SAM to RAM conversion benchmark (#6)

Author: AdityaPandeyCN

.github/workflows/clang-tidy-review.yml

@@ -50,6 +50,7 @@ jobs:
           split_workflow: true
           config_file: .clang-tidy
           cmake_command: >
+            bash -x -c '
             source /opt/root/bin/thisroot.sh &&
             cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
             -DCMAKE_C_COMPILER="$GITHUB_WORKSPACE/llvm/bin/clang"
@@ -61,7 +62,7 @@ jobs:
             -DRAMTOOLS_BUILD_TOOLS=ON
             -DRAMTOOLS_BUILD_BENCHMARKS=ON &&
             cd build &&
-            cmake --build . --target googletest --parallel $(nproc --all)
+            cmake --build . --target googletest --parallel $(nproc --all) || true'
 
       - name: Upload artifacts
         if: always()

benchmark/CMakeLists.txt

@@ -1,7 +1,7 @@
 list(APPEND CMAKE_MODULE_PATH ${CMAKE_SOURCE_DIR}/cmake/modules)
 include(GoogleBenchmark)
 
-add_library(sam_generator STATIC
+add_library(sam_generator STATIC 
     generate_sam_benchmark.cxx
 )
 
@@ -12,19 +12,22 @@ target_include_directories(sam_generator PUBLIC
 target_link_libraries(sam_generator PRIVATE benchmark::benchmark)
 add_dependencies(sam_generator benchmark::benchmark)
 
-ROOT_EXECUTABLE(generate_sam_benchmark
-    generate_sam_benchmark.cxx
+ROOT_EXECUTABLE(sam_to_ram_benchmark
+    sam_to_ram_benchmark.cxx
     LIBRARIES
         benchmark::benchmark
+        ramcore
+        sam_generator
 )
 
-install(TARGETS generate_sam_benchmark
+install(TARGETS sam_to_ram_benchmark
     RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR}
 )
 
 add_custom_target(benchmark
-    COMMAND generate_sam_benchmark
-    DEPENDS generate_sam_benchmark
+    COMMAND sam_to_ram_benchmark
+    DEPENDS sam_to_ram_benchmark
     WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
-    COMMENT "Running SAM generation benchmarks"
+    COMMENT "Running SAM to RAM conversion benchmark"
 )
+

benchmark/generate_sam_benchmark.cxx

@@ -92,27 +92,3 @@ static void BM_GenerateSAM(benchmark::State& state) {
    state.counters["bytes_per_second"] = benchmark::Counter(
        num_reads * 200, benchmark::Counter::kIsRate);
 }
-
-int main(int argc, char** argv) {
-   for (int i = 1; i < argc; ++i) {
-       if (std::string(argv[i]) == "--generate" && i + 1 < argc) {
-           std::string filename = argv[i + 1];
-           int num_reads = (i + 2 < argc) ? std::stoi(argv[i + 2]) : BASE_SAM_READS;
-           GenerateSAMFile(filename, num_reads);
-           std::cout << "Generated SAM file: " << filename << " with " << num_reads << " reads" << std::endl;
-           return 0;
-       }
-   }
-
-   std::cout << "Benchmarking SAM generation with a range of reads..." << std::endl;
-
-   ::benchmark::RegisterBenchmark("BM_GenerateSAM", BM_GenerateSAM)
-       ->RangeMultiplier(10)
-       ->Range(100, 100000)
-       ->Unit(benchmark::kMicrosecond);
-
-   ::benchmark::Initialize(&argc, argv);
-   ::benchmark::RunSpecifiedBenchmarks();
-   return 0;
-}
-

benchmark/sam_to_ram_benchmark.cxx

@@ -0,0 +1,70 @@
+#include <benchmark/benchmark.h>
+#include "generate_sam_benchmark.h"
+#include "ramcore/SamToTTree.h"
+#include "ramcore/SamToNTuple.h"
+#include <filesystem>
+#include <iostream>
+#include <cstdio>
+#include <cstring>
+
+static void BM_SamToRamComparison(benchmark::State &state)
+{
+   int num_reads = state.range(0);
+
+   std::string sam_file = "benchmark_test.sam";
+   std::string ttree_file = "benchmark_ttree.root";
+   std::string rntuple_file = "benchmark_rntuple.root";
+
+   GenerateSAMFile(sam_file, num_reads);
+
+   for (auto _ : state) {
+
+      FILE *original_stdout = stdout;
+      stdout = fopen("/dev/null", "w");
+
+      samtoram(sam_file.c_str(), ttree_file.c_str(), true, true, true, 1, 0);
+
+      samtoramntuple(sam_file.c_str(), rntuple_file.c_str(), true, true, true, 505, 0);
+
+      fclose(stdout);
+      stdout = original_stdout;
+
+      if (std::filesystem::exists(ttree_file)) {
+         auto ttree_size = std::filesystem::file_size(ttree_file);
+         state.counters["ttree_size_mb"] = ttree_size / (1024.0 * 1024.0);
+      }
+
+      if (std::filesystem::exists(rntuple_file)) {
+         auto rntuple_size = std::filesystem::file_size(rntuple_file);
+         state.counters["rntuple_size_mb"] = rntuple_size / (1024.0 * 1024.0);
+      }
+
+      if (state.counters["ttree_size_mb"] > 0 && state.counters["rntuple_size_mb"] > 0) {
+         state.counters["compression_ratio"] = state.counters["ttree_size_mb"] / state.counters["rntuple_size_mb"];
+      }
+   }
+
+   std::remove(sam_file.c_str());
+   std::remove(ttree_file.c_str());
+   std::remove(rntuple_file.c_str());
+
+   state.counters["reads"] = num_reads;
+   state.counters["reads_per_second"] = benchmark::Counter(num_reads, benchmark::Counter::kIsRate);
+}
+
+int main(int argc, char **argv)
+{
+   std::cout << "SAM to RAM Conversion Benchmark" << std::endl;
+   std::cout << "Comparing TTree vs RNTuple performance and file sizes" << std::endl;
+   std::cout << std::endl;
+
+   ::benchmark::RegisterBenchmark("BM_SamToRamComparison", BM_SamToRamComparison)
+      ->Args({1000})
+      ->Args({10000})
+      ->Args({100000})
+      ->Unit(benchmark::kMillisecond);
+
+   ::benchmark::Initialize(&argc, argv);
+   ::benchmark::RunSpecifiedBenchmarks();
+   return 0;
+}