Skip sorting if already sorted for skip list

This works around a weird error I was getting when trying to link a
translation unit that included Internal.h with libconflict-set-static.a
with clang + gnu objcopy

.gitea/workflows/ci.yml

@@ -1,322 +0,0 @@
-name: CI
-
-on: [push, pull_request]
-
-env:
-  CC: clang
-  CXX: clang++
-
-jobs:
-  pre-commit:
-    runs-on: ubuntu-latest-amd64
-    steps:
-      - uses: actions/checkout@v4
-
-      - uses: actions/cache@v4
-        with:
-          path: /var/cache/apt/archives
-          key: apt-amd64-${{ hashFiles('.gitea/workflows/ci.yml') }}
-
-      - name: Install dependencies
-        run: |
-          sudo apt-get update -qq
-          sudo apt-get install -y git nodejs pre-commit
-
-      - uses: actions/cache@v4
-        with:
-          path: ~/.cache/pre-commit
-          key: pre-commit-${{ hashFiles('.pre-commit-config.yaml') }}
-
-      - name: Run pre-commit
-        run: |
-          git config --global --add safe.directory "$PWD"
-          pre-commit run --all-files --show-diff-on-failure
-
-  test:
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - name: 64-bit-versions
-            cmake_args: -DCMAKE_CXX_FLAGS=-DUSE_64_BIT=1
-          - name: debug
-            cmake_args: -DCMAKE_BUILD_TYPE=Debug -DMSAN_TOOLCHAIN_PATH=/opt/msan
-          - name: simd-fallback
-            cmake_args: -DUSE_SIMD_FALLBACK=ON
-          - name: gcc
-            cmake_args: -DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++
-    runs-on: ubuntu-latest-amd64
-    steps:
-      - uses: actions/checkout@v4
-
-      - uses: actions/cache@v4
-        with:
-          path: /var/cache/apt/archives
-          key: apt-amd64-${{ hashFiles('.gitea/workflows/ci.yml') }}
-
-      - name: Install common dependencies
-        run: |
-          sudo apt-get update -qq
-          sudo apt-get install -y \
-            build-essential ccache cmake gcc g++ libc6-dbg \
-            ninja-build python3 valgrind zstd
-          sudo curl -Ls "https://minio.weaselab.dev/public/$(uname -m)/mc.RELEASE.2025-08-13T08-35-41Z" \
-            -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
-
-      - name: Install LLVM toolchain
-        if: matrix.name != 'gcc'
-        run: |
-          . /etc/os-release
-          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
-          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
-          sudo apt-get update -qq
-          sudo apt-get install -y \
-            clang-21 llvm-21 lld-21 mold
-          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
-            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
-          done
-
-      - name: Download MSan toolchain
-        if: matrix.name == 'debug'
-        run: |
-          curl -Ls "https://minio.weaselab.dev/public/x86_64/msan-toolchain-21.1.8.tar.zst" -o /tmp/msan-toolchain.tar.zst
-          sudo mkdir -p /opt/msan
-          sudo tar --zstd -xf /tmp/msan-toolchain.tar.zst -C /opt/msan
-
-      - name: Build
-        run: |
-          export CCACHE_DIR="$GITHUB_WORKSPACE/.ccache"
-          rm -rf build
-          cmake -S . -B build -G Ninja -DCMAKE_CXX_COMPILER_LAUNCHER=ccache ${{ matrix.cmake_args }}
-          ninja -C build
-          ccache -s
-
-      - name: Test
-        run: |
-          cd build
-          ctest --no-compress-output --test-output-size-passed 100000 --test-output-size-failed 100000 -T Test -j "$(nproc)" --timeout 90 > /dev/null
-
-      - name: Upload test results to MinIO
-        if: always()
-        env:
-          MINIO_ACCESS_KEY: ${{ secrets.MINIO_ACCESS_KEY }}
-          MC_HOST_minio: https://${{ secrets.MINIO_ACCESS_KEY }}:${{ secrets.MINIO_SECRET_KEY }}@minio.weaselab.dev
-        run: |
-          if [ -z "$MINIO_ACCESS_KEY" ]; then
-            echo "MinIO credentials not configured; skipping upload"
-            exit 0
-          fi
-          zstd build/Testing/*/Test.xml
-          mc cp build/Testing/*/Test.xml.zst "minio/jenkins/conflict-set/${{ gitea.run_number }}/${{ matrix.name }}/"
-
-      - name: Test summary
-        if: always()
-        run: |
-          python3 ctest_summary.py build/Testing/*/Test.xml \
-            --link "https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/${{ matrix.name }}/Test.xml.zst" \
-            | tee -a "$GITHUB_STEP_SUMMARY"
-
-  release:
-    strategy:
-      fail-fast: false
-      matrix:
-        include:
-          - runner: ubuntu-latest-amd64
-            arch: amd64
-            cmake_args: -DMSAN_TOOLCHAIN_PATH=/opt/msan
-          - runner: ubuntu-latest-arm64
-            arch: arm64
-            cmake_args: ""
-    runs-on: ${{ matrix.runner }}
-    steps:
-      - uses: actions/checkout@v4
-
-      - uses: actions/cache@v4
-        with:
-          path: /var/cache/apt/archives
-          key: apt-${{ matrix.arch }}-${{ hashFiles('.gitea/workflows/ci.yml') }}
-
-      - name: Install dependencies
-        run: |
-          . /etc/os-release
-          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
-          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
-          sudo apt-get update -qq
-          sudo apt-get install -y \
-            build-essential ccache clang-21 cmake devscripts \
-            libc6-dbg llvm-21 lld-21 mold ninja-build rpm \
-            valgrind zstd
-          sudo curl -Ls "https://minio.weaselab.dev/public/$(uname -m)/mc.RELEASE.2025-08-13T08-35-41Z" \
-            -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
-          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
-            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
-          done
-
-      - name: Install paper build dependencies
-        if: matrix.arch == 'amd64'
-        run: |
-          sudo apt-get install -y \
-            biber latexmk \
-            texlive-bibtex-extra texlive-fonts-recommended \
-            texlive-latex-extra texlive-pictures
-
-      - uses: actions/cache@v4
-        with:
-          path: .ccache
-          key: ccache-release-${{ matrix.arch }}-${{ gitea.sha }}
-          restore-keys: |
-            ccache-release-${{ matrix.arch }}-
-
-      - name: Download MSan toolchain
-        if: matrix.arch == 'amd64'
-        run: |
-          curl -Ls "https://minio.weaselab.dev/public/x86_64/msan-toolchain-21.1.8.tar.zst" -o /tmp/msan-toolchain.tar.zst
-          sudo mkdir -p /opt/msan
-          sudo tar --zstd -xf /tmp/msan-toolchain.tar.zst -C /opt/msan
-
-      - name: Build
-        run: |
-          export CCACHE_DIR="$GITHUB_WORKSPACE/.ccache"
-          rm -rf build
-          cmake -S . -B build -G Ninja -DCMAKE_CXX_COMPILER_LAUNCHER=ccache -DCMAKE_CXX_FLAGS=-DNVALGRIND ${{ matrix.cmake_args }}
-          ninja -C build
-          ccache -s
-
-      - name: Test
-        run: |
-          cd build
-          # On arm64, valgrind needs the MAKE_MEM_DEFINED client requests for
-          # https://git.weaselab.dev/weaselab/conflict-set/issues/39, but this
-          # build has -DNVALGRIND, which compiles them out. Skip valgrind
-          # tests here; they run annotated in the test job.
-          ctest --no-compress-output --test-output-size-passed 100000 --test-output-size-failed 100000 ${{ matrix.arch == 'arm64' && '-E valgrind' || '' }} -T Test -j "$(nproc)" --timeout 90 > /dev/null
-
-      - name: Package
-        run: |
-          cd build
-          cpack -G DEB
-          cpack -G RPM
-
-      - name: Build paper
-        if: matrix.arch == 'amd64'
-        run: |
-          cd paper
-          make
-
-      - name: Upload artifacts to MinIO
-        if: always()
-        shell: bash
-        env:
-          MINIO_ACCESS_KEY: ${{ secrets.MINIO_ACCESS_KEY }}
-          MC_HOST_minio: https://${{ secrets.MINIO_ACCESS_KEY }}:${{ secrets.MINIO_SECRET_KEY }}@minio.weaselab.dev
-        run: |
-          if [ -z "$MINIO_ACCESS_KEY" ]; then
-            echo "MinIO credentials not configured; skipping upload"
-            exit 0
-          fi
-          dest="minio/jenkins/conflict-set/${{ gitea.run_number }}/release-${{ matrix.arch }}/"
-          zstd build/Testing/*/Test.xml
-          mc cp build/Testing/*/Test.xml.zst "$dest"
-          if compgen -G "build/*.deb" > /dev/null; then
-            mc cp build/*.deb "$dest"
-          fi
-          if compgen -G "build/*.rpm" > /dev/null; then
-            mc cp build/*.rpm "$dest"
-          fi
-          if compgen -G "paper/*.pdf" > /dev/null; then
-            mc cp paper/*.pdf "$dest"
-          fi
-
-      - name: Test summary
-        if: always()
-        run: |
-          python3 ctest_summary.py build/Testing/*/Test.xml \
-            --link "https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/release-${{ matrix.arch }}/Test.xml.zst" \
-            | tee -a "$GITHUB_STEP_SUMMARY"
-
-  coverage:
-    runs-on: ubuntu-latest-amd64
-    steps:
-      - uses: actions/checkout@v4
-
-      - uses: actions/cache@v4
-        with:
-          path: /var/cache/apt/archives
-          key: apt-amd64-${{ hashFiles('.gitea/workflows/ci.yml') }}
-
-      - name: Install dependencies
-        run: |
-          . /etc/os-release
-          wget -qO- https://apt.llvm.org/llvm-snapshot.gpg.key | sudo tee /etc/apt/trusted.gpg.d/apt.llvm.org.asc
-          echo "deb http://apt.llvm.org/${VERSION_CODENAME}/ llvm-toolchain-${VERSION_CODENAME}-21 main" | sudo tee /etc/apt/sources.list.d/llvm.list
-          sudo apt-get update -qq
-          sudo apt-get install -y \
-            build-essential ccache clang-21 cmake gcovr \
-            libc6-dbg llvm-21 lld-21 mold ninja-build python3 valgrind zstd
-          sudo curl -Ls "https://minio.weaselab.dev/public/$(uname -m)/mc.RELEASE.2025-08-13T08-35-41Z" \
-            -o /usr/local/bin/mc && sudo chmod +x /usr/local/bin/mc
-          for tool in clang clang++ llvm-ar llvm-nm llvm-ranlib llvm-objcopy llvm-cov llvm-symbolizer lld ld.lld; do
-            sudo update-alternatives --install /usr/bin/${tool} ${tool} /usr/bin/${tool}-21 100
-          done
-
-      - uses: actions/cache@v4
-        with:
-          path: .ccache
-          key: ccache-coverage-${{ gitea.sha }}
-          restore-keys: |
-            ccache-coverage-
-
-      - name: Build
-        run: |
-          export CCACHE_DIR="$GITHUB_WORKSPACE/.ccache"
-          rm -rf build
-          cmake -S . -B build -G Ninja -DCMAKE_CXX_COMPILER_LAUNCHER=ccache \
-            -DCMAKE_C_FLAGS=--coverage -DCMAKE_CXX_FLAGS=--coverage \
-            -DCMAKE_BUILD_TYPE=Debug -DDISABLE_TSAN=ON
-          ninja -C build
-          ccache -s
-
-      - name: Test
-        run: |
-          cd build
-          ctest --no-compress-output --test-output-size-passed 100000 --test-output-size-failed 100000 -T Test -j "$(nproc)" --timeout 90 > /dev/null
-
-      - name: Coverage report
-        shell: bash
-        run: |
-          gcov_args=(-f ConflictSet.cpp -f LongestCommonPrefix.h -f Metrics.h
-                     --gcov-executable "llvm-cov gcov" --exclude-noncode-lines)
-          gcovr "${gcov_args[@]}" --cobertura > build/coverage.xml
-          gcovr "${gcov_args[@]}"
-          mkdir -p build/coverage_html
-          gcovr "${gcov_args[@]}" --html-details build/coverage_html/index.html
-          gcovr "${gcov_args[@]}" --fail-under-line 100 > /dev/null
-
-      - name: Upload results to MinIO
-        if: always()
-        env:
-          MINIO_ACCESS_KEY: ${{ secrets.MINIO_ACCESS_KEY }}
-          MC_HOST_minio: https://${{ secrets.MINIO_ACCESS_KEY }}:${{ secrets.MINIO_SECRET_KEY }}@minio.weaselab.dev
-        run: |
-          if [ -z "$MINIO_ACCESS_KEY" ]; then
-            echo "MinIO credentials not configured; skipping upload"
-            exit 0
-          fi
-          dest="minio/jenkins/conflict-set/${{ gitea.run_number }}/coverage/"
-          zstd build/Testing/*/Test.xml
-          mc cp build/Testing/*/Test.xml.zst "$dest"
-          if [ -e build/coverage.xml ]; then
-            mc cp build/coverage.xml "$dest"
-          fi
-          if [ -d build/coverage_html ]; then
-            mc cp -r build/coverage_html "$dest"
-          fi
-
-      - name: Test summary
-        if: always()
-        run: |
-          python3 ctest_summary.py build/Testing/*/Test.xml \
-            --link "https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/coverage/Test.xml.zst" \
-            | tee -a "$GITHUB_STEP_SUMMARY"
-          echo "" | tee -a "$GITHUB_STEP_SUMMARY"
-          echo "📊 [Coverage report](https://minio.weaselab.dev/jenkins/conflict-set/${{ gitea.run_number }}/coverage/coverage_html/index.html)" | tee -a "$GITHUB_STEP_SUMMARY"

