gem5-perf-template.yml

name: gem5 Performance Test Template (Unified)

on:
  workflow_call:
    inputs:
      config_path:
        required: true
        type: string
        description: "gem5 config file path (relative to repo root), e.g. configs/example/idealkmhv3.py"
      extra_args:
        required: false
        type: string
        default: ""
        description: "Extra arguments for gem5 (e.g., --disable-mgsc). Only works with .py config files."
      benchmark_type:
        required: true
        type: string
        description: "Benchmark type: spec06-0.3c, spec06-0.8c, spec06-1.0c, spec17-1.0c, spec06-rvv-1.0c or spec06int-rvv-0.8c"
      vector_type:
        required: false
        type: string
        description: "Benchmark type: base, simple"
      check_result:
        required: false
        type: boolean
        description: "Whether to check result (failed ckpts)"
        default: true
      pr_sha:
        required: false
        type: string
        description: "Specific commit SHA to checkout (takes precedence)"
jobs:
  run_performance_test:
    runs-on: [self-hosted, node]
    continue-on-error: false
    name: XS-GEM5 - Run performance test (${{ inputs.benchmark_type }})
    steps:
      - uses: actions/checkout@v2
        with:
          ref: ${{ inputs.pr_sha || github.sha }}
      
      - name: Set benchmark configuration
        id: config
        run: |
          case "${{ inputs.benchmark_type }}" in
              "spec06-0.3c")
              echo "checkpoint_list=/nfs/home/share/gem5_ci/spec06_cpts/spec06_0.3c_int.lst" >> $GITHUB_OUTPUT
              echo "checkpoint_root_node=/nfs/home/share/jiaxiaoyu/simpoint_checkpoint_zstd_format/spec06_rv64gcb_O3_20m_gcc12.2.0-intFpcOff-jeMalloc" >> $GITHUB_OUTPUT
              echo "score_script=gem5-score-ci.sh" >> $GITHUB_OUTPUT
              echo "cluster_config=/nfs/home/share/gem5_ci/spec06_cpts/cluster-0-0.json" >> $GITHUB_OUTPUT
              echo "artifact_name=performance-score-spec06-0.3c" >> $GITHUB_OUTPUT
              echo "comment=run 30% coverage spec06 checkpoints, 148 checkpoints" >> $GITHUB_OUTPUT
              ;;
            "spec06-0.8c")
              echo "checkpoint_list=/nfs/home/share/gem5_ci/spec06_cpts/spec_0.8c_int.lst" >> $GITHUB_OUTPUT
              echo "checkpoint_root_node=/nfs/home/share/jiaxiaoyu/simpoint_checkpoint_zstd_format/spec06_rv64gcb_O3_20m_gcc12.2.0-intFpcOff-jeMalloc" >> $GITHUB_OUTPUT
              echo "score_script=gem5-score-ci.sh" >> $GITHUB_OUTPUT
              echo "cluster_config=/nfs/home/share/gem5_ci/spec06_cpts/cluster-0-0.json" >> $GITHUB_OUTPUT
              echo "artifact_name=performance-score-spec06-0.8c" >> $GITHUB_OUTPUT
              echo "comment=run 80% coverage spec06 checkpoints, 500+ checkpoints" >> $GITHUB_OUTPUT
              ;;
            "spec06-1.0c")
              echo "checkpoint_list=/nfs/home/share/gem5_ci/spec06_cpts/checkpoint-0-0-0.lst" >> $GITHUB_OUTPUT
              echo "checkpoint_root_node=/nfs/home/share/jiaxiaoyu/simpoint_checkpoint_zstd_format/spec06_rv64gcb_O3_20m_gcc12.2.0-intFpcOff-jeMalloc" >> $GITHUB_OUTPUT
              echo "score_script=gem5-score-ci.sh" >> $GITHUB_OUTPUT
              echo "cluster_config=/nfs/home/share/gem5_ci/spec06_cpts/cluster-0-0.json" >> $GITHUB_OUTPUT
              echo "artifact_name=performance-score-spec06-1.0c" >> $GITHUB_OUTPUT
              echo "comment=run 100% coverage spec06 checkpoints" >> $GITHUB_OUTPUT
              ;;
            "spec17-1.0c")
