WIP: what if we don't check when getting capture

We still keep the error check in the function, but hide it in an
outlined function inside a .cpp file, to promote inlining of the
retrieval part.

In the future, we should explore two things

1) Skipping over the context retrieval here, allowing direct access.
   I currently do not see a way to do this while keeping the
   "greppability" of mutable vs immutable accesses that is there now,
   but it would help a lot when inlining is not enabled.
2) Removing the error check, to make the function trivially inlinable,
   and without branches.

**runtime difference**

| --------- | Debug | Release |
|:----------|------:|--------:|
| Slow path |  0.98 |    1.07 |
| Fast path |  1.04 |    1.08 |

We lost bit of performance on the assertion slow path in debug mode,
but together with the previous commit, it comes out at net zero.
For other combinations, we see 5-10% perf improvement across the
two commits.

.clang-tidy

@@ -78,5 +78,5 @@ WarningsAsErrors: >-
   readability-duplicate-include,
 HeaderFilterRegex: '.*\.(c|cxx|cpp)$'
 FormatStyle:     none
-CheckOptions: {}
+CheckOptions: []
 ...

CMakeLists.txt

@@ -20,6 +20,7 @@ cmake_dependent_option(CATCH_BUILD_TESTING "Build the SelfTest project" ON "CATC
 cmake_dependent_option(CATCH_BUILD_EXAMPLES "Build code examples" OFF "CATCH_DEVELOPMENT_BUILD" OFF)
 cmake_dependent_option(CATCH_BUILD_EXTRA_TESTS "Build extra tests" OFF "CATCH_DEVELOPMENT_BUILD" OFF)
 cmake_dependent_option(CATCH_BUILD_FUZZERS "Build fuzzers" OFF "CATCH_DEVELOPMENT_BUILD" OFF)
+cmake_dependent_option(CATCH_BUILD_BENCHMARKS "Build the benchmarks" OFF "CATCH_DEVELOPMENT_BUILD" OFF)
 cmake_dependent_option(CATCH_ENABLE_COVERAGE "Generate coverage for codecov.io" OFF "CATCH_DEVELOPMENT_BUILD" OFF)
 cmake_dependent_option(CATCH_ENABLE_WERROR "Enables Werror during build" ON "CATCH_DEVELOPMENT_BUILD" OFF)
 cmake_dependent_option(CATCH_BUILD_SURROGATES "Enable generating and building surrogate TUs for the main headers" OFF "CATCH_DEVELOPMENT_BUILD" OFF)
@@ -77,6 +78,11 @@ set(SELF_TEST_DIR ${CATCH_DIR}/tests/SelfTest)
 # We need to bring-in the variables defined there to this scope
 add_subdirectory(src)
 
+if (CATCH_BUILD_BENCHMARKS)
+  set(CMAKE_FOLDER "benchmarks")
+  add_subdirectory(benchmarks)
+endif()
+
 # Build tests only if requested
 if(BUILD_TESTING AND CATCH_BUILD_TESTING AND NOT_SUBPROJECT)
   find_package(Python3 REQUIRED COMPONENTS Interpreter)

CMakePresets.json

@@ -15,14 +15,23 @@
             }
         },
         {
-            "name": "all-tests",
+            "name": "most-tests",
             "inherits": "basic-tests",
             "displayName": "Full development build",
-            "description": "Enables development build with examples and ALL tests",
+            "description": "Enables development build with extended set of tests (still relatively cheap to build)",
             "cacheVariables": {
                 "CATCH_BUILD_EXAMPLES": "ON",
                 "CATCH_BUILD_EXTRA_TESTS": "ON",
                 "CATCH_BUILD_SURROGATES": "ON",
+                "CATCH_BUILD_BENCHMARKS": "ON"
+            }
+        },
+        {
+            "name": "all-tests",
+            "inherits": "most-tests",
+            "displayName": "Full development build",
+            "description": "Enables development build with examples and ALL tests",
+            "cacheVariables": {
                 "CATCH_ENABLE_CONFIGURE_TESTS": "ON",
                 "CATCH_ENABLE_CMAKE_HELPER_TESTS": "ON"
             }

