Add cpu benchmark

Author: OliverRietmann

.github/workflows/standalone-benchmark.yml

@@ -14,36 +14,46 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        name: [nvidia-h100, nvidia-l40s, amd-mi300x, amd-w7900]
+        name: [cpu, nvidia-h100, nvidia-l40s, amd-mi300x, amd-w7900]
         include:
+          - name: cpu
+            runner: cern-nextgen-h100
+            cmake_args: -DENABLE_CUDA=0 -DENABLE_HIP=0
+            profiler_runs: 42
+            standalone_runs: 42
+            cpu_gpu: "-c"
           - name: nvidia-h100
             runner: cern-nextgen-h100
             cmake_args: -DENABLE_CUDA=1 -DENABLE_HIP=0 -DCUDA_COMPUTETARGET=90
             profiler_runs: 21
             standalone_runs: 42
+            cpu_gpu: "-g --memSize 20000000000"
           - name: nvidia-l40s
             runner: cern-nextgen-l40s
             cmake_args: -DENABLE_CUDA=1 -DENABLE_HIP=0 -DCUDA_COMPUTETARGET=89
             profiler_runs: 42
             standalone_runs: 42
+            cpu_gpu: "-g --memSize 20000000000"
           - name: amd-mi300x
             runner: cern-nextgen-mi300x
             cmake_args: -DENABLE_CUDA=0 -DENABLE_HIP=1 -DHIP_AMDGPUTARGET=gfx942
             profiler_runs: 42
             standalone_runs: 42
+            cpu_gpu: "-g --memSize 20000000000"
           - name: amd-w7900
             runner: cern-nextgen-w7900
             cmake_args: -DENABLE_CUDA=0 -DENABLE_HIP=1 -DHIP_AMDGPUTARGET=gfx1100
             profiler_runs: 42
             standalone_runs: 42
+            cpu_gpu: "-g --memSize 20000000000"
     env:
       WORK_DIR: /cvmfs/alice.cern.ch
       ALIBUILD_ARCH_PREFIX: el9-x86_64/Packages
       MODULEPATH: /cvmfs/alice.cern.ch/etc/toolchain/modulefiles/el9-x86_64:/cvmfs/alice.cern.ch/el9-x86_64/Modules/modulefiles
       STANDALONE_DIR: /root/standalone
       BENCHMARK_CSV: ${{ matrix.name }}.csv
       PROFILER_CSV: results_${{ matrix.name }}.csv
-      TIMING_CA: ./ca -e 50kHz -g --seed 0 --memSize 15000000000 --sync --debug 1 # Add --PROCdebugMarkdown 1 --runs 42 --runsInit 2 --PROCresetTimers 1 for benchmark runs
+      TIMING_CA: ./ca -e 50kHz ${{ matrix.cpu_gpu }} --seed 0 --sync --debug 1 # Add --PROCdebugMarkdown 1 --runs 42 --runsInit 2 --PROCresetTimers 1 for benchmark runs
       LD_LIBRARY_PATH: /usr/local/cuda-13.0/compat
 
     name: ${{ matrix.name }}
@@ -75,21 +85,21 @@ jobs:
         env:
           DETERMINISTIC_MODE: GPU
 
-      - name: Test GPU Track Reconstruction
+      - name: Test Track Reconstruction
         run: |
           source /etc/profile.d/modules.sh
           module load ninja/fortran-v1.11.1.g9-15 Vc/1.4.5-10 boost/v1.83.0-alice2-57 fmt/11.1.2-14 CMake/v3.31.6-10 ms_gsl/4.2.1-3 Clang/v20.1.7-9 TBB/v2022.3.0-3 ROOT/v6-36-04-alice9-15 ONNXRuntime/v1.22.0-71 GLFW/3.3.2-25
           cd ${STANDALONE_DIR}
-          ${STANDALONE_DIR}/ca -e o2-simple -g --seed 0 --memSize 20000000000 --sync --runs 1 --RTCenable --PROCdeterministicGPUReconstruction 1 --RTCoptConstexpr 1 --RTCoptSpecialCode 1 --debug 6
-          cmp ${STANDALONE_DIR}/GPU.out ${STANDALONE_DIR}/o2-simple-GPU.out
-          rm -rf ${STANDALONE_DIR}/GPU.out ${STANDALONE_DIR}/o2-simple-GPU.out ${STANDALONE_DIR}/events/o2-simple ${STANDALONE_DIR}/build
+          ${STANDALONE_DIR}/ca -e o2-simple ${{ matrix.cpu_gpu }} --seed 0 --sync --runs 1 --RTCenable --PROCdeterministicGPUReconstruction 1 --RTCoptConstexpr 1 --RTCoptSpecialCode 1 --debug 6
+          cmp ${STANDALONE_DIR}/*.out
+          rm -rf ${STANDALONE_DIR}/*.out ${STANDALONE_DIR}/events/o2-simple ${STANDALONE_DIR}/build
 
       - name: Build Non-Deterministic
         run: *build
         env:
           DETERMINISTIC_MODE: OFF
 
-      - name: Benchmark GPU Track Reconstruction
+      - name: Benchmark Track Reconstruction
         run: |
           source /etc/profile.d/modules.sh
           module load ninja/fortran-v1.11.1.g9-15 Vc/1.4.5-10 boost/v1.83.0-alice2-57 fmt/11.1.2-14 CMake/v3.31.6-10 ms_gsl/4.2.1-3 Clang/v20.1.7-9 TBB/v2022.3.0-3 ROOT/v6-36-04-alice9-15 ONNXRuntime/v1.22.0-71 GLFW/3.3.2-25
@@ -149,3 +159,4 @@ jobs:
           echo -e "\n\n" >> ${GITHUB_STEP_SUMMARY}
           python3 ${GITHUB_WORKSPACE}/.github/scripts/csv_to_md.py --runs ${{ matrix.standalone_runs }} --baseline ${STANDALONE_DIR}/baseline/${BENCHMARK_CSV} --current /root/${BENCHMARK_CSV} >> ${GITHUB_STEP_SUMMARY}
           rm -rf ${STANDALONE_DIR}/baseline
+        if: ${{ matrix.name != 'cpu' }}