[FIRRT] Domains: Types Approach #8830
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Add a new `firrtl.domain` op which defines the existence of a kind of domain. E.g., this can be used to declare abstract notions of a clock or reset domain. Signed-off-by: Schuyler Eldridge <[email protected]>
Add a dedicate type for representing a value of a declared domain. This is working towards being able to represent user-defined domains in FIRRTL and eventually in Chisel. Signed-off-by: Schuyler Eldridge <[email protected]>
Add domain information to some FIRRTL types: UInt, SInt, and Analog. The MLIR representation of FIRRTL Dialect now supports all of the following: ``` mlir !firrtl.uint<1> !firrtl.uint<1, []> !firrtl.uint<1, [@A]> !firrtl.uint<1, [@A, @b]> ``` Note: this should be extended in the future to be on all base types or on all types. When doing this extension, we will need to figure out how to handle bundles where there are redundant encodings for types (which is a huge red flag if allowed). Signed-off-by: Schuyler Eldridge <[email protected]>
For scalar types which currently support domains (UInt, SInt, and Analog), canonicalize their domains in their backing storage. Specifically, sort and unique the domains. This means that domains will be stored alphabetically and no duplicates will be allowed. This is done to enable trivial pointer comparisons between types and not require deep introspection to know if two types are the same. This does not, yet solve the problem of canonicalization involving aggregates whose leaves may contain domains. Aside: this will not run into the same problems that the mythological "flip type canonicalization" effort ran into. There are no operations which care about the order of domains. We are free to rearrange them as we want. Signed-off-by: Schuyler Eldridge <[email protected]>
|
Closed in favor or #8843. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This is an experimental approach to add domains to FIRRTL. This approach
sees what it looks like to put domain information into the MLIR type
system.
This gets very funky, very quickly because this is doing many things which
are very against the ethos of MLIR:
(2) is specifically working around the fact that MLIR provides no built-in
way to associate a value to a type. (There are good reasons for this!)
This will be contrasted with a future approach that does an association of
domain information on a per-operation basis.
This is broken up into the following logical commits:
firrtl.domain