arm_compute v18.02
Change-Id: I7207aa488e5470f235f39b6c188b4678dc38d1a6
diff --git a/documentation/softmax__layer_8cl.xhtml b/documentation/softmax__layer_8cl.xhtml
index e9579f3..1cf3e6d 100644
--- a/documentation/softmax__layer_8cl.xhtml
+++ b/documentation/softmax__layer_8cl.xhtml
@@ -40,7 +40,7 @@
<tr style="height: 56px;">
<td style="padding-left: 0.5em;">
<div id="projectname">Compute Library
-  <span id="projectnumber">18.01</span>
+  <span id="projectnumber">18.02</span>
</div>
</td>
</tr>
@@ -152,12 +152,6 @@
</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:a9e86f72ca0d249cb88c36bd50620da8a"><td class="memItemLeft" align="right" valign="top">__kernel void </td><td class="memItemRight" valign="bottom"><a class="el" href="softmax__layer_8cl.xhtml#a9e86f72ca0d249cb88c36bd50620da8a">softmax_layer_max</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, uint width)</td></tr>
-<tr class="memdesc:a9e86f72ca0d249cb88c36bd50620da8a"><td class="mdescLeft"> </td><td class="mdescRight">Identifies the maximum value across the 1st dimension. <a href="#a9e86f72ca0d249cb88c36bd50620da8a">More...</a><br /></td></tr>
-<tr class="separator:a9e86f72ca0d249cb88c36bd50620da8a"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:aa4412bb319c67deaeab07a481d93dfdb"><td class="memItemLeft" align="right" valign="top">__kernel void </td><td class="memItemRight" valign="bottom"><a class="el" href="softmax__layer_8cl.xhtml#aa4412bb319c67deaeab07a481d93dfdb">softmax_layer_shift_exp_sum</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 *max_ptr, uint max_stride_x, uint max_step_x, uint max_stride_y, uint max_step_y, uint max_stride_z, uint max_step_z, uint max_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 *sum_ptr, uint sum_stride_x, uint sum_step_x, uint sum_stride_y, uint sum_step_y, uint sum_stride_z, uint sum_step_z, uint sum_offset_first_element_in_bytes, uint width)</td></tr>
-<tr class="memdesc:aa4412bb319c67deaeab07a481d93dfdb"><td class="mdescLeft"> </td><td class="mdescRight">Shifts the values of the input tensor by the max calculated in softmax_layer_max kernel, then gets the exponent of each element as sums all elements across each row. <a href="#aa4412bb319c67deaeab07a481d93dfdb">More...</a><br /></td></tr>
-<tr class="separator:aa4412bb319c67deaeab07a481d93dfdb"><td class="memSeparator" colspan="2"> </td></tr>
<tr class="memitem:ac4247ac0991e85965b7ded764e78f12c"><td class="memItemLeft" align="right" valign="top">__kernel void </td><td class="memItemRight" valign="bottom"><a class="el" href="softmax__layer_8cl.xhtml#ac4247ac0991e85965b7ded764e78f12c">softmax_layer_norm</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 *sum_ptr, uint sum_stride_x, uint sum_step_x, uint sum_stride_y, uint sum_step_y, uint sum_stride_z, uint sum_step_z, uint sum_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)</td></tr>
<tr class="memdesc:ac4247ac0991e85965b7ded764e78f12c"><td class="mdescLeft"> </td><td class="mdescRight">Divides all the values of the input tensor by the sum calculated from softmax_layer_shift_exp_sum kernel. <a href="#ac4247ac0991e85965b7ded764e78f12c">More...</a><br /></td></tr>
<tr class="separator:ac4247ac0991e85965b7ded764e78f12c"><td class="memSeparator" colspan="2"> </td></tr>
@@ -220,7 +214,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00044">44</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">softmax_layer_max_shift_exp_sum_parallel()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00196">softmax_layer_shift_exp_sum()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">softmax_layer_max_shift_exp_sum_parallel()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
@@ -262,7 +256,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00047">47</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00292">softmax_layer_norm()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00120">softmax_layer_norm()</a>.</p>
</div>
</div>
@@ -298,7 +292,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00048">48</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">softmax_layer_max_shift_exp_sum_parallel()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00196">softmax_layer_shift_exp_sum()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">softmax_layer_max_shift_exp_sum_parallel()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
@@ -314,7 +308,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00062">62</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">softmax_layer_max_shift_exp_sum_parallel()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">softmax_layer_max_shift_exp_sum_parallel()</a>.</p>
</div>
</div>
@@ -330,7 +324,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00080">80</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
@@ -372,7 +366,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00043">43</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00114">softmax_layer_max()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">softmax_layer_max_shift_exp_sum_parallel()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">softmax_layer_max_shift_exp_sum_parallel()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
@@ -428,7 +422,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00046">46</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">softmax_layer_max_shift_exp_sum_parallel()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00196">softmax_layer_shift_exp_sum()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">softmax_layer_max_shift_exp_sum_parallel()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
@@ -444,7 +438,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00055">55</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00114">softmax_layer_max()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">softmax_layer_max_shift_exp_sum_parallel()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00196">softmax_layer_shift_exp_sum()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">softmax_layer_max_shift_exp_sum_parallel()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
@@ -486,7 +480,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00045">45</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">softmax_layer_max_shift_exp_sum_parallel()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00196">softmax_layer_shift_exp_sum()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">softmax_layer_max_shift_exp_sum_parallel()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
@@ -502,173 +496,11 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00079">79</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="gemm_8cl_source.xhtml#l00223">gemm_interleave4x4_8bit()</a>, <a class="el" href="convolution__layer_8cl_source.xhtml#l00306">im2col_reduced()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
<h2 class="groupheader">Function Documentation</h2>
-<a class="anchor" id="a9e86f72ca0d249cb88c36bd50620da8a"></a>
-<div class="memitem">
-<div class="memproto">
- <table class="memname">
- <tr>
- <td class="memname">__kernel void softmax_layer_max </td>
- <td>(</td>
- <td class="paramtype">__global uchar * </td>
- <td class="paramname"><em>src_ptr</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_stride_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_step_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_stride_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_step_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_stride_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_step_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </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 * </td>
- <td class="paramname"><em>dst_ptr</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_stride_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_step_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_stride_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_step_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_stride_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_step_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_offset_first_element_in_bytes</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>width</em> </td>
- </tr>
- <tr>
- <td></td>
- <td>)</td>
- <td></td><td></td>
- </tr>
- </table>
-</div><div class="memdoc">
-
-<p>Identifies the maximum value across the 1st dimension. </p>
-<dl class="section note"><dt>Note</dt><dd>Datatype must be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=short </dd>
-<dd>
-Fixed point position must be given as a preprocessor argument using -DFIXED_POINT_POSITION=pos. e.g. DFIXED_POINT_POSITION=4 </dd>
-<dd>
-In case the input is not multiple of 16 -DNON_MULTIPLE_OF_16 must be passed.</dd></dl>
-<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 tensor slice. Supported data types: QS8/QS16/F16/F32 </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">src_stride_x</td><td>Stride of the source tensor 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 tensor 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_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 Z 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 tensor </td></tr>
- <tr><td class="paramdir">[out]</td><td class="paramname">dst_ptr</td><td>Pointer to the destination tensor slice. Supported data types: same as <code>src_ptr</code> </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 Z 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">width</td><td>Input image width </td></tr>
- </table>
- </dd>
-</dl>
-
-<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00114">114</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.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="softmax__layer_8cl_source.xhtml#l00087">idx16</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00043">MAX_OP</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="softmax__layer_8cl_source.xhtml#l00055">SELECT_DATA_TYPE</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00086">type_min</a>, and <a class="el" href="fixed__point_8h_source.xhtml#l00093">VEC_DATA_TYPE</a>.</p>
-<div class="fragment"><div class="line"><a name="l00118"></a><span class="lineno"> 118</span> {</div><div class="line"><a name="l00119"></a><span class="lineno"> 119</span>  <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="l00120"></a><span class="lineno"> 120</span>  <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="l00121"></a><span class="lineno"> 121</span> </div><div class="line"><a name="l00122"></a><span class="lineno"> 122</span>  <span class="comment">// Initialize local maximum</span></div><div class="line"><a name="l00123"></a><span class="lineno"> 123</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 16)</div><div class="line"><a name="l00124"></a><span class="lineno"> 124</span>  max_val = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 16))<a class="code" href="softmax__layer_8cl.xhtml#a538b4b63f40e7b12891774e03a4f0dec">type_min</a>;</div><div class="line"><a name="l00125"></a><span class="lineno"> 125</span> </div><div class="line"><a name="l00126"></a><span class="lineno"> 126</span>  <span class="comment">// Calculate max of row</span></div><div class="line"><a name="l00127"></a><span class="lineno"> 127</span>  const uint width4 = width >> 4;</div><div class="line"><a name="l00128"></a><span class="lineno"> 128</span>  for(uint i = 0; i < width4; i++)</div><div class="line"><a name="l00129"></a><span class="lineno"> 129</span>  {</div><div class="line"><a name="l00130"></a><span class="lineno"> 130</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 16)</div><div class="line"><a name="l00131"></a><span class="lineno"> 131</span>  data = vload16(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i << 4, 0));</div><div class="line"><a name="l00132"></a><span class="lineno"> 132</span>  max_val = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(data, max_val, DATA_TYPE, 16);</div><div class="line"><a name="l00133"></a><span class="lineno"> 133</span>  }</div><div class="line"><a name="l00134"></a><span class="lineno"> 134</span> </div><div class="line"><a name="l00135"></a><span class="lineno"> 135</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_16</span></div><div class="line"><a name="l00136"></a><span class="lineno"> 136</span>  <span class="comment">// Handle non multiple of 16</span></div><div class="line"><a name="l00137"></a><span class="lineno"> 137</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 16)</div><div class="line"><a name="l00138"></a><span class="lineno"> 138</span>  data = vload16(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, width4 << 4, 0));</div><div class="line"><a name="l00139"></a><span class="lineno"> 139</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a>, 16)</div><div class="line"><a name="l00140"></a><span class="lineno"> 140</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>(((uint16)(width4 << 4) + <a class="code" href="softmax__layer_8cl.xhtml#a0712735973f172ac9efc7d48a31e47ad">idx16</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, 16));</div><div class="line"><a name="l00141"></a><span class="lineno"> 141</span>  max_val = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val, select(type_min, data, widx), DATA_TYPE, 16);</div><div class="line"><a name="l00142"></a><span class="lineno"> 142</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_16 */</span><span class="preprocessor"></span></div><div class="line"><a name="l00143"></a><span class="lineno"> 143</span> </div><div class="line"><a name="l00144"></a><span class="lineno"> 144</span>  <span class="comment">// Perform max reduction</span></div><div class="line"><a name="l00145"></a><span class="lineno"> 145</span>  max_val.s01234567 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val.s01234567, max_val.s89ABCDEF, DATA_TYPE, 8);</div><div class="line"><a name="l00146"></a><span class="lineno"> 146</span>  max_val.s0123 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val.s0123, max_val.s4567, DATA_TYPE, 4);</div><div class="line"><a name="l00147"></a><span class="lineno"> 147</span>  max_val.s01 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val.s01, max_val.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00148"></a><span class="lineno"> 148</span>  max_val.s0 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val.s0, max_val.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00149"></a><span class="lineno"> 149</span> </div><div class="line"><a name="l00150"></a><span class="lineno"> 150</span>  <span class="comment">// Store result</span></div><div class="line"><a name="l00151"></a><span class="lineno"> 151</span>  *((__global DATA_TYPE *)dst.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>) = max_val.s0;</div><div class="line"><a name="l00152"></a><span class="lineno"> 152</span> }</div><div class="ttc" id="fixed__point_8h_xhtml_aa8d95ba04fc73845abc6045952cae5be"><div class="ttname"><a href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a></div><div class="ttdeci">#define CONVERT(x, type)</div><div class="ttdef"><b>Definition:</b> <a href="fixed__point_8h_source.xhtml#l00098">fixed_point.h:98</a></div></div>
-<div class="ttc" id="convolution3x3_8cl_xhtml_afb8c72ce35c4a1f4a2588d6573e54aa1"><div class="ttname"><a href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a></div><div class="ttdeci">#define DATA_TYPE</div><div class="ttdef"><b>Definition:</b> <a href="convolution3x3_8cl_source.xhtml#l00027">convolution3x3.cl:27</a></div></div>
-<div class="ttc" id="softmax__layer_8cl_xhtml_abaa48ad818c44e415fd3f9dd0f27bf01"><div class="ttname"><a href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a></div><div class="ttdeci">#define MAX_OP(x, y, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00043">softmax_layer.cl:43</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="softmax__layer_8cl_xhtml_af5b2e33e3c5fcaab3a213f26c2300170"><div class="ttname"><a href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a></div><div class="ttdeci">#define SELECT_DATA_TYPE</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00055">softmax_layer.cl:55</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#l00123">Convolution.cpp:123</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="fixed__point_8h_xhtml_a36f754c05b6fddf6df0d8d0a74f8159f"><div class="ttname"><a href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a></div><div class="ttdeci">#define VEC_DATA_TYPE(type, size)</div><div class="ttdef"><b>Definition:</b> <a href="fixed__point_8h_source.xhtml#l00093">fixed_point.h:93</a></div></div>
-<div class="ttc" id="softmax__layer_8cl_xhtml_a0712735973f172ac9efc7d48a31e47ad"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a0712735973f172ac9efc7d48a31e47ad">idx16</a></div><div class="ttdeci">__constant uint16 idx16</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00087">softmax_layer.cl:87</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="softmax__layer_8cl_xhtml_a538b4b63f40e7b12891774e03a4f0dec"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a538b4b63f40e7b12891774e03a4f0dec">type_min</a></div><div class="ttdeci">__constant DATA_TYPE16 type_min</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00086">softmax_layer.cl:86</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 & src</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_convolution_8cpp_source.xhtml#l00133">Convolution.cpp:133</a></div></div>
-</div><!-- fragment -->
-</div>
-</div>
<a class="anchor" id="a82bc360279fcf7cf8033dca6022f21f5"></a>
<div class="memitem">
<div class="memproto">
@@ -926,10 +758,10 @@
</dd>
</dl>
-<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">500</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