Bench.cpp

@@ -37,16 +37,15 @@ ConflictSet::ReadRange singleton(Arena &arena, TrivialSpan key) {
 }
 
 ConflictSet::ReadRange prefixRange(Arena &arena, TrivialSpan key) {
-  int index = key.size() - 1;
+  int index;
-  for (; index >= 0; index--)
+  for (index = key.size() - 1; index >= 0; index--)
-    if (key[index] != 255)
+    if ((key[index]) != 255)
       break;
 
   // Must not be called with a string that consists only of zero or more '\xff'
-  // bytes, or with an empty string (which has no finite upper bound).
+  // bytes.
   if (index < 0) {
     assert(false);
-    std::abort();
   }
 
   uint8_t *buf = new (arena) uint8_t[index + 1];
@@ -58,7 +57,6 @@ ConflictSet::ReadRange prefixRange(Arena &arena, TrivialSpan key) {
 
 void benchConflictSet() {
   ankerl::nanobench::Bench bench;
-  bench.minEpochIterations(10000);
   ConflictSet cs{0};
 
   bench.batch(kOpsPerTx);

CMakeLists.txt

@@ -1,11 +1,11 @@
 cmake_minimum_required(VERSION 3.18)
 project(
   conflict-set
-  VERSION 0.1.0
+  VERSION 0.0.14
   DESCRIPTION
     "A data structure for optimistic concurrency control on ranges of bitwise-lexicographically-ordered keys."
   HOMEPAGE_URL "https://git.weaselab.dev/weaselab/conflict-set"
-  LANGUAGES C CXX ASM)
+  LANGUAGES C CXX)
 set(CMAKE_CXX_STANDARD 20)
 
 file(WRITE ${CMAKE_CURRENT_BINARY_DIR}/version.txt ${PROJECT_VERSION})
@@ -32,12 +32,7 @@ if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
 endif()
 
 add_compile_options(
-  -Werror=switch-enum
+  -Werror=switch-enum -Wswitch-enum -fPIC -fdata-sections -ffunction-sections
-  -Wswitch-enum
-  -Wunused-variable
-  -fPIC
-  -fdata-sections
-  -ffunction-sections
   -fno-jump-tables # https://github.com/llvm/llvm-project/issues/54247
 )
 
@@ -53,7 +48,6 @@ endif()
 
 if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
   add_compile_options("-Wno-maybe-uninitialized")
-  add_compile_options("-Wno-maybe-musttail-local-addr")
 endif()
 
 if(NOT APPLE)
@@ -61,21 +55,16 @@ if(NOT APPLE)
   add_compile_options(-g -fno-omit-frame-pointer)
 endif()
 
-set(relro_flags "LINKER:-z,relro,-z,now,-z,noexecstack")
+set(full_relro_flags "-pie;LINKER:-z,relro,-z,now,-z,noexecstack")
-set(full_relro_flags "-pie;${relro_flags}")
 cmake_push_check_state()
 list(APPEND CMAKE_REQUIRED_LINK_OPTIONS ${full_relro_flags})
 check_cxx_source_compiles("int main(){}" HAS_FULL_RELRO FAIL_REGEX "warning:")
 if(HAS_FULL_RELRO)
-  # -pie only applies to executables; passing it when linking a shared library
+  add_link_options(${full_relro_flags})
-  # makes the driver pull in Scrt1.o, which requires main.
-  add_link_options("$<$<STREQUAL:$<TARGET_PROPERTY:TYPE>,EXECUTABLE>:-pie>"
-                   ${relro_flags})
 endif()
 cmake_pop_check_state()
 
