Mike Urbach

[PyCDE] Add new API to import existing HW modules into a System.

3

This can be used to load HW modules from an MLIR file on disk, and insert them into a fresh System.

This is WIP to demonstrate how the pass can be extended to: * Extract wires and instances it knows are only used in the verification statements * Remove things that...

[Scheduling] Explore support for SMT solvers

4

I don't know much about SMT, but I had a chat with @amaleewilson who does, and the conclusion was that it would be interesting to explore formulating the CIRCT scheduling...

[Scheduling] Import/Export support for external problem formats

2

Now that the SSP dialect is starting to land, one potentially enhancement is to import/export CIRCT scheduling problems into and out of well-defined solver formats. For example, SAT problems have...

[Python] Use the "private namespace" approach to include the upstream `mlir` packages within the `circt` package.

3

This was discussed on Discord nearly a year ago: https://discord.com/channels/636084430946959380/742572728787402763/873039553517215854. We just never got around to implementing it. The Discord points out a patch where IREE adopts a similar approach:...

[FIRRTL] Emit binds alongside the modules and interfaces they bind in

1

Emitting all of the binds into monolithic *.bindings.sv files makes it impossible for tools to work with a subset of the design. If they are only interested in a specific...

This is obviously not a huge deal, but there is a slight inconvenience in that it appears to be impossible to add new System Verilog tests at the end of...

Add a `fud` stage to go through the MLIR backend and use CIRCT to compile all the way to Verilog.

1

@rachitnigam brought this up on the CIRCT PR and issue about compiling Calyx to CIRCT's core dialects: https://github.com/llvm/circt/issues/2892#issuecomment-1128162272. Quoting that comment: > A reasonable next goal is to lower all...

Pipeline multiply and other primitives

3

This came up over here: https://github.com/cucapra/calyx/pull/909#discussion_r811391088. The multiplier and some of the other non-trivial primitives use multiple stages internally, and it should be possible to pipeline them with II=1. Now...

[FIRRTL] Add folders for IntegerMulOp

5

This is a primitive Property operation, defined by FIRRTL spec section 25.1.2. General folds for signed arbitrary precision integer arithmetic can be applied, but none were defined in the initial...