benchmarks/CMakeLists.txt

@@ -0,0 +1,16 @@
+include(CatchMiscFunctions)
+
+add_executable(AssertionsFastPath
+  runtime_assertion_benches.cpp
+)
+
+add_executable(AssertionsSlowPath
+  runtime_assertion_benches.cpp
+  assertion_listener.cpp
+)
+
+target_link_libraries(AssertionsFastPath PRIVATE Catch2::Catch2WithMain)
+target_link_libraries(AssertionsSlowPath PRIVATE Catch2::Catch2WithMain)
+
+list(APPEND CATCH_TEST_TARGETS AssertionsFastPath AssertionsSlowPath)
+set(CATCH_TEST_TARGETS ${CATCH_TEST_TARGETS} PARENT_SCOPE)

benchmarks/assertion_listener.cpp

@@ -0,0 +1,28 @@
+
+//              Copyright Catch2 Authors
+// Distributed under the Boost Software License, Version 1.0.
+//   (See accompanying file LICENSE.txt or copy at
+//        https://www.boost.org/LICENSE_1_0.txt)
+
+// SPDX-License-Identifier: BSL-1.0
+
+#include <catch2/reporters/catch_reporter_event_listener.hpp>
+#include <catch2/reporters/catch_reporter_registrars.hpp>
+
+/**
+ * Event listener that listens to all assertions, forcing assertion slow path
+ */
+class AssertionSlowPathListener : public Catch::EventListenerBase {
+public:
+    static std::string getDescription() {
+        return "Validates ordering of Catch2's listener events";
+    }
+
+    AssertionSlowPathListener(Catch::IConfig const* config) :
+        EventListenerBase(config) {
+        m_preferences.shouldReportAllAssertions = true;
+        m_preferences.shouldReportAllAssertionStarts = true;
+    }
+};
+
+CATCH_REGISTER_LISTENER( AssertionSlowPathListener )

benchmarks/runtime_assertion_benches.cpp

@@ -0,0 +1,27 @@
+
+//              Copyright Catch2 Authors
+// Distributed under the Boost Software License, Version 1.0.
+//   (See accompanying file LICENSE.txt or copy at
+//        https://www.boost.org/LICENSE_1_0.txt)
+
+// SPDX-License-Identifier: BSL-1.0
+
+#include <catch2/catch_test_macros.hpp>
+
+TEST_CASE("Simple REQUIRE - 10M") {
+    for (size_t i = 0; i < 10'000'000; ++i) {
+        REQUIRE(true);
+    }
+}
+
+TEST_CASE("Simple NOTHROW - 10M") {
+    for (size_t i = 0; i < 10'000'000; ++i) {
+        REQUIRE_NOTHROW([](){}());
+    }
+}
+
+TEST_CASE("Simple THROWS - 10M") {
+    for (size_t i = 0; i < 10'000'000; ++i) {
+        REQUIRE_THROWS([]() { throw 1; }());
+    }
+}

src/catch2/catch_message.cpp

@@ -22,7 +22,7 @@ namespace Catch {
         m_messageId( builder.m_info.sequence ) {
         MessageInfo info( CATCH_MOVE( builder.m_info ) );
         info.message = builder.m_stream.str();
-        getResultCapture().pushScopedMessage( CATCH_MOVE(info) );
+        IResultCapture::pushScopedMessage( CATCH_MOVE( info ) );
     }
 
     ScopedMessage::ScopedMessage( ScopedMessage&& old ) noexcept:
@@ -31,15 +31,14 @@ namespace Catch {
     }
 
     ScopedMessage::~ScopedMessage() {
-        if ( !m_moved ) { getResultCapture().popScopedMessage( m_messageId ); }
+        if ( !m_moved ) { IResultCapture::popScopedMessage( m_messageId ); }
     }
 
