scenesdetect 0.1.0

name: Benchmarks

on:
  push:
    branches:
      - main
    paths:
      - 'benches/**'
      - 'src/**'
      - 'Cargo.toml'
      - 'Cargo.lock'
      - '.github/workflows/benchmark.yml'
  pull_request:
    paths:
      - 'benches/**'
      - 'src/**'
      - 'Cargo.toml'
      - 'Cargo.lock'
      - '.github/workflows/benchmark.yml'
  workflow_dispatch:

env:
  CARGO_TERM_COLOR: always
  RUST_BACKTRACE: 1

jobs:
  benchmark:
    name: ${{ matrix.label }}
    strategy:
      fail-fast: false
      matrix:
        include:
          # aarch64 — exercises the NEON SIMD backend (vld3q_u8 deinterleave,
          # vabdq_u8 / vpaddlq mean-abs-diff, NEON Sobel).
          - os: macos-latest
            arch: aarch64
            tier: neon
            rustflags: ''
            label: macos-aarch64-neon

          # x86_64 default: the runtime dispatcher (`is_x86_feature_detected!`)
          # picks AVX2 on modern GH runners, falls back to SSSE3 otherwise.
          # This exercises the x86 dispatch code path as shipped.
          - os: ubuntu-latest
            arch: x86_64
            tier: default
            rustflags: ''
            label: ubuntu-x86_64-default

          # x86_64 with `-C target-cpu=native`: lets LLVM auto-vectorize the
          # non-SIMD scalar code (histogram accumulate, phash DCT, adaptive
          # rolling sum, etc.) with the full feature set of the runner's CPU.
          # Complements the default tier to show the ceiling of scalar wins.
          - os: ubuntu-latest
            arch: x86_64
            tier: native
            rustflags: '-C target-cpu=native'
            label: ubuntu-x86_64-native

          # x86_64 with SSSE3 forced on at compile time and AVX/AVX2 off:
          # exercises the SSSE3 dispatch path even when the runner CPU
          # supports AVX2. We gate on compile-time target_feature in
          # `content/arch.rs` only in the `not(feature = "std")` branch; with
          # std the dispatcher uses `is_x86_feature_detected!`, so this tier
          # primarily guards that the SSSE3 module *compiles* without AVX2.
          - os: ubuntu-latest
            arch: x86_64
            tier: ssse3-only
            rustflags: '-C target-feature=+ssse3,-avx,-avx2,-fma'
            label: ubuntu-x86_64-ssse3-only

          # Windows x86_64 — same dispatcher as Linux but validates the MSVC
          # toolchain handles the intrinsics-heavy modules.
          - os: windows-latest
            arch: x86_64
            tier: default
            rustflags: ''
            label: windows-x86_64-default

    runs-on: ${{ matrix.os }}
    env:
      RUSTFLAGS: ${{ matrix.rustflags }}
    steps:
      - uses: actions/checkout@v6

      - name: Install Rust
        run: rustup update stable --no-self-update && rustup default stable

      - name: Print CPU info (Linux)
        if: runner.os == 'Linux'
        shell: bash
        run: |
          echo "=== /proc/cpuinfo (first flags line) ==="
          grep -m1 '^flags' /proc/cpuinfo || true
          echo "=== lscpu ==="
          lscpu || true

      - name: Print CPU info (macOS)
        if: runner.os == 'macOS'
        shell: bash
        run: |
          echo "=== sysctl machdep.cpu ==="
          sysctl machdep.cpu || true
          echo "=== uname -m ==="
          uname -m

      - name: Print CPU info (Windows)
        if: runner.os == 'Windows'
        shell: pwsh
        run: |
          Get-CimInstance Win32_Processor | Select-Object Name, Manufacturer, NumberOfCores, NumberOfLogicalProcessors | Format-List

      - name: Cache cargo build and registry
        uses: actions/cache@v5
        with:
          path: |
            ~/.cargo/registry
            ~/.cargo/git
            target
          key: ${{ runner.os }}-bench-${{ matrix.tier }}-${{ hashFiles('**/Cargo.lock') }}
          restore-keys: |
            ${{ runner.os }}-bench-${{ matrix.tier }}-
            ${{ runner.os }}-bench-

      - name: Run benchmarks - histogram
        shell: bash
        run: cargo bench --bench histogram -- --output-format bencher | tee benchmark-histogram-${{ matrix.label }}.txt
        continue-on-error: true

      - name: Run benchmarks - phash
        shell: bash
        run: cargo bench --bench phash -- --output-format bencher | tee benchmark-phash-${{ matrix.label }}.txt
        continue-on-error: true

      - name: Run benchmarks - threshold
        shell: bash
        run: cargo bench --bench threshold -- --output-format bencher | tee benchmark-threshold-${{ matrix.label }}.txt
        continue-on-error: true

