arm_compute v18.02

Change-Id: I7207aa488e5470f235f39b6c188b4678dc38d1a6
diff --git a/documentation/depthwise__convolution_8cl.xhtml b/documentation/depthwise__convolution_8cl.xhtml
index 2bd7be7..0a7875d 100644
--- a/documentation/depthwise__convolution_8cl.xhtml
+++ b/documentation/depthwise__convolution_8cl.xhtml
@@ -40,7 +40,7 @@
  <tr style="height: 56px;">
   <td style="padding-left: 0.5em;">
    <div id="projectname">Compute Library
-   &#160;<span id="projectnumber">18.01</span>
+   &#160;<span id="projectnumber">18.02</span>
    </div>
   </td>
  </tr>
@@ -113,6 +113,9 @@
 </div>
 
 <div class="header">
+  <div class="summary">
+<a href="#define-members">Macros</a> &#124;
+<a href="#func-members">Functions</a>  </div>
   <div class="headertitle">
 <div class="title">depthwise_convolution.cl File Reference</div>  </div>
 </div><!--header-->
@@ -120,13 +123,578 @@
 <div class="textblock"><code>#include &quot;<a class="el" href="helpers_8h_source.xhtml">helpers.h</a>&quot;</code><br />
 </div>
 <p><a href="depthwise__convolution_8cl_source.xhtml">Go to the source code of this file.</a></p>
+<table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="define-members"></a>
+Macros</h2></td></tr>
+<tr class="memitem:aa5ab4cf85fe56de3821727cfcef327d9"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(acc,  src0,  weights_row0)</td></tr>
+<tr class="separator:aa5ab4cf85fe56de3821727cfcef327d9"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:a31cbdd6b28ee6331839432f69d87fdcb"><td class="memItemLeft" align="right" valign="top">#define&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="depthwise__convolution_8cl.xhtml#a31cbdd6b28ee6331839432f69d87fdcb">CONVOLUTION1x3_BIFROST2X1_STRIDE2</a>(acc,  src0,  src1,  weights_row0)</td></tr>
+<tr class="separator:a31cbdd6b28ee6331839432f69d87fdcb"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table><table class="memberdecls">
+<tr class="heading"><td colspan="2"><h2 class="groupheader"><a name="func-members"></a>
+Functions</h2></td></tr>
+<tr class="memitem:a2fbbf5b73260b66b310d4d857db5289e"><td class="memItemLeft" align="right" valign="top">__kernel void&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="depthwise__convolution_8cl.xhtml#a2fbbf5b73260b66b310d4d857db5289e">depthwise_convolution_3x3_stridex1_stridey1_bifrost</a> (__global uchar *src_ptr, uint src_stride_x, uint src_step_x, uint src_stride_y, uint src_step_y, uint src_stride_z, uint src_step_z, uint src_offset_first_element_in_bytes, __global uchar *dst_ptr, uint dst_stride_x, uint dst_step_x, uint dst_stride_y, uint dst_step_y, uint dst_stride_z, uint dst_step_z, uint dst_offset_first_element_in_bytes, __global uchar *weights_ptr, uint weights_stride_x, uint weights_step_x, uint weights_stride_y, uint weights_step_y, uint weights_stride_z, uint weights_step_z, uint weights_offset_first_element_in_bytes, __global uchar *biases_ptr, uint biases_stride_x, uint biases_step_x, uint biases_offset_first_element_in_bytes)</td></tr>
+<tr class="memdesc:a2fbbf5b73260b66b310d4d857db5289e"><td class="mdescLeft">&#160;</td><td class="mdescRight">This OpenCL kernel is optimized for Bifrost architectures and computes the depthwise convolution 3x3 when both stride_x and stride_y are equal to 1.  <a href="#a2fbbf5b73260b66b310d4d857db5289e">More...</a><br /></td></tr>
+<tr class="separator:a2fbbf5b73260b66b310d4d857db5289e"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:a328ab548c536089e8238fb33de617bec"><td class="memItemLeft" align="right" valign="top">__kernel void&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="depthwise__convolution_8cl.xhtml#a328ab548c536089e8238fb33de617bec">depthwise_convolution_3x3_stridex2_stridey2_bifrost</a> (__global uchar *src_ptr, uint src_stride_x, uint src_step_x, uint src_stride_y, uint src_step_y, uint src_stride_z, uint src_step_z, uint src_offset_first_element_in_bytes, __global uchar *dst_ptr, uint dst_stride_x, uint dst_step_x, uint dst_stride_y, uint dst_step_y, uint dst_stride_z, uint dst_step_z, uint dst_offset_first_element_in_bytes, __global uchar *weights_ptr, uint weights_stride_x, uint weights_step_x, uint weights_stride_y, uint weights_step_y, uint weights_stride_z, uint weights_step_z, uint weights_offset_first_element_in_bytes, __global uchar *biases_ptr, uint biases_stride_x, uint biases_step_x, uint biases_offset_first_element_in_bytes)</td></tr>
+<tr class="memdesc:a328ab548c536089e8238fb33de617bec"><td class="mdescLeft">&#160;</td><td class="mdescRight">This OpenCL kernel is optimized for Bifrost architectures and computes the depthwise convolution 3x3 when both stride_x and stride_y are equal to 2.  <a href="#a328ab548c536089e8238fb33de617bec">More...</a><br /></td></tr>
+<tr class="separator:a328ab548c536089e8238fb33de617bec"><td class="memSeparator" colspan="2">&#160;</td></tr>
+</table>
+<h2 class="groupheader">Macro Definition Documentation</h2>
+<a class="anchor" id="aa5ab4cf85fe56de3821727cfcef327d9"></a>
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define CONVOLUTION1x3_BIFROST2X1_STRIDE1</td>
+          <td>(</td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">acc, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">src0, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">weights_row0&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+<b>Value:</b><div class="fragment"><div class="line">({                                                             \</div><div class="line">        acc.s0 = fma(src0.s0, weights_row0.s0, acc.s0);            \</div><div class="line">        acc.s0 = fma(src0.s1, weights_row0.s1, acc.s0);            \</div><div class="line">        acc.s0 = fma(src0.s2, weights_row0.s2, acc.s0);            \</div><div class="line">        acc.s1 = fma(src0.s1, weights_row0.s0, acc.s1);            \</div><div class="line">        acc.s1 = fma(src0.s2, weights_row0.s1, acc.s1);            \</div><div class="line">        acc.s1 = fma(src0.s3, weights_row0.s2, acc.s1);            \</div><div class="line">    })</div></div><!-- fragment -->
+<p>Definition at line <a class="el" href="depthwise__convolution_8cl_source.xhtml#l00211">211</a> of file <a class="el" href="depthwise__convolution_8cl_source.xhtml">depthwise_convolution.cl</a>.</p>
