Fuse matching product scf.if regions

[1sgtpepper]

Summary

Fuse matching compute/constrain scf.if regions in product functions while preserving constrained signal reads, source order, and observable effects. Unsafe or ambiguous pairs remain separate.

Related issues

Part of #300. Supports identical SSA conditions and resultless constrain conditionals.

Changes

• Preserve compute results, compatible conditional/yield attributes, and combined source locations during conditional fusion.
• Retain eligible constrain-side signal reads and reject movement across observable effects or invalid dominance.
• Keep loop fusion restricted to matching induction sequences and comparison modes with safe body interleaving. Document the loop and conditional legality rules separately.

Testing

• Fork validation passed: nix build for .#debugGCC, .#debugClangCov, .#checks.x86_64-linux.llzk-installcheck-debug, and .#checks.x86_64-linux.llzk-installcheck-release; nix --option sandbox false --print-build-logs build .#docs.
• Fixture replay: test commands passed; output-comparison differences were reviewed manually (replay job remains failed).
• Style checks: git-clang-format 20.1.8 passed; clang-tidy reported no new LLZK source diagnostics.

Submission checklist

• If I am an external contributor, this PR has a linked issue marked approved; otherwise, this does not apply.
• I added or updated tests for all relevant behavior, or explained above why tests are not needed.
• I updated the relevant TableGen or other documentation, or explained above why documentation is not needed.
• I added a changelog entry describing user-visible changes. (create-changelog in the Nix development shell creates a template.)
• I enabled Allow edits from maintainers if this PR comes from a fork.

AI assistance

• No AI tools contributed to this PR.
• AI tools contributed to this PR.

Tools used: AI coding assistant

How the tools contributed: Implementation and code review.

How I verified the contribution: Reviewed the final diff and exact-head hosted results.

[iangneal]

@codex review

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: b4213c67d7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[raghav198]

The constrain use is remapped to the branch-local compute value instead of hoisting the read.

I may be misunderstanding, but I think this is not safe when the struct.readm is a constrain op:

%0 = felt.add %a, %b {source = "compute"}
struct.writem %self[@foo] = %0 {source = "compute"}
%foo = struct.readm %self[@foo] {source = "constrain"}
constrain.eq %foo, %c {source = "constrain"}

emits a constraint guaranteeing that the @foo signal equals the result of %c, whereas in the remapped

%0 = felt.add %a, %b {source = "compute"}
struct.writem %self[@foo] = %0 {source = "compute"}
constrain.eq %0, %c {source = "constrain"}

the emitted constraint doesn't refer to a struct signal at all.

[1sgtpepper]

Thanks for catching this, Raghav. You're right. Remapping the struct.readm result to the branch local compute value can remove the member signal from the emitted constraint. I'll make this fusion case conservative for now and add a regression test so we keep the constraint tied to the member read.

[Copilot]

Pull request overview

This PR extends the -llzk-fuse-product-loops pass to also fuse matching compute/constrain scf.if regions in struct product programs (Fixes #300), and adds regression tests covering fusable and non-fusable scf.if patterns.

Changes:

• Add scf.if fusion logic that clones compute + constrain branches into a single fused scf.if when conditions and structural constraints match.
• Refactor pass entry/recursion to fuse both scf.if and scf.for control flow via a shared fuseMatchingRegionControlFlow routine.
• Add lit/FileCheck tests for scf.if fusion and “do not fuse” guardrails; add an unreleased changelog entry.

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
lib/Transforms/LLZKFuseProductLoopsPass.cpp	Implements scf.if fusion and refactors recursion over region control flow.
test/Transforms/FuseProductLoops/fuse_if_no_member_read.llzk	Positive test ensuring scf.if fusion occurs in a safe case.
test/Transforms/FuseProductLoops/fuse_loop_then_if.llzk	Ensures loop fusion still works and enables nested scf.if fusion after loop fusion.
test/Transforms/FuseProductLoops/no_fuse_if_call_after_write.llzk	Negative test: blocks fusion when a call occurs after a mapped write.
test/Transforms/FuseProductLoops/no_fuse_if_condition_mismatch.llzk	Negative test: blocks fusion when conditions differ.
test/Transforms/FuseProductLoops/no_fuse_if_internal_read.llzk	Negative test: blocks fusion when constrain side reads members inside the scf.if.
test/Transforms/FuseProductLoops/no_fuse_if_read_after_write.llzk	Negative test: blocks fusion when a member read appears between the ifs.
changelogs/unreleased/fuse-product-if-regions.yaml	Changelog entry for the new scf.if fusion capability.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[iangneal]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Can't wait for the next one!

Reviewed commit: 326acc24af

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[1sgtpepper]

I seem to have found some loose ends, will update in a bit.

[1sgtpepper]

Pushed a conservative safety guard for moved constrain branches.

[iangneal]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Hooray!

Reviewed commit: 8725011bd2

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

resolved

[1sgtpepper]

I think I've found some issues with the pr will mark draft for now while I fix them

[1sgtpepper]

@iangneal I kept the legacy name to avoid breaking existing CLI and pipeline references. Should I open a follow-up issue to rename them, or keep it as-is?

[tim-hoffman]

@iangneal I kept the legacy name to avoid breaking existing CLI and pipeline references. Should I open a follow-up issue to rename them, or keep it as-is?

Go ahead and do the requested renaming in this PR.

[1sgtpepper]

@tim-hoffman Small non-blocking missed-optimization follow-up related to #619: a same-parent loop in another block can suppress valid same-block fusion. Reproducer · IR output. Result: baseline = 1 fused loop; cross-block = 1 compute + 2 constrain, 0 fused. Should this be tracked under #619?