+<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">328</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>References <a class="el" href="softmax__layer_8cl_source.xhtml#l00044">ADD_OP</a>, <a class="el" href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00112">arm_compute::test::validation::beta</a>, <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="_c_l_2_convolution_8cpp_source.xhtml#l00123">arm_compute::test::validation::dst</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00048">EXP_OP</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00062">GRID_SIZE</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00088">idx4</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00043">MAX_OP</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00046">MUL_OP</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="softmax__layer_8cl_source.xhtml#l00055">SELECT_DATA_TYPE</a>, <a class="el" href="_c_l_2_convolution_8cpp_source.xhtml#l00133">arm_compute::test::validation::src</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00045">SUB_OP</a>, <a class="el" href="reduction__operation_8cl_source.xhtml#l00052">sum()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00081">type_min_</a>, <a class="el" href="fixed__point_8h_source.xhtml#l00093">VEC_DATA_TYPE</a>, <a class="el" href="helpers_8h_source.xhtml#l00040">VLOAD</a>, and <a class="el" href="helpers_8h_source.xhtml#l00043">VSTORE</a>.</p>
-<div class="fragment"><div class="line"><a name="l00506"></a><span class="lineno"> 506</span> {</div><div class="line"><a name="l00507"></a><span class="lineno"> 507</span>  <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="l00508"></a><span class="lineno"> 508</span>  <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="l00509"></a><span class="lineno"> 509</span>  <a class="code" href="struct_image.xhtml">Image</a> maxo = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(maxo);</div><div class="line"><a name="l00510"></a><span class="lineno"> 510</span>  <a class="code" href="struct_image.xhtml">Image</a> <a class="code" href="reduction__operation_8cl.xhtml#ab0df00f5333da51860deb93deb44a782">sum</a> = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(sum);</div><div class="line"><a name="l00511"></a><span class="lineno"> 511</span> </div><div class="line"><a name="l00512"></a><span class="lineno"> 512</span>  <span class="keyword">const</span> uint lid = get_local_id(0);</div><div class="line"><a name="l00513"></a><span class="lineno"> 513</span> </div><div class="line"><a name="l00514"></a><span class="lineno"> 514</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00515"></a><span class="lineno"> 515</span>  <span class="comment">// Initialize beta</span></div><div class="line"><a name="l00516"></a><span class="lineno"> 516</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 4)</div><div class="line"><a name="l00517"></a><span class="lineno"> 517</span>  <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a> = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 4))BETA;</div><div class="line"><a name="l00518"></a><span class="lineno"> 518</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00519"></a><span class="lineno"> 519</span> </div><div class="line"><a name="l00520"></a><span class="lineno"> 520</span>  <span class="comment">// Define one temporary vector per work-item.</span></div><div class="line"><a name="l00521"></a><span class="lineno"> 521</span>  __local <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4) tmp_local[<a class="code" href="softmax__layer_8cl.xhtml#a08246606c233e7785a497c09672f366f">GRID_SIZE</a>];</div><div class="line"><a name="l00522"></a><span class="lineno"> 522</span>  __local DATA_TYPE max_local;</div><div class="line"><a name="l00523"></a><span class="lineno"> 523</span> </div><div class="line"><a name="l00524"></a><span class="lineno"> 524</span>  __constant <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4) type_min4 = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4))(<a class="code" href="softmax__layer_8cl.xhtml#a80b22c555ddadb47cc6ca338a9c49126">MINVAL</a>);</div><div class="line"><a name="l00525"></a><span class="lineno"> 525</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00526"></a><span class="lineno"> 526</span>  max_val_vec = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4))type_min4;</div><div class="line"><a name="l00527"></a><span class="lineno"> 527</span>  <span class="comment">// Number of elements per work-item.</span></div><div class="line"><a name="l00528"></a><span class="lineno"> 528</span>  const uint row = width / GRID_SIZE;</div><div class="line"><a name="l00529"></a><span class="lineno"> 529</span>  <span class="comment">// Number of iterations per work-item.</span></div><div class="line"><a name="l00530"></a><span class="lineno"> 530</span>  const uint width_ = row >> 2;</div><div class="line"><a name="l00531"></a><span class="lineno"> 531</span>  <span class="comment">// Calculate max of row</span></div><div class="line"><a name="l00532"></a><span class="lineno"> 532</span>  uint i = 0;</div><div class="line"><a name="l00533"></a><span class="lineno"> 533</span>  for(; i < width_; i++)</div><div class="line"><a name="l00534"></a><span class="lineno"> 534</span>  {</div><div class="line"><a name="l00535"></a><span class="lineno"> 535</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00536"></a><span class="lineno"> 536</span>  data_max = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00537"></a><span class="lineno"> 537</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(data_max, max_val_vec, DATA_TYPE, 4);</div><div class="line"><a name="l00538"></a><span class="lineno"> 538</span>  }</div><div class="line"><a name="l00539"></a><span class="lineno"> 539</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_GRID_SIZE</span></div><div class="line"><a name="l00540"></a><span class="lineno"> 540</span>  <span class="comment">// How many work-items needed to complete the computation.</span></div><div class="line"><a name="l00541"></a><span class="lineno"> 541</span>  <span class="keywordtype">int</span> boundary_workitems = (width % (GRID_SIZE * 4)) / 4;</div><div class="line"><a name="l00542"></a><span class="lineno"> 542</span>  <span class="keywordflow">if</span>(lid < boundary_workitems)</div><div class="line"><a name="l00543"></a><span class="lineno"> 543</span>  {</div><div class="line"><a name="l00544"></a><span class="lineno"> 544</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00545"></a><span class="lineno"> 545</span>  data_max = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00546"></a><span class="lineno"> 546</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(data_max, max_val_vec, DATA_TYPE, 4);</div><div class="line"><a name="l00547"></a><span class="lineno"> 547</span>  }</div><div class="line"><a name="l00548"></a><span class="lineno"> 548</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_VECTOR_SIZE</span></div><div class="line"><a name="l00549"></a><span class="lineno"> 549</span>  <span class="keywordflow">if</span>(boundary_workitems == 0)</div><div class="line"><a name="l00550"></a><span class="lineno"> 550</span>  {</div><div class="line"><a name="l00551"></a><span class="lineno"> 551</span>  boundary_workitems = <a class="code" href="softmax__layer_8cl.xhtml#a08246606c233e7785a497c09672f366f">GRID_SIZE</a>;</div><div class="line"><a name="l00552"></a><span class="lineno"> 552</span>  i--;</div><div class="line"><a name="l00553"></a><span class="lineno"> 553</span>  }</div><div class="line"><a name="l00554"></a><span class="lineno"> 554</span>  <span class="keywordflow">if</span>(lid == (boundary_workitems - 1))</div><div class="line"><a name="l00555"></a><span class="lineno"> 555</span>  {</div><div class="line"><a name="l00556"></a><span class="lineno"> 556</span>  <span class="comment">// Handle non multiple of 4</span></div><div class="line"><a name="l00557"></a><span class="lineno"> 557</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00558"></a><span class="lineno"> 558</span>  data_max = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, (GRID_SIZE * i * 4) + 4, 0));</div><div class="line"><a name="l00559"></a><span class="lineno"> 559</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a>, 4)</div><div class="line"><a name="l00560"></a><span class="lineno"> 560</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>(((uint4)(GRID_SIZE * i * 4) + boundary_workitems * 4 + <a class="code" href="softmax__layer_8cl.xhtml#a4884a666a1e93fbf8c27bd7d2da3c8bb">idx4</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, 4));</div><div class="line"><a name="l00561"></a><span class="lineno"> 561</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec, select(<a class="code" href="softmax__layer_8cl.xhtml#af7a08044d0e491a0ee1520a24a107a2b">type_min_</a>, data_max, widx), DATA_TYPE, 4);</div><div class="line"><a name="l00562"></a><span class="lineno"> 562</span>  }</div><div class="line"><a name="l00563"></a><span class="lineno"> 563</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_VECTOR_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00564"></a><span class="lineno"> 564</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_GRID_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00565"></a><span class="lineno"> 565</span>  tmp_local[lid] = max_val_vec;</div><div class="line"><a name="l00566"></a><span class="lineno"> 566</span> </div><div class="line"><a name="l00567"></a><span class="lineno"> 567</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00568"></a><span class="lineno"> 568</span> </div><div class="line"><a name="l00569"></a><span class="lineno"> 569</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 256)</div><div class="line"><a name="l00570"></a><span class="lineno"> 570</span>  {</div><div class="line"><a name="l00571"></a><span class="lineno"> 571</span>  <span class="keywordflow">if</span>(lid < 128)</div><div class="line"><a name="l00572"></a><span class="lineno"> 572</span>  {</div><div class="line"><a name="l00573"></a><span class="lineno"> 573</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 128], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00574"></a><span class="lineno"> 574</span>  }</div><div class="line"><a name="l00575"></a><span class="lineno"> 575</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00576"></a><span class="lineno"> 576</span>  }</div><div class="line"><a name="l00577"></a><span class="lineno"> 577</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 128)</div><div class="line"><a name="l00578"></a><span class="lineno"> 578</span>  {</div><div class="line"><a name="l00579"></a><span class="lineno"> 579</span>  <span class="keywordflow">if</span>(lid < 64)</div><div class="line"><a name="l00580"></a><span class="lineno"> 580</span>  {</div><div class="line"><a name="l00581"></a><span class="lineno"> 581</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 64], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00582"></a><span class="lineno"> 582</span>  }</div><div class="line"><a name="l00583"></a><span class="lineno"> 583</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00584"></a><span class="lineno"> 584</span>  }</div><div class="line"><a name="l00585"></a><span class="lineno"> 585</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 64)</div><div class="line"><a name="l00586"></a><span class="lineno"> 586</span>  {</div><div class="line"><a name="l00587"></a><span class="lineno"> 587</span>  <span class="keywordflow">if</span>(lid < 32)</div><div class="line"><a name="l00588"></a><span class="lineno"> 588</span>  {</div><div class="line"><a name="l00589"></a><span class="lineno"> 589</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 32], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00590"></a><span class="lineno"> 590</span>  }</div><div class="line"><a name="l00591"></a><span class="lineno"> 591</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00592"></a><span class="lineno"> 592</span>  }</div><div class="line"><a name="l00593"></a><span class="lineno"> 593</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 32)</div><div class="line"><a name="l00594"></a><span class="lineno"> 594</span>  {</div><div class="line"><a name="l00595"></a><span class="lineno"> 595</span>  <span class="keywordflow">if</span>(lid < 16)</div><div class="line"><a name="l00596"></a><span class="lineno"> 596</span>  {</div><div class="line"><a name="l00597"></a><span class="lineno"> 597</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 16], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00598"></a><span class="lineno"> 598</span>  }</div><div class="line"><a name="l00599"></a><span class="lineno"> 599</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00600"></a><span class="lineno"> 600</span>  }</div><div class="line"><a name="l00601"></a><span class="lineno"> 601</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 16)</div><div class="line"><a name="l00602"></a><span class="lineno"> 602</span>  {</div><div class="line"><a name="l00603"></a><span class="lineno"> 603</span>  <span class="keywordflow">if</span>(lid < 8)</div><div class="line"><a name="l00604"></a><span class="lineno"> 604</span>  {</div><div class="line"><a name="l00605"></a><span class="lineno"> 605</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 8], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00606"></a><span class="lineno"> 606</span>  }</div><div class="line"><a name="l00607"></a><span class="lineno"> 607</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00608"></a><span class="lineno"> 608</span>  }</div><div class="line"><a name="l00609"></a><span class="lineno"> 609</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 8)</div><div class="line"><a name="l00610"></a><span class="lineno"> 610</span>  {</div><div class="line"><a name="l00611"></a><span class="lineno"> 611</span>  <span class="keywordflow">if</span>(lid < 4)</div><div class="line"><a name="l00612"></a><span class="lineno"> 612</span>  {</div><div class="line"><a name="l00613"></a><span class="lineno"> 613</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 4], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00614"></a><span class="lineno"> 614</span>  }</div><div class="line"><a name="l00615"></a><span class="lineno"> 615</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00616"></a><span class="lineno"> 616</span>  }</div><div class="line"><a name="l00617"></a><span class="lineno"> 617</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 4)</div><div class="line"><a name="l00618"></a><span class="lineno"> 618</span>  {</div><div class="line"><a name="l00619"></a><span class="lineno"> 619</span>  <span class="keywordflow">if</span>(lid < 2)</div><div class="line"><a name="l00620"></a><span class="lineno"> 620</span>  {</div><div class="line"><a name="l00621"></a><span class="lineno"> 621</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 2], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00622"></a><span class="lineno"> 622</span>  }</div><div class="line"><a name="l00623"></a><span class="lineno"> 623</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00624"></a><span class="lineno"> 624</span>  }</div><div class="line"><a name="l00625"></a><span class="lineno"> 625</span>  <span class="keywordflow">if</span>(lid == 0)</div><div class="line"><a name="l00626"></a><span class="lineno"> 626</span>  {</div><div class="line"><a name="l00627"></a><span class="lineno"> 627</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 1], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00628"></a><span class="lineno"> 628</span>  max_val_vec.s01 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s01, max_val_vec.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00629"></a><span class="lineno"> 629</span>  max_val_vec.s0 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s0, max_val_vec.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00630"></a><span class="lineno"> 630</span>  max_local = max_val_vec.s0;</div><div class="line"><a name="l00631"></a><span class="lineno"> 631</span>  }</div><div class="line"><a name="l00632"></a><span class="lineno"> 632</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00633"></a><span class="lineno"> 633</span> </div><div class="line"><a name="l00634"></a><span class="lineno"> 634</span>  <span class="comment">/* Second section */</span></div><div class="line"><a name="l00635"></a><span class="lineno"> 635</span> </div><div class="line"><a name="l00636"></a><span class="lineno"> 636</span>  <span class="comment">// Set sum vector</span></div><div class="line"><a name="l00637"></a><span class="lineno"> 637</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00638"></a><span class="lineno"> 638</span>  sum1D = 0;</div><div class="line"><a name="l00639"></a><span class="lineno"> 639</span>  DATA_TYPE max_val = max_local;</div><div class="line"><a name="l00640"></a><span class="lineno"> 640</span> </div><div class="line"><a name="l00641"></a><span class="lineno"> 641</span>  <span class="comment">// Shift values, exp and sum</span></div><div class="line"><a name="l00642"></a><span class="lineno"> 642</span>  for(i = 0; i < width_; i++)</div><div class="line"><a name="l00643"></a><span class="lineno"> 643</span>  {</div><div class="line"><a name="l00644"></a><span class="lineno"> 644</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00645"></a><span class="lineno"> 645</span>  data = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00646"></a><span class="lineno"> 646</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, 4);</div><div class="line"><a name="l00647"></a><span class="lineno"> 647</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00648"></a><span class="lineno"> 648</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, 4);</div><div class="line"><a name="l00649"></a><span class="lineno"> 649</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00650"></a><span class="lineno"> 650</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, 4);</div><div class="line"><a name="l00651"></a><span class="lineno"> 651</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(4)</div><div class="line"><a name="l00652"></a><span class="lineno"> 652</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00653"></a><span class="lineno"> 653</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, 4);</div><div class="line"><a name="l00654"></a><span class="lineno"> 654</span>  }</div><div class="line"><a name="l00655"></a><span class="lineno"> 655</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_GRID_SIZE</span></div><div class="line"><a name="l00656"></a><span class="lineno"> 656</span>  boundary_workitems = (width % (GRID_SIZE * 4)) / 4;</div><div class="line"><a name="l00657"></a><span