-if(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64 OR CMAKE_SYSTEM_PROCESSOR STREQUAL
+if(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64)
-                                              arm64)
   add_compile_options(-mbranch-protection=standard)
 else()
   add_compile_options(-fcf-protection)
@@ -103,42 +92,37 @@ option(USE_SIMD_FALLBACK
 
 option(DISABLE_TSAN "Disable TSAN" OFF)
 
-set(MSAN_TOOLCHAIN_PATH
-    ""
-    CACHE
-      PATH
-      "Path to an MSan-instrumented libc++ toolchain. When set, an MSan fuzz_driver target is added. Must contain include/c++/v1 and lib. See build_msan_toolchain.sh to produce one."
-)
-
 # This is encouraged according to
 # https://valgrind.org/docs/manual/manual-core-adv.html#manual-core-adv.clientreq
 include_directories(SYSTEM ${CMAKE_CURRENT_SOURCE_DIR}/third_party/valgrind)
 
+add_compile_options($<$<COMPILE_LANGUAGE:CXX>:-Wno-invalid-offsetof>)
+
 if(APPLE)
   add_link_options(-Wl,-dead_strip)
 else()
   add_link_options(-Wl,--gc-sections)
 endif()
 
-if(USE_SIMD_FALLBACK)
+if(NOT USE_SIMD_FALLBACK)
-  add_compile_definitions(USE_SIMD_FALLBACK)
+  cmake_push_check_state()
-else()
+  list(APPEND CMAKE_REQUIRED_FLAGS -mavx)
-  if(CMAKE_SYSTEM_PROCESSOR STREQUAL x86_64)
+  check_include_file_cxx("immintrin.h" HAS_AVX)
+  if(HAS_AVX)
     add_compile_options(-mavx)
+    add_compile_definitions(HAS_AVX)
+  endif()
+  cmake_pop_check_state()
+
+  check_include_file_cxx("arm_neon.h" HAS_ARM_NEON)
+  if(HAS_ARM_NEON)
+    add_compile_definitions(HAS_ARM_NEON)
   endif()
 endif()
 
 set(CMAKE_CXX_IMPLICIT_LINK_LIBRARIES "")
 
-# Architecture-specific SIMD assembly. These functions operate on
+add_library(${PROJECT_NAME}-object OBJECT ConflictSet.cpp)
-# potentially-indeterminate memory, which is UB in C++ but well-defined in
-# assembly.
-set(SIMD_ASM_FILES)
-if(CMAKE_SYSTEM_PROCESSOR STREQUAL x86_64 AND NOT USE_SIMD_FALLBACK)
-  set(SIMD_ASM_FILES ${CMAKE_CURRENT_SOURCE_DIR}/simd_x86_64.S)
-endif()
-
-add_library(${PROJECT_NAME}-object OBJECT ConflictSet.cpp ${SIMD_ASM_FILES})
 target_compile_options(${PROJECT_NAME}-object PRIVATE -fno-exceptions
                                                       -fvisibility=hidden)
 target_include_directories(${PROJECT_NAME}-object
@@ -154,7 +138,6 @@ add_custom_command(
   COMMAND_EXPAND_LISTS)
 
 add_library(${PROJECT_NAME} SHARED ${CMAKE_BINARY_DIR}/${PROJECT_NAME}.o)
-add_dependencies(${PROJECT_NAME} ${PROJECT_NAME}-object)
 set_target_properties(
   ${PROJECT_NAME} PROPERTIES LIBRARY_OUTPUT_DIRECTORY
                              "${CMAKE_CURRENT_BINARY_DIR}/radix_tree")
@@ -171,7 +154,6 @@ if(HAS_VERSION_SCRIPT)
 endif()
 
 add_library(${PROJECT_NAME}-static STATIC ${CMAKE_BINARY_DIR}/${PROJECT_NAME}.o)
-add_dependencies(${PROJECT_NAME}-static ${PROJECT_NAME}-object)
 if(CMAKE_BUILD_TYPE STREQUAL Debug)
   set_target_properties(${PROJECT_NAME}-static PROPERTIES LINKER_LANGUAGE CXX)
 else()
@@ -241,7 +223,7 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   endif()
 
   # ad hoc testing
-  add_executable(conflict_set_main ConflictSet.cpp ${SIMD_ASM_FILES})
+  add_executable(conflict_set_main ConflictSet.cpp)
   target_include_directories(conflict_set_main
                              PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/include)
   target_compile_definitions(conflict_set_main PRIVATE ENABLE_MAIN)
@@ -257,7 +239,7 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
     cmake_pop_check_state()
 
     if(HAS_LIB_FUZZER)
-      add_executable(conflict_set_fuzz_test ConflictSet.cpp ${SIMD_ASM_FILES})
+      add_executable(conflict_set_fuzz_test ConflictSet.cpp)
       target_include_directories(conflict_set_fuzz_test
                                  PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/include)
       target_compile_definitions(conflict_set_fuzz_test PRIVATE ENABLE_FUZZ)
@@ -268,9 +250,8 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
     endif()
   endif()
 
-  # whitebox tests asan+ubsan
+  # whitebox tests
-  add_executable(fuzz_driver ConflictSet.cpp FuzzTestDriver.cpp
+  add_executable(fuzz_driver ConflictSet.cpp FuzzTestDriver.cpp)
-                             ${SIMD_ASM_FILES})
   target_compile_options(fuzz_driver PRIVATE ${TEST_FLAGS})
   if(NOT CMAKE_CROSSCOMPILING)
     target_compile_options(fuzz_driver PRIVATE -fsanitize=address,undefined)
@@ -284,39 +265,9 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
     add_test(NAME conflict_set_fuzz_${hash} COMMAND fuzz_driver ${TEST})
   endforeach()
 
-  # whitebox tests msan
-  if(MSAN_TOOLCHAIN_PATH)
-    add_executable(fuzz_driver_msan ConflictSet.cpp FuzzTestDriver.cpp
-                                    ${SIMD_ASM_FILES})
-    target_compile_options(fuzz_driver_msan PRIVATE ${TEST_FLAGS})
-    if(NOT CMAKE_CROSSCOMPILING)
-      target_compile_options(
-        fuzz_driver_msan
-        PRIVATE -fsanitize=memory -fsanitize-memory-track-origins=2
-                $<$<COMPILE_LANGUAGE:CXX>:-stdlib=libc++>
-                -I${MSAN_TOOLCHAIN_PATH}/include/c++/v1)
-      target_link_options(
-        fuzz_driver_msan
-        PRIVATE
-        -fsanitize=memory
-        -fsanitize-memory-track-origins=2
-        -stdlib=libc++
-        -L${MSAN_TOOLCHAIN_PATH}/lib
-        LINKER:-rpath,${MSAN_TOOLCHAIN_PATH}/lib)
-    endif()
-    target_compile_definitions(fuzz_driver_msan PRIVATE ENABLE_FUZZ)
-    target_include_directories(fuzz_driver_msan
-                               PRIVATE ${CMAKE_CURRENT_SOURCE_DIR}/include)
-    foreach(TEST ${CORPUS_TESTS})
-      get_filename_component(hash ${TEST} NAME)
-      add_test(NAME conflict_set_msan_${hash} COMMAND fuzz_driver_msan ${TEST})
-    endforeach()
-  endif()
-
   # tsan tests
   if(NOT CMAKE_CROSSCOMPILING AND NOT DISABLE_TSAN)
-    add_executable(tsan_driver ConflictSet.cpp FuzzTestDriver.cpp
+    add_executable(tsan_driver ConflictSet.cpp FuzzTestDriver.cpp)
-                               ${SIMD_ASM_FILES})
     target_compile_options(tsan_driver PRIVATE ${TEST_FLAGS} -fsanitize=thread)
     target_link_options(tsan_driver PRIVATE -fsanitize=thread)
     target_compile_definitions(tsan_driver PRIVATE ENABLE_FUZZ THREAD_TEST)
@@ -395,7 +346,8 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   # c++98
   add_executable(conflict_set_cxx_api_test conflict_set_cxx_api_test.cpp)
   target_compile_options(conflict_set_cxx_api_test PRIVATE ${TEST_FLAGS})
-  target_link_libraries(conflict_set_cxx_api_test PRIVATE ${PROJECT_NAME})
+  target_link_libraries(conflict_set_cxx_api_test
+                        PRIVATE ${PROJECT_NAME}-static)
   set_target_properties(conflict_set_cxx_api_test PROPERTIES CXX_STANDARD 98)
   set_target_properties(conflict_set_cxx_api_test
                         PROPERTIES CXX_STANDARD_REQUIRED ON)
@@ -431,30 +383,9 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   if(NOT CMAKE_CROSSCOMPILING)
     find_program(HARDENING_CHECK hardening-check)
     if(HARDENING_CHECK)
-      # Not all versions of hardening-check support the same options, so query
+      add_test(NAME hardening_check
-      # the help output before using architecture-specific skips.
+               COMMAND ${HARDENING_CHECK} $<TARGET_FILE:${PROJECT_NAME}>
-      execute_process(
+                       --nofortify --nostackprotector)
-        COMMAND ${HARDENING_CHECK} --help
-        OUTPUT_VARIABLE hardening_check_help
-        ERROR_VARIABLE hardening_check_help
-        OUTPUT_STRIP_TRAILING_WHITESPACE ERROR_STRIP_TRAILING_WHITESPACE)
-      set(hardening_check_arch_flags "")
-      # Control flow integrity (CET) is x86-only and branch protection (PAC/BTI)
-      # is arm64-only, so ignore whichever doesn't apply.
-      if(CMAKE_SYSTEM_PROCESSOR STREQUAL aarch64 OR CMAKE_SYSTEM_PROCESSOR
-                                                    STREQUAL arm64)
-        if(hardening_check_help MATCHES "nocfprotection")
-          list(APPEND hardening_check_arch_flags --nocfprotection)
-        endif()
-      else()
-        if(hardening_check_help MATCHES "nobranchprotection")
-          list(APPEND hardening_check_arch_flags --nobranchprotection)
-        endif()
-      endif()
-      add_test(
-        NAME hardening_check
-        COMMAND ${HARDENING_CHECK} $<TARGET_FILE:${PROJECT_NAME}> --nofortify
-                --nostackprotector ${hardening_check_arch_flags})
     endif()
   endif()
 

