The QOpenGLTimerQuery class wraps an OpenGL timer query object. More...
Header: | #include <QOpenGLTimerQuery> |
CMake: | find_package(Qt6 REQUIRED COMPONENTS OpenGL) target_link_libraries(mytarget PRIVATE Qt6::OpenGL) |
qmake: | QT += opengl |
Since: | Qt 5.1 |
Inherits: | QObject |
QOpenGLTimerQuery(QObject *parent = nullptr) | |
virtual | ~QOpenGLTimerQuery() |
void | begin() |
bool | create() |
void | destroy() |
void | end() |
bool | isCreated() const |
bool | isResultAvailable() const |
GLuint | objectId() const |
void | recordTimestamp() |
GLuint64 | waitForResult() const |
GLuint64 | waitForTimestamp() const |
OpenGL timer query objects are OpenGL managed resources to measure the execution times of sequences of OpenGL commands on the GPU.
OpenGL offers various levels of support for timer queries, depending on the version of OpenGL you have and the presence of the ARB_timer_query or EXT_timer_query extensions. The support can be summarized as:
OpenGL represents time with a granularity of 1 nanosecond (1e-9 seconds). As a consequence of this, 32-bit integers would only give a total possible duration of approximately 4 seconds, which would not be difficult to exceed in poorly performing or lengthy operations. OpenGL therefore uses 64 bit integer types to represent times. A GLuint64 variable has enough width to contain a duration of hundreds of years, which is plenty for real-time rendering needs.
As with the other Qt OpenGL classes, QOpenGLTimerQuery has a create() function to create the underlying OpenGL object. This is to allow the developer to ensure that there is a valid current OpenGL context at the time.
Once created, timer queries can be issued in one of several ways. The simplest method is to delimit a block of commands with calls to begin() and end(). This instructs OpenGL to measure the time taken from completing all commands issued prior to begin() until the completion of all commands issued prior to end().
At the end of a frame we can retrieve the results by calling waitForResult(). As this function's name implies, it blocks CPU execution until OpenGL notifies that the timer query result is available. To avoid blocking, you can check if the query result is available by calling isResultAvailable(). Note that modern GPUs are deeply pipelined and query results may not become available for between 1-5 frames after they were issued.
Note that OpenGL does not permit nesting or interleaving of multiple timer queries using begin() and end(). Using multiple timer queries and recordTimestamp() avoids this limitation. When using recordTimestamp() the result can be obtained at some later time using isResultAvailable() and waitForResult(). Qt provides the convenience class QOpenGLTimeMonitor that helps with using multiple query objects.
See also QOpenGLTimeMonitor.
Creates a QOpenGLTimerQuery instance with the given parent. You must call create() with a valid OpenGL context before using.
[virtual]
QOpenGLTimerQuery::~QOpenGLTimerQuery()Destroys the QOpenGLTimerQuery and the underlying OpenGL resource.
Marks the start point in the OpenGL command queue for a sequence of commands to be timed by this query object.
This is useful for simple use-cases. Usually it is better to use recordTimestamp().
See also end(), isResultAvailable(), waitForResult(), and recordTimestamp().
Creates the underlying OpenGL timer query object. There must be a valid OpenGL context that supports query objects current for this function to succeed.
Returns true
if the OpenGL timer query object was successfully created.
Destroys the underlying OpenGL timer query object. The context that was current when create() was called must be current when calling this function.
Marks the end point in the OpenGL command queue for a sequence of commands to be timed by this query object.
This is useful for simple use-cases. Usually it is better to use recordTimestamp().
See also begin(), isResultAvailable(), waitForResult(), and recordTimestamp().
Returns true
if the underlying OpenGL query object has been created. If this returns true
and the associated OpenGL context is current, then you are able to issue queries with this object.
Returns true
if the OpenGL timer query result is available.
This function is non-blocking and ideally should be used to check for the availability of the query result before calling waitForResult().
See also waitForResult().
Returns the id of the underlying OpenGL query object.
Places a marker in the OpenGL command queue for the GPU to record the timestamp when this marker is reached by the GPU. This function is non-blocking and the result will become available at some later time.
The availability of the result can be checked with isResultAvailable(). The result can be fetched with waitForResult() which will block if the result is not yet available.
See also waitForResult(), isResultAvailable(), begin(), and end().
Returns the result of the OpenGL timer query.
This function will block until the result is made available by OpenGL. It is recommended to call isResultAvailable() to ensure that the result is available to avoid unnecessary blocking and stalling.
See also isResultAvailable().
Returns the current timestamp of the GPU when all previously issued OpenGL commands have been received but not necessarily executed by the GPU.
This function blocks until the result is returned.
See also recordTimestamp().