helps with patterns like `Assert(fsm.ongoing("IDLE"))`, which
would otherwise point into nMigen internals.
2019-01-19 00:08:51 +00:00
whitequark 66466a8a0e back.rtlil: only emit each AnyConst/AnySeq cell once.
These are semantically like signals, not like constants.
2019-01-18 01:34:48 +00:00
whitequark 198efcad31 hdl.xfrm: add SampleLowerer. 2019-01-17 01:41:02 +00:00
whitequark f2425001aa back.rtlil: slightly nicer naming for $next signals. NFC. 2019-01-16 17:20:38 +00:00
whitequark 935bf2d8cf back.rtlil: rename \sig$next to $next$sig.
These used to serve a useful purpose being public, back when the RTLIL
backend was immature. Not anymore; now they merely clutter up views
in gtkwave and so on.
2019-01-16 14:51:20 +00:00
whitequark 6191760c30 Unbreak 655d02d5. 2019-01-15 23:09:10 +00:00
William D. Jones 655d02d5b8 back.rtlil: Generate $anyconst and $anyseq cells. 2019-01-15 22:52:45 +00:00
William D. Jones 77728c2dea hdl.xfrm: Add on_AnyConst and on_AnySeq abstract methods for ValueVisitor and children. 2019-01-15 22:52:45 +00:00
whitequark cbf7bd6e31 back.pysim: handle non-driven, non-port signals.
Fixes #20.
2019-01-13 08:31:38 +00:00
whitequark 06faeee357 back.verilog: better error message if Yosys is not found.
Fixes #17.
2019-01-13 08:10:23 +00:00
whitequark 307de722cb back.verilog: remove undriven check.
This check no longer finds bugs and is prone to false positives.
Instead, we should do integration tests on the entire stack, from
fragments to Verilog.

Fixes #23.
2019-01-08 22:43:09 +00:00
Adam Greig 560bb007cc Give the top level scope a name to fix VCD hierarchy. 2019-01-06 00:10:37 +00:00
whitequark ef1e0b8d55 back.rtlil: translate empty slices correctly. 2019-01-02 18:14:29 +00:00
William D. Jones f31055a4ef back.rtlil: Generate RTLIL for Assert/Assume statements. 2019-01-02 11:17:39 +00:00
William D. Jones f77dc40256 hdl.xfrm: Add Assert and Assume abstract methods for StatementVisitor, implement for children. 2019-01-02 11:17:39 +00:00
whitequark 1a9dcd2f28 back.rtlil: fix typo. 2019-01-01 08:50:28 +00:00
whitequark ae3c5834ed back.rtlil: match shape of Array elements to ArrayProxy shape.
Fixes #15.
2018-12-31 03:43:34 +00:00
whitequark cdc40eaa9b back.rtlil: fix typo. 2018-12-31 03:37:38 +00:00
whitequark 849c649259 back.pysim: warn if simulation is not run.
This would have prevented 3ea35b85.
2018-12-29 15:02:04 +00:00
whitequark 92a96e1644 hdl.rec: add basic record support. 2018-12-28 13:22:10 +00:00
whitequark fe8cb55204 lib.cdc: add tests for MultiReg. 2018-12-26 12:58:30 +00:00
whitequark b4fbef65ca back.rtlil: clarify $verilog_initial_trigger behavior. NFC. 2018-12-26 06:45:57 +00:00
whitequark 010ddb96b5 back.rtlil: unbreak d47c1f8a. 2018-12-24 19:11:07 +00:00
whitequark d47c1f8a8a back.rtlil: use one $meminit cell, not one per word.
This is *far* more efficient.
2018-12-24 11:53:58 +00:00