Dockerfile

@@ -0,0 +1,58 @@
+FROM ubuntu:rolling
+
+WORKDIR /tmp
+ENV HOME=/tmp
+RUN chmod -R 777 /tmp
+
+# Install apt dependencies
+RUN apt-get update
+RUN apt-get upgrade -y
+RUN TZ=America/Los_Angeles DEBIAN_FRONTEND=noninteractive apt-get install -y \
+    binutils-aarch64-linux-gnu \
+    build-essential \
+    ccache \
+    cmake \
+    curl \
+    devscripts \
+    g++-aarch64-linux-gnu \
+    gcovr \
+    git \
+    gnupg \
+    libc6-dbg \
+    lsb-release \
+    mold \
+    ninja-build \
+    pre-commit \
+    python3-requests \
+    qemu-user \
+    rpm \
+    software-properties-common \
+    texlive-full \
+    wget \
+    zstd
+
+# Install recent valgrind from source
+RUN curl -Ls https://sourceware.org/pub/valgrind/valgrind-3.22.0.tar.bz2 -o valgrind.tar.bz2 && \
+    echo "c811db5add2c5f729944caf47c4e7a65dcaabb9461e472b578765dd7bf6d2d4c  valgrind.tar.bz2" > valgrind-sha.txt && \
+    sha256sum --quiet -c valgrind-sha.txt && \
+    mkdir valgrind && \
+    tar --strip-components 1 --no-same-owner --no-same-permissions --directory valgrind -xjf valgrind.tar.bz2 && \
+    cd valgrind && \
+    ./configure --enable-only64bit --enable-lto && \
+    make -j`nproc` && \
+    make install && \
+    cd .. && \
+    rm -rf /tmp/*
+
+# Recent clang
+RUN wget https://apt.llvm.org/llvm.sh && chmod +x ./llvm.sh && ./llvm.sh 20
+
+RUN apt-get -y install clang llvm
+
+# Set after building valgrind, which doesn't build with clang for some reason
+ENV CC=clang
+ENV CXX=clang++
+
+# Try to have all the pre-commit hooks we'll need already initialized
+COPY .pre-commit-config.yaml /tmp/
+RUN git init && pre-commit install-hooks

HashTable.cpp

@@ -96,9 +96,7 @@ void ConflictSet::setOldestVersion(int64_t oldestVersion) {
   return impl->setOldestVersion(oldestVersion);
 }
 
-// The hash_table implementation does not track memory usage, so return 0 to
+int64_t ConflictSet::getBytes() const { return -1; }
-// satisfy the API contract that getBytes() returns a non-negative value.
-int64_t ConflictSet::getBytes() const { return 0; }
 
 void ConflictSet::getMetricsV1(MetricsV1 **metrics, int *count) const {
   *metrics = nullptr;
@@ -121,13 +119,7 @@ ConflictSet::ConflictSet(ConflictSet &&other) noexcept
     : impl(std::exchange(other.impl, nullptr)) {}
 
 ConflictSet &ConflictSet::operator=(ConflictSet &&other) noexcept {
-  if (this != &other) {
+  impl = std::exchange(other.impl, nullptr);
-    if (impl) {
-      impl->~Impl();
-      safe_free(impl, sizeof(Impl));
-    }
-    impl = std::exchange(other.impl, nullptr);
-  }
   return *this;
 }
 
@@ -163,10 +155,7 @@ __attribute__((__visibility__("default"))) void ConflictSet_destroy(void *cs) {
 }
 __attribute__((__visibility__("default"))) int64_t
 ConflictSet_getBytes(void *cs) {
-  (void)cs;
+  using Impl = ConflictSet::Impl;
-  // The hash_table implementation does not track memory usage, so return 0 to
+  return -1;
-  // satisfy the API contract that ConflictSet_getBytes returns a non-negative
-  // value.
-  return 0;
 }
 }

Internal.h

@@ -18,6 +18,7 @@ using namespace weaselab;
 #include <span>
 #include <string>
 #include <thread>
+#include <unordered_set>
 #include <utility>
 
 #include <callgrind.h>
@@ -367,6 +368,23 @@ template <class T, class C = std::less<T>> auto set(Arena &arena) {
   return Set<T, C>(ArenaAlloc<T>(&arena));
 }
 
+template <class T> struct MyHash;
+
+template <class T> struct MyHash<T *> {
+  size_t operator()(const T *t) const noexcept {
+    size_t result;
+    memcpy(&result, &t, sizeof(result));
+    return result;
+  }
+};
+
+template <class T>
+using HashSet =
+    std::unordered_set<T, MyHash<T>, std::equal_to<T>, ArenaAlloc<T>>;
+template <class T> auto hashSet(Arena &arena) {
+  return HashSet<T>(ArenaAlloc<T>(&arena));
+}
+
 template <class T, class U>
 bool operator==(const ArenaAlloc<T> &lhs, const ArenaAlloc<U> &rhs) {
   return lhs.arena == rhs.arena;