class="lineno"> 657</span>  <span class="keywordflow">if</span>(lid < boundary_workitems)</div><div class="line"><a name="l00658"></a><span class="lineno"> 658</span>  {</div><div class="line"><a name="l00659"></a><span class="lineno"> 659</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00660"></a><span class="lineno"> 660</span>  data = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00661"></a><span class="lineno"> 661</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, 4);</div><div class="line"><a name="l00662"></a><span class="lineno"> 662</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00663"></a><span class="lineno"> 663</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, 4);</div><div class="line"><a name="l00664"></a><span class="lineno"> 664</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00665"></a><span class="lineno"> 665</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, 4);</div><div class="line"><a name="l00666"></a><span class="lineno"> 666</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(4)</div><div class="line"><a name="l00667"></a><span class="lineno"> 667</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00668"></a><span class="lineno"> 668</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, 4);</div><div class="line"><a name="l00669"></a><span class="lineno"> 669</span>  }</div><div class="line"><a name="l00670"></a><span class="lineno"> 670</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_VECTOR_SIZE</span></div><div class="line"><a name="l00671"></a><span class="lineno"> 671</span>  <span class="keywordflow">if</span>(boundary_workitems == 0)</div><div class="line"><a name="l00672"></a><span class="lineno"> 672</span>  {</div><div class="line"><a name="l00673"></a><span class="lineno"> 673</span>  boundary_workitems = <a class="code" href="softmax__layer_8cl.xhtml#a08246606c233e7785a497c09672f366f">GRID_SIZE</a>;</div><div class="line"><a name="l00674"></a><span class="lineno"> 674</span>  i--;</div><div class="line"><a name="l00675"></a><span class="lineno"> 675</span>  }</div><div class="line"><a name="l00676"></a><span class="lineno"> 676</span>  <span class="keywordflow">if</span>(lid == (boundary_workitems - 1))</div><div class="line"><a name="l00677"></a><span class="lineno"> 677</span>  {</div><div class="line"><a name="l00678"></a><span class="lineno"> 678</span>  <span class="comment">// Handle non multiple of vector size ((GRID_SIZE * i * 4) + 4, 0); move 4 float positions ahead, *4 is due to the stride</span></div><div class="line"><a name="l00679"></a><span class="lineno"> 679</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00680"></a><span class="lineno"> 680</span>  data = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, (GRID_SIZE * i * 4) + 4, 0));</div><div class="line"><a name="l00681"></a><span class="lineno"> 681</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, 4);</div><div class="line"><a name="l00682"></a><span class="lineno"> 682</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00683"></a><span class="lineno"> 683</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, 4);</div><div class="line"><a name="l00684"></a><span class="lineno"> 684</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00685"></a><span class="lineno"> 685</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, 4);</div><div class="line"><a name="l00686"></a><span class="lineno"> 686</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, 4)</div><div class="line"><a name="l00687"></a><span class="lineno"> 687</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>(((uint4)(GRID_SIZE * i * 4) + boundary_workitems * 4 + <a class="code" href="softmax__layer_8cl.xhtml#a4884a666a1e93fbf8c27bd7d2da3c8bb">idx4</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, 4));</div><div class="line"><a name="l00688"></a><span class="lineno"> 688</span>  data = select(0, data, widx);</div><div class="line"><a name="l00689"></a><span class="lineno"> 689</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(4)</div><div class="line"><a name="l00690"></a><span class="lineno"> 690</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, (GRID_SIZE * i * 4) + 4, 0));</div><div class="line"><a name="l00691"></a><span class="lineno"> 691</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, 4);</div><div class="line"><a name="l00692"></a><span class="lineno"> 692</span>  }</div><div class="line"><a name="l00693"></a><span class="lineno"> 693</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_VECTOR_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00694"></a><span class="lineno"> 694</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_GRID_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00695"></a><span class="lineno"> 695</span>  tmp_local[lid] = sum1D;</div><div class="line"><a name="l00696"></a><span class="lineno"> 696</span> </div><div class="line"><a name="l00697"></a><span class="lineno"> 697</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00698"></a><span class="lineno"> 698</span> </div><div class="line"><a name="l00699"></a><span class="lineno"> 699</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 256)</div><div class="line"><a name="l00700"></a><span class="lineno"> 700</span>  {</div><div class="line"><a name="l00701"></a><span class="lineno"> 701</span>  <span class="keywordflow">if</span>(lid < 128)</div><div class="line"><a name="l00702"></a><span class="lineno"> 702</span>  {</div><div class="line"><a name="l00703"></a><span class="lineno"> 703</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 128], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00704"></a><span class="lineno"> 704</span>  }</div><div class="line"><a name="l00705"></a><span class="lineno"> 705</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00706"></a><span class="lineno"> 706</span>  }</div><div class="line"><a name="l00707"></a><span class="lineno"> 707</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 128)</div><div class="line"><a name="l00708"></a><span class="lineno"> 708</span>  {</div><div class="line"><a name="l00709"></a><span class="lineno"> 709</span>  <span class="keywordflow">if</span>(lid < 64)</div><div class="line"><a name="l00710"></a><span class="lineno"> 710</span>  {</div><div class="line"><a name="l00711"></a><span class="lineno"> 711</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 64], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00712"></a><span class="lineno"> 712</span>  }</div><div class="line"><a name="l00713"></a><span class="lineno"> 713</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00714"></a><span class="lineno"> 714</span>  }</div><div class="line"><a name="l00715"></a><span class="lineno"> 715</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 64)</div><div class="line"><a name="l00716"></a><span class="lineno"> 716</span>  {</div><div class="line"><a name="l00717"></a><span class="lineno"> 717</span>  <span class="keywordflow">if</span>(lid < 32)</div><div class="line"><a name="l00718"></a><span class="lineno"> 718</span>  {</div><div class="line"><a name="l00719"></a><span class="lineno"> 719</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 32], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00720"></a><span class="lineno"> 720</span>  }</div><div class="line"><a name="l00721"></a><span class="lineno"> 721</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00722"></a><span class="lineno"> 722</span>  }</div><div class="line"><a name="l00723"></a><span class="lineno"> 723</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 32)</div><div class="line"><a name="l00724"></a><span class="lineno"> 724</span>  {</div><div class="line"><a name="l00725"></a><span class="lineno"> 725</span>  <span class="keywordflow">if</span>(lid < 16)</div><div class="line"><a name="l00726"></a><span class="lineno"> 726</span>  {</div><div class="line"><a name="l00727"></a><span class="lineno"> 727</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 16], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00728"></a><span class="lineno"> 728</span>  }</div><div class="line"><a name="l00729"></a><span class="lineno"> 729</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00730"></a><span class="lineno"> 730</span>  }</div><div class="line"><a name="l00731"></a><span class="lineno"> 731</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 16)</div><div class="line"><a name="l00732"></a><span class="lineno"> 732</span>  {</div><div class="line"><a name="l00733"></a><span class="lineno"> 733</span>  <span class="keywordflow">if</span>(lid < 8)</div><div class="line"><a name="l00734"></a><span class="lineno"> 734</span>  {</div><div class="line"><a name="l00735"></a><span class="lineno"> 735</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 8], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00736"></a><span class="lineno"> 736</span>  }</div><div class="line"><a name="l00737"></a><span class="lineno"> 737</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00738"></a><span class="lineno"> 738</span>  }</div><div class="line"><a name="l00739"></a><span class="lineno"> 739</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 8)</div><div class="line"><a name="l00740"></a><span class="lineno"> 740</span>  {</div><div class="line"><a name="l00741"></a><span class="lineno"> 741</span>  <span class="keywordflow">if</span>(lid < 4)</div><div class="line"><a name="l00742"></a><span class="lineno"> 742</span>  {</div><div class="line"><a name="l00743"></a><span class="lineno"> 743</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 4], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00744"></a><span class="lineno"> 744</span>  }</div><div class="line"><a name="l00745"></a><span class="lineno"> 745</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00746"></a><span class="lineno"> 746</span>  }</div><div class="line"><a name="l00747"></a><span class="lineno"> 747</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 4)</div><div class="line"><a name="l00748"></a><span class="lineno"> 748</span>  {</div><div class="line"><a name="l00749"></a><span class="lineno"> 749</span>  <span class="keywordflow">if</span>(lid < 2)</div><div class="line"><a name="l00750"></a><span class="lineno"> 750</span>  {</div><div class="line"><a name="l00751"></a><span class="lineno"> 751</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 2], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00752"></a><span class="lineno"> 752</span>  }</div><div class="line"><a name="l00753"></a><span class="lineno"> 753</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00754"></a><span class="lineno"> 754</span>  }</div><div class="line"><a name="l00755"></a><span class="lineno"> 755</span>  <span class="keywordflow">if</span>(lid == 0)</div><div class="line"><a name="l00756"></a><span class="lineno"> 756</span>  {</div><div class="line"><a name="l00757"></a><span class="lineno"> 757</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 1], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00758"></a><span class="lineno"> 758</span>  <span class="comment">// Perform max reduction</span></div><div class="line"><a name="l00759"></a><span class="lineno"> 759</span>  sum1D.s01 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s01, sum1D.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00760"></a><span class="lineno"> 760</span>  sum1D.s0 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s0, sum1D.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00761"></a><span class="lineno"> 761</span>  *((__global DATA_TYPE *)sum.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>) = sum1D.s0;</div><div class="line"><a name="l00762"></a><span class="lineno"> 762</span>  }</div><div class="line"><a name="l00763"></a><span class="lineno"> 763</span> }</div><div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_ab146b9cbab6e73e7588b240dc709fe01"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">arm_compute::test::validation::beta</a></div><div class="ttdeci">beta</div><div class="ttdef"><b>Definition:</b> <a href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00112">GEMM.cpp:112</a></div></div>
+<p>References <a class="el" href="softmax__layer_8cl_source.xhtml#l00044">ADD_OP</a>, <a class="el" href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00113">arm_compute::test::validation::beta</a>, <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="_c_l_2_convolution_8cpp_source.xhtml#l00137">arm_compute::test::validation::dst</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00048">EXP_OP</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00062">GRID_SIZE</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00088">idx4</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00043">MAX_OP</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00046">MUL_OP</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="softmax__layer_8cl_source.xhtml#l00055">SELECT_DATA_TYPE</a>, <a class="el" href="_c_l_2_convolution_8cpp_source.xhtml#l00147">arm_compute::test::validation::src</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00045">SUB_OP</a>, <a class="el" href="reduction__operation_8cl_source.xhtml#l00052">sum()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00081">type_min_</a>, <a class="el" href="fixed__point_8h_source.xhtml#l00093">VEC_DATA_TYPE</a>, <a class="el" href="helpers_8h_source.xhtml#l00040">VLOAD</a>, and <a class="el" href="helpers_8h_source.xhtml#l00043">VSTORE</a>.</p>
+<div class="fragment"><div class="line"><a name="l00334"></a><span class="lineno"> 334</span> {</div><div class="line"><a name="l00335"></a><span class="lineno"> 335</span>  <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="l00336"></a><span class="lineno"> 336</span>  <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="l00337"></a><span class="lineno"> 337</span>  <a class="code" href="struct_image.xhtml">Image</a> maxo = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(maxo);</div><div class="line"><a name="l00338"></a><span class="lineno"> 338</span>  <a class="code" href="struct_image.xhtml">Image</a> <a class="code" href="reduction__operation_8cl.xhtml#ab0df00f5333da51860deb93deb44a782">sum</a> = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(sum);</div><div class="line"><a name="l00339"></a><span class="lineno"> 339</span> </div><div class="line"><a name="l00340"></a><span class="lineno"> 340</span>  <span class="keyword">const</span> uint lid = get_local_id(0);</div><div class="line"><a name="l00341"></a><span class="lineno"> 341</span> </div><div class="line"><a name="l00342"></a><span class="lineno"> 342</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00343"></a><span class="lineno"> 343</span>  <span class="comment">// Initialize beta</span></div><div class="line"><a name="l00344"></a><span class="lineno"> 344</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 4)</div><div class="line"><a name="l00345"></a><span class="lineno"> 345</span>  <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a> = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 4))BETA;</div><div class="line"><a name="l00346"></a><span class="lineno"> 346</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00347"></a><span class="lineno"> 347</span> </div><div class="line"><a name="l00348"></a><span class="lineno"> 348</span>  <span class="comment">// Define one temporary vector per work-item.</span></div><div class="line"><a name="l00349"></a><span class="lineno"> 349</span>  __local <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4) tmp_local[<a class="code" href="softmax__layer_8cl.xhtml#a08246606c233e7785a497c09672f366f">GRID_SIZE</a>];</div><div class="line"><a name="l00350"></a><span class="lineno"> 350</span>  __local DATA_TYPE max_local;</div><div class="line"><a name="l00351"></a><span class="lineno"> 351</span> </div><div class="line"><a name="l00352"></a><span class="lineno"> 352</span>  __constant <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4) type_min4 = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4))(<a class="code" href="softmax__layer_8cl.xhtml#a80b22c555ddadb47cc6ca338a9c49126">MINVAL</a>);</div><div class="line"><a name="l00353"></a><span class="lineno"> 353</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00354"></a><span class="lineno"> 354</span>  max_val_vec = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4))type_min4;</div><div class="line"><a name="l00355"></a><span class="lineno"> 355</span>  <span class="comment">// Number of elements per work-item.</span></div><div class="line"><a name="l00356"></a><span class="lineno"> 356</span>  const uint row = width / GRID_SIZE;</div><div class="line"><a name="l00357"></a><span class="lineno"> 357</span>  <span class="comment">// Number of iterations per work-item.</span></div><div class="line"><a name="l00358"></a><span class="lineno"> 358</span>  const uint width_ = row >> 2;</div><div class="line"><a name="l00359"></a><span class="lineno"> 359</span>  <span class="comment">// Calculate max of row</span></div><div class="line"><a name="l00360"></a><span class="lineno"> 360</span>  uint i = 0;</div><div class="line"><a name="l00361"></a><span class="lineno"> 361</span>  for(; i < width_; i++)</div><div class="line"><a name="l00362"></a><span class="lineno"> 362</span>  {</div><div class="line"><a name="l00363"></a><span class="lineno"> 363</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00364"></a><span class="lineno"> 364</span>  data_max = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00365"></a><span class="lineno"> 365</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(data_max, max_val_vec, DATA_TYPE, 4);</div><div class="line"><a name="l00366"></a><span class="lineno"> 366</span>  }</div><div class="line"><a name="l00367"></a><span class="lineno"> 367</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_GRID_SIZE</span></div><div class="line"><a name="l00368"></a><span class="lineno"> 368</span>  <span class="comment">// How many work-items needed to complete the computation.