| Age | Commit message (Collapse) | Author |
|---|
|
- Fix BoringUtils deduplication bug, include new tests
- Update/clarify BoringUtils scaladoc
|
|
This compresses the Scaladoc for BoringUtils slightly by using 120
character lines and removing unnecessary whitespace.
This also changes the poorly named "dedup" parameter to the what it
actually is: "disableDedup".
Signed-off-by: Schuyler Eldridge <schuyler.eldridge@ibm.com>
|
|
This adds two tests to the BoringUtilsSpec to explicitly verify that
deduplication is required when boring. This adds tests that both
verify that the test passes as expected with deduplication enabled and
that the same test fails with deduplication disabled.
Signed-off-by: Schuyler Eldridge <schuyler.eldridge@ibm.com>
|
|
This fixes a bug where BoringUtils non-hierarchical sinks would be
deduplicated even when specified that they should not be.
h/t @ucbjrl for discovering this!
Signed-off-by: Schuyler Eldridge <schuyler.eldridge@ibm.com>
|
|
|
|
|
|
Module class names (#994)
|
|
|
|
Fix scaladoc for UInt.unary_!
|
|
|
|
|
|
|
|
|
|
It performs the operation (x === 0.U), just like in C. The scaladoc
incorrectly described it as performing the operation !x(0). (Obviously, these
are equivalent for Bool, but not for UInt in general).
|
|
Generate better code for UInt.andR
|
|
|
|
In the case that the width is known, we can emit one fewer Firrtl node.
This obviously synthesizes the same way, but compiles/simulates faster.
|
|
|
|
|
|
|
|
Avoid procedural wire assignment in test resource
|
|
Verilator 4.008 dropped the hammer on procedural wire assignment to
align with the IEEE standard (first I've heard of this, though). The
VerilogVendingMachine.v test resource will error in Verilator 4.008
with a PROCASSWIRE error if you try to compile it. This fixes that
example to only assign to a register.
Signed-off-by: Schuyler Eldridge <schuyler.eldridge@ibm.com>
|
|
JDK 11 `java.lang.String#lines` conflicts with Scala `StringOps#lines`.
This has been fixed in scalac 2.12.8 but projects using 2.11 in their
cross-build need the `Predef.augmentString` patch.
[Scala bug & fix reference](https://github.com/scala/bug/issues/11125)
|
|
**NOTE**: A `publishLocal` should replace the downloaded `.ivy2/cache/...` version with a pointer to the `.ivy2/local/...` version. To force refetching of the Sonatype repository version, you should delete both `.ivy2/{cache,local}/...` versions.
|
|
|
|
* Fix width propagation of non-literals in WireInit and RegInit
* Change .getWidth to throw an exception instead of calling .get
* Add utilities for checking inferred vs. known widths
* Add tests for Wire, WireInit, Reg, and RegInit width inference
* Add ScalaDoc for Reg, Wire, RegInit, and WireInit
|
|
Improve quality of code generation for UInt.-%
|
|
This is semantically equivalent, but gets rid of a bunch of Firrtl text.
It also gets rid of a bunch of Verilog, because Firrtl is capable of
pattern-matching the new expression into SubWrap. The effect is that
we now get
wire [4:0] in;
wire [4:0] res;
assign res = 5'h0 - in;
instead of
wire [4:0] in;
wire [5:0] _T_40;
wire [5:0] _T_41;
wire [4:0] res;
assign _T_40 = 5'h0 - in;
assign _T_41 = $unsigned(_T_40);
assign res = _T_41[4:0];
|
|
|
|
asBools, asBool, and chained apply on asBools
|
|
|
|
|
|
The expanded version substituted in by the macro was misspelled, renamed
from toBools -> do_toBools as expected by the macro
|
|
- Trim stack trace to show better, reduced information to the user
- Add --full-stacktrace to FIRRTL option to show full stack trace
|
|
Now that Chisel3 has been released, you only need local Chisel to use the master branch in a project.
|
|
* Bump sbt to 1.2.6; update dependencies
* Add explanation for explicit junit library dependency
|
|
Fixes #893
|
|
|
|
- Add Chisel logo PNG and SVG to repo
|
|
|
|
Use reusable executors and reusable commands.
|
|
Add CircleCI status badge to README
|
|
|
|
Resolves #917
|
|
Be consistent with the use of /dev/null for sbt's stdin (force sbt to exit instead of bringing up a dialog).
Enforce Jenkins' scalastyle limit of 40 warnings via CHECKSTYLE_LIMIT in environment.
Force tests to run sequentially if -DminimalResources is set on the command line.
Ensure we see valid scalastyle output.
Make checkstyle dependent on one of the tests (so a style failure will fail the build only after tests pass).
|
|
|
|
* Turned off strong enum annotations because they weren't working
with Vec indexes
* Ignore annotation tests using ScalaTest's 'ignore', rather than
by commenting them out
|
|
Quickfix for LoadMemoryTransform that gets this to work with Instance
Annotations. The new Instance Annotations caused a corner case where a
LoadMemoryAnnotation would be duplicated (via update/renaming) and the
resulting annotation would differ from the original in only their
originalMemoryNameOpt field. This corrects that by having the
ChiselLoadMemoryAnnotation also emit the originalMemoryNameOpt field where
it did not previously.
First part of a fix for freechipsproject/firrtl#922.
Signed-off-by: Schuyler Eldridge <schuyler.eldridge@ibm.com>
|
|
Make BaseModule.name lazy
|
|
This wraps the evaluation of BaseModule.name in try/catch to look for a
NullPointerException that may result from trying to evaluate desiredName
before it's ready. This catches a test case of using a desiredName that
depends on a later defined eager subinstance.
h/t @jackkoenig
Signed-off-by: Schuyler Eldridge <schuyler.eldridge@ibm.com>
|
