[MLIR] Properties migration todo-list

[joker-eph]

This is a tracking tickets for the list of things to do to complete the migration to properties as the preferred way to store static informations on operations.

Some context:

• https://discourse.llvm.org/t/mlir-open-meeting-properties-design-discussion-and-next-steps/87774
• https://discourse.llvm.org/t/non-attribute-properties-and-the-inherent-discardable-attribute-split-what-needs-doing/87782

Finishing the split from discardable attributes

• prop-dict printer: should use the custom printing for the attributes it contains.
• Declarative assembly should never include properties in attr-dict (forces prop-dict to be present)
• Operation builder that consistently stops looking into attributes to define properties.

Upstream migration:

• Audits of getAttrs(), getAttrDictionary() and so on. Example: the “sym_name” should not be load bearing as a key, we should use op interface consistently.
• Start migrating upstream dialects (llvm, arith, scf as first candidates?)

Other improvements:

• “Context-free” parsing (remove the need for attribute conversion)
• C-bindings generated from ODS for properties.

We can file tracking tickets for some of the larger tasks as needed.

[llvmbot]

@llvm/issue-subscribers-mlir

Author: Mehdi Amini (joker-eph)

This is a tracking tickets for the list of things to do to complete the migration to properties as the preferred way to store static informations on operations.

Some context:

• https://discourse.llvm.org/t/mlir-open-meeting-properties-design-discussion-and-next-steps/87774
• https://discourse.llvm.org/t/non-attribute-properties-and-the-inherent-discardable-attribute-split-what-needs-doing/87782

Finishing the split from discardable attributes

• prop-dict printer: should use the custom printing for the attributes it contains.
• Declarative assembly should never include properties in attr-dict (forces prop-dict to be present)
• Operation builder that consistently stops looking into attributes to define properties.

Upstream migration:

• Audits of getAttrs(), getAttrDictionary() and so on. Example: the “sym_name” should not be load bearing as a key, we should use op interface consistently.
• Start migrating upstream dialects (llvm, arith, scf as first candidates?)

Other improvements:

• “Context-free” parsing (remove the need for attribute conversion)
• C-bindings generated from ODS for properties.

We can file tracking tickets for some of the larger tasks as needed.