</span></div><div class="line"><a name="l00369"></a><span class="lineno"> 369</span>  <span class="keywordtype">int</span> boundary_workitems = (width % (GRID_SIZE * 4)) / 4;</div><div class="line"><a name="l00370"></a><span class="lineno"> 370</span>  <span class="keywordflow">if</span>(lid < boundary_workitems)</div><div class="line"><a name="l00371"></a><span class="lineno"> 371</span>  {</div><div class="line"><a name="l00372"></a><span class="lineno"> 372</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00373"></a><span class="lineno"> 373</span>  data_max = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00374"></a><span class="lineno"> 374</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(data_max, max_val_vec, DATA_TYPE, 4);</div><div class="line"><a name="l00375"></a><span class="lineno"> 375</span>  }</div><div class="line"><a name="l00376"></a><span class="lineno"> 376</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_VECTOR_SIZE</span></div><div class="line"><a name="l00377"></a><span class="lineno"> 377</span>  <span class="keywordflow">if</span>(boundary_workitems == 0)</div><div class="line"><a name="l00378"></a><span class="lineno"> 378</span>  {</div><div class="line"><a name="l00379"></a><span class="lineno"> 379</span>  boundary_workitems = <a class="code" href="softmax__layer_8cl.xhtml#a08246606c233e7785a497c09672f366f">GRID_SIZE</a>;</div><div class="line"><a name="l00380"></a><span class="lineno"> 380</span>  i--;</div><div class="line"><a name="l00381"></a><span class="lineno"> 381</span>  }</div><div class="line"><a name="l00382"></a><span class="lineno"> 382</span>  <span class="keywordflow">if</span>(lid == (boundary_workitems - 1))</div><div class="line"><a name="l00383"></a><span class="lineno"> 383</span>  {</div><div class="line"><a name="l00384"></a><span class="lineno"> 384</span>  <span class="comment">// Handle non multiple of 4</span></div><div class="line"><a name="l00385"></a><span class="lineno"> 385</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00386"></a><span class="lineno"> 386</span>  data_max = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, (GRID_SIZE * i * 4) + 4, 0));</div><div class="line"><a name="l00387"></a><span class="lineno"> 387</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a>, 4)</div><div class="line"><a name="l00388"></a><span class="lineno"> 388</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>(((uint4)(GRID_SIZE * i * 4) + boundary_workitems * 4 + <a class="code" href="softmax__layer_8cl.xhtml#a4884a666a1e93fbf8c27bd7d2da3c8bb">idx4</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, 4));</div><div class="line"><a name="l00389"></a><span class="lineno"> 389</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec, select(<a class="code" href="softmax__layer_8cl.xhtml#af7a08044d0e491a0ee1520a24a107a2b">type_min_</a>, data_max, widx), DATA_TYPE, 4);</div><div class="line"><a name="l00390"></a><span class="lineno"> 390</span>  }</div><div class="line"><a name="l00391"></a><span class="lineno"> 391</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_VECTOR_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00392"></a><span class="lineno"> 392</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_GRID_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00393"></a><span class="lineno"> 393</span>  tmp_local[lid] = max_val_vec;</div><div class="line"><a name="l00394"></a><span class="lineno"> 394</span> </div><div class="line"><a name="l00395"></a><span class="lineno"> 395</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00396"></a><span class="lineno"> 396</span> </div><div class="line"><a name="l00397"></a><span class="lineno"> 397</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 256)</div><div class="line"><a name="l00398"></a><span class="lineno"> 398</span>  {</div><div class="line"><a name="l00399"></a><span class="lineno"> 399</span>  <span class="keywordflow">if</span>(lid < 128)</div><div class="line"><a name="l00400"></a><span class="lineno"> 400</span>  {</div><div class="line"><a name="l00401"></a><span class="lineno"> 401</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 128], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00402"></a><span class="lineno"> 402</span>  }</div><div class="line"><a name="l00403"></a><span class="lineno"> 403</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00404"></a><span class="lineno"> 404</span>  }</div><div class="line"><a name="l00405"></a><span class="lineno"> 405</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 128)</div><div class="line"><a name="l00406"></a><span class="lineno"> 406</span>  {</div><div class="line"><a name="l00407"></a><span class="lineno"> 407</span>  <span class="keywordflow">if</span>(lid < 64)</div><div class="line"><a name="l00408"></a><span class="lineno"> 408</span>  {</div><div class="line"><a name="l00409"></a><span class="lineno"> 409</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 64], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00410"></a><span class="lineno"> 410</span>  }</div><div class="line"><a name="l00411"></a><span class="lineno"> 411</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00412"></a><span class="lineno"> 412</span>  }</div><div class="line"><a name="l00413"></a><span class="lineno"> 413</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 64)</div><div class="line"><a name="l00414"></a><span class="lineno"> 414</span>  {</div><div class="line"><a name="l00415"></a><span class="lineno"> 415</span>  <span class="keywordflow">if</span>(lid < 32)</div><div class="line"><a name="l00416"></a><span class="lineno"> 416</span>  {</div><div class="line"><a name="l00417"></a><span class="lineno"> 417</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 32], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00418"></a><span class="lineno"> 418</span>  }</div><div class="line"><a name="l00419"></a><span class="lineno"> 419</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00420"></a><span class="lineno"> 420</span>  }</div><div class="line"><a name="l00421"></a><span class="lineno"> 421</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 32)</div><div class="line"><a name="l00422"></a><span class="lineno"> 422</span>  {</div><div class="line"><a name="l00423"></a><span class="lineno"> 423</span>  <span class="keywordflow">if</span>(lid < 16)</div><div class="line"><a name="l00424"></a><span class="lineno"> 424</span>  {</div><div class="line"><a name="l00425"></a><span class="lineno"> 425</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 16], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00426"></a><span class="lineno"> 426</span>  }</div><div class="line"><a name="l00427"></a><span class="lineno"> 427</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00428"></a><span class="lineno"> 428</span>  }</div><div class="line"><a name="l00429"></a><span class="lineno"> 429</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 16)</div><div class="line"><a name="l00430"></a><span class="lineno"> 430</span>  {</div><div class="line"><a name="l00431"></a><span class="lineno"> 431</span>  <span class="keywordflow">if</span>(lid < 8)</div><div class="line"><a name="l00432"></a><span class="lineno"> 432</span>  {</div><div class="line"><a name="l00433"></a><span class="lineno"> 433</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 8], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00434"></a><span class="lineno"> 434</span>  }</div><div class="line"><a name="l00435"></a><span class="lineno"> 435</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00436"></a><span class="lineno"> 436</span>  }</div><div class="line"><a name="l00437"></a><span class="lineno"> 437</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 8)</div><div class="line"><a name="l00438"></a><span class="lineno"> 438</span>  {</div><div class="line"><a name="l00439"></a><span class="lineno"> 439</span>  <span class="keywordflow">if</span>(lid < 4)</div><div class="line"><a name="l00440"></a><span class="lineno"> 440</span>  {</div><div class="line"><a name="l00441"></a><span class="lineno"> 441</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 4], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00442"></a><span class="lineno"> 442</span>  }</div><div class="line"><a name="l00443"></a><span class="lineno"> 443</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00444"></a><span class="lineno"> 444</span>  }</div><div class="line"><a name="l00445"></a><span class="lineno"> 445</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 4)</div><div class="line"><a name="l00446"></a><span class="lineno"> 446</span>  {</div><div class="line"><a name="l00447"></a><span class="lineno"> 447</span>  <span class="keywordflow">if</span>(lid < 2)</div><div class="line"><a name="l00448"></a><span class="lineno"> 448</span>  {</div><div class="line"><a name="l00449"></a><span class="lineno"> 449</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 2], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00450"></a><span class="lineno"> 450</span>  }</div><div class="line"><a name="l00451"></a><span class="lineno"> 451</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00452"></a><span class="lineno"> 452</span>  }</div><div class="line"><a name="l00453"></a><span class="lineno"> 453</span>  <span class="keywordflow">if</span>(lid == 0)</div><div class="line"><a name="l00454"></a><span class="lineno"> 454</span>  {</div><div class="line"><a name="l00455"></a><span class="lineno"> 455</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(tmp_local[lid + 1], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00456"></a><span class="lineno"> 456</span>  max_val_vec.s01 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s01, max_val_vec.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00457"></a><span class="lineno"> 457</span>  max_val_vec.s0 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s0, max_val_vec.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00458"></a><span class="lineno"> 458</span>  max_local = max_val_vec.s0;</div><div class="line"><a name="l00459"></a><span class="lineno"> 459</span>  }</div><div class="line"><a name="l00460"></a><span class="lineno"> 460</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00461"></a><span class="lineno"> 461</span> </div><div class="line"><a name="l00462"></a><span class="lineno"> 462</span>  <span class="comment">/* Second section */</span></div><div class="line"><a name="l00463"></a><span class="lineno"> 463</span> </div><div class="line"><a name="l00464"></a><span class="lineno"> 464</span>  <span class="comment">// Set sum vector</span></div><div class="line"><a name="l00465"></a><span class="lineno"> 465</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00466"></a><span class="lineno"> 466</span>  sum1D = 0;</div><div class="line"><a name="l00467"></a><span class="lineno"> 467</span>  DATA_TYPE max_val = max_local;</div><div class="line"><a name="l00468"></a><span class="lineno"> 468</span> </div><div class="line"><a name="l00469"></a><span class="lineno"> 469</span>  <span class="comment">// Shift values, exp and sum</span></div><div class="line"><a name="l00470"></a><span class="lineno"> 470</span>  for(i = 0; i < width_; i++)</div><div class="line"><a name="l00471"></a><span class="lineno"> 471</span>  {</div><div class="line"><a name="l00472"></a><span class="lineno"> 472</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00473"></a><span class="lineno"> 473</span>  data = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00474"></a><span class="lineno"> 474</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, 4);</div><div class="line"><a name="l00475"></a><span class="lineno"> 475</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00476"></a><span class="lineno"> 476</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, 4);</div><div class="line"><a name="l00477"></a><span class="lineno"> 477</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00478"></a><span class="lineno"> 478</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, 4);</div><div class="line"><a name="l00479"></a><span class="lineno"> 479</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(4)</div><div class="line"><a name="l00480"></a><span class="lineno"> 480</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00481"></a><span class="lineno"> 481</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, 4);</div><div class="line"><a name="l00482"></a><span class="lineno"> 482</span>  }</div><div class="line"><a name="l00483"></a><span class="lineno"> 483</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_GRID_SIZE</span></div><div class="line"><a name="l00484"></a><span class="lineno"> 484</span>  boundary_workitems = (width % (GRID_SIZE * 4)) / 4;</div><div class="line"><a name="l00485"></a><span class="lineno"> 485</span>  <span class="keywordflow">if</span>(lid < boundary_workitems)</div><div class="line"><a name="l00486"></a><span class="lineno"> 486</span>  {</div><div class="line"><a name="l00487"></a><span class="lineno"> 487</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00488"></a><span class="lineno"> 488</span>  data = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00489"></a><span class="lineno"> 489</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, 4);</div><div class="line"><a name="l00490"></a><span class="lineno"> 490</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00491"></a><span class="lineno"> 491</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, 4);</div><div class="line"><a name="l00492"></a><span class="lineno"> 492</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00493"></a><span class="lineno"> 493</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, 4);</div><div class="line"><a name="l00494"></a><span class="lineno"> 494</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(4)</div><div class="line"><a name="l00495"></a><span class="lineno"> 495</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, i * GRID_SIZE * 4, 0));</div><div class="line"><a name="l00496"></a><span class="lineno"> 496</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, 4);</div><div class="line"><a name="l00497"></a><span class="lineno"> 497</span>  }</div><div class="line"><a name="l00498"></a><span class="lineno"> 498</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_VECTOR_SIZE</span></div><div class="line"><a name="l00499"></a><span class="lineno"> 499</span>  <span class="keywordflow">if</span>(boundary_workitems == 0)</div><div class="line"><a name="l00500"></a><span class="lineno"> 500</span>  {</div><div class="line"><a name="l00501"></a><span class="lineno"> 501</span>  boundary_workitems = <a class="code" href="softmax__layer_8cl.xhtml#a08246606c233e7785a497c09672f366f">GRID_SIZE</a>;</div><div class="line"><a name="l00502"></a><span class="lineno"> 502</span>  i--;</div><div class="line"><a name="l00503"></a><span class="lineno"> 503</span>  }</div><div class="line"><a name="l00504"></a><span class="lineno"> 504</span>  <span class="keywordflow">if</span>(lid == (boundary_workitems - 1))</div><div class="line"><a name="l00505"></a><span class="lineno"> 505</span>  {</div><div class="line"><a name="l00506"></a><span class="lineno"> 506</span>  <span class="comment">// Handle non multiple of vector size ((GRID_SIZE * i * 4) + 4, 0); move 4 float positions ahead, *4 is due to the stride</span></div><div class="line"><a name="l00507"></a><span class="lineno"> 507</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 4)</div><div class="line"><a name="l00508"></a><span class="lineno"> 508</span>  data = VLOAD(4)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, (GRID_SIZE * i * 4) + 4, 0));</div><div class="line"><a name="l00509"></a><span class="lineno"> 509</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, 4);</div><div class="line"><a name="l00510"></a><span class="lineno"> 510</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00511"></a><span class="lineno"> 511</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, 4);</div><div class="line"><a name="l00512"></a><span class="lineno"> 512</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00513"></a><span class="lineno"> 513</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, 4);</div><div class="line"><a name="l00514"></a><span class="lineno"> 514</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, 4)</div><div class="line"><a name="l00515"></a><span class="lineno"> 515</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>(((uint4)(GRID_SIZE * i * 4) + boundary_workitems * 4 + <a class="code" href="softmax__layer_8cl.xhtml#a4884a666a1e93fbf8c27bd7d2da3c8bb">idx4</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, 4));</div><div class="line"><a name="l00516"></a><span class="lineno"> 516</span>  data = select(0, data, widx);</div><div class="line"><a name="l00517"></a><span class="lineno"> 517</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(4)</div><div class="line"><a name="l00518"></a><span class="lineno"> 518</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, (GRID_SIZE * i * 4) + 4, 0));</div><div class="line"><a name="l00519"></a><span class="lineno"> 519</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, 4);</div><div class="line"><a name="l00520"></a><span class="lineno"> 520</span>  }</div><div class="line"><a name="l00521"></a><span class="lineno"> 521</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_VECTOR_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00522"></a><span class="lineno"> 522</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_GRID_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00523"></a><span class="lineno"> 523</span>  tmp_local[lid] = sum1D;</div><div class="line"><a name="l00524"></a><span class="lineno"> 524</span> </div><div class="line"><a name="l00525"></a><span class="lineno"> 525</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00526"></a><span class="lineno"> 526</span> </div><div class="line"><a name="l00527"></a><span class="lineno"> 527</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 256)</div><div class="line"><a name="l00528"></a><span class="lineno"> 528</span>  {</div><div class="line"><a name="l00529"></a><span class="lineno"> 529</span>  <span class="keywordflow">if</span>(lid < 128)</div><div class="line"><a name="l00530"></a><span class="lineno"> 530</span>  {</div><div