Add cycles_total

If numChildren + entryPresent is enough, we don't have to copy even if
it would fit in a smaller node.

If we have to copy, we might as well use the smallest acceptable node
type.

Bench.cpp

@@ -358,30 +358,28 @@ void benchWorstCaseForRadixRangeRead() {
   // }
 }
 
-void benchMetrics() {
-  ankerl::nanobench::Bench bench;
-  ConflictSet cs{0};
-
-  int count;
-  ConflictSet::MetricsV1 *m;
-  cs.getMetricsV1(&m, &count);
-  bench.batch(count);
-  bench.run("fetch metric", [&]() {
-    for (int i = 0; i < count; ++i) {
-      m[i].getValue();
-    }
-  });
-}
-
 void benchCreateAndDestroy() {
   ankerl::nanobench::Bench bench;
 
-  bench.run("create and destroy", [&]() { ConflictSet cs{0}; });
+  bench.run("create and destroy", [&]() {
+    ConflictSet cs{0};
+    ConflictSet::WriteRange w;
+    uint8_t b[9];
+    b[8] = 0;
+    for (int64_t i = 0; i < 1000; i += 7) {
+      auto x = __builtin_bswap64(i);
+      memcpy(b, &x, 8);
+      w.begin.p = b;
+      w.begin.len = 8;
+      w.end.len = 0;
+      w.end.p = b;
+      cs.addWrites(&w, 1, 1);
+    }
+  });
 }
 
 int main(void) {
   benchConflictSet();
   benchWorstCaseForRadixRangeRead();
-  benchMetrics();
   benchCreateAndDestroy();
 }

CMakeLists.txt

@@ -1,7 +1,7 @@
 cmake_minimum_required(VERSION 3.18)
 project(
   conflict-set
-  VERSION 0.0.9
+  VERSION 0.0.12
   DESCRIPTION
     "A data structure for optimistic concurrency control on ranges of bitwise-lexicographically-ordered keys."
   HOMEPAGE_URL "https://git.weaselab.dev/weaselab/conflict-set"
@@ -19,12 +19,6 @@ include(CheckCXXSourceCompiles)
 
 set(DEFAULT_BUILD_TYPE "Release")
 
-if(EMSCRIPTEN OR CMAKE_SYSTEM_NAME STREQUAL WASI)
-  set(WASM ON)
-else()
-  set(WASM OFF)
-endif()
-
 if(NOT CMAKE_BUILD_TYPE AND NOT CMAKE_CONFIGURATION_TYPES)
   message(
     STATUS
@@ -39,6 +33,10 @@ endif()
 
 add_compile_options(-fdata-sections -ffunction-sections -Wswitch-enum
                     -Werror=switch-enum -fPIC)
+if(NOT APPLE)
+  # This causes some versions of clang to crash on macos
+  add_compile_options(-g -fno-omit-frame-pointer)
+endif()
 
 set(full_relro_flags "-pie;LINKER:-z,relro,-z,now,-z,noexecstack")
 cmake_push_check_state()
@@ -81,15 +79,6 @@ if(EMSCRIPTEN)
 endif()
 
 if(NOT USE_SIMD_FALLBACK)
-  cmake_push_check_state()
-  list(APPEND CMAKE_REQUIRED_FLAGS -msimd128)
-  check_include_file_cxx("wasm_simd128.h" HAS_WASM_SIMD)
-  if(HAS_WASM_SIMD)
-    add_compile_options(-msimd128)
-    add_compile_definitions(HAS_WASM_SIMD)
-  endif()
-  cmake_pop_check_state()
-
   cmake_push_check_state()
   list(APPEND CMAKE_REQUIRED_FLAGS -mavx)
   check_include_file_cxx("immintrin.h" HAS_AVX)
@@ -162,47 +151,40 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
 
   file(GLOB CORPUS_TESTS ${CMAKE_CURRENT_SOURCE_DIR}/corpus/*)
 
-  # extra testing that relies on shared libraries, which aren't available with
-  # wasm
-  if(NOT WASM)
-    # Shared library version of FoundationDB's skip list implementation
-    add_library(skip_list SHARED SkipList.cpp)
-    target_compile_options(skip_list PRIVATE -fno-exceptions
-                                             -fvisibility=hidden)
-    target_include_directories(skip_list
-                               PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include)
-    set_target_properties(
-      skip_list PROPERTIES LIBRARY_OUTPUT_DIRECTORY
-                           "${CMAKE_CURRENT_BINARY_DIR}/skip_list")
-    set_target_properties(skip_list PROPERTIES OUTPUT_NAME ${PROJECT_NAME})
-    set_target_properties(
-      skip_list PROPERTIES VERSION ${PROJECT_VERSION} SOVERSION
-                                                      ${PROJECT_VERSION_MAJOR})
+  # Shared library version of FoundationDB's skip list implementation
+  add_library(skip_list SHARED SkipList.cpp)
+  target_compile_options(skip_list PRIVATE -fno-exceptions -fvisibility=hidden)
+  target_include_directories(skip_list
+                             PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include)
+  set_target_properties(
+    skip_list PROPERTIES LIBRARY_OUTPUT_DIRECTORY
+                         "${CMAKE_CURRENT_BINARY_DIR}/skip_list")
+  set_target_properties(skip_list PROPERTIES OUTPUT_NAME ${PROJECT_NAME})
+  set_target_properties(skip_list PROPERTIES VERSION ${PROJECT_VERSION}
+                                             SOVERSION ${PROJECT_VERSION_MAJOR})
 
-    # Shared library version of a std::unordered_map-based conflict set (point
-    # queries only)
-    add_library(hash_table SHARED HashTable.cpp)
-    target_compile_options(hash_table PRIVATE -fno-exceptions
-                                              -fvisibility=hidden)
-    target_include_directories(hash_table
-                               PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include)
-    set_target_properties(
-      hash_table PROPERTIES LIBRARY_OUTPUT_DIRECTORY
-                            "${CMAKE_CURRENT_BINARY_DIR}/hash_table")
-    set_target_properties(hash_table PROPERTIES OUTPUT_NAME ${PROJECT_NAME})
-    set_target_properties(
-      hash_table PROPERTIES VERSION ${PROJECT_VERSION} SOVERSION
-                                                       ${PROJECT_VERSION_MAJOR})
+  # Shared library version of a std::unordered_map-based conflict set (point
+  # queries only)
+  add_library(hash_table SHARED HashTable.cpp)
+  target_compile_options(hash_table PRIVATE -fno-exceptions -fvisibility=hidden)
+  target_include_directories(hash_table
+                             PUBLIC ${CMAKE_CURRENT_SOURCE_DIR}/include)
+  set_target_properties(
+    hash_table PROPERTIES LIBRARY_OUTPUT_DIRECTORY
+                          "${CMAKE_CURRENT_BINARY_DIR}/hash_table")
+  set_target_properties(hash_table PROPERTIES OUTPUT_NAME ${PROJECT_NAME})
+  set_target_properties(
+    hash_table PROPERTIES VERSION ${PROJECT_VERSION} SOVERSION
+                                                     ${PROJECT_VERSION_MAJOR})
 