+
+<p>Referenced by <a class="el" href="depthwise__convolution_8cl_source.xhtml#l00263">depthwise_convolution_3x3_stridex1_stridey1_bifrost()</a>.</p>
+
+</div>
+</div>
+<a class="anchor" id="a31cbdd6b28ee6331839432f69d87fdcb"></a>
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">#define CONVOLUTION1x3_BIFROST2X1_STRIDE2</td>
+          <td>(</td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">acc, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">src0, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">src1, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">&#160;</td>
+          <td class="paramname">weights_row0&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+<b>Value:</b><div class="fragment"><div class="line">({                                                                   \</div><div class="line">        acc.s0 = fma(src0.s0, weights_row0.s0, acc.s0);                  \</div><div class="line">        acc.s0 = fma(src0.s1, weights_row0.s1, acc.s0);                  \</div><div class="line">        acc.s0 = fma(src0.s2, weights_row0.s2, acc.s0);                  \</div><div class="line">        acc.s1 = fma(src0.s2, weights_row0.s0, acc.s1);                  \</div><div class="line">        acc.s1 = fma(src0.s3, weights_row0.s1, acc.s1);                  \</div><div class="line">        acc.s1 = fma(src1.s0, weights_row0.s2, acc.s1);                  \</div><div class="line">    })</div></div><!-- fragment -->
+<p>Definition at line <a class="el" href="depthwise__convolution_8cl_source.xhtml#l00221">221</a> of file <a class="el" href="depthwise__convolution_8cl_source.xhtml">depthwise_convolution.cl</a>.</p>
+
+<p>Referenced by <a class="el" href="depthwise__convolution_8cl_source.xhtml#l00360">depthwise_convolution_3x3_stridex2_stridey2_bifrost()</a>.</p>
+
+</div>
+</div>
+<h2 class="groupheader">Function Documentation</h2>
+<a class="anchor" id="a2fbbf5b73260b66b310d4d857db5289e"></a>
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">__kernel void depthwise_convolution_3x3_stridex1_stridey1_bifrost </td>
+          <td>(</td>
+          <td class="paramtype">__global uchar *&#160;</td>
+          <td class="paramname"><em>src_ptr</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_stride_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_step_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_stride_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_step_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_stride_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_step_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_offset_first_element_in_bytes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">__global uchar *&#160;</td>
+          <td class="paramname"><em>dst_ptr</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_stride_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_step_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_stride_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_step_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_stride_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_step_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_offset_first_element_in_bytes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">__global uchar *&#160;</td>
+          <td class="paramname"><em>weights_ptr</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_stride_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_step_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_stride_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_step_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_stride_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_step_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_offset_first_element_in_bytes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">__global uchar *&#160;</td>
+          <td class="paramname"><em>biases_ptr</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>biases_stride_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>biases_step_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>biases_offset_first_element_in_bytes</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>This OpenCL kernel is optimized for Bifrost architectures and computes the depthwise convolution 3x3 when both stride_x and stride_y are equal to 1. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_ptr</td><td>Pointer to the source image. Supported data types: F32 </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_stride_x</td><td>Stride of the source image in X dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_step_x</td><td>src_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_stride_y</td><td>Stride of the source image in Y dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_step_y</td><td>src_stride_y * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_offset_first_element_in_bytes</td><td>The offset of the first element in the source image </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_stride_z</td><td>Stride of the source tensor in Z dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_step_z</td><td>src_stride_z * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_ptr</td><td>Pointer to the destination tensor. Supported data types: F32 </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_stride_x</td><td>Stride of the destination tensor in X dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_step_x</td><td>dst_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_stride_y</td><td>Stride of the destination tensor in Y dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_step_y</td><td>dst_stride_y * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_stride_z</td><td>Stride of the destination tensor in Z dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_step_z</td><td>dst_stride_z * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_offset_first_element_in_bytes</td><td>The offset of the first element in the destination tensor </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_ptr</td><td>Pointer to the weights tensor. Supported data types: F32 </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_stride_x</td><td>Stride of the weights