              echo "checkpoint_list=/nfs/home/yanyue/spec17_cpts/checkpoint-0-0-0/checkpoint.lst" >> $GITHUB_OUTPUT
              echo "checkpoint_root_node=/nfs/home/yanyue/spec17_cpts/checkpoint-0-0-0/" >> $GITHUB_OUTPUT
              echo "score_script=gem5-score-ci-17.sh" >> $GITHUB_OUTPUT
              echo "cluster_config=/nfs/home/yanyue/spec17_cpts/checkpoint-0-0-0/cluster-0-0.json" >> $GITHUB_OUTPUT
              echo "artifact_name=performance-score-spec17-1.0c" >> $GITHUB_OUTPUT
              echo "comment=run 100% coverage spec17 checkpoints" >> $GITHUB_OUTPUT
              ;;
            "spec06-rvv-1.0c")
              echo "checkpoint_list=/nfs/home/xutongqiao/GEM5-CI/spec06_gcc15_rv64gcbv_O3_lto_base_nemu_single_core_NEMU_archgroup_2024-10-12-16-05/checkpoint06_rv64gcbv/checkpoint.lst" >> $GITHUB_OUTPUT
              echo "checkpoint_root_node=/nfs/home/xutongqiao/GEM5-CI/spec06_gcc15_rv64gcbv_O3_lto_base_nemu_single_core_NEMU_archgroup_2024-10-12-16-05/checkpoint06_rv64gcbv/" >> $GITHUB_OUTPUT
              echo "score_script=gem5-score-ci.sh" >> $GITHUB_OUTPUT
              echo "cluster_config=/nfs/home/xutongqiao/GEM5-CI/spec06_gcc15_rv64gcbv_O3_lto_base_nemu_single_core_NEMU_archgroup_2024-10-12-16-05/checkpoint06_rv64gcbv/cluster-0-0.json" >> $GITHUB_OUTPUT
              echo "artifact_name=performance-score-spec06-1.0c-with-rvv-extension" >> $GITHUB_OUTPUT
              echo "comment=run 100% coverage spec06 rvv checkpoints" >> $GITHUB_OUTPUT
              ;;
            "spec06int-rvv-0.8c")
              echo "checkpoint_list=/nfs/home/xutongqiao/GEM5-CI/spec06_gcc15_rv64gcbv_O3_lto_base_nemu_single_core_NEMU_archgroup_2024-10-12-16-05/checkpoint06_rv64gcbv/checkpoint_0.8c_int.lst" >> $GITHUB_OUTPUT
              echo "checkpoint_root_node=/nfs/home/xutongqiao/GEM5-CI/spec06_gcc15_rv64gcbv_O3_lto_base_nemu_single_core_NEMU_archgroup_2024-10-12-16-05/checkpoint06_rv64gcbv/" >> $GITHUB_OUTPUT
              echo "score_script=gem5-score-ci.sh" >> $GITHUB_OUTPUT
              echo "cluster_config=/nfs/home/xutongqiao/GEM5-CI/spec06_gcc15_rv64gcbv_O3_lto_base_nemu_single_core_NEMU_archgroup_2024-10-12-16-05/checkpoint06_rv64gcbv/cluster_0.8c_int.json" >> $GITHUB_OUTPUT
              echo "artifact_name=performance-score-spec06int-0.8c-with-rvv-extension" >> $GITHUB_OUTPUT
              echo "comment=run 80% coverage spec06 int rvv checkpoints" >> $GITHUB_OUTPUT
              ;;
            *)
              echo "Error: Invalid benchmark_type '${{ inputs.benchmark_type }}'. Must be one of: spec06-0.8c, spec06-1.0c, spec17-1.0c, spec06-rvv-1.0c or spec06int-rvv-0.8c"
              exit 1
              ;;
          esac
      - uses: ./.github/actions/build-dramsim
      - name: Build GEM5 fast
        run: |
          export GEM5_HOME=$(pwd)
          export GCBV_REF_SO="/nfs/home/share/gem5_ci/ref/normal/riscv64-nemu-notama-tvalref-so"
          case "${{ inputs.vector_type }}" in
            "base")
              # use pgo profile to build gem5
              bash util/pgo/basic_pgo_new.sh
              ;;
            "simple")
              # use pgo profile to build gem5, use simple vector decode strategy in scons
              bash util/pgo/basic_pgo_new_vector.sh
              ;;
            "")
              # use pgo profile to build gem5
              bash util/pgo/basic_pgo_new.sh
              ;;
            *)
              echo "Error: Invalid vector_type '${{ inputs.vector_type }}'. Must be one of: base, simple"
              exit 1
              ;;
          esac