-    add_executable(driver_skip_list TestDriver.cpp)
-    target_compile_options(driver_skip_list PRIVATE ${TEST_FLAGS})
-    target_link_libraries(driver_skip_list PRIVATE skip_list)
+  add_executable(driver_skip_list TestDriver.cpp)
+  target_compile_options(driver_skip_list PRIVATE ${TEST_FLAGS})
+  target_link_libraries(driver_skip_list PRIVATE skip_list)
 
-    foreach(TEST ${CORPUS_TESTS})
-      get_filename_component(hash ${TEST} NAME)
-      add_test(NAME skip_list_${hash} COMMAND driver_skip_list ${TEST})
-    endforeach()
-  endif()
+  foreach(TEST ${CORPUS_TESTS})
+    get_filename_component(hash ${TEST} NAME)
+    add_test(NAME skip_list_${hash} COMMAND driver_skip_list ${TEST})
+  endforeach()
 
   # ad hoc testing
   add_executable(conflict_set_main ConflictSet.cpp)
@@ -320,7 +302,7 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   add_test(NAME conflict_set_cxx_api_test COMMAND conflict_set_cxx_api_test)
 
   # symbol visibility tests
-  if(NOT WASM AND NOT CMAKE_BUILD_TYPE STREQUAL Debug)
+  if(NOT CMAKE_BUILD_TYPE STREQUAL Debug)
     if(APPLE)
       set(symbol_exports ${CMAKE_CURRENT_SOURCE_DIR}/apple-symbol-exports.txt)
       set(symbol_imports ${CMAKE_CURRENT_SOURCE_DIR}/apple-symbol-imports.txt)
@@ -358,6 +340,11 @@ if(CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR AND BUILD_TESTING)
   add_executable(driver_perf TestDriver.cpp)
   target_compile_definitions(driver_perf PRIVATE PERF_TEST=1)
   target_link_libraries(driver_perf PRIVATE ${PROJECT_NAME})
+
+  # server bench
+  add_executable(server_bench ServerBench.cpp)
+  target_link_libraries(server_bench PRIVATE ${PROJECT_NAME})
+  set_target_properties(server_bench PROPERTIES SKIP_BUILD_RPATH ON)
 endif()
 
 # packaging

HashTable.cpp

@@ -98,6 +98,13 @@ void ConflictSet::setOldestVersion(int64_t oldestVersion) {
 
 int64_t ConflictSet::getBytes() const { return -1; }
 
+void ConflictSet::getMetricsV1(MetricsV1 **metrics, int *count) const {
+  *metrics = nullptr;
+  *count = 0;
+}
+
+double ConflictSet::MetricsV1::getValue() const { return 0; }
+
 ConflictSet::ConflictSet(int64_t oldestVersion)
     : impl(new(safe_malloc(sizeof(Impl))) Impl{oldestVersion}) {}
 

Internal.h

@@ -273,6 +273,16 @@ template <class T> struct Vector {
     size_ += slice.size();
   }
 
+  // Caller must write to the returned slice
+  std::span<T> unsafePrepareAppend(int appendSize) {
+    if (size_ + appendSize > capacity) {
+      grow(std::max<int>(size_ + appendSize, capacity * 2));
+    }
+    auto result = std::span<T>(t + size_, appendSize);
+    size_ += appendSize;
+    return result;
+  }
+
   void push_back(const T &t) { append(std::span<const T>(&t, 1)); }
 
   T *begin() { return t; }
@@ -580,11 +590,14 @@ namespace {
 
 template <class ConflictSetImpl, bool kEnableAssertions = true>
 struct TestDriver {
-  Arbitrary arbitrary;
-  explicit TestDriver(const uint8_t *data, size_t size)
-      : arbitrary({data, size}) {}
+  Arbitrary *arbitrary;
+  explicit TestDriver(Arbitrary &a) : arbitrary(&a) {
+#if DEBUG_VERBOSE && !defined(NDEBUG)
+    fprintf(stderr, "%p Initial version: {%" PRId64 "}\n", this, writeVersion);
+#endif
+  }
 
-  int64_t oldestVersion = arbitrary.next();
+  int64_t oldestVersion = arbitrary->next();
   int64_t writeVersion = oldestVersion;
   ConflictSetImpl cs{oldestVersion};
   ReferenceImpl refImpl{oldestVersion};
@@ -593,33 +606,34 @@ struct TestDriver {
 
   bool ok = true;
 
-  const int prefixLen = arbitrary.bounded(512);
-  const int prefixByte = arbitrary.randT<uint8_t>();
+  const int prefixLen = arbitrary->bounded(512);
+  const int prefixByte = arbitrary->randT<uint8_t>();
 
   // Call until it returns true, for "done". Check internal invariants etc
   // between calls to next.
   bool next() {
     assert(cs.getBytes() >= 0);
-    if (!arbitrary.hasEntropy()) {
+    if (!arbitrary->hasEntropy()) {
       return true;
     }
     Arena arena;
     {
-      int numPointWrites = arbitrary.bounded(100);
-      int numRangeWrites = arbitrary.bounded(100);
-      int64_t v = (writeVersion += arbitrary.bounded(10) ? arbitrary.bounded(10)
-                                                         : arbitrary.next());
+      int numPointWrites = arbitrary->bounded(100);
+      int numRangeWrites = arbitrary->bounded(100);
+      int64_t v =
+          (writeVersion +=
+           arbitrary->bounded(10) ? arbitrary->bounded(10) : arbitrary->next());
       auto *writes =
           new (arena) ConflictSet::WriteRange[numPointWrites + numRangeWrites];
       auto keys = set<std::string_view>(arena);
       while (int(keys.size()) < numPointWrites + numRangeWrites * 2) {
-        if (!arbitrary.hasEntropy()) {
+        if (!arbitrary->hasEntropy()) {
           return true;
         }
-        int keyLen = prefixLen + arbitrary.bounded(kMaxKeySuffixLen);
+        int keyLen = prefixLen + arbitrary->bounded(kMaxKeySuffixLen);
         auto *begin = new (arena) uint8_t[keyLen];
         memset(begin, prefixByte, prefixLen);
-        arbitrary.randomBytes(begin + prefixLen, keyLen - prefixLen);
+        arbitrary->randomBytes(begin + prefixLen, keyLen - prefixLen);
         keys.insert(std::string_view((const char *)begin, keyLen));
       }
 
