Merge pull request #384 from eseiler/feature/clang

[FEATURE] Clang support

.github/workflows/ci_linux.yml

@@ -31,23 +31,29 @@ jobs:
       fail-fast: false
       matrix:
         include:
+          - name: "clang17"
+            compiler: "clang-17"
+            build: unit
+            build_type: Release
+            cxx_flags: "-std=c++20"
+
           - name: "gcc13"
             compiler: "gcc-13"
             build: unit
             build_type: Release
-            cxx_flags: "-std=c++23"
+            cxx_flags: "-std=c++23 -Wno-interference-size"
 
           - name: "gcc12"
             compiler: "gcc-12"
             build: unit
             build_type: Release
-            cxx_flags: "-std=c++23"
+            cxx_flags: "-std=c++23 -Wno-interference-size"
 
           - name: "gcc11"
             compiler: "gcc-11"
             build: unit
             build_type: Release
-            cxx_flags: "-std=c++23"
+            cxx_flags: "-std=c++23 -Wno-interference-size"
 
     steps:
       - name: Checkout
@@ -63,6 +69,10 @@ jobs:
           compiler: ${{ matrix.compiler }}
           ccache_size: 150M
 
+      - name: Install OpenMP
+        if: contains(matrix.name, 'clang')
+        run: install libomp-17-dev
+
       - name: Install CMake
         uses: seqan/actions/setup-cmake@main
         with:
@@ -73,7 +83,7 @@ jobs:
           mkdir build
           cd build
           cmake ../raptor/test/${{ matrix.build }} -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} \
-                                                   -DCMAKE_CXX_FLAGS="${{ matrix.cxx_flags }} -Wno-interference-size" \
+                                                   -DCMAKE_CXX_FLAGS="${{ matrix.cxx_flags }}" \
                                                    -DRAPTOR_NATIVE_BUILD=OFF
           make -j2 gtest_build
 

.github/workflows/ci_sanitizer.yml

@@ -13,6 +13,7 @@ env:
   SHARG_NO_VERSION_CHECK: 1
   TZ: Europe/Berlin
   ASAN_OPTIONS: strict_string_checks=1:detect_stack_use_after_return=1:check_initialization_order=1:strict_init_order=1
+  TSAN_OPTIONS: ignore_noninstrumented_modules=1
   UBSAN_OPTIONS: print_stacktrace=1
 
 defaults:
@@ -32,13 +33,16 @@ jobs:
         build_type: [Release, Debug]
         include:
           - name: "AddressSanitizer"
-            cxx_flags: "-std=c++23 -fno-omit-frame-pointer -fsanitize=address -Wno-maybe-uninitialized"
+            compiler: gcc-13
+            cxx_flags: "-std=c++23 -fno-omit-frame-pointer -fsanitize=address -Wno-maybe-uninitialized -Wno-interference-size"
 
           - name: "ThreadSanitizer"
-            cxx_flags: "-std=c++23 -fno-omit-frame-pointer -fsanitize=thread"
+            compiler: clang-17
+            cxx_flags: "-std=c++20 -fno-omit-frame-pointer -fsanitize=thread"
 
           - name: "UndefinedBehaviorSanitizer"
-            cxx_flags: "-std=c++23 -fno-omit-frame-pointer -fsanitize=undefined"
+            compiler: gcc-13
+            cxx_flags: "-std=c++23 -fno-omit-frame-pointer -fsanitize=undefined -Wno-interference-size"
 
     steps:
       - name: Checkout
@@ -53,6 +57,12 @@ jobs:
         with:
           compiler: gcc-13
 
+      - name: Install OpenMP
+        if: contains(matrix.name, 'clang')
+        run: |
+          install libomp-17-dev
+          echo "OMP_TOOL_LIBRARIES=/usr/lib/llvm-17/lib/libarcher.so" >> "$GITHUB_ENV"
+
       - name: Configure OS
         run: |
           sudo bash -c "echo 0 > /proc/sys/vm/overcommit_memory"
@@ -68,7 +78,7 @@ jobs:
           mkdir build
           cd build
           cmake ../raptor/test/unit -DCMAKE_BUILD_TYPE=${{ matrix.build_type }} \
-                                    -DCMAKE_CXX_FLAGS="${{ matrix.cxx_flags }} -Wno-interference-size"
+                                    -DCMAKE_CXX_FLAGS="${{ matrix.cxx_flags }}"
           make -j2 gtest_build
 