tensor in X dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_step_x</td><td>weights_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_stride_y</td><td>Stride of the weights tensor in Y dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_step_y</td><td>weights_stride_y * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_stride_z</td><td>Stride of the weights tensor in Z dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_step_z</td><td>weights_stride_z * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_offset_first_element_in_bytes</td><td>The offset of the first element in the biases vector </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">biases_ptr</td><td>(Optional) Pointer to the biases vector. Supported data types: F32 </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">biases_stride_x</td><td>(Optional) Stride of the biases vector in X dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">biases_step_x</td><td>(Optional) biases_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">biases_offset_first_element_in_bytes</td><td>(Optional) The offset of the first element in the biases vector </td></tr>
+  </table>
+  </dd>
+</dl>
+
+<p>Definition at line <a class="el" href="depthwise__convolution_8cl_source.xhtml#l00263">263</a> of file <a class="el" href="depthwise__convolution_8cl_source.xhtml">depthwise_convolution.cl</a>.</p>
+
+<p>References <a class="el" href="helpers_8h_source.xhtml#l00114">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>, <a class="el" href="helpers_8h_source.xhtml#l00117">CONVERT_TO_TENSOR3D_STRUCT</a>, <a class="el" href="helpers_8h_source.xhtml#l00099">CONVERT_TO_VECTOR_STRUCT_NO_STEP</a>, <a class="el" href="depthwise__convolution_8cl_source.xhtml#l00211">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>, <a class="el" href="helpers_8h_source.xhtml#l00301">offset()</a>, <a class="el" href="helpers_8h_source.xhtml#l00142">Image::ptr</a>, <a class="el" href="helpers_8h_source.xhtml#l00151">Tensor3D::ptr</a>, and <a class="el" href="helpers_8h_source.xhtml#l00290">vector_offset()</a>.</p>
+<div class="fragment"><div class="line"><a name="l00272"></a><span class="lineno">  272</span>&#160;{</div><div class="line"><a name="l00273"></a><span class="lineno">  273</span>&#160;    <a class="code" href="struct_image.xhtml">Image</a>    <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a6743f0a130e8311e6f5b1a23df102472">src</a>     = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(src);</div><div class="line"><a name="l00274"></a><span class="lineno">  274</span>&#160;    <a class="code" href="struct_image.xhtml">Image</a>    <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#adbf67dcee294e673cf796f1ed8aeb6a4">dst</a>     = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(dst);</div><div class="line"><a name="l00275"></a><span class="lineno">  275</span>&#160;    <a class="code" href="struct_tensor3_d.xhtml">Tensor3D</a> weights = <a class="code" href="helpers_8h.xhtml#a31c8c760f08fb1a331b16b7c204321dc">CONVERT_TO_TENSOR3D_STRUCT</a>(weights);</div><div class="line"><a name="l00276"></a><span class="lineno">  276</span>&#160;</div><div class="line"><a name="l00277"></a><span class="lineno">  277</span>&#160;    float2 pixels0 = 0.0f;</div><div class="line"><a name="l00278"></a><span class="lineno">  278</span>&#160;    float2 pixels1 = 0.0f;</div><div class="line"><a name="l00279"></a><span class="lineno">  279</span>&#160;    float2 pixels2 = 0.0f;</div><div class="line"><a name="l00280"></a><span class="lineno">  280</span>&#160;    float2 pixels3 = 0.0f;</div><div class="line"><a name="l00281"></a><span class="lineno">  281</span>&#160;</div><div class="line"><a name="l00282"></a><span class="lineno">  282</span>&#160;    __global uchar *weights_addr = (__global uchar *)weights.<a class="code" href="struct_tensor3_d.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>;</div><div class="line"><a name="l00283"></a><span class="lineno">  283</span>&#160;    __global uchar *src_addr     = (__global uchar *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&amp;src, 0, 0);</div><div class="line"><a name="l00284"></a><span class="lineno">  284</span>&#160;</div><div class="line"><a name="l00285"></a><span class="lineno">  285</span>&#160;    <span class="comment">// Load the weights</span></div><div class="line"><a name="l00286"></a><span class="lineno">  286</span>&#160;    float3 weights_row0 = vload3(0, (__global <span class="keywordtype">float</span> *)(weights_addr + 0 * weights_stride_y));</div><div class="line"><a name="l00287"></a><span class="lineno">  287</span>&#160;    float3 weights_row1 = vload3(0, (__global <span class="keywordtype">float</span> *)(weights_addr + 1 * weights_stride_y));</div><div class="line"><a name="l00288"></a><span class="lineno">  288</span>&#160;    float3 weights_row2 = vload3(0, (__global <span class="keywordtype">float</span> *)(weights_addr + 2 * weights_stride_y));</div><div class="line"><a name="l00289"></a><span class="lineno">  289</span>&#160;</div><div class="line"><a name="l00290"></a><span class="lineno">  290</span>&#160;    <span class="comment">// Note: Since each work-item computes 4x2 elements, we need to load 4 rows from the input tensor</span></div><div class="line"><a name="l00291"></a><span class="lineno">  291</span>&#160;    float4 src00 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 0 * src_stride_y)); <span class="comment">// Row0</span></div><div class="line"><a name="l00292"></a><span class="lineno">  292</span>&#160;    float4 src10 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 1 * src_stride_y)); <span class="comment">// Row1</span></div><div class="line"><a name="l00293"></a><span class="lineno">  293</span>&#160;    float4 src20 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 2 * src_stride_y)); <span class="comment">// Row2</span></div><div class="line"><a name="l00294"></a><span class="lineno">  294</span>&#160;    float4 src30 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 3 * src_stride_y)); <span class="comment">// Row3</span></div><div class="line"><a name="l00295"></a><span class="lineno">  295</span>&#160;    float4 src40 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 