@@ -629,7 +643,7 @@ struct TestDriver {
                rangesRemaining = numRangeWrites;
            pointsRemaining > 0 || rangesRemaining > 0; ++i) {
         bool pointRead = pointsRemaining > 0 && rangesRemaining > 0
-                             ? bool(arbitrary.bounded(2))
+                             ? bool(arbitrary->bounded(2))
                              : pointsRemaining > 0;
         if (pointRead) {
           assert(pointsRemaining > 0);
@@ -648,33 +662,20 @@ struct TestDriver {
           ++iter;
           --rangesRemaining;
         }
-#if DEBUG_VERBOSE && !defined(NDEBUG)
-        if (writes[i].end.len == 0) {
-          fprintf(stderr, "Write: {%s}\n", printable(writes[i].begin).c_str());
-        } else {
-          fprintf(stderr, "Write: [%s, %s)\n",
-                  printable(writes[i].begin).c_str(),
-                  printable(writes[i].end).c_str());
-        }
-#endif
       }
       assert(iter == keys.end());
       assert(i == numPointWrites + numRangeWrites);
 
-#if DEBUG_VERBOSE && !defined(NDEBUG)
-      fprintf(stderr, "Write @ %" PRId64 "\n", v);
-#endif
-
       // Test non-canonical writes
       if (numPointWrites > 0) {
-        int overlaps = arbitrary.bounded(numPointWrites);
+        int overlaps = arbitrary->bounded(numPointWrites);
         for (int i = 0; i < numPointWrites + numRangeWrites && overlaps > 0;
              ++i) {
           if (writes[i].end.len == 0) {
-            int keyLen = prefixLen + arbitrary.bounded(kMaxKeySuffixLen);
+            int keyLen = prefixLen + arbitrary->bounded(kMaxKeySuffixLen);
             auto *begin = new (arena) uint8_t[keyLen];
             memset(begin, prefixByte, prefixLen);
-            arbitrary.randomBytes(begin + prefixLen, keyLen - prefixLen);
+            arbitrary->randomBytes(begin + prefixLen, keyLen - prefixLen);
             writes[i].end.len = keyLen;
             writes[i].end.p = begin;
             auto c =
@@ -692,10 +693,10 @@ struct TestDriver {
           }
         }
       }
-      if (arbitrary.bounded(2)) {
+      if (arbitrary->bounded(2)) {
         // Shuffle writes
         for (int i = numPointWrites + numRangeWrites - 1; i > 0; --i) {
-          int j = arbitrary.bounded(i + 1);
+          int j = arbitrary->bounded(i + 1);
           if (i != j) {
             using std::swap;
             swap(writes[i], writes[j]);
@@ -704,7 +705,7 @@ struct TestDriver {
       }
 
       oldestVersion +=
-          arbitrary.bounded(10) ? arbitrary.bounded(10) : arbitrary.next();
+          arbitrary->bounded(10) ? arbitrary->bounded(10) : arbitrary->next();
       oldestVersion = std::min(oldestVersion, writeVersion);
 
 #ifdef THREAD_TEST
@@ -721,6 +722,20 @@ struct TestDriver {
       ready.wait();
 #endif
 
+#if DEBUG_VERBOSE && !defined(NDEBUG)
+      for (int i = 0; i < numPointWrites + numRangeWrites; ++i) {
+        if (writes[i].end.len == 0) {
+          fprintf(stderr, "%p Write: {%s}\n", this,
+                  printable(writes[i].begin).c_str());
+        } else {
+          fprintf(stderr, "%p Write: [%s, %s)\n", this,
+                  printable(writes[i].begin).c_str(),
+                  printable(writes[i].end).c_str());
+        }
+      }
+      fprintf(stderr, "%p Write @ %" PRId64 "\n", this, v);
+#endif
+
       CALLGRIND_START_INSTRUMENTATION;
       cs.addWrites(writes, numPointWrites + numRangeWrites, v);
       CALLGRIND_STOP_INSTRUMENTATION;
@@ -729,6 +744,10 @@ struct TestDriver {
         refImpl.addWrites(writes, numPointWrites + numRangeWrites, v);
       }
 
+#if DEBUG_VERBOSE && !defined(NDEBUG)
+      fprintf(stderr, "%p Set oldest version: %" PRId64 "\n", this,
+              oldestVersion);
+#endif
       cs.setOldestVersion(oldestVersion);
       if constexpr (kEnableAssertions) {
         refImpl.setOldestVersion(oldestVersion);
@@ -739,24 +758,24 @@ struct TestDriver {
 #endif
     }
     {
-      int numPointReads = arbitrary.bounded(100);
-      int numRangeReads = arbitrary.bounded(100);
+      int numPointReads = arbitrary->bounded(100);
+      int numRangeReads = arbitrary->bounded(100);
 
-      int64_t v = std::max<int64_t>(writeVersion - (arbitrary.bounded(10)
-                                                        ? arbitrary.bounded(10)
-                                                        : arbitrary.next()),
+      int64_t v = std::max<int64_t>(writeVersion - (arbitrary->bounded(10)
+                                                        ? arbitrary->bounded(10)
+                                                        : arbitrary->next()),
                                     0);
       auto *reads =
           new (arena) ConflictSet::ReadRange[numPointReads + numRangeReads];
       auto keys = set<std::string_view>(arena);
       while (int(keys.size()) < numPointReads + numRangeReads * 2) {
-        if (!arbitrary.hasEntropy()) {
+        if (!arbitrary->hasEntropy()) {
           return true;
         }
-        int keyLen = prefixLen + arbitrary.bounded(kMaxKeySuffixLen);
+        int keyLen = prefixLen + arbitrary->bounded(kMaxKeySuffixLen);
         auto *begin = new (arena) uint8_t[keyLen];
         memset(begin, prefixByte, prefixLen);
-        arbitrary.randomBytes(begin + prefixLen, keyLen - prefixLen);
+        arbitrary->randomBytes(begin + prefixLen, keyLen - prefixLen);
         keys.insert(std::string_view((const char *)begin, keyLen));
       }
 
@@ -765,7 +784,7 @@ struct TestDriver {
       for (int pointsRemaining = numPointReads, rangesRemaining = numRangeReads;
            pointsRemaining > 0 || rangesRemaining > 0; ++i) {
         bool pointRead = pointsRemaining > 0 && rangesRemaining > 0
-                             ? bool(arbitrary.bounded(2))
+                             ? bool(arbitrary->bounded(2))
                              : pointsRemaining > 0;
         if (pointRead) {
           assert(pointsRemaining > 0);
@@ -787,10 +806,10 @@ struct TestDriver {
         reads[i].readVersion = v;
 #if DEBUG_VERBOSE && !defined(NDEBUG)
         if (reads[i].end.len == 0) {
-          fprintf(stderr, "Read: {%s} @ %" PRId64 "\n",
+          fprintf(stderr, "%p Read: {%s} @ %" PRId64 "\n", this,
                   printable(reads[i].begin).c_str(), reads[i].readVersion);
         } else {
-          fprintf(stderr, "Read: [%s, %s) @ %" PRId64 "\n",
+          fprintf(stderr, "%p Read: [%s, %s) @ %" PRId64 "\n", this,
                   printable(reads[i].begin).c_str(),
                   printable(reads[i].end).c_str(), reads[i].readVersion);
         }
@@ -839,24 +858,27 @@ struct TestDriver {
         refImpl.check(reads, results2, numPointReads + numRangeReads);
       }
 