class="line"><a name="l00531"></a><span class="lineno"> 531</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 128], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00532"></a><span class="lineno"> 532</span>  }</div><div class="line"><a name="l00533"></a><span class="lineno"> 533</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00534"></a><span class="lineno"> 534</span>  }</div><div class="line"><a name="l00535"></a><span class="lineno"> 535</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 128)</div><div class="line"><a name="l00536"></a><span class="lineno"> 536</span>  {</div><div class="line"><a name="l00537"></a><span class="lineno"> 537</span>  <span class="keywordflow">if</span>(lid < 64)</div><div class="line"><a name="l00538"></a><span class="lineno"> 538</span>  {</div><div class="line"><a name="l00539"></a><span class="lineno"> 539</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 64], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00540"></a><span class="lineno"> 540</span>  }</div><div class="line"><a name="l00541"></a><span class="lineno"> 541</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00542"></a><span class="lineno"> 542</span>  }</div><div class="line"><a name="l00543"></a><span class="lineno"> 543</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 64)</div><div class="line"><a name="l00544"></a><span class="lineno"> 544</span>  {</div><div class="line"><a name="l00545"></a><span class="lineno"> 545</span>  <span class="keywordflow">if</span>(lid < 32)</div><div class="line"><a name="l00546"></a><span class="lineno"> 546</span>  {</div><div class="line"><a name="l00547"></a><span class="lineno"> 547</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 32], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00548"></a><span class="lineno"> 548</span>  }</div><div class="line"><a name="l00549"></a><span class="lineno"> 549</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00550"></a><span class="lineno"> 550</span>  }</div><div class="line"><a name="l00551"></a><span class="lineno"> 551</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 32)</div><div class="line"><a name="l00552"></a><span class="lineno"> 552</span>  {</div><div class="line"><a name="l00553"></a><span class="lineno"> 553</span>  <span class="keywordflow">if</span>(lid < 16)</div><div class="line"><a name="l00554"></a><span class="lineno"> 554</span>  {</div><div class="line"><a name="l00555"></a><span class="lineno"> 555</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 16], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00556"></a><span class="lineno"> 556</span>  }</div><div class="line"><a name="l00557"></a><span class="lineno"> 557</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00558"></a><span class="lineno"> 558</span>  }</div><div class="line"><a name="l00559"></a><span class="lineno"> 559</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 16)</div><div class="line"><a name="l00560"></a><span class="lineno"> 560</span>  {</div><div class="line"><a name="l00561"></a><span class="lineno"> 561</span>  <span class="keywordflow">if</span>(lid < 8)</div><div class="line"><a name="l00562"></a><span class="lineno"> 562</span>  {</div><div class="line"><a name="l00563"></a><span class="lineno"> 563</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 8], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00564"></a><span class="lineno"> 564</span>  }</div><div class="line"><a name="l00565"></a><span class="lineno"> 565</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00566"></a><span class="lineno"> 566</span>  }</div><div class="line"><a name="l00567"></a><span class="lineno"> 567</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 8)</div><div class="line"><a name="l00568"></a><span class="lineno"> 568</span>  {</div><div class="line"><a name="l00569"></a><span class="lineno"> 569</span>  <span class="keywordflow">if</span>(lid < 4)</div><div class="line"><a name="l00570"></a><span class="lineno"> 570</span>  {</div><div class="line"><a name="l00571"></a><span class="lineno"> 571</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 4], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00572"></a><span class="lineno"> 572</span>  }</div><div class="line"><a name="l00573"></a><span class="lineno"> 573</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00574"></a><span class="lineno"> 574</span>  }</div><div class="line"><a name="l00575"></a><span class="lineno"> 575</span>  <span class="keywordflow">if</span>(GRID_SIZE >= 4)</div><div class="line"><a name="l00576"></a><span class="lineno"> 576</span>  {</div><div class="line"><a name="l00577"></a><span class="lineno"> 577</span>  <span class="keywordflow">if</span>(lid < 2)</div><div class="line"><a name="l00578"></a><span class="lineno"> 578</span>  {</div><div class="line"><a name="l00579"></a><span class="lineno"> 579</span>  tmp_local[lid] = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 2], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00580"></a><span class="lineno"> 580</span>  }</div><div class="line"><a name="l00581"></a><span class="lineno"> 581</span>  barrier(CLK_LOCAL_MEM_FENCE);</div><div class="line"><a name="l00582"></a><span class="lineno"> 582</span>  }</div><div class="line"><a name="l00583"></a><span class="lineno"> 583</span>  <span class="keywordflow">if</span>(lid == 0)</div><div class="line"><a name="l00584"></a><span class="lineno"> 584</span>  {</div><div class="line"><a name="l00585"></a><span class="lineno"> 585</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(tmp_local[lid + 1], tmp_local[lid], DATA_TYPE, 4);</div><div class="line"><a name="l00586"></a><span class="lineno"> 586</span>  <span class="comment">// Perform max reduction</span></div><div class="line"><a name="l00587"></a><span class="lineno"> 587</span>  sum1D.s01 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s01, sum1D.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00588"></a><span class="lineno"> 588</span>  sum1D.s0 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s0, sum1D.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00589"></a><span class="lineno"> 589</span>  *((__global DATA_TYPE *)sum.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>) = sum1D.s0;</div><div class="line"><a name="l00590"></a><span class="lineno"> 590</span>  }</div><div class="line"><a name="l00591"></a><span class="lineno"> 591</span> }</div><div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_ab146b9cbab6e73e7588b240dc709fe01"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">arm_compute::test::validation::beta</a></div><div class="ttdeci">beta</div><div class="ttdef"><b>Definition:</b> <a href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00113">GEMM.cpp:113</a></div></div>
<div class="ttc" id="fixed__point_8h_xhtml_aa8d95ba04fc73845abc6045952cae5be"><div class="ttname"><a href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a></div><div class="ttdeci">#define CONVERT(x, type)</div><div class="ttdef"><b>Definition:</b> <a href="fixed__point_8h_source.xhtml#l00098">fixed_point.h:98</a></div></div>
<div class="ttc" id="helpers_8h_xhtml_a287e2fc366c312b468382c95bb90f91f"><div class="ttname"><a href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a></div><div class="ttdeci">#define VLOAD(size)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00040">helpers.h:40</a></div></div>
<div class="ttc" id="softmax__layer_8cl_xhtml_a93cf800667317d96574477b9f0a75234"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a></div><div class="ttdeci">#define EXP_OP(x, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00048">softmax_layer.cl:48</a></div></div>
@@ -942,7 +774,7 @@
<div class="ttc" id="helpers_8h_xhtml_acb282042d1edeeaa3cc979a206f78b54"><div class="ttname"><a href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a></div><div class="ttdeci">#define VSTORE(size)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00043">helpers.h:43</a></div></div>
<div class="ttc" id="softmax__layer_8cl_xhtml_af5b2e33e3c5fcaab3a213f26c2300170"><div class="ttname"><a href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a></div><div class="ttdeci">#define SELECT_DATA_TYPE</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00055">softmax_layer.cl:55</a></div></div>
<div class="ttc" id="softmax__layer_8cl_xhtml_a4884a666a1e93fbf8c27bd7d2da3c8bb"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a4884a666a1e93fbf8c27bd7d2da3c8bb">idx4</a></div><div class="ttdeci">__constant uint4 idx4</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00088">softmax_layer.cl:88</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#l00123">Convolution.cpp:123</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="softmax__layer_8cl_xhtml_a22303c4047ec5027c1538d53964b9d0d"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a></div><div class="ttdeci">#define MUL_OP(x, y, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00046">softmax_layer.cl:46</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>
@@ -951,7 +783,7 @@
<div class="ttc" id="softmax__layer_8cl_xhtml_a80b22c555ddadb47cc6ca338a9c49126"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a80b22c555ddadb47cc6ca338a9c49126">MINVAL</a></div><div class="ttdeci">#define MINVAL</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00054">softmax_layer.cl:54</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="softmax__layer_8cl_xhtml_a08246606c233e7785a497c09672f366f"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a08246606c233e7785a497c09672f366f">GRID_SIZE</a></div><div class="ttdeci">#define GRID_SIZE</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00062">softmax_layer.cl:62</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 & src</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_convolution_8cpp_source.xhtml#l00133">Convolution.cpp:133</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 & 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>
@@ -1212,11 +1044,11 @@
</dd>
</dl>
-<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">350</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
+<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">178</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>References <a class="el" href="softmax__layer_8cl_source.xhtml#l00044">ADD_OP</a>, <a class="el" href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00112">arm_compute::test::validation::beta</a>, <a class="el" href="helpers_8h_source.xhtml#l00049">CL_VEC_DATA_TYPE</a>, <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="_c_l_2_convolution_8cpp_source.xhtml#l00123">arm_compute::test::validation::dst</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00048">EXP_OP</a>, <a class="el" href="helpers_8h_source.xhtml#l00035">EXPAND</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00082">idx__</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00080">LOG_VECTOR_SIZE</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00043">MAX_OP</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00046">MUL_OP</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="softmax__layer_8cl_source.xhtml#l00055">SELECT_DATA_TYPE</a>, <a class="el" href="_c_l_2_convolution_8cpp_source.xhtml#l00133">arm_compute::test::validation::src</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00045">SUB_OP</a>, <a class="el" href="reduction__operation_8cl_source.xhtml#l00052">sum()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00081">type_min_</a>, <a class="el" href="fixed__point_8h_source.xhtml#l00093">VEC_DATA_TYPE</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00079">VECTOR_SIZE</a>, <a class="el" href="helpers_8h_source.xhtml#l00040">VLOAD</a>, and <a class="el" href="helpers_8h_source.xhtml#l00043">VSTORE</a>.</p>
-<div class="fragment"><div class="line"><a name="l00356"></a><span class="lineno"> 356</span> {</div><div class="line"><a name="l00357"></a><span class="lineno"> 357</span>  <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="l00358"></a><span class="lineno"> 358</span>  <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="l00359"></a><span class="lineno"> 359</span>  <a class="code" href="struct_image.xhtml">Image</a> maxo = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(maxo);</div><div class="line"><a name="l00360"></a><span class="lineno"> 360</span>  <a class="code" href="struct_image.xhtml">Image</a> <a class="code" href="reduction__operation_8cl.xhtml#ab0df00f5333da51860deb93deb44a782">sum</a> = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(sum);</div><div class="line"><a name="l00361"></a><span class="lineno"> 361</span> </div><div class="line"><a name="l00362"></a><span class="lineno"> 362</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00363"></a><span class="lineno"> 363</span>  <span class="comment">// Initialize beta</span></div><div class="line"><a name="l00364"></a><span class="lineno"> 364</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>)</div><div class="line"><a name="l00365"></a><span class="lineno"> 365</span>  <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a> = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>))BETA;</div><div class="line"><a name="l00366"></a><span class="lineno"> 366</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00367"></a><span class="lineno"> 367</span> </div><div class="line"><a name="l00368"></a><span class="lineno"> 368</span>  <span class="comment">// Initialize local maximum</span></div><div class="line"><a name="l00369"></a><span class="lineno"> 369</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>)</div><div class="line"><a name="l00370"></a><span class="lineno"> 370</span>  max_val_vec = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>))<a class="code" href="softmax__layer_8cl.xhtml#af7a08044d0e491a0ee1520a24a107a2b">type_min_</a>;</div><div class="line"><a name="l00371"></a><span class="lineno"> 371</span> </div><div class="line"><a name="l00372"></a><span class="lineno"> 372</span>  <span class="comment">// Calculate max of row</span></div><div class="line"><a name="l00373"></a><span class="lineno"> 373</span>  const uint width_ = width >> <a class="code" href="softmax__layer_8cl.xhtml#a372393c380805985b813dbb16d589a64">LOG_VECTOR_SIZE</a>;</div><div class="line"><a name="l00374"></a><span class="lineno"> 374</span>  for(uint i = 0; i < width_; i++)</div><div class="line"><a name="l00375"></a><span class="lineno"> 375</span>  {</div><div class="line"><a name="l00376"></a><span class="lineno"> 376</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>)</div><div class="line"><a name="l00377"></a><span class="lineno"> 377</span>  data_max = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(<a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00378"></a><span class="lineno"> 378</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(data_max, max_val_vec, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00379"></a><span class="lineno"> 379</span>  }</div><div class="line"><a name="l00380"></a><span class="lineno"> 380</span> </div><div class="line"><a name="l00381"></a><span class="lineno"> 381</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_VECTOR_SIZE</span></div><div class="line"><a name="l00382"></a><span class="lineno"> 382</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, VECTOR_SIZE)</div><div class="line"><a name="l00383"></a><span class="lineno"> 383</span>  data_max = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(VECTOR_SIZE)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, width_ << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00384"></a><span class="lineno"> 384</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a>, VECTOR_SIZE)</div><div class="line"><a name="l00385"></a><span class="lineno"> 385</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>((<a class="code" href="helpers_8h.xhtml#ae4b532a93c757194ec73b6790a3e6b1f">EXPAND</a>((<a class="code" href="helpers_8h.xhtml#a0f6ad555d097377151003a5119ccee45">CL_VEC_DATA_TYPE</a>(uint, VECTOR_SIZE)))(width_ << LOG_VECTOR_SIZE) + <a class="code" href="softmax__layer_8cl.xhtml#aa1dd94b8d98f1c6d790bdf0fc5de29e9">idx__</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, VECTOR_SIZE));</div><div class="line"><a name="l00386"></a><span class="lineno"> 386</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec, select(type_min_, data_max, widx), DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00387"></a><span class="lineno"> 387</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_VECTOR_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00388"></a><span class="lineno"> 388</span> </div><div class="line"><a name="l00389"></a><span class="lineno"> 389</span>  <span class="comment">// Perform max reduction</span></div><div class="line"><a name="l00390"></a><span class="lineno"> 390</span> <span class="preprocessor">#if VECTOR_SIZE == 16</span></div><div class="line"><a name="l00391"></a><span class="lineno"> 391</span>  max_val_vec.s01234567 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s01234567, max_val_vec.s89ABCDEF, DATA_TYPE, 8);</div><div class="line"><a name="l00392"></a><span class="lineno"> 392</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 16 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00393"></a><span class="lineno"> 393</span> <span class="preprocessor">#if VECTOR_SIZE >= 8</span></div><div class="line"><a name="l00394"></a><span class="lineno"> 394</span>  max_val_vec.s0123 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s0123, max_val_vec.s4567, DATA_TYPE, 4);</div><div class="line"><a name="l00395"></a><span class="lineno"> 395</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 8 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00396"></a><span class="lineno"> 396</span> <span class="preprocessor">#if VECTOR_SIZE >= 4</span></div><div class="line"><a name="l00397"></a><span class="lineno"> 397</span>  max_val_vec.s01 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s01, max_val_vec.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00398"></a><span class="lineno"> 398</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 4 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00399"></a><span class="lineno"> 399</span>  max_val_vec.s0 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s0, max_val_vec.