chore(ssa): Initial validation module

[vezenovm]

Description

Problem*

Working towards #8705. It has multiple sub-issues so once those are resolved I think we could close the issue.

Summary*

This PR simply makes an initial validation module that moves over some of our assertions on Function to a separate Validator struct. This felt like the simplest solution for our current validation plans. We can then expand this validator in the future as needed (e.g. extra state like the CFG and DominatorTree).

Additional Context

Documentation*

Check one:

• No documentation needed.
• Documentation included in this PR.
• [For Experimental Features] Documentation to be submitted in a separate PR.

PR Checklist*

• I have tested the changes locally.
• I have formatted the changes with Prettier and/or cargo fmt on default settings.

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Test Suite Duration'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 8d6e7a2	Previous: bf1dbdb	Ratio
test_report_zkpassport_noir_rsa_	2 s	1 s	2

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench

[aakoshh]

Would it make sense to run this between SSA passes if some CLI flag is on? Maybe in integration tests?

[TomAFrench]

One thing to note here is that there's various validations which make sense to check on the initial SSA (even if we're starting halfway through the SSA optimization pipeline) and validations which we expect to hold after each and every SSA pass.

I mention this as we can have cases where we assert that the initial SSA must follow rules like "all arithmetic must be checked" however SSA optimizations may convert checked arithmetic to unchecked later on so it would be good to have the validator flexible enough to handle this.

[vezenovm]

Would it make sense to run this between SSA passes if some CLI flag is on? Maybe in integration tests?

Yeah a CLI command for this would be useful. Although this verifier is just on the initial SSA at the moment. I think we can extend the CLI in a follow-up. To start I think this on its own provides a nice benefit to fuzzing as it helps check that the SSA generated is well formed.

[vezenovm]

One thing to note here is that there's various validations which make sense to check on the initial SSA (even if we're starting halfway through the SSA optimization pipeline) and validations which we expect to hold after each and every SSA pass

I will add configurations to the Validator to enable handling validation of different stages of the compilation pipeline. I will build off of this branch. For now, this PR is ready for review again.

[vezenovm]

One thing to note here is that there's various validations which make sense to check on the initial SSA (even if we're starting halfway through the SSA optimization pipeline) and validations which we expect to hold after each and every SSA pass.

I mention this as we can have cases where we assert that the initial SSA must follow rules like "all arithmetic must be checked" however SSA optimizations may convert checked arithmetic to unchecked later on so it would be good to have the validator flexible enough to handle this.

I was starting to add some configuration settings but at the moment I wasn't sure what invariants we apply to the initial SSA that we don't apply after other passes. There are per pass invariants (e.g. no more function types in parameters after defunctionalize) and I started on marking different stages of the compilation pipeline in the validator. However, I was thinking this may be good to keep localized in each pass. We can then have a hybrid approach where we apply per-pass localized checks, while this Validator can be for invariants that should hold about the SSA form no matter the pass. I feel this may be a better separation of concerns.

[asterite]

Looks good! I left a comment for something that might need a fix.

[github-actions]

⚠️ Performance Alert ⚠️

Possible performance regression was detected for benchmark 'Execution Time'.
Benchmark result of this commit is worse than the previous benchmark result exceeding threshold 1.20.

Benchmark suite	Current: 84c9529	Previous: bf1dbdb	Ratio
rollup-merge	0.004 s	0.003 s	1.33

This comment was automatically generated by workflow using github-action-benchmark.

CC: @TomAFrench