Jenkinsfile

@@ -0,0 +1,151 @@
+def CleanBuildAndTest(String cmakeArgs) {
+    sh """
+    export CCACHE_DIR=/ccache
+    rm -rf build
+    mkdir build
+    cd build
+    cmake .. -G Ninja -DCMAKE_CXX_COMPILER_LAUNCHER=ccache ${cmakeArgs} 
+    ninja
+    ccache -s
+    """
+    catchError {
+        sh '''
+        cd build
+        ctest --no-compress-output --test-output-size-passed 100000 --test-output-size-failed 100000 -T Test -j `nproc` --timeout 90 > /dev/null
+        zstd Testing/*/Test.xml
+        '''
+    }
+    xunit tools: [CTest(pattern: 'build/Testing/*/Test.xml')], skipPublishingChecks: false
+    minio bucket: 'jenkins', credentialsId: 'jenkins-minio', excludes: '', host: 'minio.weaselab.dev', includes: 'build/Testing/*/Test.xml.zst', targetFolder: '${JOB_NAME}/${BUILD_NUMBER}/${STAGE_NAME}/'
+}
+
+pipeline {
+    agent any
+    stages {
+        stage('Pre-commit') {
+            agent {
+                dockerfile {
+                    args '-v /home/jenkins/ccache:/ccache'
+                    reuseNode true
+                }
+            }
+            steps {
+                script {
+                    env.HOME = env.WORKSPACE
+                }
+                sh 'pre-commit run --all-files --show-diff-on-failure'
+            }
+        }
+        stage('64 bit versions') {
+            agent {
+                dockerfile {
+                    args '-v /home/jenkins/ccache:/ccache'
+                    reuseNode true
+                }
+            }
+            steps {
+                CleanBuildAndTest("-DCMAKE_CXX_FLAGS=-DUSE_64_BIT=1")
+            }
+        }
+        stage('Debug') {
+            agent {
+                dockerfile {
+                    args '-v /home/jenkins/ccache:/ccache'
+                    reuseNode true
+                }
+            }
+            steps {
+                CleanBuildAndTest("-DCMAKE_BUILD_TYPE=Debug")
+            }
+        }
+        stage('SIMD fallback') {
+            agent {
+                dockerfile {
+                    args '-v /home/jenkins/ccache:/ccache'
+                    reuseNode true
+                }
+            }
+            steps {
+                CleanBuildAndTest("-DUSE_SIMD_FALLBACK=ON")
+            }
+        }
+        stage('Release [clang]') {
+            agent {
+                dockerfile {
+                    args '-v /home/jenkins/ccache:/ccache'
+                    reuseNode true
+                }
+            }
+            steps {
+                CleanBuildAndTest("-DCMAKE_CXX_FLAGS=-DNVALGRIND")
+                recordIssues(tools: [clang()])
+                sh '''
+                cd build
+                cpack -G DEB
+                cpack -G RPM
+                '''
+                sh '''
+                cd paper
+                make
+                '''
+                minio bucket: 'jenkins', credentialsId: 'jenkins-minio', excludes: '', host: 'minio.weaselab.dev', includes: 'build/*.deb,build/*.rpm,paper/*.pdf', targetFolder: '${JOB_NAME}/${BUILD_NUMBER}/${STAGE_NAME}/'
+            }
+        }
+        stage('Release [gcc]') {
+            agent {
+                dockerfile {
+                    args '-v /home/jenkins/ccache:/ccache'
+                    reuseNode true
+                }
+            }
+            steps {
+                CleanBuildAndTest("-DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++ -DCMAKE_CXX_FLAGS=-DNVALGRIND")
+                recordIssues(tools: [gcc()])
+            }
+        }
+        stage('Release [clang,aarch64]') {
+            agent {
+                dockerfile {
+                    args '-v /home/jenkins/ccache:/ccache'
+                    reuseNode true
+                }
+            }
+            steps {
+                CleanBuildAndTest("-DCMAKE_TOOLCHAIN_FILE=../aarch64-toolchain.cmake -DCMAKE_CXX_FLAGS=-DNVALGRIND")
+                sh '''
+                cd build
+                cpack -G DEB
+                cpack -G RPM
+                '''
+                minio bucket: 'jenkins', credentialsId: 'jenkins-minio', excludes: '', host: 'minio.weaselab.dev', includes: 'build/*.deb,build/*.rpm', targetFolder: '${JOB_NAME}/${BUILD_NUMBER}/${STAGE_NAME}'
+            }
+        }
+        stage('Coverage') {
+            agent {
+                dockerfile {
+                    args '-v /home/jenkins/ccache:/ccache'
+                    reuseNode true
+                }
+            }
+            steps {
+                script {
+                    gcov_args = "-f ConflictSet.cpp -f LongestCommonPrefix.h -f Metrics.h --gcov-executable 'llvm-cov gcov' --exclude-noncode-lines"
+                }
+                CleanBuildAndTest("-DCMAKE_C_FLAGS=--coverage -DCMAKE_CXX_FLAGS=--coverage -DCMAKE_BUILD_TYPE=Debug -DDISABLE_TSAN=ON")
+                sh """
+                gcovr ${gcov_args} --cobertura > build/coverage.xml
+                """
+                recordCoverage qualityGates: [[criticality: 'NOTE', metric: 'MODULE']], tools: [[parser: 'COBERTURA', pattern: 'build/coverage.xml']]
+                sh """
+                gcovr ${gcov_args}
+                gcovr ${gcov_args} --fail-under-line 100 > /dev/null
+                """
+            }
+        }
+    }
+    post {
+        always {
+            emailext mimeType: 'text/html', body: '${SCRIPT, template="groovy-html.template"}', subject: "${env.JOB_NAME} - Build# ${env.BUILD_NUMBER} - ${currentBuild.currentResult}", to: 'andrew@weaselab.dev'
+        }
+    }
+}

README.md

@@ -4,14 +4,7 @@ Intended as an alternative to FoundationDB's skip list.
 
 Hardware for all benchmarks is an AMD Ryzen 9 7900 with (2x32GB) 5600MT/s CL28-34-34-89 1.35V RAM.
 
-```
+Compiler is `Ubuntu clang version 20.0.0 (++20241029082144+7544d3af0e28-1~exp1~20241029082307.506)`.
-$ clang++ --version
-
-Ubuntu clang version 21.1.8 (6ubuntu1)
-Target: x86_64-pc-linux-gnu
-Thread model: posix
-InstalledDir: /usr/lib/llvm-21/bin
-```
 
 # Microbenchmark
 
@@ -19,45 +12,44 @@ InstalledDir: /usr/lib/llvm-21/bin
 
 |               ns/op |                op/s |    err% |          ins/op |          cyc/op |    IPC |         bra/op |   miss% |     total | benchmark
 |--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
-|              164.29 |        6,086,873.38 |    0.0% |        3,107.03 |          604.19 |  5.142 |         558.59 |    0.0% |      1.96 | `point reads`
+|              159.65 |        6,263,576.52 |    1.6% |        2,972.36 |          820.37 |  3.623 |         504.59 |    0.0% |      0.01 | `point reads`
-|              161.05 |        6,209,395.38 |    0.1% |        3,036.76 |          592.21 |  5.128 |         539.35 |    0.0% |      1.93 | `prefix reads`
+|              156.32 |        6,397,320.65 |    0.7% |        2,913.62 |          806.87 |  3.611 |         490.19 |    0.0% |      0.01 | `prefix reads`
-|              239.55 |        4,174,539.38 |    0.1% |        3,722.71 |          880.68 |  4.227 |         692.00 |    0.0% |      2.86 | `range reads`
+|              229.18 |        4,363,293.65 |    1.2% |        3,541.05 |        1,219.75 |  2.903 |         629.33 |    0.0% |      0.01 | `range reads`
-|              354.75 |        2,818,919.14 |    0.7% |        4,523.64 |        1,304.75 |  3.467 |         720.22 |    2.0% |      4.23 | `point writes`
+|              363.37 |        2,752,026.30 |    0.3% |        5,273.63 |        1,951.54 |  2.702 |         851.66 |    1.7% |      0.01 | `point writes`
-|              345.32 |        2,895,878.47 |    0.1% |        4,484.57 |        1,270.31 |  3.530 |         705.00 |    1.8% |      4.12 | `prefix writes`
+|              364.99 |        2,739,787.02 |    0.3% |        5,250.92 |        1,958.54 |  2.681 |         839.24 |    1.7% |      0.01 | `prefix writes`
-|              193.48 |        5,168,547.42 |    0.1% |        2,224.10 |          711.72 |  3.125 |         377.17 |    3.3% |      2.32 | `range writes`
+|              242.26 |        4,127,796.58 |    2.9% |        3,117.33 |        1,304.41 |  2.390 |         541.07 |    2.8% |      0.02 | `range writes`
-|              404.89 |        2,469,777.50 |    2.4% |        6,855.96 |        1,489.70 |  4.602 |       1,227.82 |    1.3% |      0.05 | `monotonic increasing point writes`
+|              562.48 |        1,777,855.27 |    0.8% |        7,305.21 |        3,034.34 |  2.408 |       1,329.30 |    1.3% |      0.01 | `monotonic increasing point writes`
-|          134,231.80 |            7,449.80 |    1.9% |      812,045.25 |      495,770.40 |  1.638 |     151,246.50 |    0.9% |      0.01 | `worst case for radix tree`
+|          122,688.57 |            8,150.72 |    0.7% |      798,766.00 |      666,842.00 |  1.198 |     144,584.50 |    0.1% |      0.01 | `worst case for radix tree`
-|               37.80 |       26,454,311.17 |    0.4% |          701.00 |          139.14 |  5.038 |         102.00 |    0.0% |      0.01 | `create and destroy`
+|               41.71 |       23,976,459.34 |    1.7% |          885.00 |          219.17 |  4.038 |         132.00 |    0.0% |      0.01 | `create and destroy`
 