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00400"></a><span class="lineno"> 400</span>  <span class="comment">// Store result</span></div><div class="line"><a name="l00401"></a><span class="lineno"> 401</span>  *((__global DATA_TYPE *)maxo.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>) = max_val_vec.s0;</div><div class="line"><a name="l00402"></a><span class="lineno"> 402</span> </div><div class="line"><a name="l00403"></a><span class="lineno"> 403</span>  <span class="comment">/* Second section */</span></div><div class="line"><a name="l00404"></a><span class="lineno"> 404</span> </div><div class="line"><a name="l00405"></a><span class="lineno"> 405</span>  <span class="comment">// Load max value of 1D logits vector (row)</span></div><div class="line"><a name="l00406"></a><span class="lineno"> 406</span>  DATA_TYPE max_val = *((__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&maxo, 0, 0));</div><div class="line"><a name="l00407"></a><span class="lineno"> 407</span> </div><div class="line"><a name="l00408"></a><span class="lineno"> 408</span>  <span class="comment">// Set sum vector</span></div><div class="line"><a name="l00409"></a><span class="lineno"> 409</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, VECTOR_SIZE)</div><div class="line"><a name="l00410"></a><span class="lineno"> 410</span>  sum1D = 0;</div><div class="line"><a name="l00411"></a><span class="lineno"> 411</span> </div><div class="line"><a name="l00412"></a><span class="lineno"> 412</span>  <span class="comment">// Shift values, exp and sum</span></div><div class="line"><a name="l00413"></a><span class="lineno"> 413</span>  for(uint i = 0; i < width_; i++)</div><div class="line"><a name="l00414"></a><span class="lineno"> 414</span>  {</div><div class="line"><a name="l00415"></a><span class="lineno"> 415</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, VECTOR_SIZE)</div><div class="line"><a name="l00416"></a><span class="lineno"> 416</span>  data = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(VECTOR_SIZE)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00417"></a><span class="lineno"> 417</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00418"></a><span class="lineno"> 418</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00419"></a><span class="lineno"> 419</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00420"></a><span class="lineno"> 420</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00421"></a><span class="lineno"> 421</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00422"></a><span class="lineno"> 422</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(VECTOR_SIZE)</div><div class="line"><a name="l00423"></a><span class="lineno"> 423</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, i << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00424"></a><span class="lineno"> 424</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00425"></a><span class="lineno"> 425</span>  }</div><div class="line"><a name="l00426"></a><span class="lineno"> 426</span> </div><div class="line"><a name="l00427"></a><span class="lineno"> 427</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_VECTOR_SIZE</span></div><div class="line"><a name="l00428"></a><span class="lineno"> 428</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, VECTOR_SIZE)</div><div class="line"><a name="l00429"></a><span class="lineno"> 429</span>  data = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(VECTOR_SIZE)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, width_ << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00430"></a><span class="lineno"> 430</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00431"></a><span class="lineno"> 431</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00432"></a><span class="lineno"> 432</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00433"></a><span class="lineno"> 433</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00434"></a><span class="lineno"> 434</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00435"></a><span class="lineno"> 435</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>((<a class="code" href="helpers_8h.xhtml#ae4b532a93c757194ec73b6790a3e6b1f">EXPAND</a>((<a class="code" href="helpers_8h.xhtml#a0f6ad555d097377151003a5119ccee45">CL_VEC_DATA_TYPE</a>(uint, VECTOR_SIZE)))(width_ << LOG_VECTOR_SIZE) + <a class="code" href="softmax__layer_8cl.xhtml#aa1dd94b8d98f1c6d790bdf0fc5de29e9">idx__</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, VECTOR_SIZE));</div><div class="line"><a name="l00436"></a><span class="lineno"> 436</span>  data = select(0, data, widx);</div><div class="line"><a name="l00437"></a><span class="lineno"> 437</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(VECTOR_SIZE)</div><div class="line"><a name="l00438"></a><span class="lineno"> 438</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, width_ << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00439"></a><span class="lineno"> 439</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00440"></a><span class="lineno"> 440</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_VECTOR_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00441"></a><span class="lineno"> 441</span> </div><div class="line"><a name="l00442"></a><span class="lineno"> 442</span>  <span class="comment">// Perform sum reduction</span></div><div class="line"><a name="l00443"></a><span class="lineno"> 443</span> <span class="preprocessor">#if VECTOR_SIZE == 16</span></div><div class="line"><a name="l00444"></a><span class="lineno"> 444</span>  sum1D.s01234567 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s01234567, sum1D.s89ABCDEF, DATA_TYPE, 8);</div><div class="line"><a name="l00445"></a><span class="lineno"> 445</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 16 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00446"></a><span class="lineno"> 446</span> <span class="preprocessor">#if VECTOR_SIZE >= 8</span></div><div class="line"><a name="l00447"></a><span class="lineno"> 447</span>  sum1D.s0123 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s0123, sum1D.s4567, DATA_TYPE, 4);</div><div class="line"><a name="l00448"></a><span class="lineno"> 448</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 8 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00449"></a><span class="lineno"> 449</span> <span class="preprocessor">#if VECTOR_SIZE >= 4</span></div><div class="line"><a name="l00450"></a><span class="lineno"> 450</span>  sum1D.s01 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s01, sum1D.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00451"></a><span class="lineno"> 451</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 4 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00452"></a><span class="lineno"> 452</span>  sum1D.s0 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s0, sum1D.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00453"></a><span class="lineno"> 453</span> </div><div class="line"><a name="l00454"></a><span class="lineno"> 454</span>  <span class="comment">// Calculate and store result</span></div><div class="line"><a name="l00455"></a><span class="lineno"> 455</span>  *((__global DATA_TYPE *)sum.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>) = sum1D.s0;</div><div class="line"><a name="l00456"></a><span class="lineno"> 456</span> }</div><div class="ttc" id="softmax__layer_8cl_xhtml_aa1dd94b8d98f1c6d790bdf0fc5de29e9"><div class="ttname"><a href="softmax__layer_8cl.xhtml#aa1dd94b8d98f1c6d790bdf0fc5de29e9">idx__</a></div><div class="ttdeci">__constant uint16 idx__</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00082">softmax_layer.cl:82</a></div></div>
-<div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_ab146b9cbab6e73e7588b240dc709fe01"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">arm_compute::test::validation::beta</a></div><div class="ttdeci">beta</div><div class="ttdef"><b>Definition:</b> <a href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00112">GEMM.cpp:112</a></div></div>
+<p>References <a class="el" href="softmax__layer_8cl_source.xhtml#l00044">ADD_OP</a>, <a class="el" href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00113">arm_compute::test::validation::beta</a>, <a class="el" href="helpers_8h_source.xhtml#l00049">CL_VEC_DATA_TYPE</a>, <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="_c_l_2_convolution_8cpp_source.xhtml#l00137">arm_compute::test::validation::dst</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00048">EXP_OP</a>, <a class="el" href="helpers_8h_source.xhtml#l00035">EXPAND</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00082">idx__</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00080">LOG_VECTOR_SIZE</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00043">MAX_OP</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00046">MUL_OP</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="softmax__layer_8cl_source.xhtml#l00055">SELECT_DATA_TYPE</a>, <a class="el" href="_c_l_2_convolution_8cpp_source.xhtml#l00147">arm_compute::test::validation::src</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00045">SUB_OP</a>, <a class="el" href="reduction__operation_8cl_source.xhtml#l00052">sum()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00081">type_min_</a>, <a class="el" href="fixed__point_8h_source.xhtml#l00093">VEC_DATA_TYPE</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00079">VECTOR_SIZE</a>, <a class="el" href="helpers_8h_source.xhtml#l00040">VLOAD</a>, and <a class="el" href="helpers_8h_source.xhtml#l00043">VSTORE</a>.</p>
+<div class="fragment"><div class="line"><a name="l00184"></a><span class="lineno"> 184</span> {</div><div class="line"><a name="l00185"></a><span class="lineno"> 185</span>  <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="l00186"></a><span class="lineno"> 186</span>  <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="l00187"></a><span class="lineno"> 187</span>  <a class="code" href="struct_image.xhtml">Image</a> maxo = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(maxo);</div><div class="line"><a name="l00188"></a><span class="lineno"> 188</span>  <a class="code" href="struct_image.xhtml">Image</a> <a class="code" href="reduction__operation_8cl.xhtml#ab0df00f5333da51860deb93deb44a782">sum</a> = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(sum);</div><div class="line"><a name="l00189"></a><span class="lineno"> 189</span> </div><div class="line"><a name="l00190"></a><span class="lineno"> 190</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00191"></a><span class="lineno"> 191</span>  <span class="comment">// Initialize beta</span></div><div class="line"><a name="l00192"></a><span class="lineno"> 192</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>)</div><div class="line"><a name="l00193"></a><span class="lineno"> 193</span>  <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a> = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>))BETA;</div><div class="line"><a name="l00194"></a><span class="lineno"> 194</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00195"></a><span class="lineno"> 195</span> </div><div class="line"><a name="l00196"></a><span class="lineno"> 196</span>  <span class="comment">// Initialize local maximum</span></div><div class="line"><a name="l00197"></a><span class="lineno"> 197</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>)</div><div class="line"><a name="l00198"></a><span class="lineno"> 198</span>  max_val_vec = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>))<a class="code" href="softmax__layer_8cl.xhtml#af7a08044d0e491a0ee1520a24a107a2b">type_min_</a>;</div><div class="line"><a name="l00199"></a><span class="lineno"> 199</span> </div><div class="line"><a name="l00200"></a><span class="lineno"> 200</span>  <span class="comment">// Calculate max of row</span></div><div class="line"><a name="l00201"></a><span class="lineno"> 201</span>  const uint width_ = width >> <a class="code" href="softmax__layer_8cl.xhtml#a372393c380805985b813dbb16d589a64">LOG_VECTOR_SIZE</a>;</div><div class="line"><a name="l00202"></a><span class="lineno"> 202</span>  for(uint i = 0; i < width_; i++)</div><div class="line"><a name="l00203"></a><span class="lineno"> 203</span>  {</div><div class="line"><a name="l00204"></a><span class="lineno"> 204</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, <a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>)</div><div class="line"><a name="l00205"></a><span class="lineno"> 205</span>  data_max = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(<a class="code" href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a>)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00206"></a><span class="lineno"> 206</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(data_max, max_val_vec, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00207"></a><span class="lineno"> 207</span>  }</div><div class="line"><a name="l00208"></a><span class="lineno"> 208</span> </div><div class="line"><a name="l00209"></a><span class="lineno"> 209</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_VECTOR_SIZE</span></div><div class="line"><a name="l00210"></a><span class="lineno"> 210</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, VECTOR_SIZE)</div><div class="line"><a name="l00211"></a><span class="lineno"> 211</span>  data_max = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(VECTOR_SIZE)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, width_ << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00212"></a><span class="lineno"> 212</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a>, VECTOR_SIZE)</div><div class="line"><a name="l00213"></a><span class="lineno"> 213</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>((<a class="code" href="helpers_8h.xhtml#ae4b532a93c757194ec73b6790a3e6b1f">EXPAND</a>((<a class="code" href="helpers_8h.xhtml#a0f6ad555d097377151003a5119ccee45">CL_VEC_DATA_TYPE</a>(uint, VECTOR_SIZE)))(width_ << LOG_VECTOR_SIZE) + <a class="code" href="softmax__layer_8cl.xhtml#aa1dd94b8d98f1c6d790bdf0fc5de29e9">idx__</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, VECTOR_SIZE));</div><div class="line"><a name="l00214"></a><span class="lineno"> 214</span>  max_val_vec = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec, select(type_min_, data_max, widx), DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00215"></a><span class="lineno"> 215</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_VECTOR_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00216"></a><span class="lineno"> 216</span> </div><div class="line"><a name="l00217"></a><span class="lineno"> 217</span>  <span class="comment">// Perform max reduction</span></div><div class="line"><a name="l00218"></a><span class="lineno"> 218</span> <span class="preprocessor">#if VECTOR_SIZE == 16</span></div><div class="line"><a name="l00219"></a><span class="lineno"> 219</span>  max_val_vec.s01234567 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s01234567, max_val_vec.s89ABCDEF, DATA_TYPE, 8);</div><div class="line"><a name="l00220"></a><span class="lineno"> 220</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 16 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00221"></a><span class="lineno"> 221</span> <span class="preprocessor">#if VECTOR_SIZE >= 8</span></div><div class="line"><a name="l00222"></a><span class="lineno"> 222</span>  max_val_vec.s0123 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s0123, max_val_vec.s4567, DATA_TYPE, 4);</div><div class="line"><a name="l00223"></a><span class="lineno"> 223</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 8 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00224"></a><span class="lineno"> 224</span> <span class="preprocessor">#if VECTOR_SIZE >= 4</span></div><div class="line"><a name="l00225"></a><span class="lineno"> 225</span>  max_val_vec.s01 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s01, max_val_vec.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00226"></a><span class="lineno"> 226</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 4 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00227"></a><span class="lineno"> 227</span>  max_val_vec.s0 = <a class="code" href="softmax__layer_8cl.xhtml#abaa48ad818c44e415fd3f9dd0f27bf01">MAX_OP</a>(max_val_vec.s0, max_val_vec.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00228"></a><span class="lineno"> 228</span>  <span class="comment">// Store result</span></div><div class="line"><a name="l00229"></a><span class="lineno"> 229</span>  *((__global DATA_TYPE *)maxo.