 
     Capturer::Capturer( StringRef macroName,
                         SourceLineInfo const& lineInfo,
                         ResultWas::OfType resultType,
-                        StringRef names ):
+                        StringRef names ) {
-        m_resultCapture( getResultCapture() ) {
         auto trimmed = [&] (size_t start, size_t end) {
             while (names[start] == ',' || isspace(static_cast<unsigned char>(names[start]))) {
                 ++start;
@@ -101,14 +100,14 @@ namespace Catch {
     Capturer::~Capturer() {
         assert( m_captured == m_messages.size() );
         for (auto const& message : m_messages) {
-            m_resultCapture.popScopedMessage( message.sequence );
+            IResultCapture::popScopedMessage( message.sequence );
         }
     }
 
     void Capturer::captureValue( size_t index, std::string const& value ) {
         assert( index < m_messages.size() );
         m_messages[index].message += value;
-        m_resultCapture.pushScopedMessage( CATCH_MOVE(m_messages[index]) );
+        IResultCapture::pushScopedMessage( CATCH_MOVE( m_messages[index] ) );
         m_captured++;
     }
 

src/catch2/catch_message.hpp

@@ -63,7 +63,6 @@ namespace Catch {
 
     class Capturer {
         std::vector<MessageInfo> m_messages;
-        IResultCapture& m_resultCapture;
         size_t m_captured = 0;
     public:
         Capturer( StringRef macroName, SourceLineInfo const& lineInfo, ResultWas::OfType resultType, StringRef names );
@@ -111,7 +110,7 @@ namespace Catch {
 
 ///////////////////////////////////////////////////////////////////////////////
 #define INTERNAL_CATCH_UNSCOPED_INFO( macroName, log ) \
-    Catch::getResultCapture().emplaceUnscopedMessage( Catch::MessageBuilder( macroName##_catch_sr, CATCH_INTERNAL_LINEINFO, Catch::ResultWas::Info ) << log )
+    Catch::IResultCapture::emplaceUnscopedMessage( Catch::MessageBuilder( macroName##_catch_sr, CATCH_INTERNAL_LINEINFO, Catch::ResultWas::Info ) << log )
 
 
 #if defined(CATCH_CONFIG_PREFIX_MESSAGES) && !defined(CATCH_CONFIG_DISABLE)

src/catch2/catch_registry_hub.cpp

@@ -96,7 +96,6 @@ namespace Catch {
     }
     void cleanUp() {
         cleanupSingletons();
-        cleanUpContext();
     }
     std::string translateActiveException() {
         return getRegistryHub().getExceptionTranslatorRegistry().translateActiveException();

src/catch2/interfaces/catch_interfaces_capture.cpp

@@ -7,7 +7,14 @@
 // SPDX-License-Identifier: BSL-1.0
 
 #include <catch2/interfaces/catch_interfaces_capture.hpp>
+#include <catch2/internal/catch_enforce.hpp>
 
 namespace Catch {
-    IResultCapture::~IResultCapture() = default;
+    namespace Detail {
+        void missingCaptureInstance() {
+            CATCH_INTERNAL_ERROR( "No result capture instance" );
         }
+    } // namespace Detail
+
+    IResultCapture::~IResultCapture() = default;
+} // namespace Catch

src/catch2/interfaces/catch_interfaces_capture.hpp

@@ -10,6 +10,7 @@
 
 #include <string>
 
+#include <catch2/internal/catch_context.hpp>
 #include <catch2/internal/catch_stringref.hpp>
 #include <catch2/internal/catch_result_type.hpp>
 #include <catch2/internal/catch_unique_ptr.hpp>
@@ -62,10 +63,9 @@ namespace Catch {
         virtual void benchmarkEnded( BenchmarkStats<> const& stats ) = 0;
         virtual void benchmarkFailed( StringRef error ) = 0;
 
-        virtual void pushScopedMessage( MessageInfo&& message ) = 0;
+        static void pushScopedMessage( MessageInfo&& message );
-        virtual void popScopedMessage( unsigned int messageId ) = 0;
+        static void popScopedMessage( unsigned int messageId );
-
+        static void emplaceUnscopedMessage( MessageBuilder&& builder );
-        virtual void emplaceUnscopedMessage( MessageBuilder&& builder ) = 0;
 
         virtual void handleFatalErrorCondition( StringRef message ) = 0;
 
@@ -101,7 +101,14 @@ namespace Catch {
         virtual void exceptionEarlyReported() = 0;
     };
 
