sdk/2.0/docs/man/sub_group_barrier.xml

<?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>Synchronization, barrier</keyword>
        </keywordset>
    </refentryinfo>

    <refmeta>
        <refentrytitle>sub_group_barrier</refentrytitle>

        <refmiscinfo>
            <copyright>
                <year>2007-2013</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="sub_group_barrier">
        <refname>sub_group_barrier</refname>

        <refpurpose>
            All work-items in a sub-group executing the kernel on a 
processor must execute this function before any are 
allowed to continue execution beyond the subgroup 
barrier
        </refpurpose>
    </refnamediv>

    <refsynopsisdiv xmlns:xlink="http://www.w3.org/1999/xlink"><title></title>

        <funcsynopsis>
            <funcprototype>
                <funcdef>
                    <link xlink:href="scalarDataTypes.html">void</link>
                    <function>
                        sub_group_barrier
                    </function>
                </funcdef>
                <paramdef><link xlink:href="enums.html#cl_mem_fence_flags">cl_mem_fence_flags</link><parameter>flags</parameter></paramdef>
            </funcprototype>
        </funcsynopsis>

        <funcsynopsis>
            <funcprototype>
                <funcdef>
                    <link xlink:href="scalarDataTypes.html">void</link>
                    <function>
                        sub_group_barrier
                    </function>
                </funcdef>
                <paramdef><link xlink:href="enums.html#cl_mem_fence_flags">cl_mem_fence_flags</link><parameter>flags</parameter></paramdef>
                <paramdef><link xlink:href="memory_order.html">memory_scope</link><parameter>scope</parameter></paramdef>

            </funcprototype>
        </funcsynopsis>

    </refsynopsisdiv>

<!-- ================================ DESCRIPTION  -->

    <refsect1 id="description"><title>Description</title>
        <para>
            All work-items in a sub-group executing the kernel on a 
            processor must execute this function before any are 
            allowed to continue execution beyond the subgroup 
            barrier. This function must be encountered by all 
            workitems in a sub-group executing the kernel. These rules 
            apply to ND-ranges implemented with uniform and 
            nonuniform work-groups. 
        </para>

        <para>
          If <function>sub_group_barrier</function> is inside 
          a conditional statement, then all work-items
          must enter the conditional if any work-item 
          enters the conditional statement and
          executes the subgroup barrier.
        </para>

        <para>
          If <function>sub_group_barrier</function> is inside a loop, 
          all work-items within the sub-group must execute the
          <function>sub_group_barrier</function> for each 
          iteration of the loop before any are allowed
          to continue execution beyond the <function>sub_group_barrier</function>.
        </para>

        <para>
          The <function>sub_group_barrier</function> function 
          also queues a memory fence (reads and writes)
          to ensure correct ordering of memory operations to local or global memory.
        </para>

        <para>
          The <varname>flags</varname> argument specifies 
          the memory address space and can be
          set to a combination of the following values.
        </para>

        <para>
          <constant>CLK_LOCAL_MEM_FENCE</constant> - The 
          <function>sub_group_barrier</function> function
          will either flush any variables stored in local memory or queue a memory fence to
          ensure correct ordering of memory operations to local memory.
        </para>

        <para>
          <constant>CLK_GLOBAL_MEM_FENCE</constant> - The 
          <function>sub_group_barrier</function> function
          will queue a memory fence to ensure correct ordering of memory operations to global
          memory. This can be useful when work-items, for example, write to buffer or image
          objects and then want to read the updated data from these buffer objects.
        </para>

        <para>
          <constant>CLK_IMAGE_MEM_FENCE</constant> - The 
            <function>sub_group_barrier</function> function
            will queue a memory fence to ensure correct 
            ordering of memory operations to image objects. This 
            can be useful when work-items, for example, write to 
            image objects and then want to read the updated data 
            from these image objects.
        </para>
    </refsect1>

<!-- ================================ NOTES
    <refsect1 id="notes"><title>Notes</title>
        <para>

        </para>
    </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="sub_group_barrier">OpenCL Specification</olink>
        </para>
    </refsect1>

<!-- ================================ ALSO SEE  -->

    <refsect1 id="seealso"><title>Also see</title>
        <para>
            <citerefentry><refentrytitle>cl_khr_subgroups</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>

<!-- 23-Dec-2013, rev. 19 -->
</refentry>