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>) = max_val_vec.s0;</div><div class="line"><a name="l00230"></a><span class="lineno"> 230</span> </div><div class="line"><a name="l00231"></a><span class="lineno"> 231</span>  <span class="comment">/* Second section */</span></div><div class="line"><a name="l00232"></a><span class="lineno"> 232</span> </div><div class="line"><a name="l00233"></a><span class="lineno"> 233</span>  <span class="comment">// Load max value of 1D logits vector (row)</span></div><div class="line"><a name="l00234"></a><span class="lineno"> 234</span>  DATA_TYPE max_val = *((__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&maxo, 0, 0));</div><div class="line"><a name="l00235"></a><span class="lineno"> 235</span> </div><div class="line"><a name="l00236"></a><span class="lineno"> 236</span>  <span class="comment">// Set sum vector</span></div><div class="line"><a name="l00237"></a><span class="lineno"> 237</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, VECTOR_SIZE)</div><div class="line"><a name="l00238"></a><span class="lineno"> 238</span>  sum1D = 0;</div><div class="line"><a name="l00239"></a><span class="lineno"> 239</span> </div><div class="line"><a name="l00240"></a><span class="lineno"> 240</span>  <span class="comment">// Shift values, exp and sum</span></div><div class="line"><a name="l00241"></a><span class="lineno"> 241</span>  for(uint i = 0; i < width_; i++)</div><div class="line"><a name="l00242"></a><span class="lineno"> 242</span>  {</div><div class="line"><a name="l00243"></a><span class="lineno"> 243</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, VECTOR_SIZE)</div><div class="line"><a name="l00244"></a><span class="lineno"> 244</span>  data = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(VECTOR_SIZE)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00245"></a><span class="lineno"> 245</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00246"></a><span class="lineno"> 246</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00247"></a><span class="lineno"> 247</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00248"></a><span class="lineno"> 248</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00249"></a><span class="lineno"> 249</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00250"></a><span class="lineno"> 250</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(VECTOR_SIZE)</div><div class="line"><a name="l00251"></a><span class="lineno"> 251</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, i << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00252"></a><span class="lineno"> 252</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00253"></a><span class="lineno"> 253</span>  }</div><div class="line"><a name="l00254"></a><span class="lineno"> 254</span> </div><div class="line"><a name="l00255"></a><span class="lineno"> 255</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_VECTOR_SIZE</span></div><div class="line"><a name="l00256"></a><span class="lineno"> 256</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, VECTOR_SIZE)</div><div class="line"><a name="l00257"></a><span class="lineno"> 257</span>  data = <a class="code" href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a>(VECTOR_SIZE)(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, width_ << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00258"></a><span class="lineno"> 258</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00259"></a><span class="lineno"> 259</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00260"></a><span class="lineno"> 260</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00261"></a><span class="lineno"> 261</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00262"></a><span class="lineno"> 262</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00263"></a><span class="lineno"> 263</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>((<a class="code" href="helpers_8h.xhtml#ae4b532a93c757194ec73b6790a3e6b1f">EXPAND</a>((<a class="code" href="helpers_8h.xhtml#a0f6ad555d097377151003a5119ccee45">CL_VEC_DATA_TYPE</a>(uint, VECTOR_SIZE)))(width_ << LOG_VECTOR_SIZE) + <a class="code" href="softmax__layer_8cl.xhtml#aa1dd94b8d98f1c6d790bdf0fc5de29e9">idx__</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(SELECT_DATA_TYPE, VECTOR_SIZE));</div><div class="line"><a name="l00264"></a><span class="lineno"> 264</span>  data = select(0, data, widx);</div><div class="line"><a name="l00265"></a><span class="lineno"> 265</span>  <a class="code" href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a>(VECTOR_SIZE)</div><div class="line"><a name="l00266"></a><span class="lineno"> 266</span>  (data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, width_ << LOG_VECTOR_SIZE, 0));</div><div class="line"><a name="l00267"></a><span class="lineno"> 267</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, VECTOR_SIZE);</div><div class="line"><a name="l00268"></a><span class="lineno"> 268</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_VECTOR_SIZE */</span><span class="preprocessor"></span></div><div class="line"><a name="l00269"></a><span class="lineno"> 269</span> </div><div class="line"><a name="l00270"></a><span class="lineno"> 270</span>  <span class="comment">// Perform sum reduction</span></div><div class="line"><a name="l00271"></a><span class="lineno"> 271</span> <span class="preprocessor">#if VECTOR_SIZE == 16</span></div><div class="line"><a name="l00272"></a><span class="lineno"> 272</span>  sum1D.s01234567 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s01234567, sum1D.s89ABCDEF, DATA_TYPE, 8);</div><div class="line"><a name="l00273"></a><span class="lineno"> 273</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 16 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00274"></a><span class="lineno"> 274</span> <span class="preprocessor">#if VECTOR_SIZE >= 8</span></div><div class="line"><a name="l00275"></a><span class="lineno"> 275</span>  sum1D.s0123 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s0123, sum1D.s4567, DATA_TYPE, 4);</div><div class="line"><a name="l00276"></a><span class="lineno"> 276</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 8 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00277"></a><span class="lineno"> 277</span> <span class="preprocessor">#if VECTOR_SIZE >= 4</span></div><div class="line"><a name="l00278"></a><span class="lineno"> 278</span>  sum1D.s01 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s01, sum1D.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00279"></a><span class="lineno"> 279</span> <span class="preprocessor">#endif </span><span class="comment">/* VECTOR SIZE 4 END */</span><span class="preprocessor"></span></div><div class="line"><a name="l00280"></a><span class="lineno"> 280</span>  sum1D.s0 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s0, sum1D.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00281"></a><span class="lineno"> 281</span> </div><div class="line"><a name="l00282"></a><span class="lineno"> 282</span>  <span class="comment">// Calculate and store result</span></div><div class="line"><a name="l00283"></a><span class="lineno"> 283</span>  *((__global DATA_TYPE *)sum.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>) = sum1D.s0;</div><div class="line"><a name="l00284"></a><span class="lineno"> 284</span> }</div><div class="ttc" id="softmax__layer_8cl_xhtml_aa1dd94b8d98f1c6d790bdf0fc5de29e9"><div class="ttname"><a href="softmax__layer_8cl.xhtml#aa1dd94b8d98f1c6d790bdf0fc5de29e9">idx__</a></div><div class="ttdeci">__constant uint16 idx__</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00082">softmax_layer.cl:82</a></div></div>
+<div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_ab146b9cbab6e73e7588b240dc709fe01"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">arm_compute::test::validation::beta</a></div><div class="ttdeci">beta</div><div class="ttdef"><b>Definition:</b> <a href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00113">GEMM.cpp:113</a></div></div>
<div class="ttc" id="fixed__point_8h_xhtml_aa8d95ba04fc73845abc6045952cae5be"><div class="ttname"><a href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a></div><div class="ttdeci">#define CONVERT(x, type)</div><div class="ttdef"><b>Definition:</b> <a href="fixed__point_8h_source.xhtml#l00098">fixed_point.h:98</a></div></div>
<div class="ttc" id="helpers_8h_xhtml_a287e2fc366c312b468382c95bb90f91f"><div class="ttname"><a href="helpers_8h.xhtml#a287e2fc366c312b468382c95bb90f91f">VLOAD</a></div><div class="ttdeci">#define VLOAD(size)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00040">helpers.h:40</a></div></div>
<div class="ttc" id="softmax__layer_8cl_xhtml_a93cf800667317d96574477b9f0a75234"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a></div><div class="ttdeci">#define EXP_OP(x, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00048">softmax_layer.cl:48</a></div></div>
@@ -1230,7 +1062,7 @@
<div class="ttc" id="helpers_8h_xhtml_acb282042d1edeeaa3cc979a206f78b54"><div class="ttname"><a href="helpers_8h.xhtml#acb282042d1edeeaa3cc979a206f78b54">VSTORE</a></div><div class="ttdeci">#define VSTORE(size)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00043">helpers.h:43</a></div></div>
<div class="ttc" id="softmax__layer_8cl_xhtml_a7c78836761fa3b5b124efea237dac70f"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a7c78836761fa3b5b124efea237dac70f">VECTOR_SIZE</a></div><div class="ttdeci">#define VECTOR_SIZE</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00079">softmax_layer.cl:79</a></div></div>
<div class="ttc" id="softmax__layer_8cl_xhtml_af5b2e33e3c5fcaab3a213f26c2300170"><div class="ttname"><a href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a></div><div class="ttdeci">#define SELECT_DATA_TYPE</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00055">softmax_layer.cl:55</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#l00123">Convolution.cpp:123</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="softmax__layer_8cl_xhtml_a22303c4047ec5027c1538d53964b9d0d"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a></div><div class="ttdeci">#define MUL_OP(x, y, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00046">softmax_layer.cl:46</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>
@@ -1239,7 +1071,7 @@
<div class="ttc" id="fixed__point_8h_xhtml_a36f754c05b6fddf6df0d8d0a74f8159f"><div class="ttname"><a href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a></div><div class="ttdeci">#define VEC_DATA_TYPE(type, size)</div><div class="ttdef"><b>Definition:</b> <a href="fixed__point_8h_source.xhtml#l00093">fixed_point.h:93</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="helpers_8h_xhtml_ae4b532a93c757194ec73b6790a3e6b1f"><div class="ttname"><a href="helpers_8h.xhtml#ae4b532a93c757194ec73b6790a3e6b1f">EXPAND</a></div><div class="ttdeci">#define EXPAND(x)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00035">helpers.h:35</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 & src</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_convolution_8cpp_source.xhtml#l00133">Convolution.cpp:133</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 & 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>
@@ -1433,300 +1265,19 @@
</dd>
</dl>
-<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00292">292</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
+<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00120">120</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.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#l00111">CONVERT_TENSOR3D_TO_IMAGE_STRUCT_NO_STEP</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00047">DIV_OP</a>, <a class="el" href="_c_l_2_convolution_8cpp_source.xhtml#l00123">arm_compute::test::validation::dst</a>, <a class="el" href="helpers_8h_source.xhtml#l00301">offset()</a>, <a class="el" href="_c_l_2_convolution_8cpp_source.xhtml#l00133">arm_compute::test::validation::src</a>, <a class="el" href="reduction__operation_8cl_source.xhtml#l00052">sum()</a>, and <a class="el" href="fixed__point_8h_source.xhtml#l00093">VEC_DATA_TYPE</a>.</p>
-<div class="fragment"><div class="line"><a name="l00296"></a><span class="lineno"> 296</span> {</div><div class="line"><a name="l00297"></a><span class="lineno"> 297</span>  <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="l00298"></a><span class="lineno"> 298</span>  <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="l00299"></a><span class="lineno"> 299</span>  <a class="code" href="struct_image.xhtml">Image</a> <a class="code" href="reduction__operation_8cl.xhtml#ab0df00f5333da51860deb93deb44a782">sum</a> = <a class="code" href="helpers_8h.xhtml#a071aa45af973feac43b14f62e54a6fce">CONVERT_TENSOR3D_TO_IMAGE_STRUCT_NO_STEP</a>(sum);</div><div class="line"><a name="l00300"></a><span class="lineno"> 300</span> </div><div class="line"><a name="l00301"></a><span class="lineno"> 301</span>  <span class="comment">// Load max value of 1D logits vector (row)</span></div><div class="line"><a name="l00302"></a><span class="lineno"> 302</span>  <a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a> sum_val = *((__global <a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a> *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&sum, 0, get_global_id(1)));</div><div class="line"><a name="l00303"></a><span class="lineno"> 303</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 16)</div><div class="line"><a name="l00304"></a><span class="lineno"> 304</span>  data = vload16(0, (__global <a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a> *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, 0, 0));</div><div class="line"><a name="l00305"></a><span class="lineno"> 305</span>  vstore16(<a class="code" href="softmax__layer_8cl.xhtml#a8cde99b1ce0f3c1dacd49261b0cf03d8">DIV_OP</a>(data, sum_val, <a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 16), 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, 0, 0));</div><div class="line"><a name="l00306"></a><span class="lineno"> 306</span> }</div><div class="ttc" id="helpers_8h_xhtml_a071aa45af973feac43b14f62e54a6fce"><div class="ttname"><a href="helpers_8h.xhtml#a071aa45af973feac43b14f62e54a6fce">CONVERT_TENSOR3D_TO_IMAGE_STRUCT_NO_STEP</a></div><div class="ttdeci">#define CONVERT_TENSOR3D_TO_IMAGE_STRUCT_NO_STEP(name)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00111">helpers.h:111</a></div></div>
+<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#l00111">CONVERT_TENSOR3D_TO_IMAGE_STRUCT_NO_STEP</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00047">DIV_OP</a>, <a class="el" href="helpers_8h_source.xhtml#l00301">offset()</a>, and <a class="el" href="fixed__point_8h_source.xhtml#l00093">VEC_DATA_TYPE</a>.</p>
+<div class="fragment"><div class="line"><a name="l00124"></a><span class="lineno"> 124</span> {</div><div class="line"><a name="l00125"></a><span class="lineno"> 125</span>  <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="l00126"></a><span class="lineno"> 126</span>  <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="l00127"></a><span class="lineno"> 127</span>  <a class="code" href="struct_image.xhtml">Image</a> <a class="code" href="reduction__operation_8cl.xhtml#ab0df00f5333da51860deb93deb44a782">sum</a> = <a class="code" href="helpers_8h.xhtml#a071aa45af973feac43b14f62e54a6fce">CONVERT_TENSOR3D_TO_IMAGE_STRUCT_NO_STEP</a>(sum);</div><div class="line"><a name="l00128"></a><span class="lineno"> 128</span> </div><div class="line"><a name="l00129"></a><span class="lineno"> 129</span>  <span class="comment">// Load max value of 1D logits vector (row)</span></div><div class="line"><a name="l00130"></a><span class="lineno"> 130</span>  <a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a> sum_val = *((__global <a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a> *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&sum, 0, get_global_id(1)));</div><div class="line"><a name="l00131"></a><span class="lineno"> 131</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 16)</div><div class="line"><a name="l00132"></a><span class="lineno"> 132</span>  data = vload16(0, (__global <a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a> *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, 0, 0));</div><div class="line"><a name="l00133"></a><span class="lineno"> 133</span>  vstore16(<a class="code" href="softmax__layer_8cl.xhtml#a8cde99b1ce0f3c1dacd49261b0cf03d8">DIV_OP</a>(data, sum_val, <a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 16), 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, 0, 0));</div><div class="line"><a name="l00134"></a><span class="lineno"> 134</span> }</div><div class="ttc" id="helpers_8h_xhtml_a071aa45af973feac43b14f62e54a6fce"><div class="ttname"><a href="helpers_8h.xhtml#a071aa45af973feac43b14f62e54a6fce">CONVERT_TENSOR3D_TO_IMAGE_STRUCT_NO_STEP</a></div><div class="ttdeci">#define CONVERT_TENSOR3D_TO_IMAGE_STRUCT_NO_STEP(name)</div><div class="ttdef"><b>Definition:</b> <a href="helpers_8h_source.xhtml#l00111">helpers.h:111</a></div></div>
<div class="ttc" id="convolution3x3_8cl_xhtml_afb8c72ce35c4a1f4a2588d6573e54aa1"><div class="ttname"><a href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a></div><div class="ttdeci">#define DATA_TYPE</div><div class="ttdef"><b>Definition:</b> <a href="convolution3x3_8cl_source.xhtml#l00027">convolution3x3.cl:27</a></div></div>
<div class="ttc" id="reduction__operation_8cl_xhtml_ab0df00f5333da51860deb93deb44a782"><div class="ttname"><a href="reduction__operation_8cl.xhtml#ab0df00f5333da51860deb93deb44a782">sum</a></div><div class="ttdeci">DATA_TYPE sum(__global const DATA_TYPE *input)</div><div class="ttdoc">Calculate sum of a vector. </div><div class="ttdef"><b>Definition:</b> <a href="reduction__operation_8cl_source.xhtml#l00052">reduction_operation.cl:52</a></div></div>
<div class="ttc" id="softmax__layer_8cl_xhtml_a8cde99b1ce0f3c1dacd49261b0cf03d8"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a8cde99b1ce0f3c1dacd49261b0cf03d8">DIV_OP</a></div><div class="ttdeci">#define DIV_OP(x, y, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00047">softmax_layer.cl:47</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#l00123">Convolution.cpp:123</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="fixed__point_8h_xhtml_a36f754c05b6fddf6df0d8d0a74f8159f"><div class="ttname"><a href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a></div><div class="ttdeci">#define VEC_DATA_TYPE(type, size)</div><div class="ttdef"><b>Definition:</b> <a href="fixed__point_8h_source.xhtml#l00093">fixed_point.h:93</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="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 & src</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_convolution_8cpp_source.xhtml#l00133">Convolution.cpp:133</a></div></div>
-</div><!-- fragment -->
-</div>
-</div>
-<a class="anchor" id="aa4412bb319c67deaeab07a481d93dfdb"></a>
-<div class="memitem">
-<div class="memproto">
- <table class="memname">
- <tr>
- <td class="memname">__kernel void softmax_layer_shift_exp_sum </td>
- <td>(</td>
- <td class="paramtype">__global uchar * </td>
- <td class="paramname"><em>src_ptr</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_stride_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_step_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_stride_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_step_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_stride_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>src_step_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </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 * </td>
