blob: b7e828a02cc4b7ebf2545967cc2af2f894ad953f [file] [log] [blame]
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE book PUBLIC "-//OASIS//DTD DocBook MathML Module V1.1b1//EN"
"http://www.oasis-open.org/docbook/xml/mathml/1.1CR1/dbmathml.dtd">
<refentry>
<refentryinfo>
<keywordset>
<keyword>clEnqueueNDRangeKernel</keyword>
</keywordset>
</refentryinfo>
<refmeta>
<refentrytitle>clEnqueueNDRangeKernel</refentrytitle>
<refmiscinfo>
<copyright>
<year>2007-2011</year>
<holder>The Khronos Group Inc.
Permission is hereby granted, free of charge, to any person obtaining a
copy of this software and/or associated documentation files (the
"Materials"), to deal in the Materials without restriction, including
without limitation the rights to use, copy, modify, merge, publish,
distribute, sublicense, and/or sell copies of the Materials, and to
permit persons to whom the Materials are furnished to do so, subject to
the condition that this copyright notice and permission notice shall be included
in all copies or substantial portions of the Materials.</holder>
</copyright>
</refmiscinfo>
<manvolnum>3</manvolnum>
</refmeta>
<!-- ================================ SYNOPSIS -->
<refnamediv id="clEnqueueNDRangeKernel">
<refname>clEnqueueNDRangeKernel</refname>
<refpurpose>
Enqueues a command to execute a kernel on a device.
</refpurpose>
</refnamediv>
<refsynopsisdiv xmlns:xlink="http://www.w3.org/1999/xlink"><title></title>
<funcsynopsis>
<funcprototype>
<funcdef>
<link xlink:href="scalarDataTypes.html">cl_int</link>
<function>
clEnqueueNDRangeKernel
</function>
</funcdef>
<paramdef><link xlink:href="abstractDataTypes.html">cl_command_queue</link><parameter>command_queue</parameter></paramdef>
<paramdef><link xlink:href="abstractDataTypes.html">cl_kernel</link><parameter>kernel</parameter></paramdef>
<paramdef><link xlink:href="scalarDataTypes.html">cl_uint</link><parameter>work_dim</parameter></paramdef>
<paramdef>const <link xlink:href="scalarDataTypes.html">size_t</link><parameter>*global_work_offset</parameter></paramdef>
<paramdef>const <link xlink:href="scalarDataTypes.html">size_t</link><parameter>*global_work_size</parameter></paramdef>
<paramdef>const <link xlink:href="scalarDataTypes.html">size_t</link><parameter>*local_work_size</parameter></paramdef>
<paramdef><link xlink:href="scalarDataTypes.html">cl_uint</link><parameter>num_events_in_wait_list</parameter></paramdef>
<paramdef>const <link xlink:href="abstractDataTypes.html">cl_event</link><parameter>*event_wait_list</parameter></paramdef>
<paramdef><link xlink:href="abstractDataTypes.html">cl_event</link><parameter>*event</parameter></paramdef>
</funcprototype>
</funcsynopsis>
</refsynopsisdiv>
<!-- ================================ PARAMETERS -->
<refsect1 id="parameters">
<title>Parameters</title>
<variablelist>
<varlistentry>
<term> <varname> command_queue </varname> </term>
<listitem>
<para> <!-- parameter description -->
A valid command-queue. The kernel will be queued for execution on
the device associated with <varname>command_queue</varname>.
</para>
</listitem>
</varlistentry>
<varlistentry>
<term> <varname> kernel </varname> </term>
<listitem>
<para>
<!-- parameter description --> A valid kernel object. The
OpenCL context associated with <varname>kernel</varname> and
<varname>command_queue</varname> must be the same.
</para>
</listitem>
</varlistentry>
<varlistentry>
<term> <varname> work_dim </varname> </term>
<listitem>
<para>
<!-- parameter description --> The number of dimensions
used to specify the global work-items and work-items in the
work-group. <varname>work_dim</varname> must be greater than zero and less
than or equal to <constant>CL_DEVICE_MAX_WORK_ITEM_DIMENSIONS</constant>.
</para>
</listitem>
</varlistentry>
<varlistentry>
<term> <varname> global_work_offset </varname> </term>
<listitem>
<para>
<!-- parameter description --> <varname>global_work_offset</varname>
can be used to specify an array of <varname>work_dim</varname> unsigned
values that describe the offset used to calculate the global ID of a
work-item. If <varname>global_work_offset</varname> is NULL, the global
IDs start at offset (0, 0, ... 0).