-      auto compareResults = [reads](ConflictSet::Result *results1,
-                                    ConflictSet::Result *results2, int count) {
+      auto compareResults = [reads, this](ConflictSet::Result *results1,
+                                          ConflictSet::Result *results2,
+                                          int count) {
         for (int i = 0; i < count; ++i) {
           if (results1[i] != results2[i]) {
             if (reads[i].end.len == 0) {
-              fprintf(stderr,
-                      "Expected %s, got %s for read of {%s} at version %" PRId64
-                      "\n",
-                      resultToStr(results2[i]), resultToStr(results1[i]),
-                      printable(reads[i].begin).c_str(), reads[i].readVersion);
-            } else {
               fprintf(
                   stderr,
-                  "Expected %s, got %s for read of [%s, %s) at version %" PRId64
+                  "%p Expected %s, got %s for read of {%s} at version %" PRId64
                   "\n",
-                  resultToStr(results2[i]), resultToStr(results1[i]),
-                  printable(reads[i].begin).c_str(),
-                  printable(reads[i].end).c_str(), reads[i].readVersion);
+                  (void *)this, resultToStr(results2[i]),
+                  resultToStr(results1[i]), printable(reads[i].begin).c_str(),
+                  reads[i].readVersion);
+            } else {
+              fprintf(stderr,
+                      "%p Expected %s, got %s for read of [%s, %s) at version "
+                      "%" PRId64 "\n",
+                      (void *)this, resultToStr(results2[i]),
+                      resultToStr(results1[i]),
+                      printable(reads[i].begin).c_str(),
+                      printable(reads[i].end).c_str(), reads[i].readVersion);
             }
             return false;
           }

Jenkinsfile

@@ -117,15 +117,18 @@ pipeline {
                 }
             }
             steps {
+                script {
+                    filter_args = "-f ConflictSet.cpp -f LongestCommonPrefix.h"
+                }
                 CleanBuildAndTest("-DCMAKE_C_COMPILER=gcc -DCMAKE_CXX_COMPILER=g++ -DCMAKE_C_FLAGS=--coverage -DCMAKE_CXX_FLAGS=--coverage -DCMAKE_BUILD_TYPE=Debug -DDISABLE_TSAN=ON")
-                sh '''
-                gcovr -f ConflictSet.cpp --cobertura > build/coverage.xml
-                '''
+                sh """
+                gcovr ${filter_args} --cobertura > build/coverage.xml
+                """
                 recordCoverage qualityGates: [[criticality: 'NOTE', metric: 'MODULE']], tools: [[parser: 'COBERTURA', pattern: 'build/coverage.xml']]
-                sh '''
-                # Suppress again, because we haven't dealt with function multi-versioning for x86 yet
-                # gcovr -f ConflictSet.cpp --fail-under-line 100 > /dev/null
-                '''
+                sh """
+                gcovr ${filter_args}
+                gcovr ${filter_args} --fail-under-line 100 > /dev/null
+                """
             }
         }
     }