 ## Radix tree (this implementation)
 
-
 |               ns/op |                op/s |    err% |          ins/op |          cyc/op |    IPC |         bra/op |   miss% |     total | benchmark
 |--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
-|               12.89 |       77,565,115.56 |    0.1% |          244.55 |           47.43 |  5.155 |          34.21 |    0.6% |      0.15 | `point reads`
+|               12.63 |       79,186,868.18 |    1.4% |          241.61 |           64.76 |  3.731 |          31.64 |    0.8% |      0.01 | `point reads`
-|               15.11 |       66,162,047.76 |    0.1% |          297.79 |           55.60 |  5.356 |          43.23 |    0.4% |      0.18 | `prefix reads`
+|               14.48 |       69,078,073.40 |    0.3% |          292.42 |           74.69 |  3.915 |          41.49 |    0.5% |      0.01 | `prefix reads`
-|               36.29 |       27,559,358.29 |    0.1% |          783.16 |          133.44 |  5.869 |         109.52 |    0.2% |      0.43 | `range reads`
+|               34.37 |       29,094,694.11 |    0.2% |          759.53 |          179.77 |  4.225 |         100.38 |    0.2% |      0.01 | `range reads`
-|               20.53 |       48,719,405.55 |    0.1% |          381.81 |           75.51 |  5.057 |          51.04 |    0.5% |      0.25 | `point writes`
+|               19.34 |       51,713,896.36 |    0.7% |          369.70 |          101.81 |  3.631 |          47.88 |    0.6% |      0.01 | `point writes`
-|               39.37 |       25,402,042.40 |    0.1% |          685.00 |          144.83 |  4.730 |         106.72 |    0.3% |      0.47 | `prefix writes`
+|               39.16 |       25,538,968.61 |    0.2% |          653.16 |          206.77 |  3.159 |          89.62 |    0.8% |      0.01 | `prefix writes`
-|               43.78 |       22,843,841.63 |    0.1% |          800.40 |          161.06 |  4.970 |         127.36 |    0.1% |      0.53 | `range writes`
+|               40.58 |       24,642,681.12 |    4.7% |          718.44 |          216.44 |  3.319 |          99.28 |    0.6% |      0.01 | `range writes`
-|               78.37 |       12,760,008.75 |    1.0% |        1,452.61 |          288.24 |  5.040 |         278.69 |    0.1% |      0.01 | `monotonic increasing point writes`
+|               78.77 |       12,694,520.69 |    3.8% |        1,395.55 |          421.73 |  3.309 |         249.81 |    0.1% |      0.01 | `monotonic increasing point writes`
-|          322,885.50 |            3,097.07 |    1.5% |    4,362,382.00 |    1,183,852.00 |  3.685 |     765,301.00 |    0.1% |      0.01 | `worst case for radix tree`
+|          287,760.50 |            3,475.11 |    0.5% |    3,929,266.50 |    1,550,225.50 |  2.535 |     639,064.00 |    0.0% |      0.01 | `worst case for radix tree`
-|               99.99 |       10,000,718.79 |    0.4% |        1,775.00 |          367.93 |  4.824 |         288.00 |    0.0% |      0.01 | `create and destroy`
+|              104.76 |        9,545,250.65 |    3.1% |        2,000.00 |          552.82 |  3.618 |         342.00 |    0.0% |      0.01 | `create and destroy`
 
 # "Real data" test
 
-Point queries only. Gc ratio is the ratio of time spent doing garbage collection to time spent adding writes or doing garbage collection. Lower is better.
+Point queries only, best of three runs. Gc ratio is the ratio of time spent doing garbage collection to time spent adding writes or doing garbage collection. Lower is better.
 
 ## skip list
 
 ```
-Check: 4.62967 seconds, 352.195 MB/s, Add: 3.34177 seconds, 167.771 MB/s, Gc ratio: 37.9399%, Peak idle memory: 5.51852e+06
+Check: 4.39702 seconds, 370.83 MB/s, Add: 4.50025 seconds, 124.583 MB/s, Gc ratio: 29.1333%, Peak idle memory: 5.51852e+06
 ```
 
 ## radix tree
 
 ```
-Check: 1.00477 seconds, 1622.8 MB/s, Add: 1.21142 seconds, 462.808 MB/s, Gc ratio: 39.4716%, Peak idle memory: 2.0226e+06
+Check: 0.987757 seconds, 1650.76 MB/s, Add: 1.24815 seconds, 449.186 MB/s, Gc ratio: 41.4675%, Peak idle memory: 2.02872e+06
 ```
 
 ## hash table
@@ -65,6 +57,5 @@ Check: 1.00477 seconds, 1622.8 MB/s, Add: 1.21142 seconds, 462.808 MB/s, Gc rati
 (The hash table implementation doesn't work on range queries, and its purpose is to provide an idea of how fast point queries can be)
 
 ```
-Check: 0.854254 seconds, 1908.74 MB/s, Add: 0.632626 seconds, 886.232 MB/s, Gc ratio: 41.0827%, Peak idle memory: 0
+Check: 0.84256 seconds, 1935.23 MB/s, Add: 0.697204 seconds, 804.146 MB/s, Gc ratio: 35.4091%
 ```
-

RealDataBench.cpp

@@ -1,12 +1,11 @@
 #include <ConflictSet.h>
 
-#include <algorithm>
 #include <cerrno>
 #include <chrono>
 #include <cstdio>
 #include <cstring>
 #include <fcntl.h>
-#include <span>
+#include <string_view>
 #include <sys/mman.h>
 #include <sys/stat.h>
 #include <unistd.h>
@@ -65,7 +64,7 @@ int main(int argc, const char **argv) {
     auto *const mapOriginal = begin;
     const auto sizeOriginal = size;
 
-    using StringView = std::span<const uint8_t>;
+    using StringView = std::basic_string_view<uint8_t>;
 
     StringView write;
     std::vector<StringView> reads;
@@ -78,10 +77,10 @@ int main(int argc, const char **argv) {
       begin = end + 1;
       end = (uint8_t *)memchr(begin, '\n', size);
 
-      if (line.size() >= 2 && line[0] == 'P') {
+      if (line.size() > 0 && line[0] == 'P') {
-        write = line.subspan(2, line.size() - 2);
+        write = line.substr(2, line.size());
-      } else if (line.size() >= 2 && line[0] == 'L') {
+      } else if (line.size() > 0 && line[0] == 'L') {
-        reads.push_back(line.subspan(2, line.size() - 2));
+        reads.push_back(line.substr(2, line.size()));
       } else if (line.empty()) {
         {
           readRanges.resize(reads.size());
@@ -91,7 +90,7 @@ int main(int argc, const char **argv) {
             iter->begin.len = read.size();
             checkBytes += read.size();
             iter->end.len = 0;
-            iter->readVersion = std::max<int64_t>(0, version - 100);
+            iter->readVersion = version - 100;
             ++iter;
           }
         }
@@ -122,7 +121,7 @@ int main(int argc, const char **argv) {
         }
 
         timer = now();
-        cs.setOldestVersion(std::max<int64_t>(0, version - 10000));
+        cs.setOldestVersion(version - 10000);
         gcTime += now() - timer;
       }
     }
@@ -134,10 +133,10 @@ int main(int argc, const char **argv) {
   int metricsCount;
   cs.getMetricsV1(&metrics, &metricsCount);
   for (int i = 0; i < metricsCount; ++i) {
-    fprintf(stderr, "# HELP %s %s\n", metrics[i].name, metrics[i].help);
+    printf("# HELP %s %s\n", metrics[i].name, metrics[i].help);
-    fprintf(stderr, "# TYPE %s %s\n", metrics[i].name,
+    printf("# TYPE %s %s\n", metrics[i].name,
-            metrics[i].type == metrics[i].Counter ? "counter" : "gauge");
+           metrics[i].type == metrics[i].Counter ? "counter" : "gauge");
-    fprintf(stderr, "%s %g\n", metrics[i].name, metrics[i].getValue());
+    printf("%s %g\n", metrics[i].name, metrics[i].getValue());
   }
 
   printf("Check: %g seconds, %g MB/s, Add: %g seconds, %g MB/s, Gc ratio: "

ServerBench.cpp

@@ -12,7 +12,6 @@
 #include <sys/ioctl.h>
 #include <sys/resource.h>
 #include <sys/socket.h>
-#include <sys/syscall.h>
 #include <sys/types.h>
 #include <sys/uio.h>
 #include <thread>
@@ -24,97 +23,6 @@
 #include "Internal.h"
 #include "third_party/nadeau.h"
 