</para>
</listitem>
</varlistentry>
<varlistentry>
<term> <varname> global_work_size </varname> </term>
<listitem>
<para>
<!-- parameter description --> Points to an array of
<varname>work_dim</varname> unsigned values that describe the number
of global work-items in <varname>work_dim</varname> dimensions
that will execute the kernel function. The total number of global
work-items is computed as <varname>global_work_size</varname>[0] *...*
<varname>global_work_size</varname>[<varname>work_dim</varname> - 1].
</para>
</listitem>
</varlistentry>
<varlistentry>
<term> <varname> local_work_size </varname> </term>
<listitem>
<para>
<!-- parameter description --> Points to an array of
<varname>work_dim</varname> unsigned values that describe the number
of work-items that make up a work-group (also referred to as the
size of the work-group) that will execute the kernel specified
by <varname>kernel</varname>. The total number of work-items in
a work-group is computed as <varname>local_work_size</varname>[0]
*... * <varname>local_work_size</varname>[<varname>work_dim</varname> -
1]. The total number of work-items in the work-group must be less than
or equal to the <constant>CL_DEVICE_MAX_WORK_GROUP_SIZE</constant>
value specified in table of OpenCL Device Queries for
<citerefentry><refentrytitle>clGetDeviceInfo</refentrytitle></citerefentry>
and the number of work-items specified
in <varname>local_work_size</varname>[0],...
<varname>local_work_size</varname>[<varname>work_dim</varname> - 1]
must be less than or equal to the corresponding values specified
by <constant>CL_DEVICE_MAX_WORK_ITEM_SIZES</constant>[0],....
<constant>CL_DEVICE_MAX_WORK_ITEM_SIZES</constant>[<varname>work_dim</varname>
- 1]. The explicitly specified <varname>local_work_size</varname>
will be used to determine how to break the global work-items specified
by <varname>global_work_size</varname> into appropriate work-group
instances. If <varname>local_work_size</varname> is specified,
the values specified in <varname>global_work_size</varname>[0],...
<varname>global_work_size</varname>[<varname>work_dim</varname>
- 1] must be evenly divisible by the corresponding values
specified in <varname>local_work_size</varname>[0],...
<varname>local_work_size</varname>[<varname>work_dim</varname> - 1].
</para>
<para>
<varname>local_work_size</varname> can also be a NULL value in which
case the OpenCL implementation will determine how to be break the global
work-items into appropriate work-group instances.
</para>
</listitem>
</varlistentry>
<varlistentry>
<term>
<varname> <!-- parameter name -->
event_wait_list
</varname> and
<varname> <!-- parameter name -->
num_events_in_wait_list
</varname>
</term>
<listitem>
<para>
<!-- parameter description --> Specify events that need
to complete before this particular command can be executed. If
<varname>event_wait_list</varname> is NULL, then this particular command does
not wait on any event to complete. If <varname>event_wait_list</varname>
is NULL, <varname>num_events_in_wait_list</varname> must be 0. If
<varname>event_wait_list</varname> is not NULL, the list of events
pointed to by <varname>event_wait_list</varname> must be valid
and <varname>num_events_in_wait_list</varname> must be greater
than 0. The events specified in <varname>event_wait_list</varname>
act as synchronization points. The context associated with events in
<varname>event_wait_list</varname> and <varname>command_queue</varname> must
be the same. The memory associated with <varname>event_wait_list</varname>
can be reused or freed after the function returns.
</para>
</listitem>
</varlistentry>
<varlistentry>
<term> <varname> event </varname> </term>
<listitem>
<para>
<!-- parameter description --> Returns an event object that identifies
this particular kernel execution instance. Event objects are unique and
can be used to identify a particular kernel execution instance later on. If
<varname>event</varname> is NULL, no event will be created for this kernel
execution instance and therefore it will not be possible for the application
to query or queue a wait for this particular kernel execution instance.
</para>
</listitem>
</varlistentry>
</variablelist>
</refsect1>
<!-- ================================ NOTES -->
<refsect1 id="notes"><title>Notes</title>
<para>
The work-group size to be used for <varname>kernel</varname> can also
be specified in the program source using the <code><citerefentry
href="attribute"><refentrytitle>__attribute__</refentrytitle></citerefentry>
((reqd_work_group_size(X, Y, Z)))</code> qualifier. In this case the size
of work group specified by <varname>local_work_size</varname> must match the
value specified by the <code>reqd_work_group_size</code> <citerefentry
href="attribute"><refentrytitle>__attribute__</refentrytitle></citerefentry> qualifier.
</para>
<para>
These work-group instances are executed in parallel across multiple compute units or
concurrently on the same compute unit.
</para>
<para>
Each work-item is uniquely identified by a global identifier. The global ID,
which can be read inside the kernel, is computed using the value given by
<varname>global_work_size</varname> and <varname>global_work_offset</varname>.