LongestCommonPrefix.h

@@ -0,0 +1,185 @@
+#pragma once
+
+#include <assert.h>
+#include <bit>
+#include <stdint.h>
+#include <string.h>
+
+#ifdef HAS_AVX
+#include <immintrin.h>
+#elif HAS_ARM_NEON
+#include <arm_neon.h>
+#endif
+
+#ifndef __SANITIZE_THREAD__
+#if defined(__has_feature)
+#if __has_feature(thread_sanitizer)
+#define __SANITIZE_THREAD__
+#endif
+#endif
+#endif
+
+#if defined(HAS_AVX) || defined(HAS_ARM_NEON)
+constexpr int kStride = 64;
+#else
+constexpr int kStride = 16;
+#endif
+
+constexpr int kUnrollFactor = 4;
+
+inline bool compareStride(const uint8_t *ap, const uint8_t *bp) {
+#if defined(HAS_ARM_NEON)
+  static_assert(kStride == 64);
+  uint8x16_t x[4]; // GCOVR_EXCL_LINE
+  for (int i = 0; i < 4; ++i) {
+    x[i] = vceqq_u8(vld1q_u8(ap + i * 16), vld1q_u8(bp + i * 16));
+  }
+  auto results = vreinterpretq_u16_u8(
+      vandq_u8(vandq_u8(x[0], x[1]), vandq_u8(x[2], x[3])));
+  bool eq = vget_lane_u64(vreinterpret_u64_u8(vshrn_n_u16(results, 4)), 0) ==
+            uint64_t(-1);
+#elif defined(HAS_AVX)
+  static_assert(kStride == 64);
+  __m128i x[4]; // GCOVR_EXCL_LINE
+  for (int i = 0; i < 4; ++i) {
+    x[i] = _mm_cmpeq_epi8(_mm_loadu_si128((__m128i *)(ap + i * 16)),
+                          _mm_loadu_si128((__m128i *)(bp + i * 16)));
+  }
+  auto eq =
+      _mm_movemask_epi8(_mm_and_si128(_mm_and_si128(x[0], x[1]),
+                                      _mm_and_si128(x[2], x[3]))) == 0xffff;
+#else
+  // Hope it gets vectorized
+  auto eq = memcmp(ap, bp, kStride) == 0;
+#endif
+  return eq;
+}
+
+// Precondition: ap[:kStride] != bp[:kStride]
+inline int firstNeqStride(const uint8_t *ap, const uint8_t *bp) {
+#if defined(HAS_AVX)
+  static_assert(kStride == 64);
+  uint64_t c[kStride / 16]; // GCOVR_EXCL_LINE
+  for (int i = 0; i < kStride; i += 16) {
+    const auto a = _mm_loadu_si128((__m128i *)(ap + i));
+    const auto b = _mm_loadu_si128((__m128i *)(bp + i));
+    const auto compared = _mm_cmpeq_epi8(a, b);
+    c[i / 16] = _mm_movemask_epi8(compared) & 0xffff;
+  }
+  return std::countr_zero(~(c[0] | c[1] << 16 | c[2] << 32 | c[3] << 48));
+#elif defined(HAS_ARM_NEON)
+  static_assert(kStride == 64);
+  for (int i = 0; i < kStride; i += 16) {
+    // 0xff for each match
+    uint16x8_t results =
+        vreinterpretq_u16_u8(vceqq_u8(vld1q_u8(ap + i), vld1q_u8(bp + i)));
+    // 0xf for each mismatch
+    uint64_t bitfield =
+        ~vget_lane_u64(vreinterpret_u64_u8(vshrn_n_u16(results, 4)), 0);
+    if (bitfield) {
+      return i + (std::countr_zero(bitfield) >> 2);
+    }
+  }
+  __builtin_unreachable(); // GCOVR_EXCL_LINE
+#else
+  int i = 0;
+  for (; i < kStride - 1; ++i) {
+    if (*ap++ != *bp++) {
+      break;
+    }
+  }
+  return i;
+#endif
+}
+
+// This gets covered in local development
+// GCOVR_EXCL_START
+#if defined(HAS_AVX) && !defined(__SANITIZE_THREAD__)
+__attribute__((target("avx512f,avx512bw"))) inline int
+longestCommonPrefix(const uint8_t *ap, const uint8_t *bp, int cl) {
+  int i = 0;
+  int end = cl & ~63;
+  while (i < end) {
+    const uint64_t eq =
+        _mm512_cmpeq_epi8_mask(_mm512_loadu_epi8(ap), _mm512_loadu_epi8(bp));
+    if (eq != uint64_t(-1)) {
+      return i + std::countr_one(eq);
+    }
+    i += 64;
+    ap += 64;
+    bp += 64;
+  }
+  if (i < cl) {
+    const uint64_t mask = (uint64_t(1) << (cl - i)) - 1;
+    const uint64_t eq = _mm512_cmpeq_epi8_mask(
+        _mm512_maskz_loadu_epi8(mask, ap), _mm512_maskz_loadu_epi8(mask, bp));
+    return i + std::countr_one(eq & mask);
+  }
+  assert(i == cl);
+  return i;
+}
+__attribute__((target("default")))
+#endif
+// GCOVR_EXCL_STOP
+
+inline int
+longestCommonPrefix(const uint8_t *ap, const uint8_t *bp, int cl) {
+  if (!(cl >= 0)) {
+    __builtin_unreachable(); // GCOVR_EXCL_LINE
+  }
+
+  int i = 0;
+  int end;
+
+  // kStride * kUnrollCount at a time
+  end = cl & ~(kStride * kUnrollFactor - 1);
+  while (i < end) {
+    for (int j = 0; j < kUnrollFactor; ++j) {
+      if (!compareStride(ap, bp)) {
+        return i + firstNeqStride(ap, bp);
+      }
+      i += kStride;
+      ap += kStride;
+      bp += kStride;
+    }
+  }
+
+  // kStride at a time
+  end = cl & ~(kStride - 1);
+  while (i < end) {
+    if (!compareStride(ap, bp)) {
+      return i + firstNeqStride(ap, bp);
+    }
+    i += kStride;
+    ap += kStride;
+    bp += kStride;
+  }
+
+  // word at a time
+  end = cl & ~(sizeof(uint64_t) - 1);
+  while (i < end) {
+    uint64_t a; // GCOVR_EXCL_LINE
+    uint64_t b; // GCOVR_EXCL_LINE
+    memcpy(&a, ap, 8);
+    memcpy(&b, bp, 8);
+    const auto mismatched = a ^ b;
+    if (mismatched) {
+      return i + std::countr_zero(mismatched) / 8;
+    }
+    i += 8;
+    ap += 8;
+    bp += 8;
+  }
+
+  // byte at a time
+  while (i < cl) {
+    if (*ap != *bp) {
+      break;
+    }
+    ++ap;
+    ++bp;
+    ++i;
+  }
+
+  return i;
+}

README.md

@@ -1,86 +1,39 @@
 A data structure for optimistic concurrency control on ranges of bitwise-lexicographically-ordered keys.
 
-Intended to replace FoundationDB's skip list.
+Intended as an alternative to FoundationDB's skip list.
 
-Hardware for all benchmarks is a mac m1 2020.
+Hardware for all benchmarks is an AMD Ryzen 9 7900 with (2x32GB) 5600MT/s CL28-34-34-89 1.35V RAM
 
-# FoundationDB's benchmark
+# Microbenchmark
 
 ## Skip list
 
-```
-New conflict set: 1.957 sec
-                  0.639 Mtransactions/sec
-                  2.555 Mkeys/sec
-Detect only:      1.845 sec
-                  0.678 Mtransactions/sec
-                  2.710 Mkeys/sec
-Skiplist only:    1.263 sec
-                  0.990 Mtransactions/sec
-                  3.960 Mkeys/sec
-Performance counters:
-               Build: 0.0546
-                 Add: 0.0563
-              Detect: 1.84
-              D.Sort: 0.412
-           D.Combine: 0.0141
-         D.CheckRead: 0.671
-   D.CheckIntraBatch: 0.0068
-        D.MergeWrite: 0.592
-      D.RemoveBefore: 0.146
-```
+|               ns/op |                op/s |    err% |          ins/op |          cyc/op |    IPC |         bra/op |   miss% |     total | benchmark
+|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
+|              172.03 |        5,812,791.77 |    0.4% |        3,130.62 |          879.00 |  3.562 |         509.23 |    0.0% |      0.01 | `point reads`
+|              167.44 |        5,972,130.71 |    0.2% |        3,065.14 |          862.27 |  3.555 |         494.30 |    0.0% |      0.01 | `prefix reads`
+|              238.77 |        4,188,130.84 |    0.9% |        3,589.93 |        1,259.30 |  2.851 |         637.12 |    0.0% |      0.01 | `range reads`
+|              424.01 |        2,358,426.70 |    0.2% |        5,620.05 |        2,242.35 |  2.506 |         854.80 |    1.7% |      0.01 | `point writes`
+|              418.45 |        2,389,780.56 |    0.4% |        5,525.07 |        2,211.05 |  2.499 |         831.71 |    1.7% |      0.01 | `prefix writes`
+|              254.87 |        3,923,568.88 |    2.6% |        3,187.01 |        1,366.50 |  2.332 |         529.11 |    2.7% |      0.02 | `range writes`
+|              675.96 |        1,479,374.50 |    3.3% |        7,735.41 |        3,468.60 |  2.230 |       1,386.02 |    1.8% |      0.01 | `monotonic increasing point writes`
+|          137,986.20 |            7,247.10 |    0.6% |      789,752.33 |      699,462.00 |  1.129 |     144,824.14 |    0.0% |      0.01 | `worst case for radix tree`
+|               21.63 |       46,231,564.03 |    1.0% |          448.00 |          107.14 |  4.181 |          84.00 |    0.0% |      0.01 | `create and destroy`
 