-constexpr int kCacheLine = 64; // TODO mac m1 is 128
-
-template <class T> struct TxQueue {
-
-  explicit TxQueue(int lgSlotCount)
-      : slotCount(1 << lgSlotCount), slotCountMask(slotCount - 1),
-        slots(new T[slotCount]) {
-    // Otherwise we can't tell the difference between full and empty.
-    assert(!(slotCountMask & 0x80000000));
-  }
-
-  /// Call from producer thread, after ensuring consumer is no longer accessing
-  /// it somehow
-  ~TxQueue() { delete[] slots; }
-
-  /// Must be called from the producer thread
-  void push(T t) {
-    if (wouldBlock()) {
-      // Wait for pops to change and try again
-      consumer.pops.wait(producer.lastPopRead, std::memory_order_relaxed);
-      producer.lastPopRead = consumer.pops.load(std::memory_order_acquire);
-    }
-    slots[producer.pushesNonAtomic++ & slotCountMask] = std::move(t);
-    // seq_cst so that the notify can't be ordered before the store
-    producer.pushes.store(producer.pushesNonAtomic, std::memory_order_seq_cst);
-    // We have to notify every time, since we don't know if this is the last
-    // push ever
-    producer.pushes.notify_one();
-  }
-
-  /// Must be called from the producer thread
-  uint32_t outstanding() {
-    return producer.pushesNonAtomic -
-           consumer.pops.load(std::memory_order_relaxed);
-  }
-
-  /// Returns true if a call to push might block. Must be called from the
-  /// producer thread.
-  bool wouldBlock() {
-    // See if we can determine that overflow won't happen entirely from state
-    // local to the producer
-    if (producer.pushesNonAtomic - producer.lastPopRead == slotCount - 1) {
-      // Re-read pops with memory order
-      producer.lastPopRead = consumer.pops.load(std::memory_order_acquire);
-      return producer.pushesNonAtomic - producer.lastPopRead == slotCount - 1;
-    }
-    return false;
-  }
-
-  /// Valid until the next pop, or until this queue is destroyed.
-  T *pop() {
-    // See if we can determine that there's an entry we can pop entirely from
-    // state local to the consumer
-    if (consumer.lastPushRead - consumer.popsNonAtomic == 0) {
-      // Re-read pushes with memory order and try again
-      consumer.lastPushRead = producer.pushes.load(std::memory_order_acquire);
-      if (consumer.lastPushRead - consumer.popsNonAtomic == 0) {
-        // Wait for pushes to change and try again
-        producer.pushes.wait(consumer.lastPushRead, std::memory_order_relaxed);
-        consumer.lastPushRead = producer.pushes.load(std::memory_order_acquire);
-      }
-    }
-    auto result = &slots[consumer.popsNonAtomic++ & slotCountMask];
-    // We only have to write pops with memory order if we've run out of items.
-    // We know that we'll eventually run out.
-    if (consumer.lastPushRead - consumer.popsNonAtomic == 0) {
-      // seq_cst so that the notify can't be ordered before the store
-      consumer.pops.store(consumer.popsNonAtomic, std::memory_order_seq_cst);
-      consumer.pops.notify_one();
-    }
-    return result;
-  }
-
-private:
-  const uint32_t slotCount;
-  const uint32_t slotCountMask;
-  T *slots;
-  struct alignas(kCacheLine) ProducerState {
-    std::atomic<uint32_t> pushes{0};
-    uint32_t pushesNonAtomic{0};
-    uint32_t lastPopRead{0};
-  };
-  struct alignas(kCacheLine) ConsumerState {
-    std::atomic<uint32_t> pops{0};
-    uint32_t popsNonAtomic{0};
-    uint32_t lastPushRead{0};
-  };
-  ProducerState producer;
-  ConsumerState consumer;
-};
-
 std::atomic<int64_t> transactions;
 
 int64_t safeUnaryMinus(int64_t x) {
@@ -139,17 +47,13 @@ void tupleAppend(std::string &output, int64_t value) {
 
 void tupleAppend(std::string &output, std::string_view value) {
   output.push_back('\x02');
-  if (memchr(value.data(), '\x00', value.size()) != nullptr) {
+  for (auto c : value) {
-    for (auto c : value) {
+    if (c == '\x00') {
-      if (c == '\x00') {
+      output.push_back('\x00');
-        output.push_back('\x00');
+      output.push_back('\xff');
-        output.push_back('\xff');
+    } else {
-      } else {
+      output.push_back(c);
-        output.push_back(c);
-      }
     }
-  } else {
-    output.insert(output.end(), value.begin(), value.end());
   }
   output.push_back('\x00');
 }
@@ -162,69 +66,49 @@ template <class... Ts> std::string tupleKey(const Ts &...ts) {
 
 constexpr int kTotalKeyRange = 1'000'000'000;
 constexpr int kWindowSize = 1'000'000;
-constexpr int kNumReadKeysPerTx = 5;
+constexpr int kNumKeys = 10;
-constexpr int kNumWriteKeysPerTx = 10;
 
-struct Transaction {
+void workload(weaselab::ConflictSet *cs) {
-  std::vector<std::string> keys;
+  int64_t version = kWindowSize;
-  std::vector<weaselab::ConflictSet::ReadRange> reads;
+  for (;; transactions.fetch_add(1, std::memory_order_relaxed)) {
-  std::vector<weaselab::ConflictSet::WriteRange> writes;
-  int64_t version;
-  int64_t oldestVersion;
-  Transaction() = default;
-  explicit Transaction(int64_t version)
-      : version(version), oldestVersion(version - kWindowSize) {
     std::vector<int64_t> keyIndices;
-    for (int i = 0; i < std::max(kNumReadKeysPerTx, kNumWriteKeysPerTx); ++i) {
+    for (int i = 0; i < kNumKeys; ++i) {
       keyIndices.push_back(rand() % kTotalKeyRange);
     }
     std::sort(keyIndices.begin(), keyIndices.end());
+    std::vector<std::string> keys;
     constexpr std::string_view fullString =
         "this is a string, where a prefix of it is used as an element of the "
         "tuple forming the key";
-    for (int i = 0; i < int(keyIndices.size()); ++i) {
+    for (int i = 0; i < kNumKeys; ++i) {
       keys.push_back(
           tupleKey(0x100, keyIndices[i] / fullString.size(),
                    fullString.substr(0, keyIndices[i] % fullString.size())));
       // printf("%s\n", printable(keys.back()).c_str());
     }
-    for (int i = 0; i < kNumWriteKeysPerTx; ++i) {
+
+    std::vector<weaselab::ConflictSet::ReadRange> reads;
+    std::vector<weaselab::ConflictSet::WriteRange> writes;
+    std::vector<weaselab::ConflictSet::Result> results;
+    for (int i = 0; i < kNumKeys; ++i) {
       writes.push_back({{(const uint8_t *)keys[i].data(), int(keys[i].size())},
                         {nullptr, 0}});
-    }
-    reads.push_back({{(const uint8_t *)keys[0].data(), int(keys[0].size())},
-                     {(const uint8_t *)keys[1].data(), int(keys[1].size())},
-                     version - std::min(10, kWindowSize)});
-    static_assert(kNumReadKeysPerTx >= 3);
-    for (int i = 2; i < kNumReadKeysPerTx; ++i) {
       reads.push_back({{(const uint8_t *)keys[i].data(), int(keys[i].size())},
                        {nullptr, 0},
                        version - kWindowSize});
     }
+    results.resize(reads.size());
+
+    cs->check(reads.data(), results.data(), reads.size());
+    bool ok = true;
+    for (auto result : results) {
+      ok &= result == weaselab::ConflictSet::Commit;
+    }
+    cs->addWrites(writes.data(), ok ? writes.size() : 0, version);
+    cs->setOldestVersion(version - kWindowSize);
+    ++version;
   }
-
+}
-  Transaction(Transaction &&) = default;
-  Transaction &operator=(Transaction &&) = default;
-  Transaction(Transaction const &) = delete;
-  Transaction const &operator=(Transaction const &) = delete;
-};
-
-struct Resolver {
-
-  void resolve(const weaselab::ConflictSet::ReadRange *reads, int readCount,
-               const weaselab::ConflictSet::WriteRange *writes, int writeCount,
-               int64_t newVersion, int64_t newOldestVersion) {
-    results.resize(readCount);
-    cs.check(reads, results.data(), readCount);
-    cs.addWrites(writes, writeCount, newVersion);
-    cs.setOldestVersion(newOldestVersion);
-  }
-
-  ConflictSet cs{0};
-
-private:
-  std::vector<weaselab::ConflictSet::Result> results;
-};
 
 // Adapted from getaddrinfo man page
 int getListenFd(const char *node, const char *service) {
@@ -367,8 +251,7 @@ int main(int argc, char **argv) {
   {
     int listenFd = getListenFd(argv[1], argv[2]);
 
-    Resolver resolver;
+    weaselab::ConflictSet cs{0};
-    auto &cs = resolver.cs;
     weaselab::ConflictSet::MetricsV1 *metrics;
     int metricsCount;
     cs.getMetricsV1(&metrics, &metricsCount);
@@ -417,22 +300,7 @@ int main(int argc, char **argv) {
     }
 #endif
 
