Skip to content
GitLab
  1. Jul 26, 2020
    • Schuyler Eldridge's avatar
      Add missing ResolveKinds/ResolveFlows invalidates · f5e7c2ef
      Schuyler Eldridge authored 
      

Update the dependencies of transforms which empirically invalidate
ResolveKinds or ResolveFlows.

This was discovered with some to-be-committed work that automates
looking for incorrect transform dependencies. Specifically, the
deficient transforms were found to not invalidate ResolveKinds or
ResolveFlows, yet running ReslveKinds or ResolveFlows after these
transforms would result in modifications to the circuit.

The LoweringCompilersSpec was updated with patches to indicate the
location of these new transforms.

Signed-off-by: default avatarSchuyler Eldridge <schuyler.eldridge@ibm.com>
      f5e7c2ef
  3. Jul 25, 2020
    • Schuyler Eldridge's avatar
      Integrate new transforms with firrtl.stage.Forms (#1754) · d4e1a466
      Schuyler Eldridge authored 
      

Move new transforms, recently added, into existing or new sets of
transforms (defined in firrtl.stage.Forms).

One transform is a mandatory low FIRRTL optimization:

  - firrtl.transforms.LegalizeAndReductionsTransform

Previously, this was included as a prerequisite of all Verilog
emitters (minimum, normal, and SystemVerilog).

Two transforms associated with converting and removing the new
verification statements are moved into a new set of transforms,
AssertsRemoved:

  - firrtl.transforms.formal.ConvertAsserts
  - firrtl.transforms.formal.RemoveVerificationStatements

Previously, these transforms were directly added as prerequisites to
the minimum Verilog and normal Verilog emitter, but not the
SystemVerilog emitter.

The designation of inputForm=LowForm for legacy, custom transforms is
updated to include assertion removal transforms as part of their
optionalPrerequisites. This has the effect of continuing to cause
inputForm=LowForm transforms to run as late as possible (right before
the low FIRRTL, minimum Verilog, Verilog, or SystemVeriog emitter).

Tests are updated to reflect the new order in both CustomTransformSpec
and LoweringCompilersSpec.

Signed-off-by: default avatarSchuyler Eldridge <schuyler.eldridge@ibm.com>

Co-authored-by: default avatarmergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
      d4e1a466
    • Albert Chen's avatar
      Fix sign extension issue in Emitter (#1785) · b24b9a01
      Albert Chen authored 
      * add sign-extend const-prop test

* Emitter: don't wrap Neg operand in concat
      b24b9a01
  5. Jul 24, 2020
  7. Jul 20, 2020
  9. Jul 18, 2020
    • Kevin Laeufer's avatar
      Faster dedup instance graph (#1732) · 1b9f4ddf
      Kevin Laeufer authored 
      

* dedup: add faster InstanceGraph implementation and use it in dedup

The new implementation takes care not to hash the instance
types contained in DefInstance nodes.
This should make dedup considerably faster.

* FastInstanceGraph: cache vertices for faster findInstancesInHierarchy

* FastInstanceGraph: remove the parent name field since it isn't actually necessary

* FastInstanceGraph -> InstanceKeyGraph

* InstanceGraph: describe performance problems.

* InstanceKeyGraph: turn moduleMap into a def instead of a val

This will make changing implementation details much easier
in the future.

* InstanceKeyGraph: return childInstances as Seq instead of Map

This ensures a deterministic iteration order and it
can easily be turned into a Map for O(1) accesses.

* InstanceKeyGraph: add tests for public methods

* InstanceKeyGraph: group public methods together

* InstanceKeyGraphSpec: fix wording of a comment

Co-authored-by: default avatarmergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
      1b9f4ddf
    • Schuyler Eldridge's avatar
      Cross-build unidoc in CI (#1772) · 5f70175d
      Schuyler Eldridge authored 
      

Add cross-building of unidoc to test issues if a user later does a
+publishLocal (like with Chisel CI).

Signed-off-by: default avatarSchuyler Eldridge <schuyler.eldridge@ibm.com>
      5f70175d
    • Schuyler Eldridge's avatar
      Merge pull request #1771 from freechipsproject/fuzzer-2.11-fixes · 59607657
      Schuyler Eldridge authored 
      Fix Fuzzer for 2.11
      59607657
  11. Jul 17, 2020
  13. Jul 16, 2020
    • Kevin Laeufer's avatar
      ir: store FileInfo string in escaped format (#1690) · 24be0ac3
      Kevin Laeufer authored 
      This should speed up the common case
as the compiler never operates on the
unescaped string.
The new escape function also fixes a bug
where ']' was not escaped even though it
is the delimiting character for FileInfo.
In order to maintain backwards
compatibility for the ProtoBuf format,
this patch adds escape/unescape calls
when going from/to protobuf format.
For better performance we should consider
changing the protobuf format.
      24be0ac3
  15. Jul 15, 2020
  17. Jul 14, 2020
  19. Jul 11, 2020
  21. Jul 10, 2020
  23. Jul 09, 2020
    • Kevin Laeufer's avatar
      dedup: use structural sha256 hash instead of agnostify and serialize (#1731) · 3fee5f87
      Kevin Laeufer authored 
      

* benchmark: add hot.DedupBenchmark

* dedup: use structural md5 hash instead of agnostify and serialize

* StructuralHash: generate PrimOp LUT

* StructuralHash: change MessageDigestHasher to not be a case class

* StructuralHash: we want Blocks and EmptyStmt to be ignored

* StructuralHash: use SHA-256 instead of MD5

* StructuralHash: clarify extmodule port name agnistification

* StructuralHash: hash the name of width vars instead of trying to agnostify

This should be in line with the old Dedup behavior.
The prior use of n(..) was incorrect since the namespace
of these vars is different from the normal module scope namespace.

* StructuralHash: address Schuyler's review comments

Co-authored-by: default avatarmergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
      3fee5f87
    • Kevin Laeufer's avatar
      ir: add faster serializer (#1694) · 05ba1c9d
      Kevin Laeufer authored 
      

This Serializer which is implemented
external to the IR node definition
uses a StringBuilder to achieve about a
1.7x performance improvement when serializing.

Eventually, all implementations of the
`serialize` methd should be replaced with
a call to `Serializer.serialize`.

However, for this PR we keep the old
code in place in order to allow for easy
regression testing with the benchmark JAR
like this:
> java -cp utils/bin/firrtl-benchmark.jar \
  firrtl.benchmark.hot.SerializationBenchmark \
  ~/benchmarks/medium.pb 2 5 test

Co-authored-by: default avatarmergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
      05ba1c9d
  25. Jul 08, 2020
  27. Jul 02, 2020
  29. Jun 27, 2020
  31. Jun 26, 2020