In addition, a work-item is also identified within a work-group by a unique local ID.
The local ID, which can also be read by the kernel, is computed using the value given
by <varname>local_work_size</varname>. The starting local ID is always (0, 0, ... 0).
</para>
</refsect1>
<!-- ================================ ERRORS -->
<refsect1 id="errors"><title>Errors</title>
<para>
Returns <errorname>CL_SUCCESS</errorname> if the kernel execution was successfully queued.
Otherwise, it returns one of the following errors:
</para>
<itemizedlist mark="disc">
<listitem>
<errorname>CL_INVALID_PROGRAM_EXECUTABLE</errorname> if there is no successfully built program
executable available for device associated with <varname>command_queue</varname>.
</listitem>
<listitem>
<errorname>CL_INVALID_COMMAND_QUEUE</errorname> if <varname>command_queue</varname> is not a
valid command-queue.
</listitem>
<listitem>
<errorname>CL_INVALID_KERNEL</errorname> if <varname>kernel</varname> is
not a valid kernel object.
</listitem>
<listitem>
<errorname>CL_INVALID_CONTEXT</errorname> if context associated with
<varname>command_queue</varname> and <varname>kernel</varname> is not the same
or if the context associated with <varname>command_queue</varname> and events in
<varname>event_wait_list</varname> are not the same.
</listitem>
<listitem>
<errorname>CL_INVALID_KERNEL_ARGS</errorname> if the kernel argument values
have not been specified.
</listitem>
<listitem>
<errorname>CL_INVALID_WORK_DIMENSION</errorname> if <varname>work_dim</varname>
is not a valid value (i.e. a value between 1 and 3).
</listitem>
<listitem>
<errorname>CL_INVALID_GLOBAL_WORK_SIZE</errorname> if
<varname>global_work_size</varname> is NULL, or if any of the values specified in
<varname>global_work_size</varname>[0], ...<varname>global_work_size</varname>
[<varname>work_dim</varname> - 1] are 0 or exceed the range given by the
<code>sizeof(size_t)</code> for the device on which the kernel execution will
be enqueued.
</listitem>
<listitem>
<errorname>CL_INVALID_GLOBAL_OFFSET</errorname> if the value specified
in <varname>global_work_size</varname> + the corresponding values in
<varname>global_work_offset</varname> for any dimensions is greater than the
<code>sizeof(size_t)</code> for the device on which the kernel execution will
be enqueued.
</listitem>
<listitem>
<errorname>CL_INVALID_WORK_GROUP_SIZE</errorname> if
<varname>local_work_size</varname> is specified and number of work-items specified by
<varname>global_work_size</varname> is not evenly divisable by size of work-group
given by <varname>local_work_size</varname> or does not match the work-group
size specified for <varname>kernel</varname> using the <code> <citerefentry
href="functionQualifiers"><refentrytitle>__attribute__</refentrytitle></citerefentry>
((reqd_work_group_size(X, Y, Z)))</code> qualifier in program source.
</listitem>
<listitem>
<errorname>CL_INVALID_WORK_GROUP_SIZE</errorname> if
<varname>local_work_size</varname> is specified and the total number of
work-items in the work-group computed as <varname>local_work_size</varname>[0] *...
<varname>local_work_size</varname>[<varname>work_dim</varname> - 1] is greater than
the value specified by <constant>CL_DEVICE_MAX_WORK_GROUP_SIZE</constant>
in the table of OpenCL Device Queries for
<citerefentry><refentrytitle>clGetDeviceInfo</refentrytitle></citerefentry>.
</listitem>
<listitem>
<errorname>CL_INVALID_WORK_GROUP_SIZE</errorname>
if <varname>local_work_size</varname> is NULL and
the <code><citerefentry href="functionQualifiers">
<refentrytitle>__attribute__</refentrytitle></citerefentry> ((reqd_work_group_size(X,
Y, Z)))</code> qualifier is used to declare the work-group size for
<varname>kernel</varname> in the program source.
</listitem>
<listitem>
<errorname>CL_INVALID_WORK_ITEM_SIZE</errorname> if the number of
work-items specified in any of <varname>local_work_size</varname>[0],
... <varname>local_work_size</varname>[<varname>work_dim</varname>
- 1] is greater than the corresponding values specified by
<constant>CL_DEVICE_MAX_WORK_ITEM_SIZES</constant>[0], ....
<constant>CL_DEVICE_MAX_WORK_ITEM_SIZES</constant>[<varname>work_dim</varname>
- 1].