-    IResultCapture& getResultCapture();
+    namespace Detail {
+        [[noreturn]]
+        void missingCaptureInstance();
+    }
+    inline IResultCapture& getResultCapture() {
+        return *getCurrentContext().getResultCapture();
+    }
+
 }
 
 #endif // CATCH_INTERFACES_CAPTURE_HPP_INCLUDED

src/catch2/internal/catch_context.cpp

@@ -6,25 +6,14 @@
 
 // SPDX-License-Identifier: BSL-1.0
 #include <catch2/internal/catch_context.hpp>
-#include <catch2/internal/catch_noncopyable.hpp>
 #include <catch2/internal/catch_random_number_generator.hpp>
 
 namespace Catch {
 
-    Context* Context::currentContext = nullptr;
+    Context Context::currentContext;
-
-    void cleanUpContext() {
-        delete Context::currentContext;
-        Context::currentContext = nullptr;
-    }
-    void Context::createContext() {
-        currentContext = new Context();
-    }
 
     Context& getCurrentMutableContext() {
-        if ( !Context::currentContext ) { Context::createContext(); }
+        return Context::currentContext;
-        // NOLINTNEXTLINE(clang-analyzer-core.uninitialized.UndefReturn)
-        return *Context::currentContext;
     }
 
     SimplePcg32& sharedRng() {

src/catch2/internal/catch_context.hpp

@@ -19,11 +19,9 @@ namespace Catch {
         IConfig const* m_config = nullptr;
         IResultCapture* m_resultCapture = nullptr;
 
-        CATCH_EXPORT static Context* currentContext;
+        CATCH_EXPORT static Context currentContext;
         friend Context& getCurrentMutableContext();
         friend Context const& getCurrentContext();
-        static void createContext();
-        friend void cleanUpContext();
 
     public:
         constexpr IResultCapture* getResultCapture() const {
@@ -34,21 +32,14 @@ namespace Catch {
             m_resultCapture = resultCapture;
         }
         constexpr void setConfig( IConfig const* config ) { m_config = config; }
-
     };
 
     Context& getCurrentMutableContext();
 
     inline Context const& getCurrentContext() {
-        // We duplicate the logic from `getCurrentMutableContext` here,
+        return Context::currentContext;
-        // to avoid paying the call overhead in debug mode.
-        if ( !Context::currentContext ) { Context::createContext(); }
-        // NOLINTNEXTLINE(clang-analyzer-core.uninitialized.UndefReturn)
-        return *Context::currentContext;
     }
 
-    void cleanUpContext();
-
     class SimplePcg32;
     SimplePcg32& sharedRng();
 }

src/catch2/internal/catch_fatal_condition_handler.cpp

@@ -86,23 +86,27 @@ namespace Catch {
         { EXCEPTION_INT_DIVIDE_BY_ZERO, "Divide by zero error" },
     };
 
+    // Since we do not support multiple instantiations, we put these
+    // into global variables and rely on cleaning them up in outlined
+    // constructors/destructors
+    static LPTOP_LEVEL_EXCEPTION_FILTER previousTopLevelExceptionFilter = nullptr;
+
+
     static LONG CALLBACK topLevelExceptionFilter(PEXCEPTION_POINTERS ExceptionInfo) {
         for (auto const& def : signalDefs) {
             if (ExceptionInfo->ExceptionRecord->ExceptionCode == def.id) {
                 reportFatal(def.name);
             }
         }
-        // If its not an exception we care about, pass it along.
+        // If a filter was previously registered, invoke it
+        if (previousTopLevelExceptionFilter) {
+            return previousTopLevelExceptionFilter(ExceptionInfo);
+        }
+        // Otherwise, pass along all exceptions.
         // This stops us from eating debugger breaks etc.
         return EXCEPTION_CONTINUE_SEARCH;
     }
 
