[CI] add LLVM IR bloat regression check

[arrayka]

This pull request introduces an automated check in the CI workflow to monitor and report on LLVM IR code size growth, helping to detect regressions in monomorphization cost. The main changes include adding a new CI job and a supporting shell script for comparing LLVM IR line counts between the current branch and the baseline.

CI workflow enhancements:

• Added a new llvm-lines job to .github/workflows/ci.yml that compares the LLVM IR line counts of the current branch against the main branch using cargo-llvm-lines. This job is not required for PR merges but provides regression reports and enforces a configurable growth threshold.

Supporting scripts:

• Introduced .github/scripts/compare-llvm-lines.sh, a Bash script that parses and compares the output of cargo-llvm-lines for the baseline and current builds, generating a markdown-formatted regression report for easier review.

[Copilot]

Pull request overview

Adds a CI job that compares LLVM IR line counts for diskann-benchmark between the PR branch and main to detect monomorphization bloat, plus a standalone bash script that produces a markdown delta report from two cargo llvm-lines outputs.

Changes:

• New llvm-lines job in .github/workflows/ci.yml (gated by needs: basics) that installs cargo-llvm-lines, runs it on both checkouts, computes overall growth, and writes a delta report to $GITHUB_STEP_SUMMARY.
• New .github/scripts/compare-llvm-lines.sh that parses two cargo llvm-lines outputs and emits a per-function markdown comparison table sorted by delta.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
.github/workflows/ci.yml	Adds an advisory llvm-lines job that builds baseline/current cargo llvm-lines output, checks total growth against LLVM_LINES_GROWTH_THRESHOLD, and surfaces a report to the job summary.
.github/scripts/compare-llvm-lines.sh	New bash script that diffs two cargo llvm-lines outputs and prints a sorted markdown table of per-function regressions.

Comments suppressed due to low confidence (2)

.github/workflows/ci.yml:529

• With GitHub Actions' default shell options (bash --noprofile --norc -eo pipefail, which defaults.run.shell: bash at the top of this file selects), cargo llvm-lines ... | head -100 | tee ... is very likely to fail the step. Once head has read 100 lines it closes its stdin; cargo llvm-lines then dies with SIGPIPE (exit code 141), and pipefail propagates that failure even though head and tee succeeded. Capture the full output first (e.g. cargo llvm-lines ... | tee ../current-llvm-lines.txt) and only apply head later when feeding the report, or disable pipefail locally for this command.

The same issue applies to the head -100 on line 541.

        run: cargo llvm-lines --package diskann-benchmark --release | head -100 | tee ../baseline-llvm-lines.txt

      - name: Generate current LLVM lines
        working-directory: diskann_rust
        run: cargo llvm-lines --package diskann-benchmark --release | head -100 | tee ../current-llvm-lines.txt

.github/workflows/ci.yml:539

• If baseline-llvm-lines.txt does not contain a (TOTAL) line (e.g. cargo llvm-lines output format changed, the build was warning/error-filtered, or the SIGPIPE issue above produces an empty file), baseline_total will be the empty string and $(( ... / baseline_total )) will fail with a bash arithmetic error, taking down the step with a confusing message. compare-llvm-lines.sh defends against baseline_total -eq 0 on line 23 but this inline computation does not. Consider either delegating the threshold check to the script or adding the same guard here (and defaulting empty values to 0).

          baseline_total=$(awk '/\(TOTAL\)/{print $1}' baseline-llvm-lines.txt)
          current_total=$(awk '/\(TOTAL\)/{print $1}' current-llvm-lines.txt)
          growth=$(( (current_total - baseline_total) * 100 / baseline_total ))

          if [ "$growth" -gt "$LLVM_LINES_GROWTH_THRESHOLD" ]; then
            echo "::warning::LLVM IR grew ${growth}% ($baseline_total → $current_total lines)"
          fi

---

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

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.45%. Comparing base (205aad7) to head (f9296de).
⚠️ Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1083      +/-   ##
==========================================
- Coverage   89.46%   89.45%   -0.01%     
==========================================
  Files         459      458       -1     
  Lines       85482    85398      -84     
==========================================
- Hits        76475    76392      -83     
+ Misses       9007     9006       -1     

Flag	Coverage Δ
miri	89.45% <ø> (-0.01%)	⬇️
unittests	89.08% <ø> (-0.01%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.
see 18 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.