 ## Radix tree (this implementation)
 
-```
-New conflict set: 1.366 sec
-                  0.915 Mtransactions/sec
-                  3.660 Mkeys/sec
-Detect only:      1.248 sec
-                  1.002 Mtransactions/sec
-                  4.007 Mkeys/sec
-Skiplist only:    0.573 sec
-                  2.182 Mtransactions/sec
-                  8.730 Mkeys/sec
-Performance counters:
-               Build: 0.0594
-                 Add: 0.0572
-              Detect: 1.25
-              D.Sort: 0.418
-           D.Combine: 0.0149
-         D.CheckRead: 0.232
-   D.CheckIntraBatch: 0.0067
-        D.MergeWrite: 0.341
-      D.RemoveBefore: 0.232
-```
+|               ns/op |                op/s |    err% |          ins/op |          cyc/op |    IPC |         bra/op |   miss% |     total | benchmark
+|--------------------:|--------------------:|--------:|----------------:|----------------:|-------:|---------------:|--------:|----------:|:----------
+|               10.80 |       92,600,541.52 |    0.6% |          180.38 |           54.49 |  3.310 |          41.51 |    0.4% |      0.01 | `point reads`
+|               15.00 |       66,687,691.68 |    0.4% |          278.44 |           76.44 |  3.642 |          55.56 |    0.3% |      0.01 | `prefix reads`
+|               36.81 |       27,163,394.61 |    0.4% |          795.06 |          187.91 |  4.231 |         142.67 |    0.2% |      0.01 | `range reads`
+|               18.14 |       55,137,674.01 |    1.2% |          338.19 |           92.86 |  3.642 |          42.81 |    0.4% |      0.01 | `point writes`
+|               33.19 |       30,127,119.71 |    0.1% |          681.03 |          170.05 |  4.005 |          98.68 |    0.2% |      0.01 | `prefix writes`
+|               37.37 |       26,759,432.70 |    1.9% |          779.70 |          195.45 |  3.989 |         114.21 |    0.0% |      0.01 | `range writes`
+|               74.36 |       13,448,582.47 |    1.9% |        1,425.68 |          389.08 |  3.664 |         258.88 |    0.1% |      0.01 | `monotonic increasing point writes`
+|          316,928.00 |            3,155.29 |    1.5% |    3,992,986.00 |    1,699,813.00 |  2.349 |     806,226.50 |    0.0% |      0.01 | `worst case for radix tree`
+|               75.26 |       13,286,517.16 |    0.5% |        1,590.01 |          386.67 |  4.112 |         258.00 |    0.0% |      0.01 | `create and destroy`
 
-# Our benchmark
-
-## Skip list
-
-|               ns/op |                op/s |    err% |     total | benchmark
-|--------------------:|--------------------:|--------:|----------:|:----------
-|              245.99 |        4,065,232.81 |    0.3% |      0.01 | `point reads`
-|              265.93 |        3,760,430.49 |    0.2% |      0.01 | `prefix reads`
-|              485.30 |        2,060,569.50 |    0.2% |      0.01 | `range reads`
-|              449.60 |        2,224,195.17 |    0.4% |      0.01 | `point writes`
-|              441.76 |        2,263,688.18 |    1.1% |      0.01 | `prefix writes`
-|              245.42 |        4,074,647.54 |    2.4% |      0.02 | `range writes`
-|              572.80 |        1,745,810.06 |    1.3% |      0.01 | `monotonic increasing point writes`
-|          154,819.33 |            6,459.14 |    0.9% |      0.01 | `worst case for radix tree`
-
-## Radix tree (this implementation)
-
-|               ns/op |                op/s |    err% |     total | benchmark
-|--------------------:|--------------------:|--------:|----------:|:----------
-|               19.17 |       52,163,930.66 |    0.1% |      0.01 | `point reads`
-|               23.68 |       42,224,388.21 |    0.7% |      0.01 | `prefix reads`
-|               63.30 |       15,797,506.06 |    0.9% |      0.01 | `range reads`
-|               29.66 |       33,720,994.74 |    0.3% |      0.01 | `point writes`
-|               43.50 |       22,987,781.25 |    1.0% |      0.01 | `prefix writes`
-|               50.00 |       20,000,000.00 |    0.8% |      0.01 | `range writes`
-|              103.25 |        9,684,786.47 |    2.9% |      0.01 | `monotonic increasing point writes`
-|        1,181,500.00 |              846.38 |    2.3% |      0.01 | `worst case for radix tree`
 
 # "Real data" test
 
@@ -89,13 +42,13 @@ Point queries only, best of three runs. Gc ratio is the ratio of time spent doin
 ## skip list
 
 ```
-Check: 11.3385 seconds, 329.718 MB/s, Add: 5.35612 seconds, 131.072 MB/s, Gc ratio: 45.7173%
+Check: 4.47891 seconds, 364.05 MB/s, Add: 4.55599 seconds, 123.058 MB/s, Gc ratio: 37.1145%
 ```
 
 ## radix tree
 
 ```
-Check: 2.48583 seconds, 1503.93 MB/s, Add: 2.12768 seconds, 329.954 MB/s, Gc ratio: 41.7943%
+Check: 0.910234 seconds, 1791.35 MB/s, Add: 1.25908 seconds, 445.287 MB/s, Gc ratio: 44.0415%
 ```
 