-    TxQueue<Transaction> queue{10};
+    auto w = std::thread{workload, &cs};
-
-    auto workloadThread = std::thread{[&]() {
-      for (int64_t version = kWindowSize;;
-           ++version, transactions.fetch_add(1, std::memory_order_relaxed)) {
-        queue.push(Transaction(version));
-      }
-    }};
-
-    auto resolverThread = std::thread{[&]() {
-      for (;;) {
-        auto tx = queue.pop();
-        resolver.resolve(tx->reads.data(), tx->reads.size(), tx->writes.data(),
-                         tx->writes.size(), tx->version, tx->oldestVersion);
-      }
-    }};
 
     for (;;) {
       struct sockaddr_storage peer_addr = {};

SkipList.cpp

@@ -38,9 +38,6 @@ std::span<const uint8_t> keyAfter(Arena &arena, std::span<const uint8_t> key) {
 
 std::span<const uint8_t> copyToArena(Arena &arena,
                                      std::span<const uint8_t> key) {
-  if (key.size() == 0) {
-    return std::span<const uint8_t>();
-  }
   auto result = std::span<uint8_t>(new (arena) uint8_t[key.size()], key.size());
   memcpy(result.data(), key.data(), key.size());
   return result;
@@ -100,7 +97,7 @@ force_inline bool getCharacter(const KeyInfo &ki, int character,
 
 bool operator<(const KeyInfo &lhs, const KeyInfo &rhs) {
   int i = std::min(lhs.key.size(), rhs.key.size());
-  int c = i > 0 ? memcmp(lhs.key.data(), rhs.key.data(), i) : 0;
+  int c = memcmp(lhs.key.data(), rhs.key.data(), i);
   if (c != 0)
     return c < 0;
 
@@ -290,8 +287,7 @@ private:
 
   static force_inline bool less(const uint8_t *a, int aLen, const uint8_t *b,
                                 int bLen) {
-    int n = std::min(aLen, bLen);
+    int c = memcmp(a, b, std::min(aLen, bLen));
-    int c = n > 0 ? memcmp(a, b, n) : 0;
     if (c < 0)
       return true;
     if (c > 0)
@@ -758,8 +754,7 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
 
   void addWrites(const ConflictSet::WriteRange *writes, int count,
                  int64_t writeVersion) {
-    auto points = std::vector<KeyInfo>();
+    auto points = std::vector<KeyInfo>(count * 2);
-    points.reserve(count * 2);
     Arena arena;
 
     for (int r = 0; r < count; r++) {
@@ -801,6 +796,7 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
     int temp[stripeSize];
     int stripes = (stringCount + stripeSize - 1) / stripeSize;
     StringRef values[stripeSize];
+    int64_t writeVersions[stripeSize / 2];
     int ss = stringCount - (stripes - 1) * stripeSize;
     int64_t entryDelta = 0;
     for (int s = stripes - 1; s >= 0; s--) {
@@ -986,12 +982,7 @@ ConflictSet::ConflictSet(ConflictSet &&other) noexcept
     : impl(std::exchange(other.impl, nullptr)) {}
 
 ConflictSet &ConflictSet::operator=(ConflictSet &&other) noexcept {
-  if (this != &other) {
+  impl = std::exchange(other.impl, nullptr);
-    if (impl) {
-      internal_destroy(impl);
-    }
-    impl = std::exchange(other.impl, nullptr);
-  }
   return *this;
 }
 

apple-symbol-imports.txt

@@ -1,4 +1,3 @@
-___chkstk_darwin
 ___stack_chk_fail
 ___stack_chk_guard
 __tlv_bootstrap
@@ -6,7 +5,6 @@ _abort
 _bzero
 _free
 _malloc
-_memcmp
 _memcpy
 _memmove
 dyld_stub_binder

build_msan_toolchain.sh

@@ -1,43 +0,0 @@
-#!/bin/bash
-
-set -euxo pipefail
-
-LLVM_VERSION="${LLVM_VERSION:-21}"
-MSAN_PREFIX="${MSAN_PREFIX:-$PWD/msan}"
-JOBS="${JOBS:-$(nproc)}"
-
-cd /tmp
-rm -rf libcxx-msan
-mkdir libcxx-msan
-cd libcxx-msan
-
-git clone --depth=1 "https://github.com/llvm/llvm-project.git" -b "release/${LLVM_VERSION}.x"
-
-cmake -S llvm-project/runtimes -B build_msan \
-  -DCMAKE_BUILD_TYPE=Release \
-  -DCMAKE_CXX_COMPILER="clang++-${LLVM_VERSION}" \
-  -DCMAKE_C_COMPILER="clang-${LLVM_VERSION}" \
-  -DCMAKE_INSTALL_PREFIX="${MSAN_PREFIX}" \
-  -DLLVM_ENABLE_RUNTIMES="libcxx;libcxxabi;libunwind" \
-  -DLLVM_USE_SANITIZER=MemoryWithOrigins
-
-cmake --build build_msan -j"${JOBS}"
-cmake --install build_msan
-
-cmake -S llvm-project/runtimes -B build_nomsan \
-  -DCMAKE_BUILD_TYPE=Release \
-  -DCMAKE_CXX_COMPILER="clang++-${LLVM_VERSION}" \
-  -DCMAKE_C_COMPILER="clang-${LLVM_VERSION}" \
-  -DCMAKE_INSTALL_PREFIX="${MSAN_PREFIX}" \
-  -DLLVM_ENABLE_RUNTIMES="libunwind"
-
-cmake --build build_nomsan -j"${JOBS}"
-cmake --install build_nomsan
-
-VERSION="$(clang-"${LLVM_VERSION}" --version | head -n1 | sed -E 's/.*clang version ([0-9.]+).*/\1/')"
-TARBALL="msan-toolchain-${VERSION}.tar.zst"
-
-tar --zstd -cf "${TARBALL}" -C "${MSAN_PREFIX}" .
-
-TARBALL_PATH="$(pwd)/${TARBALL}"
-echo "Created: ${TARBALL_PATH}"

conflict_set.py

@@ -27,37 +27,23 @@ class Result(enum.Enum):
     TOO_OLD = 2
 
 
-def _make_key(buf: bytes) -> tuple[_Key, bytearray]:
-    """Create a _Key and a backing bytearray that must be kept alive."""
-    backing = bytearray(buf)
-    array = (ctypes.c_ubyte * len(backing)).from_buffer(backing)
-    return _Key(array, len(array)), backing
-
-
 def write(begin: bytes, end: Optional[bytes] = None) -> WriteRange:
-    begin_key, begin_buf = _make_key(begin)
+    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
+
     if end is None:
-        end_key = _Key((ctypes.c_ubyte * 0)(), 0)
+        e = (ctypes.c_ubyte * 0)()
-        end_buf = None
     else:
-        end_key, end_buf = _make_key(end)
+        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
-    result = WriteRange(begin_key, end_key)
+    return WriteRange(_Key(b, len(b)), _Key(e, len(e)))
-    result._begin_buf = begin_buf
-    result._end_buf = end_buf
-    return result
 
 
 def read(version: int, begin: bytes, end: Optional[bytes] = None) -> ReadRange:
-    begin_key, begin_buf = _make_key(begin)
+    b = (ctypes.c_ubyte * len(begin)).from_buffer(bytearray(begin))
     if end is None:
-        end_key = _Key((ctypes.c_ubyte * 0)(), 0)
+        e = (ctypes.c_ubyte * 0)()
-        end_buf = None
     else:
-        end_key, end_buf = _make_key(end)
+        e = (ctypes.c_ubyte * len(end)).from_buffer(bytearray(end))
-    result = ReadRange(begin_key, end_key, version)
+    return ReadRange(_Key(b, len(b)), _Key(e, len(e)), version)
-    result._begin_buf = begin_buf
-    result._end_buf = end_buf
-    return result
 
 
 class ConflictSet:
@@ -102,7 +88,6 @@ class ConflictSet:
             ctypes.POINTER(ctypes.c_int),
             ctypes.c_int,
         )
-        self._lib.ConflictSet_check.restype = None
 
         self._lib.ConflictSet_addWrites.argtypes = (
             ctypes.c_void_p,
@@ -110,16 +95,13 @@ class ConflictSet:
             ctypes.c_int,
             ctypes.c_int64,
         )
-        self._lib.ConflictSet_addWrites.restype = None
 
         self._lib.ConflictSet_setOldestVersion.argtypes = (
             ctypes.c_void_p,
             ctypes.c_int64,
         )
-        self._lib.ConflictSet_setOldestVersion.restype = None
 
         self._lib.ConflictSet_destroy.argtypes = (ctypes.c_void_p,)
-        self._lib.ConflictSet_destroy.restype = None
 
         self._lib.ConflictSet_getBytes.argtypes = (ctypes.c_void_p,)
         self._lib.ConflictSet_getBytes.restype = ctypes.c_int64