4 * src_stride_y)); <span class="comment">// Row3</span></div><div class="line"><a name="l00296"></a><span class="lineno">  296</span>&#160;    float4 src50 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 5 * src_stride_y)); <span class="comment">// Row3</span></div><div class="line"><a name="l00297"></a><span class="lineno">  297</span>&#160;</div><div class="line"><a name="l00298"></a><span class="lineno">  298</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels0, src00, weights_row0);</div><div class="line"><a name="l00299"></a><span class="lineno">  299</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels0, src10, weights_row1);</div><div class="line"><a name="l00300"></a><span class="lineno">  300</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels0, src20, weights_row2);</div><div class="line"><a name="l00301"></a><span class="lineno">  301</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels1, src10, weights_row0);</div><div class="line"><a name="l00302"></a><span class="lineno">  302</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels1, src20, weights_row1);</div><div class="line"><a name="l00303"></a><span class="lineno">  303</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels1, src30, weights_row2);</div><div class="line"><a name="l00304"></a><span class="lineno">  304</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels2, src20, weights_row0);</div><div class="line"><a name="l00305"></a><span class="lineno">  305</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels2, src30, weights_row1);</div><div class="line"><a name="l00306"></a><span class="lineno">  306</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels2, src40, weights_row2);</div><div class="line"><a name="l00307"></a><span class="lineno">  307</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels3, src30, weights_row0);</div><div class="line"><a name="l00308"></a><span class="lineno">  308</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels3, src40, weights_row1);</div><div class="line"><a name="l00309"></a><span class="lineno">  309</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a>(pixels3, src50, weights_row2);</div><div class="line"><a name="l00310"></a><span class="lineno">  310</span>&#160;</div><div class="line"><a name="l00311"></a><span class="lineno">  311</span>&#160;<span class="preprocessor">#ifdef HAS_BIAS</span></div><div class="line"><a name="l00312"></a><span class="lineno">  312</span>&#160;    <a class="code" href="struct_vector.xhtml">Vector</a> biases = <a class="code" href="helpers_8h.xhtml#a64d779f80eeb923e0ab2313433f7b40b">CONVERT_TO_VECTOR_STRUCT_NO_STEP</a>(biases);</div><div class="line"><a name="l00313"></a><span class="lineno">  313</span>&#160;</div><div class="line"><a name="l00314"></a><span class="lineno">  314</span>&#160;    <span class="keywordtype">float</span> bias = *((__global <span class="keywordtype">float</span> *)(<a class="code" href="helpers_8h.xhtml#a7e4940407322d6f0ccb8b6b86b856019">vector_offset</a>(&amp;biases, get_global_id(2))));</div><div class="line"><a name="l00315"></a><span class="lineno">  315</span>&#160;</div><div class="line"><a name="l00316"></a><span class="lineno">  316</span>&#160;    pixels0 += (float2)bias;</div><div class="line"><a name="l00317"></a><span class="lineno">  317</span>&#160;    pixels1 += (float2)bias;</div><div class="line"><a name="l00318"></a><span class="lineno">  318</span>&#160;    pixels2 += (float2)bias;</div><div class="line"><a name="l00319"></a><span class="lineno">  319</span>&#160;    pixels3 += (float2)bias;</div><div class="line"><a name="l00320"></a><span class="lineno">  320</span>&#160;<span class="preprocessor">#endif </span><span class="comment">/* defined(HAS_BIAS) */</span><span class="preprocessor"></span></div><div class="line"><a name="l00321"></a><span class="lineno">  321</span>&#160;</div><div class="line"><a name="l00322"></a><span class="lineno">  322</span>&#160;    vstore2(pixels0, 0, (__global <span class="keywordtype">float</span> *)(dst.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a> + 0 * dst_stride_y));</div><div class="line"><a name="l00323"></a><span class="lineno">  323</span>&#160;    vstore2(pixels1, 0, (__global <span class="keywordtype">float</span> *)(dst.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a> + 1 * dst_stride_y));</div><div class="line"><a name="l00324"></a><span class="lineno">  324</span>&#160;    vstore2(pixels2, 0, (__global <span class="keywordtype">float</span> *)(dst.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a> + 2 * dst_stride_y));</div><div class="line"><a name="l00325"></a><span class="lineno">  325</span>&#160;    vstore2(pixels3, 0, (__global <span class="keywordtype">float</span> *)(dst.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a> + 3 * dst_stride_y));</div><div class="line"><a name="l00326"></a><span class="lineno">  326</span>&#160;}</div><div class="ttc" id="struct_vector_xhtml"><div class="ttname"><a href="struct_vector.xhtml">Vector</a></div><div class="ttdoc">Structure to hold Vector information. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00132">helpers.h:132</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a31c8c760f08fb1a331b16b7c204321dc"><div class="ttname"><a href="helpers_8h.xhtml#a31c8c760f08fb1a331b16b7c204321dc">CONVERT_TO_TENSOR3D_STRUCT</a></div><div class="ttdeci">#define CONVERT_TO_TENSOR3D_STRUCT(name)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00117">helpers.h:117</a></div></div>
+<div class="ttc" id="struct_tensor3_d_xhtml"><div class="ttname"><a href="struct_tensor3_d.xhtml">Tensor3D</a></div><div class="ttdoc">Structure to hold 3D tensor information. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00149">helpers.h:149</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a009469e4d9b8fce3b6d5e97d2077827d"><div class="ttname"><a href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a></div><div class="ttdeci">__global uchar * offset(const Image *img, int x, int y)</div><div class="ttdoc">Get the pointer position of a Image. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00301">helpers.h:301</a></div></div>