 ## hash table
@@ -103,5 +56,5 @@ Check: 2.48583 seconds, 1503.93 MB/s, Add: 2.12768 seconds, 329.954 MB/s, Gc rat
 (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: 1.83386 seconds, 2038.6 MB/s, Add: 0.601411 seconds, 1167.32 MB/s, Gc ratio: 48.9776%
+Check: 0.804094 seconds, 2027.81 MB/s, Add: 0.652952 seconds, 858.645 MB/s, Gc ratio: 35.3885%
 ```

ServerBench.cpp

@@ -0,0 +1,310 @@
+#include <atomic>
+#include <errno.h>
+#include <netdb.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <string>
+#include <string_view>
+#include <sys/resource.h>
+#include <sys/socket.h>
+#include <sys/uio.h>
+#include <thread>
+#include <unistd.h>
+
+#include "ConflictSet.h"
+#include "third_party/nadeau.h"
+
+std::atomic<int64_t> transactions;
+
+constexpr int kBaseSearchDepth = 32;
+constexpr int kWindowSize = 10000000;
+
+std::basic_string<uint8_t> numToKey(int64_t num) {
+  std::basic_string<uint8_t> result;
+  result.resize(kBaseSearchDepth + sizeof(int64_t));
+  memset(result.data(), 0, kBaseSearchDepth);
+  int64_t be = __builtin_bswap64(num);
+  memcpy(result.data() + kBaseSearchDepth, &be, sizeof(int64_t));
+  return result;
+}
+
+void workload(weaselab::ConflictSet *cs) {
+  int64_t version = kWindowSize;
+  cs->addWrites(nullptr, 0, version);
+  for (;; transactions.fetch_add(1, std::memory_order_relaxed)) {
+    // Reads
+    {
+      auto beginK = numToKey(version - kWindowSize);
+      auto endK = numToKey(version - 1);
+      auto pointRv = version - kWindowSize + rand() % kWindowSize + 1;
+      auto pointK = numToKey(pointRv);
+      weaselab::ConflictSet::ReadRange reads[] = {
+          {
+              {pointK.data(), int(pointK.size())},
+              {nullptr, 0},
+              pointRv,
+          },
+          {
+              {beginK.data(), int(beginK.size())},
+              {endK.data(), int(endK.size())},
+              version - 2,
+          },
+      };
+      weaselab::ConflictSet::Result result[sizeof(reads) / sizeof(reads[0])];
+      cs->check(reads, result, sizeof(reads) / sizeof(reads[0]));
+      // for (int i = 0; i < sizeof(reads) / sizeof(reads[0]); ++i) {
+      //   if (result[i] != weaselab::ConflictSet::Commit) {
+      //     fprintf(stderr, "Unexpected conflict: [%s, %s) @ %" PRId64 "\n",
+      //             printable(reads[i].begin).c_str(),
+      //             printable(reads[i].end).c_str(), reads[i].readVersion);
+      //     abort();
+      //   }
+      // }
+    }
+    // Writes
+    {
+      weaselab::ConflictSet::WriteRange w;
+      auto k = numToKey(version);
+      w.begin.p = k.data();
+      w.end.len = 0;
+      if (version % (kWindowSize / 2) == 0) {
+        for (int l = 0; l <= k.size(); ++l) {
+          w.begin.len = l;
+          cs->addWrites(&w, 1, version);
+        }
+      } else {
+        w.begin.len = k.size();
+        cs->addWrites(&w, 1, version);
+        int64_t beginN = version - kWindowSize + rand() % kWindowSize;
+        auto b = numToKey(beginN);
+        auto e = numToKey(beginN + 1000);
+        w.begin.p = b.data();
+        w.begin.len = b.size();
+        w.end.p = e.data();
+        w.end.len = e.size();
+        cs->addWrites(&w, 1, version);
+      }
+    }
+    // GC
+    cs->setOldestVersion(version - kWindowSize);
+    ++version;
+  }
+}
+
+// Adapted from getaddrinfo man page
+int getListenFd(const char *node, const char *service) {
+
+  struct addrinfo hints;
+  struct addrinfo *result, *rp;
+  int sfd, s;
+
+  memset(&hints, 0, sizeof(hints));
+  hints.ai_family = AF_UNSPEC;     /* Allow IPv4 or IPv6 */
+  hints.ai_socktype = SOCK_STREAM; /* stream socket */
+  hints.ai_flags = AI_PASSIVE;     /* For wildcard IP address */
+  hints.ai_protocol = 0;           /* Any protocol */
+  hints.ai_canonname = nullptr;
+  hints.ai_addr = nullptr;
+  hints.ai_next = nullptr;
+
+  s = getaddrinfo(node, service, &hints, &result);
+  if (s != 0) {
+    fprintf(stderr, "getaddrinfo: %s\n", gai_strerror(s));
+    abort();
+  }
+
+  /* getaddrinfo() returns a list of address structures.
+     Try each address until we successfully bind(2).
+     If socket(2) (or bind(2)) fails, we (close the socket
+     and) try the next address. */
+
+  for (rp = result; rp != nullptr; rp = rp->ai_next) {
+    sfd = socket(rp->ai_family, rp->ai_socktype, rp->ai_protocol);
+    if (sfd == -1) {
+      continue;
+    }
+
+    int val = 1;
+    setsockopt(sfd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val));
+
+    if (bind(sfd, rp->ai_addr, rp->ai_addrlen) == 0) {
+      break; /* Success */
+    }
+
+    close(sfd);
+  }
+
+  freeaddrinfo(result); /* No longer needed */
+
+  if (rp == nullptr) { /* No address succeeded */
+    fprintf(stderr, "Could not bind\n");
+    abort();
+  }
+
+  int rv = listen(sfd, SOMAXCONN);
+  if (rv) {
+    perror("listen()");
+    abort();
+  }
+
+  return sfd;
+}
+
+// HTTP response
+//
+std::string_view part1 =
+    "HTTP/1.1 200 OK \r\nContent-type: text/plain; version=0.0.4; "
+    "charset=utf-8; escaping=values\r\nContent-Length: ";
+// Decimal content length
+std::string_view part2 = "\r\n\r\n";
+// Body
+
+double toSeconds(timeval t) {
+  return double(t.tv_sec) + double(t.tv_usec) * 1e-6;
+}
+
+#include <linux/perf_event.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <sys/ioctl.h>
+#include <sys/syscall.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#ifdef __linux__
+struct PerfCounter {
+  explicit PerfCounter(int event) {
+    struct perf_event_attr pe;
+
+    memset(&pe, 0, sizeof(pe));
+    pe.type = PERF_TYPE_HARDWARE;
+    pe.size = sizeof(pe);
+    pe.config = event;
+    pe.inherit = 1;
+    pe.exclude_kernel = 1;
+    pe.exclude_hv = 1;
+
+    fd = perf_event_open(&pe, 0, -1, -1, 0);
+    if (fd == -1) {
+      fprintf(stderr, "Error opening leader %llx\n", pe.config);
+      exit(EXIT_FAILURE);
+    }
+  }
+
+  int64_t total() {
+    int64_t count;
+    if (read(fd, &count, sizeof(count)) != sizeof(count)) {
+      perror("read instructions from perf");
+      abort();
+    }
+    return count;
+  }
+
+  ~PerfCounter() { close(fd); }
+
+private:
+  int fd;
+  static long perf_event_open(struct perf_event_attr *hw_event, pid_t pid,
+                              int cpu, int group_fd, unsigned long flags) {
+    int ret;
+
+    ret = syscall(SYS_perf_event_open, hw_event, pid, cpu, group_fd, flags);
+    return ret;
+  }
+};
+#else
+struct PerfCounter {
+  explicit PerPerfCounter(int) {}
+  int64_t total() { return 0; }