      - name: XS-GEM5 - Run performance test
        # ${{ steps.config.outputs.comment }}
        env:
          GCBV_REF_SO: "/nfs/home/share/gem5_ci/ref/normal/riscv64-nemu-notama-tvalref-so"
          GCB_RESTORER: ""
          GEM5_HOME: ${{ github.workspace }}
          GEM5_BUILD_TYPE: fast
        run: |
          mkdir -p $GEM5_HOME/util/xs_scripts/test
          cd $GEM5_HOME/util/xs_scripts/test

          CONFIG_PATH="${{ inputs.config_path }}"
          if [[ "$CONFIG_PATH" == /* ]]; then
            echo "Error: config_path must be repo-root-relative, got absolute path: '$CONFIG_PATH'"
            exit 1
          fi
          if [[ "$CONFIG_PATH" != *.py ]]; then
            echo "Error: config_path must be a .py gem5 config file path, got: '$CONFIG_PATH'"
            exit 1
          fi

          CONFIG_PATH="$GEM5_HOME/$CONFIG_PATH"

          bash ../parallel_sim.sh "$(realpath "$CONFIG_PATH")" \
            ${{ steps.config.outputs.checkpoint_list }} \
            ${{ steps.config.outputs.checkpoint_root_node}} \
            spec_all \
            "${{ inputs.extra_args }}"
      - name: Setup gem5_data_proc environment
        run: |
          # 使用本地数据处理仓库，避免GitHub网络问题
          cd $GITHUB_WORKSPACE
          cp -r /nfs/home/share/gem5_ci/gem5_data_proc .
      - name: Check performance test result
        run: |
          # 添加yanyue的 Python 包路径, 包含pandas等
          export PYTHONPATH=/nfs/home/yanyue/.local/lib/python3.10/site-packages:$PYTHONPATH
          # 每次run 会进入新的shell, 需要重新设置GEM5_HOME
          export GEM5_HOME=$GITHUB_WORKSPACE
          cd $GITHUB_WORKSPACE/gem5_data_proc
          # 使用已有的数据spec_all生成测试报告
          bash example-scripts/${{ steps.config.outputs.score_script }} \
            $GEM5_HOME/util/xs_scripts/test/spec_all \
            ${{ steps.config.outputs.cluster_config }} \
            > $GITHUB_WORKSPACE/score.txt
          # 提取最后42行score信息
          echo "### performance test result :rocket:" >> $GITHUB_STEP_SUMMARY
          echo '```' >> $GITHUB_STEP_SUMMARY
          tail -n 42 $GITHUB_WORKSPACE/score.txt >> $GITHUB_STEP_SUMMARY
          echo '```' >> $GITHUB_STEP_SUMMARY
          # 提取关键指标
          FINAL_SCORE=$(grep "Estimated Int score per GHz:" $GITHUB_WORKSPACE/score.txt | awk '{print $NF}')
          echo "### Key indicators" >> $GITHUB_STEP_SUMMARY
          echo "- Final Int score per GHz: **${FINAL_SCORE}**" >> $GITHUB_STEP_SUMMARY
          