+<div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_adbf67dcee294e673cf796f1ed8aeb6a4"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#adbf67dcee294e673cf796f1ed8aeb6a4">arm_compute::test::validation::dst</a></div><div class="ttdeci">CLTensor dst</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_convolution_8cpp_source.xhtml#l00137">Convolution.cpp:137</a></div></div>
+<div class="ttc" id="struct_image_xhtml"><div class="ttname"><a href="struct_image.xhtml">Image</a></div><div class="ttdoc">Structure to hold Image information. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00140">helpers.h:140</a></div></div>
+<div class="ttc" id="depthwise__convolution_8cl_xhtml_aa5ab4cf85fe56de3821727cfcef327d9"><div class="ttname"><a href="depthwise__convolution_8cl.xhtml#aa5ab4cf85fe56de3821727cfcef327d9">CONVOLUTION1x3_BIFROST2X1_STRIDE1</a></div><div class="ttdeci">#define CONVOLUTION1x3_BIFROST2X1_STRIDE1(acc, src0, weights_row0)</div><div class="ttdef"><b>Definition:</b> <a href="depthwise__convolution_8cl_source.xhtml#l00211">depthwise_convolution.cl:211</a></div></div>
+<div class="ttc" id="struct_image_xhtml_acf52c23cbd7424606c10a606524e3e32"><div class="ttname"><a href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">Image::ptr</a></div><div class="ttdeci">__global uchar * ptr</div><div class="ttdoc">Pointer to the starting postion of the buffer. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00142">helpers.h:142</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a64d779f80eeb923e0ab2313433f7b40b"><div class="ttname"><a href="helpers_8h.xhtml#a64d779f80eeb923e0ab2313433f7b40b">CONVERT_TO_VECTOR_STRUCT_NO_STEP</a></div><div class="ttdeci">#define CONVERT_TO_VECTOR_STRUCT_NO_STEP(name)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00099">helpers.h:99</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a7e4940407322d6f0ccb8b6b86b856019"><div class="ttname"><a href="helpers_8h.xhtml#a7e4940407322d6f0ccb8b6b86b856019">vector_offset</a></div><div class="ttdeci">__global const uchar * vector_offset(const Vector *vec, int x)</div><div class="ttdoc">Get the pointer position of a Vector. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00290">helpers.h:290</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a541f8db866a0fa93ee67d58ea31a7d0c"><div class="ttname"><a href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a></div><div class="ttdeci">#define CONVERT_TENSOR3D_TO_IMAGE_STRUCT(name)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00114">helpers.h:114</a></div></div>
+<div class="ttc" id="struct_tensor3_d_xhtml_acf52c23cbd7424606c10a606524e3e32"><div class="ttname"><a href="struct_tensor3_d.xhtml#acf52c23cbd7424606c10a606524e3e32">Tensor3D::ptr</a></div><div class="ttdeci">__global uchar * ptr</div><div class="ttdoc">Pointer to the starting postion of the buffer. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00151">helpers.h:151</a></div></div>
+<div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_a6743f0a130e8311e6f5b1a23df102472"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#a6743f0a130e8311e6f5b1a23df102472">arm_compute::test::validation::src</a></div><div class="ttdeci">convolution configure &amp; src</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_convolution_8cpp_source.xhtml#l00147">Convolution.cpp:147</a></div></div>
+</div><!-- fragment -->
+</div>
+</div>
+<a class="anchor" id="a328ab548c536089e8238fb33de617bec"></a>
+<div class="memitem">
+<div class="memproto">
+      <table class="memname">
+        <tr>
+          <td class="memname">__kernel void depthwise_convolution_3x3_stridex2_stridey2_bifrost </td>
+          <td>(</td>
+          <td class="paramtype">__global uchar *&#160;</td>
+          <td class="paramname"><em>src_ptr</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_stride_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_step_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_stride_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_step_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_stride_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_step_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>src_offset_first_element_in_bytes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">__global uchar *&#160;</td>
+          <td class="paramname"><em>dst_ptr</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_stride_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_step_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_stride_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_step_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_stride_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_step_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>dst_offset_first_element_in_bytes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">__global uchar *&#160;</td>
+          <td class="paramname"><em>weights_ptr</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_stride_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_step_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_stride_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_step_y</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_stride_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_step_z</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>weights_offset_first_element_in_bytes</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">__global uchar *&#160;</td>
+          <td class="paramname"><em>biases_ptr</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>biases_stride_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>biases_step_x</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">uint&#160;</td>
+          <td class="paramname"><em>biases_offset_first_element_in_bytes</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+</div><div class="memdoc">
+