      - name: Run benchmarks - content
        shell: bash
        run: cargo bench --bench content -- --output-format bencher | tee benchmark-content-${{ matrix.label }}.txt
        continue-on-error: true

      - name: Run benchmarks - adaptive
        shell: bash
        run: cargo bench --bench adaptive -- --output-format bencher | tee benchmark-adaptive-${{ matrix.label }}.txt
        continue-on-error: true

      - name: Collect benchmark summary
        shell: bash
        run: |
          summary="benchmark-summary-${{ matrix.label }}.md"
          echo "## Benchmark Results for ${{ matrix.label }}" > "$summary"
          echo "" >> "$summary"
          echo "### System Information" >> "$summary"
          echo "- OS: ${{ matrix.os }}" >> "$summary"
          echo "- Arch: ${{ matrix.arch }}" >> "$summary"
          echo "- SIMD tier: ${{ matrix.tier }}" >> "$summary"
          echo "- Runner: ${{ runner.name }}" >> "$summary"
          echo "- Runner arch (GH): ${{ runner.arch }}" >> "$summary"
          echo "- RUSTFLAGS: \`${{ matrix.rustflags }}\`" >> "$summary"
          echo "- Date: $(date -u +"%Y-%m-%d %H:%M:%S UTC")" >> "$summary"
          echo "" >> "$summary"

          for bench in histogram phash threshold content adaptive; do
            file="benchmark-${bench}-${{ matrix.label }}.txt"
            if [ -f "$file" ]; then
              echo "### ${bench}" >> "$summary"
              echo "" >> "$summary"
              echo "\`\`\`" >> "$summary"
              grep "^test " "$file" >> "$summary" || echo "No results" >> "$summary"
              echo "\`\`\`" >> "$summary"
              echo "" >> "$summary"
            fi
          done

          cat "$summary"

      - name: Create benchmark archive
        shell: bash
        run: |
          mkdir -p benchmark-results
          mv benchmark-*.txt benchmark-results/ 2>/dev/null || true
          mv benchmark-summary-${{ matrix.label }}.md benchmark-results/ 2>/dev/null || true
          if [ -d "target/criterion" ]; then
            cp -r target/criterion benchmark-results/criterion-${{ matrix.label }} || true
          fi

      - name: Upload benchmark results
        uses: actions/upload-artifact@v7
        with:
          name: benchmark-results-${{ matrix.label }}
          path: benchmark-results/
          retention-days: 90

      - name: Upload Criterion detailed results
        uses: actions/upload-artifact@v7
        if: always()
        with:
          name: criterion-detailed-${{ matrix.label }}
          path: target/criterion/
          retention-days: 90
        continue-on-error: true

  # Aggregate results from all platforms and SIMD tiers.
  aggregate-results:
    name: Aggregate benchmark results
    needs: benchmark
    runs-on: ubuntu-latest
    if: always()
    steps:
      - name: Download all benchmark results
        uses: actions/download-artifact@v6
        with:
          path: all-results

      - name: Create combined summary
        shell: bash
        run: |
          echo "# Benchmark Results Summary" > BENCHMARK_SUMMARY.md
          echo "" >> BENCHMARK_SUMMARY.md
          echo "Date: $(date -u +"%Y-%m-%d %H:%M:%S UTC")" >> BENCHMARK_SUMMARY.md
          echo "" >> BENCHMARK_SUMMARY.md

          for os_dir in all-results/benchmark-results-*/; do
            if [ -d "$os_dir" ]; then
              for summary in "$os_dir"benchmark-summary-*.md; do
                if [ -f "$summary" ]; then
                  echo "" >> BENCHMARK_SUMMARY.md
                  cat "$summary" >> BENCHMARK_SUMMARY.md
                  echo "" >> BENCHMARK_SUMMARY.md
                  echo "---" >> BENCHMARK_SUMMARY.md
                fi
              done
            fi
          done

          cat BENCHMARK_SUMMARY.md

      - name: Upload combined results
        uses: actions/upload-artifact@v7
        with:
          name: benchmark-results-combined
          path: |
            BENCHMARK_SUMMARY.md
            all-results/
          retention-days: 90

      - name: Comment PR with benchmark results
        if: github.event_name == 'pull_request'
        uses: actions/github-script@v9
        with:
          github-token: ${{ secrets.GITHUB_TOKEN }}
          script: |
            const fs = require('fs');
            const summary = fs.readFileSync('BENCHMARK_SUMMARY.md', 'utf8');

            const comment = `## Benchmark Results\n\n${summary}\n\n<details>\n<summary>View detailed results</summary>\n\nDetailed Criterion results have been uploaded as artifacts. Download them from the workflow run to view charts and detailed statistics.\n\n</details>`;

            github.rest.issues.createComment({
              issue_number: context.issue.number,
              owner: context.repo.owner,
              repo: context.repo.repo,
              body: comment
            });
        continue-on-error: true