</listitem>
<listitem>
<errorname>CL_MISALIGNED_SUB_BUFFER_OFFSET</errorname> if a sub-buffer object
is specified as the value for an argument that is a buffer object and the
<varname>offset</varname> specified when the sub-buffer object is created is not
aligned to <constant>CL_DEVICE_MEM_BASE_ADDR_ALIGN</constant> value for device
associated with <varname>queue</varname>.
</listitem>
<listitem>
<errorname>CL_INVALID_IMAGE_SIZE</errorname> if an image object is specified
as an argument value and the image dimensions (image width, height, specified
or compute row and/or slice pitch) are not supported by device associated with
<varname>queue</varname>.
</listitem>
<listitem>
<errorname>CL_INVALID_IMAGE_FORMAT</errorname> if an image object is specified
as an argument value and the image format (image channel order and data type)
is not supported by device associated with <varname>queue</varname>.
</listitem>
<listitem>
<errorname>CL_OUT_OF_RESOURCES</errorname> if there is a failure to queue the
execution instance of <varname>kernel</varname> on the command-queue because of
insufficient resources needed to execute the kernel. For example, the explicitly
specified <varname>local_work_size</varname> causes a failure to execute the kernel
because of insufficient resources such as registers or local memory. Another example
would be the number of read-only image args used in <varname>kernel</varname>
exceed the <constant>CL_DEVICE_MAX_READ_IMAGE_ARGS</constant> value for
device or the number of write-only image args used in <varname>kernel</varname>
exceed the <constant>CL_DEVICE_MAX_WRITE_IMAGE_ARGS</constant> value for
device or the number of samplers used in <varname>kernel</varname> exceed
<constant>CL_DEVICE_MAX_SAMPLERS</constant> for device.
</listitem>
<listitem>
<errorname>CL_MEM_OBJECT_ALLOCATION_FAILURE</errorname> if there is a failure to
allocate memory for data store associated with image or buffer objects specified
as arguments to <varname>kernel</varname>.
</listitem>
<listitem>
<errorname>CL_INVALID_EVENT_WAIT_LIST</errorname>
if <varname>event_wait_list</varname> is NULL and
<varname>num_events_in_wait_list</varname> &gt; 0,
or <varname>event_wait_list</varname> is not NULL and
<varname>num_events_in_wait_list</varname> is 0, or if event objects in
<varname>event_wait_list</varname> are not valid events.
</listitem>
<listitem>
<errorname>CL_OUT_OF_RESOURCES</errorname> if there is a failure to allocate
resources required by the OpenCL implementation on the device.
</listitem>
<listitem>
<errorname>CL_OUT_OF_HOST_MEMORY</errorname> if there is a failure to allocate
resources required by the OpenCL implementation on the host.
</listitem>
</itemizedlist>
</refsect1>
<!-- ================================ EXAMPLE -->
<!-- DO NOT DELETE IN CASE AN EXAMPLE IS ADDED IN THE FUTURE -->
<!--
<refsect2 id="example1">
<title>
Example
</title>
<informaltable frame="none">
<tgroup cols="1" align="left" colsep="0" rowsep="0">
<colspec colname="col1" colnum="1" />
<tbody>
<row>
<entry>
Example goes here - it will be set in "code" type with white space preserved.
</entry>
</row>
</tbody>
</tgroup>
</informaltable>
</refsect2>
-->
<!-- ================================ SPECIFICATION -->
<!-- Set the "uri" attribute in the <olink /> element to the "named destination" for the PDF page
-->
<refsect1 id="specification"><title>Specification</title>
<para>
<imageobject>
<imagedata fileref="pdficon_small1.gif" format="gif" />
</imageobject>
<olink uri="clEnqueueNDRangeKernel">OpenCL Specification</olink>
</para>
</refsect1>
<!-- ================================ ALSO SEE -->
<refsect1 id="seealso"><title>Also see</title>
<para>
<citerefentry><refentrytitle>clCreateCommandQueue</refentrytitle>
</citerefentry>, <citerefentry><refentrytitle>clGetDeviceInfo</refentrytitle></citerefentry>,
<citerefentry><refentrytitle>clEnqueueNativeKernel</refentrytitle></citerefentry>,
<citerefentry><refentrytitle>clEnqueueTask</refentrytitle></citerefentry>,
<citerefentry href="workItemFunctions"><refentrytitle>Work-Item Functions</refentrytitle></citerefentry>
</para>
</refsect1>
<!-- ============================== COPYRIGHT -->
<!-- Content included from copyright.inc.xsl -->
<refsect3 id="Copyright"><title></title>
<imageobject>
<imagedata fileref="KhronosLogo.jpg" format="jpg" />
</imageobject>
<para />
</refsect3>
<!-- 20-Oct-2011 -->
</refentry>