+<p>This OpenCL kernel is optimized for Bifrost architectures and computes the depthwise convolution 3x3 when both stride_x and stride_y are equal to 2. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_ptr</td><td>Pointer to the source image. Supported data types: F32 </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_stride_x</td><td>Stride of the source image in X dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_step_x</td><td>src_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_stride_y</td><td>Stride of the source image in Y dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_step_y</td><td>src_stride_y * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_offset_first_element_in_bytes</td><td>The offset of the first element in the source image </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_stride_z</td><td>Stride of the source tensor in Z dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">src_step_z</td><td>src_stride_z * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_ptr</td><td>Pointer to the destination tensor. Supported data types: F32 </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_stride_x</td><td>Stride of the destination tensor in X dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_step_x</td><td>dst_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_stride_y</td><td>Stride of the destination tensor in Y dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_step_y</td><td>dst_stride_y * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_stride_z</td><td>Stride of the destination tensor in Z dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_step_z</td><td>dst_stride_z * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">dst_offset_first_element_in_bytes</td><td>The offset of the first element in the destination tensor </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_ptr</td><td>Pointer to the weights tensor. Supported data types: F32 </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_stride_x</td><td>Stride of the weights tensor in X dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_step_x</td><td>weights_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_stride_y</td><td>Stride of the weights tensor in Y dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_step_y</td><td>weights_stride_y * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_stride_z</td><td>Stride of the weights tensor in Z dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_step_z</td><td>weights_stride_z * number of elements along Y processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">weights_offset_first_element_in_bytes</td><td>The offset of the first element in the biases vector </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">biases_ptr</td><td>(Optional) Pointer to the biases vector. Supported data types: F32 </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">biases_stride_x</td><td>(Optional) Stride of the biases vector in X dimension (in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">biases_step_x</td><td>(Optional) biases_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
+    <tr><td class="paramdir">[in]</td><td class="paramname">biases_offset_first_element_in_bytes</td><td>(Optional) The offset of the first element in the biases vector </td></tr>
+  </table>
+  </dd>
+</dl>
+
+<p>Definition at line <a class="el" href="depthwise__convolution_8cl_source.xhtml#l00360">360</a> of file <a class="el" href="depthwise__convolution_8cl_source.xhtml">depthwise_convolution.cl</a>.</p>
+
+<p>References <a class="el" href="fixed__point_8h_source.xhtml#l00098">CONVERT</a>, <a class="el" href="helpers_8h_source.xhtml#l00114">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>, <a class="el" href="helpers_8h_source.xhtml#l00117">CONVERT_TO_TENSOR3D_STRUCT</a>, <a class="el" href="helpers_8h_source.xhtml#l00096">CONVERT_TO_VECTOR_STRUCT</a>, <a class="el" href="helpers_8h_source.xhtml#l00099">CONVERT_TO_VECTOR_STRUCT_NO_STEP</a>, <a class="el" href="depthwise__convolution_8cl_source.xhtml#l00221">CONVOLUTION1x3_BIFROST2X1_STRIDE2</a>, <a class="el" href="convolution3x3_8cl_source.xhtml#l00027">DATA_TYPE</a>, <a class="el" href="helpers_8h_source.xhtml#l00066">IMAGE_DECLARATION</a>, <a class="el" href="helpers_8h_source.xhtml#l00301">offset()</a>, <a class="el" href="helpers_8h_source.xhtml#l00134">Vector::ptr</a>, <a class="el" href="helpers_8h_source.xhtml#l00142">Image::ptr</a>, <a class="el" href="helpers_8h_source.xhtml#l00151">Tensor3D::ptr</a>, <a class="el" href="helpers_8h_source.xhtml#l00074">TENSOR3D_DECLARATION</a>, <a class="el" href="helpers_8h_source.xhtml#l00060">VECTOR_DECLARATION</a>, and <a class="el" href="helpers_8h_source.xhtml#l00290">vector_offset()</a>.</p>
+<div class="fragment"><div class="line"><a name="l00369"></a><span class="lineno">  369</span>&#160;{</div><div class="line"><a name="l00370"></a><span class="lineno">  370</span>&#160;    <a class="code" href="struct_image.xhtml">Image</a>    <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#a6743f0a130e8311e6f5b1a23df102472">src</a>     = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(src);</div><div class="line"><a name="l00371"></a><span class="lineno">  371</span>&#160;    <a class="code" href="struct_image.xhtml">Image</a>    <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#adbf67dcee294e673cf796f1ed8aeb6a4">dst</a>     = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(dst);</div><div class="line"><a name="l00372"></a><span class="lineno">  372</span>&#160;    <a class="code" href="struct_tensor3_d.xhtml">Tensor3D</a> weights = <a class="code" href="helpers_8h.xhtml#a31c8c760f08fb1a331b16b7c204321dc">CONVERT_TO_TENSOR3D_STRUCT</a>(weights);</div><div class="line"><a name="l00373"></a><span class="lineno">  373</span>&#160;</div><div class="line"><a name="l00374"></a><span class="lineno">  374</span>&#160;    float2 pixels0 = 0.0f;</div><div class="line"><a name="l00375"></a><span class="lineno">  375</span>&#160;    float2 pixels1 = 0.0f;</div><div class="line"><a name="l00376"></a><span class="lineno">  376</span>&#160;</div><div class="line"><a name="l00377"></a><span class="lineno">  377</span>&#160;    __global uchar *weights_addr = (__global uchar *)weights.<a class="code" href="struct_tensor3_d.