+};
+#endif
+
+int main(int argc, char **argv) {
+  if (argc != 3) {
+    goto fail;
+  }
+  {
+    int listenFd = getListenFd(argv[1], argv[2]);
+
+    weaselab::ConflictSet cs{0};
+    weaselab::ConflictSet::MetricsV1 *metrics;
+    int metricsCount;
+    cs.getMetricsV1(&metrics, &metricsCount);
+
+    PerfCounter instructions{PERF_COUNT_HW_INSTRUCTIONS};
+    PerfCounter cycles{PERF_COUNT_HW_CPU_CYCLES};
+    auto w = std::thread{workload, &cs};
+
+    for (;;) {
+      struct sockaddr_storage peer_addr = {};
+      socklen_t peer_addr_len = sizeof(peer_addr);
+      const int connfd =
+          accept(listenFd, (struct sockaddr *)&peer_addr, &peer_addr_len);
+
+      std::string body;
+
+      rusage r;
+      getrusage(RUSAGE_SELF, &r);
+      body += "# HELP process_cpu_seconds_total Total user and system CPU time "
+              "spent in seconds.\n# TYPE process_cpu_seconds_total counter\n"
+              "process_cpu_seconds_total ";
+      body += std::to_string(toSeconds(r.ru_utime) + toSeconds(r.ru_stime));
+      body += "\n";
+      body += "# HELP process_resident_memory_bytes Resident memory size in "
+              "bytes.\n# TYPE process_resident_memory_bytes gauge\n"
+              "process_resident_memory_bytes ";
+      body += std::to_string(getCurrentRSS());
+      body += "\n";
+      body += "# HELP transactions_total Total number of transactions\n"
+              "# TYPE transactions_total counter\n"
+              "transactions_total ";
+      body += std::to_string(transactions.load(std::memory_order_relaxed));
+      body += "\n";
+      body += "# HELP instructions_total Total number of instructions\n"
+              "# TYPE instructions_total counter\n"
+              "instructions_total ";
+      body += std::to_string(instructions.total());
+      body += "\n";
+      body += "# HELP cycles_total Total number of cycles\n"
+              "# TYPE cycles_total counter\n"
+              "cycles_total ";
+      body += std::to_string(cycles.total());
+      body += "\n";
+
+      for (int i = 0; i < metricsCount; ++i) {
+        body += "# HELP ";
+        body += metrics[i].name;
+        body += " ";
+        body += metrics[i].help;
+        body += "\n";
+        body += "# TYPE ";
+        body += metrics[i].name;
+        body += " ";
+        body += metrics[i].type == metrics[i].Counter ? "counter" : "gauge";
+        body += "\n";
+        body += metrics[i].name;
+        body += " ";
+        body += std::to_string(metrics[i].getValue());
+        body += "\n";
+      }
+
+      auto len = std::to_string(body.size());
+      iovec iov[] = {
+          {(void *)part1.data(), part1.size()},
+          {(void *)len.data(), len.size()},
+          {(void *)part2.data(), part2.size()},
+          {(void *)body.data(), body.size()},
+      };
+      int written;
+      do {
+        written = writev(connfd, iov, sizeof(iov) / sizeof(iov[0]));
+      } while (written < 0 && errno == EINTR);
+      close(connfd);
+    }
+  }
+fail:
+  fprintf(stderr, "Expected ./%s <host> <port>\n", argv[0]);
+  return 1;
+}

SkipList.cpp

@@ -778,10 +778,6 @@ struct __attribute__((visibility("hidden"))) ConflictSet::Impl {
     for (int s = stripes - 1; s >= 0; s--) {
       for (int i = 0; i * 2 < ss; ++i) {
         const auto &w = combinedWriteConflictRanges[s * stripeSize / 2 + i];
-#if DEBUG_VERBOSE
-        printf("Write begin: %s\n", printable(w.begin).c_str());
-        fflush(stdout);
-#endif
         values[i * 2] = w.first;
         values[i * 2 + 1] = w.second;
         keyUpdates += 3;

TestDriver.cpp

@@ -13,25 +13,58 @@ int main(int argc, char **argv) {
     std::stringstream buffer;
     buffer << t.rdbuf();
     auto str = buffer.str();
-    TestDriver<ConflictSet, !PERF_TEST> driver{(const uint8_t *)str.data(),
-                                               str.size()};
-    while (!driver.next())
-      ;
-    if (!driver.ok) {
-      abort();
+    Arbitrary arbitrary({(const uint8_t *)str.data(), str.size()});
+    TestDriver<ConflictSet, !PERF_TEST> driver1{arbitrary};
+    TestDriver<ConflictSet, !PERF_TEST> driver2{arbitrary};
+    bool done1 = false;
+    bool done2 = false;
+    for (;;) {
+      if (!done1) {
+        done1 = driver1.next();
+        if (!driver1.ok) {
+          abort();
+        }
+      }
+      if (!done2) {
+        done2 = driver2.next();
+        if (!driver2.ok) {
+          abort();
+        }
+      }
+      if (done1 && done2) {
+        break;
+      }
     }
 
-    ConflictSet::MetricsV1 *metrics;
-    int metricsCount;
-    driver.cs.getMetricsV1(&metrics, &metricsCount);
-    printf("#################### METRICS FOR %s ####################\n",
-           argv[i]);
-    for (int i = 0; i < metricsCount; ++i) {
-      printf("# HELP %s %s\n", metrics[i].name, metrics[i].help);
-      printf("# TYPE %s %s\n", metrics[i].name,
-             metrics[i].type == metrics[i].Counter ? "counter" : "gauge");
-      printf("%s %g\n", metrics[i].name, metrics[i].getValue());
+    {
+      ConflictSet::MetricsV1 *metrics;
+      int metricsCount;
+      driver1.cs.getMetricsV1(&metrics, &metricsCount);
+      printf("#################### METRICS for ConflictSet 1 for %s "
+             "####################\n",
+             argv[i]);
+      for (int i = 0; i < metricsCount; ++i) {
+        printf("# HELP %s %s\n", metrics[i].name, metrics[i].help);
+        printf("# TYPE %s %s\n", metrics[i].name,
+               metrics[i].type == metrics[i].Counter ? "counter" : "gauge");
+        printf("%s %g\n", metrics[i].name, metrics[i].getValue());
+      }
+      puts("");
+    }
+    {
+      ConflictSet::MetricsV1 *metrics;
+      int metricsCount;
+      driver2.cs.getMetricsV1(&metrics, &metricsCount);
+      printf("#################### METRICS for ConflictSet 2 for %s "
+             "####################\n",
+             argv[i]);
+      for (int i = 0; i < metricsCount; ++i) {
+        printf("# HELP %s %s\n", metrics[i].name, metrics[i].help);
+        printf("# TYPE %s %s\n", metrics[i].name,
+               metrics[i].type == metrics[i].Counter ? "counter" : "gauge");
+        printf("%s %g\n", metrics[i].name, metrics[i].getValue());
+      }
+      puts("");
     }
-    puts("");
   }
 }

aarch64-symbol-imports.txt

@@ -1,4 +1,4 @@
-__aarch64_cas8_relax
+__aarch64_ldadd8_relax
 __getauxval@GLIBC_2.17
 __stack_chk_fail@GLIBC_2.17
 __stack_chk_guard@GLIBC_2.17