          # 最后检查是否存在abort文件， 如果存在，打出前10个错误名字
          if find $GEM5_HOME/util/xs_scripts/test/spec_all -type f -name "abort" | grep -q .; then
            echo "### :x: Test Failures Detected!" >> $GITHUB_STEP_SUMMARY
            echo "Failed test count: $(find $GEM5_HOME/util/xs_scripts/test/spec_all -type f -name "abort" | wc -l)" >> $GITHUB_STEP_SUMMARY
            echo "First 10 failed tests:" >> $GITHUB_STEP_SUMMARY
            find $GEM5_HOME/util/xs_scripts/test/spec_all -type f -name "abort" | 
              sed 's|.*/\([^/]*\)/abort|\1|' |
              head -n 10 |
              while read -r line; do
                echo "- $line" >> $GITHUB_STEP_SUMMARY
              done
              if [ "${{ inputs.check_result }}" == "true" ]; then
                exit 1
              fi
          fi
      - name: Archive performance data
        if: always()
        run: |
          # Create archive directory with timestamp and commit
          ARCHIVE_ROOT="/nfs/home/share/gem5_ci/performance_data"
          BENCHMARK_DIR="${ARCHIVE_ROOT}/${{ inputs.benchmark_type }}"
          TIMESTAMP=$(date +"%Y%m%d_%H%M%S")
          COMMIT_SHORT=$(git rev-parse --short HEAD)
          RUN_NUMBER="${{ github.run_number }}"
          TARGET_DIR="${BENCHMARK_DIR}/${TIMESTAMP}_${COMMIT_SHORT}_run${RUN_NUMBER}"
          
          # Create target directory
          mkdir -p "$TARGET_DIR"
          
          # Archive entire spec_all directory with compression
          echo "Archiving performance data to $TARGET_DIR"
          tar -czf "$TARGET_DIR/spec_all.tar.gz" -C "$GITHUB_WORKSPACE/util/xs_scripts/test" spec_all
          
          # Also save score.txt for quick reference
          if [ -f "$GITHUB_WORKSPACE/score.txt" ]; then
            cp "$GITHUB_WORKSPACE/score.txt" "$TARGET_DIR/"
          fi
          
          # Save metadata
          echo "timestamp: $TIMESTAMP" > "$TARGET_DIR/metadata.txt"
          echo "commit: $(git rev-parse HEAD)" >> "$TARGET_DIR/metadata.txt"
          echo "commit_short: $COMMIT_SHORT" >> "$TARGET_DIR/metadata.txt"
          echo "branch: ${GITHUB_REF#refs/heads/}" >> "$TARGET_DIR/metadata.txt"
          echo "run_number: $RUN_NUMBER" >> "$TARGET_DIR/metadata.txt"
          echo "benchmark_type: ${{ inputs.benchmark_type }}" >> "$TARGET_DIR/metadata.txt"
          echo "workflow_run_id: ${{ github.run_id }}" >> "$TARGET_DIR/metadata.txt"
          
          # Auto cleanup: keep only last 200 runs for this benchmark type
          if [ -d "$BENCHMARK_DIR" ] && [ $(ls -1 "$BENCHMARK_DIR" 2>/dev/null | wc -l) -gt 200 ]; then
            echo "Cleaning up old performance data, keeping last 200 runs"
            # Use absolute paths to avoid accidental deletion if cd fails
            ls -1t "$BENCHMARK_DIR" | tail -n +201 | while read -r old_run; do
              if [ -n "$old_run" ] && [ -d "$BENCHMARK_DIR/$old_run" ]; then
                rm -rf "$BENCHMARK_DIR/$old_run"
                echo "Deleted: $BENCHMARK_DIR/$old_run"
              fi
            done
            echo "Cleanup completed"
          fi
          
          echo "Performance data archived successfully"
          echo "Archive size: $(du -sh "$TARGET_DIR/spec_all.tar.gz" | cut -f1)"
          {
            echo "### Archive output"
            echo "- Path: \`$TARGET_DIR\`"
            echo "- Size: $(du -sh "$TARGET_DIR/spec_all.tar.gz" | cut -f1)"
          } >> $GITHUB_STEP_SUMMARY
      - name: Upload score artifact
        if: always()
        continue-on-error: true
        uses: actions/upload-artifact@v4
        with:
          name: ${{ steps.config.outputs.artifact_name }}
          path: score.txt
          retention-days: 30
      
      - name: Fallback - Save score to shared storage
        if: always()
        run: |
          # Fallback: 如果 artifact upload 失败，保存到共享存储
          ARTIFACT_DIR="/nfs/home/share/gem5_ci/artifacts/${{ github.run_id }}"
          mkdir -p "$ARTIFACT_DIR"
          cp score.txt "$ARTIFACT_DIR/${{ steps.config.outputs.artifact_name }}.txt" 2>/dev/null || true
          echo "✅ Score also saved to: $ARTIFACT_DIR"