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>;</div><div class="line"><a name="l00378"></a><span class="lineno">  378</span>&#160;    __global uchar *src_addr     = (__global uchar *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&amp;src, 0, 0);</div><div class="line"><a name="l00379"></a><span class="lineno">  379</span>&#160;</div><div class="line"><a name="l00380"></a><span class="lineno">  380</span>&#160;    <span class="comment">// Load the weights</span></div><div class="line"><a name="l00381"></a><span class="lineno">  381</span>&#160;    float3 weights_row0 = vload3(0, (__global <span class="keywordtype">float</span> *)(weights_addr + 0 * weights_stride_y));</div><div class="line"><a name="l00382"></a><span class="lineno">  382</span>&#160;    float3 weights_row1 = vload3(0, (__global <span class="keywordtype">float</span> *)(weights_addr + 1 * weights_stride_y));</div><div class="line"><a name="l00383"></a><span class="lineno">  383</span>&#160;    float3 weights_row2 = vload3(0, (__global <span class="keywordtype">float</span> *)(weights_addr + 2 * weights_stride_y));</div><div class="line"><a name="l00384"></a><span class="lineno">  384</span>&#160;</div><div class="line"><a name="l00385"></a><span class="lineno">  385</span>&#160;    <span class="comment">// Note: Since each work-item computes 4x2 elements, we need to load 5 rows from the input tensor</span></div><div class="line"><a name="l00386"></a><span class="lineno">  386</span>&#160;    float4 src00 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 0 * src_stride_y)); <span class="comment">// Row0</span></div><div class="line"><a name="l00387"></a><span class="lineno">  387</span>&#160;    float2 src01 = vload2(2, (__global <span class="keywordtype">float</span> *)(src_addr + 0 * src_stride_y)); <span class="comment">// Row0</span></div><div class="line"><a name="l00388"></a><span class="lineno">  388</span>&#160;    float4 src10 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 1 * src_stride_y)); <span class="comment">// Row1</span></div><div class="line"><a name="l00389"></a><span class="lineno">  389</span>&#160;    float2 src11 = vload2(2, (__global <span class="keywordtype">float</span> *)(src_addr + 1 * src_stride_y)); <span class="comment">// Row1</span></div><div class="line"><a name="l00390"></a><span class="lineno">  390</span>&#160;    float4 src20 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 2 * src_stride_y)); <span class="comment">// Row2</span></div><div class="line"><a name="l00391"></a><span class="lineno">  391</span>&#160;    float2 src21 = vload2(2, (__global <span class="keywordtype">float</span> *)(src_addr + 2 * src_stride_y)); <span class="comment">// Row2</span></div><div class="line"><a name="l00392"></a><span class="lineno">  392</span>&#160;    float4 src30 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 3 * src_stride_y)); <span class="comment">// Row3</span></div><div class="line"><a name="l00393"></a><span class="lineno">  393</span>&#160;    float2 src31 = vload2(2, (__global <span class="keywordtype">float</span> *)(src_addr + 3 * src_stride_y)); <span class="comment">// Row3</span></div><div class="line"><a name="l00394"></a><span class="lineno">  394</span>&#160;    float4 src40 = vload4(0, (__global <span class="keywordtype">float</span> *)(src_addr + 4 * src_stride_y)); <span class="comment">// Row4</span></div><div class="line"><a name="l00395"></a><span class="lineno">  395</span>&#160;    float2 src41 = vload2(2, (__global <span class="keywordtype">float</span> *)(src_addr + 4 * src_stride_y)); <span class="comment">// Row4</span></div><div class="line"><a name="l00396"></a><span class="lineno">  396</span>&#160;</div><div class="line"><a name="l00397"></a><span class="lineno">  397</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#a31cbdd6b28ee6331839432f69d87fdcb">CONVOLUTION1x3_BIFROST2X1_STRIDE2</a>(pixels0, src00, src01, weights_row0);</div><div class="line"><a name="l00398"></a><span class="lineno">  398</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#a31cbdd6b28ee6331839432f69d87fdcb">CONVOLUTION1x3_BIFROST2X1_STRIDE2</a>(pixels0, src10, src11, weights_row1);</div><div class="line"><a name="l00399"></a><span class="lineno">  399</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#a31cbdd6b28ee6331839432f69d87fdcb">CONVOLUTION1x3_BIFROST2X1_STRIDE2</a>(pixels0, src20, src21, weights_row2);</div><div class="line"><a name="l00400"></a><span class="lineno">  400</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#a31cbdd6b28ee6331839432f69d87fdcb">CONVOLUTION1x3_BIFROST2X1_STRIDE2</a>(pixels1, src20, src21, weights_row0);</div><div class="line"><a name="l00401"></a><span class="lineno">  401</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#a31cbdd6b28ee6331839432f69d87fdcb">CONVOLUTION1x3_BIFROST2X1_STRIDE2</a>(pixels1, src30, src31, weights_row1);</div><div class="line"><a name="l00402"></a><span class="lineno">  402</span>&#160;    <a class="code" href="depthwise__convolution_8cl.xhtml#a31cbdd6b28ee6331839432f69d87fdcb">CONVOLUTION1x3_BIFROST2X1_STRIDE2</a>(pixels1, src40, src41, weights_row2);</div><div class="line"><a name="l00403"></a><span class="lineno">  403</span>&#160;</div><div class="line"><a name="l00404"></a><span class="lineno">  404</span>&#160;<span class="preprocessor">#ifdef HAS_BIAS</span></div><div class="line"><a name="l00405"></a><span class="lineno">  405</span>&#160;    <a class="code" href="struct_vector.xhtml">Vector</a> biases = <a class="code" href="helpers_8h.xhtml#a64d779f80eeb923e0ab2313433f7b40b">CONVERT_TO_VECTOR_STRUCT_NO_STEP</a>(biases);</div><div class="line"><a name="l00406"></a><span class="lineno">  406</span>&#160;</div><div class="line"><a name="l00407"></a><span class="lineno">  407</span>&#160;    <span class="keywordtype">float</span> bias = *((__global <span class="keywordtype">float</span> *)(<a class="code" href="helpers_8h.xhtml#a7e4940407322d6f0ccb8b6b86b856019">vector_offset</a>(&amp;biases, get_global_id(2))));</div><div class="line"><a name="l00408"></a><span class="lineno">  408</span>&#160;</div><div class="line"><a name="l00409"></a><span class="lineno">  409</span>&#160;    pixels0 += (float2)bias;</div><div class="line"><a name="l00410"></a><span class="lineno">  410</span>&#160;    pixels1 += (float2)bias;</div><div class="line"><a name="l00411"></a><span class="lineno">  411</span>&#160;<span class="preprocessor">#endif </span><span class="comment">/* defined(HAS_BIAS) */</span><span class="preprocessor"></span></div><div class="line"><a name="l00412"></a><span class="lineno">  412</span>&#160;</div><div class="line"><a name="l00413"></a><span class="lineno">  413</span>&#160;    vstore2(pixels0, 0, (__global <span class="keywordtype">float</span> *)(dst.