       - name: Build tests

build_system/raptor-config.cmake

@@ -247,7 +247,8 @@ endif ()
 
 check_cxx_compiler_flag ("-fopenmp-simd" RAPTOR_HAS_OPENMP_SIMD)
 if (RAPTOR_HAS_OPENMP_SIMD)
-    set (RAPTOR_CXX_FLAGS "${RAPTOR_CXX_FLAGS} -fopenmp-simd -DSIMDE_ENABLE_OPENMP")
+    set (RAPTOR_CXX_FLAGS "${RAPTOR_CXX_FLAGS} -fopenmp-simd")
+    set (RAPTOR_DEFINITIONS ${RAPTOR_DEFINITIONS} "-DSIMDE_ENABLE_OPENMP")
     raptor_config_print ("SIMD-OpenMP Support:        via -fopenmp-simd")
 else ()
     raptor_config_print ("SIMD-OpenMP Support:        not found")
@@ -290,10 +291,12 @@ endif ()
 if (RAPTOR_IS_DEBUG)
     raptor_config_print ("Link Time Optimization:     disabled")
 else ()
-    if ("${CMAKE_SYSTEM_NAME}" MATCHES "Darwin")
-        set (RAPTOR_LTO_FLAGS "-flto=auto")
+    # CMake's check_ipo_supported uses hardcoded lto flags
+    # macOS GCC supports -flto-auto, but not the hardcoded flag "-fno-fat-lto-objects"
+    if ("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU" AND NOT "${CMAKE_SYSTEM_NAME}" MATCHES "Darwin")
+        set (RAPTOR_LTO_FLAGS "-flto=auto -ffat-lto-objects")
     else ()
-        set (RAPTOR_LTO_FLAGS "-flto=auto -fno-fat-lto-objects")
+        set (RAPTOR_LTO_FLAGS "-flto=auto")
     endif ()
 
     set (LTO_CMAKE_SOURCE
@@ -326,7 +329,7 @@ else ()
 endif ()
 
 option (RAPTOR_STRIP_BINARY "Enable binary-stripping. Not supported on macOS." ON)
-if ("${CMAKE_SYSTEM_NAME}" MATCHES "Darwin")
+if ("${CMAKE_SYSTEM_NAME}" MATCHES "Darwin" OR "${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
     raptor_config_print ("Binary-stripping:           not supported")
 elseif (RAPTOR_IS_DEBUG OR NOT RAPTOR_STRIP_BINARY)
     raptor_config_print ("Binary-stripping:           disabled")

include/raptor/argument_parsing/search_arguments.hpp

@@ -81,15 +81,15 @@ struct search_arguments
 
     raptor::threshold::threshold_parameters make_threshold_parameters() const noexcept
     {
-        return {.window_size{window_size},
-                .shape{shape},
-                .query_length{query_length},
-                .errors{errors},
-                .percentage{threshold},
-                .p_max{p_max},
-                .tau{tau},
-                .cache_thresholds{cache_thresholds},
-                .output_directory{index_file.parent_path()}};
+        return {.window_size = window_size,
+                .shape = shape,
+                .query_length = query_length,
+                .errors = errors,
+                .percentage = threshold,
+                .p_max = p_max,
+                .tau = tau,
+                .cache_thresholds = cache_thresholds,
+                .output_directory = index_file.parent_path()};
     }
 };
 

include/raptor/build/index_factory.hpp

@@ -31,8 +31,8 @@ class index_factory
     index_factory() = default;
     index_factory(index_factory const &) = default;
     index_factory(index_factory &&) = default;
-    index_factory & operator=(index_factory const &) = default;
-    index_factory & operator=(index_factory &&) = default;
+    index_factory & operator=(index_factory const &) = delete; // const member
+    index_factory & operator=(index_factory &&) = delete;      // const member
     ~index_factory() = default;
 
     explicit index_factory(build_arguments const & args) : arguments{std::addressof(args)}
@@ -76,11 +76,14 @@ class index_factory
             seqan::hibf::serial_timer local_timer{};
             auto & ibf = index.ibf();
             local_timer.start();
-            for (auto && [file_names, bin_number] : zipped_view)
+            // https://godbolt.org/z/PeKnxzjn1
+            for (auto && zipped : zipped_view)
             {
                 std::visit(
                     [&](auto const & reader)
                     {
+                        auto && [file_names, bin_number] = zipped;
+
                         if (config == nullptr)
                             reader.hash_into(file_names, emplacer(ibf, seqan::hibf::bin_index{bin_number}));
                         else

include/raptor/search/search_singular_ibf.hpp

@@ -51,7 +51,11 @@ void search_singular_ibf(search_arguments const & arguments, index_t && index)
         seqan::hibf::serial_timer local_query_ibf_timer{};
         seqan::hibf::serial_timer local_generate_results_timer{};
 