-    // Since we do not support multiple instantiations, we put these
-    // into global variables and rely on cleaning them up in outlined
-    // constructors/destructors
-    static LPTOP_LEVEL_EXCEPTION_FILTER previousTopLevelExceptionFilter = nullptr;
-
-
     // For MSVC, we reserve part of the stack memory for handling
     // memory overflow structured exception.
     FatalConditionHandler::FatalConditionHandler() {

src/catch2/internal/catch_reusable_string_stream.cpp

@@ -12,6 +12,7 @@
 
 #include <cstdio>
 #include <sstream>
+#include <tuple>
 #include <vector>
 
 namespace Catch {
@@ -23,16 +24,16 @@ namespace Catch {
         std::ostringstream m_referenceStream; // Used for copy state/ flags from
         Detail::Mutex m_mutex;
 
-        auto add() -> std::size_t {
+        auto add() -> std::pair<std::size_t, std::ostringstream*> {
             Detail::LockGuard _( m_mutex );
             if( m_unused.empty() ) {
                 m_streams.push_back( Detail::make_unique<std::ostringstream>() );
-                return m_streams.size()-1;
+                return { m_streams.size()-1, m_streams.back().get() };
             }
             else {
                 auto index = m_unused.back();
                 m_unused.pop_back();
-                return index;
+                return { index, m_streams[index].get() };
             }
         }
 
@@ -46,10 +47,10 @@ namespace Catch {
         }
     };
 
-    ReusableStringStream::ReusableStringStream()
+    ReusableStringStream::ReusableStringStream() {
-    :   m_index( Singleton<StringStreams>::getMutable().add() ),
+        std::tie( m_index, m_oss ) =
-        m_oss( Singleton<StringStreams>::getMutable().m_streams[m_index].get() )
+            Singleton<StringStreams>::getMutable().add();
-    {}
+    }
 
     ReusableStringStream::~ReusableStringStream() {
         static_cast<std::ostringstream*>( m_oss )->str("");

src/catch2/internal/catch_run_context.cpp

@@ -483,28 +483,6 @@ namespace Catch {
         m_reporter->benchmarkFailed( error );
     }
 
-    void RunContext::pushScopedMessage( MessageInfo&& message ) {
-        Detail::g_messages.push_back( CATCH_MOVE(message) );
-    }
-
-    void RunContext::popScopedMessage( unsigned int messageId ) {
-        // Note: On average, it would probably be better to look for the message
-        //       backwards. However, we do not expect to have to deal with more
-        //       messages than low single digits, so the optimization is tiny,
-        //       and we would have to hand-write the loop to avoid terrible
-        //       codegen of reverse iterators in debug mode.
-        Detail::g_messages.erase(
-            std::find_if( Detail::g_messages.begin(),
-                          Detail::g_messages.end(),
-                          [=]( MessageInfo const& msg ) {
-                              return msg.sequence == messageId;
-                          } ) );
-    }
-
-    void RunContext::emplaceUnscopedMessage( MessageBuilder&& builder ) {
-        Detail::g_messageScopes.emplace_back( CATCH_MOVE(builder) );
-    }
-
     std::string RunContext::getCurrentTestName() const {
         return m_activeTestCase
             ? m_activeTestCase->getTestCaseInfo().name
@@ -831,11 +809,26 @@ namespace Catch {
         }
     }
 
-    IResultCapture& getResultCapture() {
+    void IResultCapture::pushScopedMessage( MessageInfo&& message ) {
-        if (auto* capture = getCurrentContext().getResultCapture())
+        Detail::g_messages.push_back( CATCH_MOVE( message ) );
-            return *capture;
+    }
-        else
+
-            CATCH_INTERNAL_ERROR("No result capture instance");
+    void IResultCapture::popScopedMessage( unsigned int messageId ) {
+        // Note: On average, it would probably be better to look for the message
+        //       backwards. However, we do not expect to have to deal with more
+        //       messages than low single digits, so the optimization is tiny,
+        //       and we would have to hand-write the loop to avoid terrible
+        //       codegen of reverse iterators in debug mode.
+        Detail::g_messages.erase( std::find_if( Detail::g_messages.begin(),
+                                                Detail::g_messages.end(),
+                                                [=]( MessageInfo const& msg ) {
+                                                    return msg.sequence ==
+                                                           messageId;
+                                                } ) );
+    }
+
+    void IResultCapture::emplaceUnscopedMessage( MessageBuilder&& builder ) {
+        Detail::g_messageScopes.emplace_back( CATCH_MOVE( builder ) );
     }
 
     void seedRng(IConfig const& config) {

src/catch2/internal/catch_run_context.hpp

@@ -94,11 +94,6 @@ namespace Catch {
         void benchmarkEnded( BenchmarkStats<> const& stats ) override;
         void benchmarkFailed( StringRef error ) override;
 