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a> + 0 * dst_stride_y));</div><div class="line"><a name="l00414"></a><span class="lineno">  414</span>&#160;    vstore2(pixels1, 0, (__global <span class="keywordtype">float</span> *)(dst.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a> + 1 * dst_stride_y));</div><div class="line"><a name="l00415"></a><span class="lineno">  415</span>&#160;}</div><div class="ttc" id="struct_vector_xhtml"><div class="ttname"><a href="struct_vector.xhtml">Vector</a></div><div class="ttdoc">Structure to hold Vector information. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00132">helpers.h:132</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a31c8c760f08fb1a331b16b7c204321dc"><div class="ttname"><a href="helpers_8h.xhtml#a31c8c760f08fb1a331b16b7c204321dc">CONVERT_TO_TENSOR3D_STRUCT</a></div><div class="ttdeci">#define CONVERT_TO_TENSOR3D_STRUCT(name)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00117">helpers.h:117</a></div></div>
+<div class="ttc" id="struct_tensor3_d_xhtml"><div class="ttname"><a href="struct_tensor3_d.xhtml">Tensor3D</a></div><div class="ttdoc">Structure to hold 3D tensor information. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00149">helpers.h:149</a></div></div>
+<div class="ttc" id="depthwise__convolution_8cl_xhtml_a31cbdd6b28ee6331839432f69d87fdcb"><div class="ttname"><a href="depthwise__convolution_8cl.xhtml#a31cbdd6b28ee6331839432f69d87fdcb">CONVOLUTION1x3_BIFROST2X1_STRIDE2</a></div><div class="ttdeci">#define CONVOLUTION1x3_BIFROST2X1_STRIDE2(acc, src0, src1, weights_row0)</div><div class="ttdef"><b>Definition:</b> <a href="depthwise__convolution_8cl_source.xhtml#l00221">depthwise_convolution.cl:221</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a009469e4d9b8fce3b6d5e97d2077827d"><div class="ttname"><a href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a></div><div class="ttdeci">__global uchar * offset(const Image *img, int x, int y)</div><div class="ttdoc">Get the pointer position of a Image. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00301">helpers.h:301</a></div></div>
+<div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_adbf67dcee294e673cf796f1ed8aeb6a4"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#adbf67dcee294e673cf796f1ed8aeb6a4">arm_compute::test::validation::dst</a></div><div class="ttdeci">CLTensor dst</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_convolution_8cpp_source.xhtml#l00137">Convolution.cpp:137</a></div></div>
+<div class="ttc" id="struct_image_xhtml"><div class="ttname"><a href="struct_image.xhtml">Image</a></div><div class="ttdoc">Structure to hold Image information. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00140">helpers.h:140</a></div></div>
+<div class="ttc" id="struct_image_xhtml_acf52c23cbd7424606c10a606524e3e32"><div class="ttname"><a href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">Image::ptr</a></div><div class="ttdeci">__global uchar * ptr</div><div class="ttdoc">Pointer to the starting postion of the buffer. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00142">helpers.h:142</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a64d779f80eeb923e0ab2313433f7b40b"><div class="ttname"><a href="helpers_8h.xhtml#a64d779f80eeb923e0ab2313433f7b40b">CONVERT_TO_VECTOR_STRUCT_NO_STEP</a></div><div class="ttdeci">#define CONVERT_TO_VECTOR_STRUCT_NO_STEP(name)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00099">helpers.h:99</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a7e4940407322d6f0ccb8b6b86b856019"><div class="ttname"><a href="helpers_8h.xhtml#a7e4940407322d6f0ccb8b6b86b856019">vector_offset</a></div><div class="ttdeci">__global const uchar * vector_offset(const Vector *vec, int x)</div><div class="ttdoc">Get the pointer position of a Vector. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00290">helpers.h:290</a></div></div>
+<div class="ttc" id="helpers_8h_xhtml_a541f8db866a0fa93ee67d58ea31a7d0c"><div class="ttname"><a href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a></div><div class="ttdeci">#define CONVERT_TENSOR3D_TO_IMAGE_STRUCT(name)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00114">helpers.h:114</a></div></div>
+<div class="ttc" id="struct_tensor3_d_xhtml_acf52c23cbd7424606c10a606524e3e32"><div class="ttname"><a href="struct_tensor3_d.xhtml#acf52c23cbd7424606c10a606524e3e32">Tensor3D::ptr</a></div><div class="ttdeci">__global uchar * ptr</div><div class="ttdoc">Pointer to the starting postion of the buffer. </div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00151">helpers.h:151</a></div></div>
+<div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_a6743f0a130e8311e6f5b1a23df102472"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#a6743f0a130e8311e6f5b1a23df102472">arm_compute::test::validation::src</a></div><div class="ttdeci">convolution configure &amp; src</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_convolution_8cpp_source.xhtml#l00147">Convolution.cpp:147</a></div></div>
+</div><!-- fragment -->
+</div>
+</div>
 </div><!-- contents -->
 </div><!-- doc-content -->
 <!-- start footer part -->
 <div id="nav-path" class="navpath"><!-- id is needed for treeview function! -->
   <ul>
     <li class="navelem"><a class="el" href="dir_68267d1309a1af8e8297ef4c3efbcdba.xhtml">src</a></li><li class="navelem"><a class="el" href="dir_aebb8dcc11953d78e620bbef0b9e2183.xhtml">core</a></li><li class="navelem"><a class="el" href="dir_8c278f79c760e5c5fbd911f9870614c1.xhtml">CL</a></li><li class="navelem"><a class="el" href="dir_25885286e9dad4fa105b7b25a8031bbf.xhtml">cl_kernels</a></li><li class="navelem"><a class="el" href="depthwise__convolution_8cl.xhtml">depthwise_convolution.cl</a></li>
-    <li class="footer">Generated on Wed Jan 24 2018 14:30:43 for Compute Library by
+    <li class="footer">Generated on Thu Feb 22 2018 15:45:22 for Compute Library by
     <a href="http://www.doxygen.org/index.html">
     <img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.8.11 </li>
   </ul>