+#if defined(__clang__)
+        auto counter = [&index]()
+#else
         auto counter = [&index, is_ibf]()
+#endif
         {
             if constexpr (is_ibf)
                 return index.ibf().template counting_agent<uint16_t>();

include/raptor/search/sync_out.hpp

@@ -26,10 +26,10 @@ class sync_out
 {
 public:
     sync_out() = default;
-    sync_out(sync_out const &) = default;
-    sync_out & operator=(sync_out const &) = default;
-    sync_out(sync_out &&) = default;
-    sync_out & operator=(sync_out &&) = default;
+    sync_out(sync_out const &) = delete;             // std::ofstream
+    sync_out & operator=(sync_out const &) = delete; // std::ofstream
+    sync_out(sync_out &&) = delete;                  // std::mutex
+    sync_out & operator=(sync_out &&) = delete;      // std::mutex
     ~sync_out() = default;
 
     sync_out(search_arguments const & arguments) : file{arguments.out_file}

src/layout/CMakeLists.txt

@@ -2,6 +2,6 @@ cmake_minimum_required (VERSION 3.15)
 
 if (NOT TARGET raptor_layout)
     add_library ("raptor_layout" STATIC raptor_layout.cpp)
-
+    target_compile_options (chopper_layout_lib PRIVATE "-w")
     target_link_libraries ("raptor_layout" PUBLIC "raptor_interface" "chopper_layout_lib" xxHash::xxhash)
 endif ()

test/include/raptor/test/cli_test.hpp

@@ -155,7 +155,7 @@ struct raptor_base : public cli_test
         using strong_bool::value;
     };
 
-    static inline auto const get_repeated_bins(size_t const repetitions) noexcept
+    static inline auto get_repeated_bins(size_t const repetitions) noexcept
     {
         using vec_t = std::vector<std::string>;
 

test/include/raptor/test/tmp_test_file.hpp

@@ -16,10 +16,10 @@ class tmp_test_file
 {
 public:
     tmp_test_file() = default;
-    tmp_test_file(tmp_test_file const &) = default;
-    tmp_test_file & operator=(tmp_test_file const &) = default;
-    tmp_test_file(tmp_test_file &&) = default;
-    tmp_test_file & operator=(tmp_test_file &&) = default;
+    tmp_test_file(tmp_test_file const &) = delete;
+    tmp_test_file & operator=(tmp_test_file const &) = delete;
+    tmp_test_file(tmp_test_file &&) = delete;
+    tmp_test_file & operator=(tmp_test_file &&) = delete;
     ~tmp_test_file() = default;
 
     template <typename... t>

test/unit/cli/build/build_hibf_test.cpp

@@ -7,22 +7,12 @@
 
 #include <raptor/test/cli_test.hpp>
 
-#if defined(__SANITIZE_THREAD__)
-#    define RAPTOR_USES_TSAN true
-#else
-#    define RAPTOR_USES_TSAN false
-#endif
-
 struct build_hibf : public raptor_base, public testing::WithParamInterface<std::tuple<size_t, size_t, bool>>
 {};
 
 TEST_P(build_hibf, with_file)
 {
     auto const [number_of_repeated_bins, window_size, run_parallel_tmp] = GetParam();
-
-    if (RAPTOR_USES_TSAN && run_parallel_tmp)
-        GTEST_SKIP() << "Threadsantizier + OpenMP causes false positives.";
-
     bool const run_parallel = run_parallel_tmp && number_of_repeated_bins >= 32;
 
     cli_test_result const result = execute_app("raptor",
@@ -47,10 +37,6 @@ TEST_P(build_hibf, with_file)
 TEST_P(build_hibf, with_shape)
 {
     auto const [number_of_repeated_bins, window_size, run_parallel_tmp] = GetParam();
-
-    if (RAPTOR_USES_TSAN && run_parallel_tmp)
-        GTEST_SKIP() << "Threadsantizier + OpenMP causes false positives.";
-
     bool const run_parallel = run_parallel_tmp && number_of_repeated_bins >= 32;
 
     cli_test_result const result = execute_app("raptor",

test/unit/cli/search/search_hibf_preprocessing_test.cpp

@@ -9,12 +9,6 @@
 
 #include <raptor/test/cli_test.hpp>
 
-#if defined(__SANITIZE_THREAD__)
-#    define RAPTOR_USES_TSAN true
-#else
-#    define RAPTOR_USES_TSAN false
-#endif
-
 struct search_hibf_preprocessing :
     public raptor_base,
     public testing::WithParamInterface<std::tuple<size_t, size_t, bool, size_t>>
@@ -74,7 +68,7 @@ TEST_P(search_hibf_preprocessing, pipeline)
     cli_test_result const result2 = execute_app("raptor",
                                                 "build",
                                                 "--threads ",
-                                                (run_parallel && !RAPTOR_USES_TSAN) ? "2" : "1",
+                                                run_parallel ? "2" : "1",
                                                 "--output raptor.index",
                                                 "--quiet",
                                                 "--input raptor_cli_test.layout");