Chapter 5: Writing a Benchmark

This final demonstrates how to write benchmarks using Qt Test.

Writing a Benchmark

To create a benchmark we extend a test function with a QBENCHMARK macro. A benchmark test function will then typically consist of setup code and a QBENCHMARK macro that contains the code to be measured. This test function benchmarks QString::localeAwareCompare().

void TestBenchmark::simple()
{
    QString str1 = QLatin1String("This is a test string");
    QString str2 = QLatin1String("This is a test string");

    QCOMPARE(str1.localeAwareCompare(str2), 0);

    QBENCHMARK {
        str1.localeAwareCompare(str2);
    }
}

Setup can be done at the beginning of the function. At this point, the clock is not running. The code inside the QBENCHMARK macro will be measured, and possibly repeated several times in order to get an accurate measurement.

Several back-ends are available and can be selected on the command line.

Data Functions

Data functions are useful for creating benchmarks that compare multiple data inputs, for example locale aware compare against standard compare.

void TestBenchmark::multiple_data()
{
    QTest::addColumn<bool>("useLocaleCompare");
    QTest::newRow("locale-aware-compare") << true;
    QTest::newRow("standard-compare") << false;
}

The test function then uses the data to determine what to benchmark.

void TestBenchmark::multiple()
{
    QFETCH(bool, useLocaleCompare);
    QString str1 = QLatin1String("This is a test string");
    QString str2 = QLatin1String("This is a test string");

    int result;
    if (useLocaleCompare) {
        QBENCHMARK {
            result = str1.localeAwareCompare(str2);
        }
    } else {
        QBENCHMARK {
            result = (str1 == str2);
        }
    }
    Q_UNUSED(result);
}

The if (useLocaleCompare) switch is placed outside the QBENCHMARK macro to avoid measuring its overhead. Each benchmark test function can have one active QBENCHMARK macro.

External Tools

Tools for handling and visualizing test data are available as part of the qtestlib-tools project. These include a tool for comparing performance data obtained from test runs and a utility to generate Web-based graphs of performance data.

See the qtestlib-tools announcement for more information on these tools and a simple graphing example.

Building the Executable

You can build the test case executable using CMake or qmake.

Building with CMake

Configure your build settings in your CMakeLists.txt file:

cmake_minimum_required(VERSION 3.16)
project(tutorial5 LANGUAGES CXX)

set(CMAKE_AUTOMOC ON)

if(NOT DEFINED INSTALL_EXAMPLESDIR)
    set(INSTALL_EXAMPLESDIR "examples")
endif()

set(INSTALL_EXAMPLEDIR "${INSTALL_EXAMPLESDIR}/qtestlib/tutorial5")

find_package(Qt6 REQUIRED COMPONENTS Core Gui Test Widgets)

qt_add_executable(tutorial5
    benchmarking.cpp
)

set_target_properties(tutorial5 PROPERTIES
    WIN32_EXECUTABLE TRUE
    MACOSX_BUNDLE TRUE
)

target_link_libraries(tutorial5 PUBLIC
    Qt::Core
    Qt::Gui
    Qt::Test
    Qt::Widgets
)

install(TARGETS tutorial5
    RUNTIME DESTINATION "${INSTALL_EXAMPLEDIR}"
    BUNDLE DESTINATION "${INSTALL_EXAMPLEDIR}"
    LIBRARY DESTINATION "${INSTALL_EXAMPLEDIR}"
)

Next, from the command line, run either cmake or use the qt-cmake convenience script located in Qt-prefix/<version>/<platform>/bin/qt-cmake:

<Qt-prefix>/<version>/<platform>/bin/qt-cmake <source-dir> <build-dir> -G Ninja

Then, run your preferred generator tool to build the executable. Here, we're using Ninja:

ninja

Building with qmake

Configure your build settings in your .pro file:

QT += widgets testlib

SOURCES = benchmarking.cpp

# install
target.path = $$[QT_INSTALL_EXAMPLES]/qtestlib/tutorial5
INSTALLS += target

Next, run qmake, and, finally, run make to build your executable:

qmake
make

Running the Executable

Running the resulting executable should give you the following output:

********* Start testing of TestBenchmark *********
Config: Using QtTest library %VERSION%, Qt %VERSION%
PASS   : TestBenchmark::initTestCase()
PASS   : TestBenchmark::simple()
RESULT : TestBenchmark::simple():
     0.00030 msecs per iteration (total: 79, iterations: 262144)
PASS   : TestBenchmark::multiple(locale-aware-compare)
RESULT : TestBenchmark::multiple():"locale-aware-compare":
     0.00029 msecs per iteration (total: 78, iterations: 262144)
.....
PASS   : TestBenchmark::series(locale-aware-compare:8001)
RESULT : TestBenchmark::series():"locale-aware-compare:8001":
     0.039 msecs per iteration (total: 81, iterations: 2048)
Totals: 15 passed, 0 failed, 0 skipped, 0 blacklisted, 3971ms
********* Finished testing of TestBenchmark *********

Example project @ code.qt.io