-        void pushScopedMessage( MessageInfo&& message ) override;
-        void popScopedMessage( unsigned int messageId ) override;
-
-        void emplaceUnscopedMessage( MessageBuilder&& builder ) override;
-
         std::string getCurrentTestName() const override;
 
         const AssertionResult* getLastResult() const override;

tests/SelfTest/IntrospectiveTests/ToString.tests.cpp

@@ -6,13 +6,11 @@
 
 // SPDX-License-Identifier: BSL-1.0
 
-#include <catch2/benchmark/catch_benchmark.hpp>
-#include <catch2/catch_template_test_macros.hpp>
-#include <catch2/catch_test_macros.hpp>
-#include <catch2/generators/catch_generators.hpp>
 #include <catch2/internal/catch_enum_values_registry.hpp>
 #include <catch2/matchers/catch_matchers_string.hpp>
 #include <catch2/matchers/catch_matchers_vector.hpp>
+#include <catch2/catch_test_macros.hpp>
+#include <catch2/catch_template_test_macros.hpp>
 
 #include <chrono>
 
@@ -141,70 +139,3 @@ TEST_CASE( "Exception thrown inside stringify does not fail the test", "[toStrin
     ThrowsOnStringification tos;
     CHECK( tos == tos );
 }
-
-namespace {
-    [[maybe_unused]]
-    std::string convertIntoString( Catch::StringRef string,
-                                   bool escapeInvisibles ) {
-        std::string ret;
-        // This is enough for the "don't escape invisibles" case, and a good
-        // lower bound on the "escape invisibles" case.
-        ret.reserve( string.size() + 2 );
-
-        if ( !escapeInvisibles ) {
-            ret += '"';
-            ret += string;
-            ret += '"';
-            return ret;
-        }
-
-        size_t last_start = 0;
-        auto write_to = [&]( size_t idx ) {
-            if ( last_start < idx ) {
-                //ret.append( string.data() + last_start, idx - last_start );
-                ret += string.substr( last_start, idx - last_start );
-            }
-            last_start = idx + 1;
-        };
-
-        ret += '"';
-        for ( size_t i = 0; i < string.size(); ++i ) {
-            const char c = string[i];
-            if (c == '\r' || c == '\n' || c == '\t' || c == '\f') {
-                write_to( i );
-                if ( c == '\r' ) { ret.append( "\\r" ); }
-                if ( c == '\n' ) { ret.append( "\\n" ); }
-                if ( c == '\t' ) { ret.append( "\\t" ); }
-                if ( c == '\f' ) { ret.append( "\\f" ); }
-            }
-        }
-        write_to( string.size() );
-        ret += '"';
-
-        return ret;
-    }
-}
-
-
-TEST_CASE( "string escaping benchmark", "[toString][!benchmark]" ) {
-    const auto input_length = GENERATE( as<size_t>{}, 10, 100, 10'000, 100'000 );
-    std::string test_input( input_length, 'a' );
-    BENCHMARK( "no-escape string, no-escaping, len=" +
-               std::to_string( input_length ) ) {
-        return Catch::Detail::convertIntoString( test_input, false );
-    };
-    BENCHMARK( "no-escape string, escaping, len=" +
-               std::to_string( input_length ) ) {
-        return Catch::Detail::convertIntoString( test_input, true );
-    };
-
-    std::string escape_input( input_length, '\r' );
-    BENCHMARK( "full escape string, no-escaping, len=" +
-               std::to_string( input_length ) ) {
-        return Catch::Detail::convertIntoString( escape_input, false );
-    };
-    BENCHMARK( "full escape string, escaping, len=" +
-               std::to_string( input_length ) ) {
-        return Catch::Detail::convertIntoString( escape_input, true );
-    };
-}