- <td class="paramname"><em>max_ptr</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>max_stride_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>max_step_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>max_stride_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>max_step_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>max_stride_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>max_step_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>max_offset_first_element_in_bytes</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">__global uchar * </td>
- <td class="paramname"><em>dst_ptr</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_stride_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_step_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_stride_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_step_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_stride_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>dst_step_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </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 * </td>
- <td class="paramname"><em>sum_ptr</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>sum_stride_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>sum_step_x</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>sum_stride_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>sum_step_y</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>sum_stride_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>sum_step_z</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>sum_offset_first_element_in_bytes</em>, </td>
- </tr>
- <tr>
- <td class="paramkey"></td>
- <td></td>
- <td class="paramtype">uint </td>
- <td class="paramname"><em>width</em> </td>
- </tr>
- <tr>
- <td></td>
- <td>)</td>
- <td></td><td></td>
- </tr>
- </table>
-</div><div class="memdoc">
-
-<p>Shifts the values of the input tensor by the max calculated in softmax_layer_max kernel, then gets the exponent of each element as sums all elements across each row. </p>
-<dl class="section note"><dt>Note</dt><dd>Datatype must be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=short </dd>
-<dd>
-Fixed point position must be given as a preprocessor argument using -DFIXED_POINT_POSITION=pos. e.g. DFIXED_POINT_POSITION=4 </dd>
-<dd>
-In case the input is not multiple of 16 -DNON_MULTIPLE_OF_16 must be passed. </dd>
-<dd>
-Beta can be optionally passed at compile time using -DBETA (if undefined, assume it equals 1.0)</dd></dl>
-<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 tensor slice. Supported data types: QS8/QS16/F16/F32 </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">src_stride_x</td><td>Stride of the source tensor 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 tensor 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_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 Z 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 tensor </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">max_ptr</td><td>Pointer to the max values tensor slice. Supported data types: same as <code>src_ptr</code> </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">max_stride_x</td><td>Stride of the max values tensor in X dimension (in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">max_step_x</td><td>max_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">max_stride_y</td><td>Stride of the max values tensor in Y dimension (in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">max_step_y</td><td>max_stride_y * number of elements along Y processed per workitem(in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">max_stride_z</td><td>Stride of the max values tensor in Z dimension (in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">max_step_z</td><td>max_stride_z * number of elements along Z processed per workitem(in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">max_offset_first_element_in_bytes</td><td>The offset of the first element in the max values tensor </td></tr>
- <tr><td class="paramdir">[out]</td><td class="paramname">dst_ptr</td><td>Pointer to the destination tensor slice. Supported data types: same as <code>src_ptr</code> </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 Z 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">[out]</td><td class="paramname">sum_ptr</td><td>Pointer to the sum values tensor slice. Supported data types: same as <code>src_ptr</code> </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">sum_stride_x</td><td>Stride of the sum values tensor in X dimension (in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">sum_step_x</td><td>sum_stride_x * number of elements along X processed per workitem(in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">sum_stride_y</td><td>Stride of the sum values tensor in Y dimension (in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">sum_step_y</td><td>sum_stride_z * number of elements along Z processed per workitem(in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">sum_stride_z</td><td>Stride of the sum values tensor in Z dimension (in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">sum_step_z</td><td>sum_stride_z * number of elements along Z processed per workitem(in bytes) </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">sum_offset_first_element_in_bytes</td><td>The offset of the first element in the sum values tensor </td></tr>
- <tr><td class="paramdir">[in]</td><td class="paramname">width</td><td>Input image width </td></tr>
- </table>
- </dd>
-</dl>
-
-<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00196">196</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-
-<p>References <a class="el" href="softmax__layer_8cl_source.xhtml#l00044">ADD_OP</a>, <a class="el" href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00112">arm_compute::test::validation::beta</a>, <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="_c_l_2_convolution_8cpp_source.xhtml#l00123">arm_compute::test::validation::dst</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00048">EXP_OP</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00087">idx16</a>, <a class="el" href="tests_2validation_2_fixed_point_8h_source.xhtml#l00889">arm_compute::test::fixed_point_arithmetic::detail::max()</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00046">MUL_OP</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="softmax__layer_8cl_source.xhtml#l00055">SELECT_DATA_TYPE</a>, <a class="el" href="_c_l_2_convolution_8cpp_source.xhtml#l00133">arm_compute::test::validation::src</a>, <a class="el" href="softmax__layer_8cl_source.xhtml#l00045">SUB_OP</a>, <a class="el" href="reduction__operation_8cl_source.xhtml#l00052">sum()</a>, and <a class="el" href="fixed__point_8h_source.xhtml#l00093">VEC_DATA_TYPE</a>.</p>
-<div class="fragment"><div class="line"><a name="l00202"></a><span class="lineno"> 202</span> {</div><div class="line"><a name="l00203"></a><span class="lineno"> 203</span>  <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="l00204"></a><span class="lineno"> 204</span>  <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="l00205"></a><span class="lineno"> 205</span>  <a class="code" href="struct_image.xhtml">Image</a> <a class="code" href="namespacearm__compute_1_1test_1_1fixed__point__arithmetic_1_1detail.xhtml#ad91bb73431b4de1f4946ed949d444849">max</a> = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(max);</div><div class="line"><a name="l00206"></a><span class="lineno"> 206</span>  <a class="code" href="struct_image.xhtml">Image</a> <a class="code" href="reduction__operation_8cl.xhtml#ab0df00f5333da51860deb93deb44a782">sum</a> = <a class="code" href="helpers_8h.xhtml#a541f8db866a0fa93ee67d58ea31a7d0c">CONVERT_TENSOR3D_TO_IMAGE_STRUCT</a>(sum);</div><div class="line"><a name="l00207"></a><span class="lineno"> 207</span> </div><div class="line"><a name="l00208"></a><span class="lineno"> 208</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00209"></a><span class="lineno"> 209</span>  <span class="comment">// Initialize beta</span></div><div class="line"><a name="l00210"></a><span class="lineno"> 210</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 16)</div><div class="line"><a name="l00211"></a><span class="lineno"> 211</span>  <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a> = (<a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a>, 16))BETA;</div><div class="line"><a name="l00212"></a><span class="lineno"> 212</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00213"></a><span class="lineno"> 213</span> </div><div class="line"><a name="l00214"></a><span class="lineno"> 214</span>  <span class="comment">// Load max value of 1D logits vector (row)</span></div><div class="line"><a name="l00215"></a><span class="lineno"> 215</span>  DATA_TYPE max_val = *((__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&max, 0, 0));</div><div class="line"><a name="l00216"></a><span class="lineno"> 216</span> </div><div class="line"><a name="l00217"></a><span class="lineno"> 217</span>  <span class="comment">// Set sum vector</span></div><div class="line"><a name="l00218"></a><span class="lineno"> 218</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 16)</div><div class="line"><a name="l00219"></a><span class="lineno"> 219</span>  sum1D = 0;</div><div class="line"><a name="l00220"></a><span class="lineno"> 220</span> </div><div class="line"><a name="l00221"></a><span class="lineno"> 221</span>  <span class="comment">// Shift values, exp and sum</span></div><div class="line"><a name="l00222"></a><span class="lineno"> 222</span>  const uint width4 = width >> 4;</div><div class="line"><a name="l00223"></a><span class="lineno"> 223</span>  for(uint i = 0; i < width4; i++)</div><div class="line"><a name="l00224"></a><span class="lineno"> 224</span>  {</div><div class="line"><a name="l00225"></a><span class="lineno"> 225</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 16)</div><div class="line"><a name="l00226"></a><span class="lineno"> 226</span>  data = vload16(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, i << 4, 0));</div><div class="line"><a name="l00227"></a><span class="lineno"> 227</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, 16);</div><div class="line"><a name="l00228"></a><span class="lineno"> 228</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00229"></a><span class="lineno"> 229</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, 16);</div><div class="line"><a name="l00230"></a><span class="lineno"> 230</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00231"></a><span class="lineno"> 231</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, 16);</div><div class="line"><a name="l00232"></a><span class="lineno"> 232</span>  vstore16(data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, i << 4, 0));</div><div class="line"><a name="l00233"></a><span class="lineno"> 233</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, 16);</div><div class="line"><a name="l00234"></a><span class="lineno"> 234</span>  }</div><div class="line"><a name="l00235"></a><span class="lineno"> 235</span> </div><div class="line"><a name="l00236"></a><span class="lineno"> 236</span> <span class="preprocessor">#ifdef NON_MULTIPLE_OF_16</span></div><div class="line"><a name="l00237"></a><span class="lineno"> 237</span>  <span class="comment">// Handle non multiple of 16</span></div><div class="line"><a name="l00238"></a><span class="lineno"> 238</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(DATA_TYPE, 16)</div><div class="line"><a name="l00239"></a><span class="lineno"> 239</span>  data = vload16(0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&src, width4 << 4, 0));</div><div class="line"><a name="l00240"></a><span class="lineno"> 240</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a>(data, max_val, DATA_TYPE, 16);</div><div class="line"><a name="l00241"></a><span class="lineno"> 241</span> <span class="preprocessor">#ifdef BETA</span></div><div class="line"><a name="l00242"></a><span class="lineno"> 242</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a>(data, <a class="code" href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">beta</a>, DATA_TYPE, 16);</div><div class="line"><a name="l00243"></a><span class="lineno"> 243</span> <span class="preprocessor">#endif </span><span class="comment">/* BETA */</span><span class="preprocessor"></span></div><div class="line"><a name="l00244"></a><span class="lineno"> 244</span>  data = <a class="code" href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a>(data, DATA_TYPE, 16);</div><div class="line"><a name="l00245"></a><span class="lineno"> 245</span>  <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a>, 16)</div><div class="line"><a name="l00246"></a><span class="lineno"> 246</span>  widx = <a class="code" href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a>(((uint16)(width4 << 4) + <a class="code" href="softmax__layer_8cl.xhtml#a0712735973f172ac9efc7d48a31e47ad">idx16</a>) < width, <a class="code" href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a>(<a class="code" href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a>, 16));</div><div class="line"><a name="l00247"></a><span class="lineno"> 247</span>  data = select(0, data, widx);</div><div class="line"><a name="l00248"></a><span class="lineno"> 248</span>  vstore16(data, 0, (__global DATA_TYPE *)<a class="code" href="helpers_8h.xhtml#a009469e4d9b8fce3b6d5e97d2077827d">offset</a>(&dst, width4 << 4, 0));</div><div class="line"><a name="l00249"></a><span class="lineno"> 249</span>  sum1D = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D, data, DATA_TYPE, 16);</div><div class="line"><a name="l00250"></a><span class="lineno"> 250</span> <span class="preprocessor">#endif </span><span class="comment">/* NON_MULTIPLE_OF_16 */</span><span class="preprocessor"></span></div><div class="line"><a name="l00251"></a><span class="lineno"> 251</span> </div><div class="line"><a name="l00252"></a><span class="lineno"> 252</span>  <span class="comment">// Perform min/max reduction</span></div><div class="line"><a name="l00253"></a><span class="lineno"> 253</span>  sum1D.s01234567 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s01234567, sum1D.s89ABCDEF, DATA_TYPE, 8);</div><div class="line"><a name="l00254"></a><span class="lineno"> 254</span>  sum1D.s0123 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s0123, sum1D.s4567, DATA_TYPE, 4);</div><div class="line"><a name="l00255"></a><span class="lineno"> 255</span>  sum1D.s01 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s01, sum1D.s23, DATA_TYPE, 2);</div><div class="line"><a name="l00256"></a><span class="lineno"> 256</span>  sum1D.s0 = <a class="code" href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a>(sum1D.s0, sum1D.s1, DATA_TYPE, 1);</div><div class="line"><a name="l00257"></a><span class="lineno"> 257</span> </div><div class="line"><a name="l00258"></a><span class="lineno"> 258</span>  <span class="comment">// Calculate and store result</span></div><div class="line"><a name="l00259"></a><span class="lineno"> 259</span>  *((__global DATA_TYPE *)sum.<a class="code" href="struct_image.xhtml#acf52c23cbd7424606c10a606524e3e32">ptr</a>) = sum1D.s0;</div><div class="line"><a name="l00260"></a><span class="lineno"> 260</span> }</div><div class="ttc" id="namespacearm__compute_1_1test_1_1validation_xhtml_ab146b9cbab6e73e7588b240dc709fe01"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1validation.xhtml#ab146b9cbab6e73e7588b240dc709fe01">arm_compute::test::validation::beta</a></div><div class="ttdeci">beta</div><div class="ttdef"><b>Definition:</b> <a href="validation_2_c_l_2_g_e_m_m_8cpp_source.xhtml#l00112">GEMM.cpp:112</a></div></div>
-<div class="ttc" id="fixed__point_8h_xhtml_aa8d95ba04fc73845abc6045952cae5be"><div class="ttname"><a href="fixed__point_8h.xhtml#aa8d95ba04fc73845abc6045952cae5be">CONVERT</a></div><div class="ttdeci">#define CONVERT(x, type)</div><div class="ttdef"><b>Definition:</b> <a href="fixed__point_8h_source.xhtml#l00098">fixed_point.h:98</a></div></div>
-<div class="ttc" id="softmax__layer_8cl_xhtml_a93cf800667317d96574477b9f0a75234"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a93cf800667317d96574477b9f0a75234">EXP_OP</a></div><div class="ttdeci">#define EXP_OP(x, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00048">softmax_layer.cl:48</a></div></div>
-<div class="ttc" id="convolution3x3_8cl_xhtml_afb8c72ce35c4a1f4a2588d6573e54aa1"><div class="ttname"><a href="convolution3x3_8cl.xhtml#afb8c72ce35c4a1f4a2588d6573e54aa1">DATA_TYPE</a></div><div class="ttdeci">#define DATA_TYPE</div><div class="ttdef"><b>Definition:</b> <a href="convolution3x3_8cl_source.xhtml#l00027">convolution3x3.cl:27</a></div></div>
-<div class="ttc" id="reduction__operation_8cl_xhtml_ab0df00f5333da51860deb93deb44a782"><div class="ttname"><a href="reduction__operation_8cl.xhtml#ab0df00f5333da51860deb93deb44a782">sum</a></div><div class="ttdeci">DATA_TYPE sum(__global const DATA_TYPE *input)</div><div class="ttdoc">Calculate sum of a vector. </div><div class="ttdef"><b>Definition:</b> <a href="reduction__operation_8cl_source.xhtml#l00052">reduction_operation.cl:52</a></div></div>
-<div class="ttc" id="softmax__layer_8cl_xhtml_a44206a4e5783c7aabacec88aad878c88"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a44206a4e5783c7aabacec88aad878c88">ADD_OP</a></div><div class="ttdeci">#define ADD_OP(x, y, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00044">softmax_layer.cl:44</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="softmax__layer_8cl_xhtml_af5b2e33e3c5fcaab3a213f26c2300170"><div class="ttname"><a href="softmax__layer_8cl.xhtml#af5b2e33e3c5fcaab3a213f26c2300170">SELECT_DATA_TYPE</a></div><div class="ttdeci">#define SELECT_DATA_TYPE</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00055">softmax_layer.cl:55</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#l00123">Convolution.cpp:123</a></div></div>
-<div class="ttc" id="softmax__layer_8cl_xhtml_a22303c4047ec5027c1538d53964b9d0d"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a22303c4047ec5027c1538d53964b9d0d">MUL_OP</a></div><div class="ttdeci">#define MUL_OP(x, y, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00046">softmax_layer.cl:46</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="softmax__layer_8cl_xhtml_ac3af2d18008cbbf7247ae48fcd6e0c4e"><div class="ttname"><a href="softmax__layer_8cl.xhtml#ac3af2d18008cbbf7247ae48fcd6e0c4e">SUB_OP</a></div><div class="ttdeci">#define SUB_OP(x, y, type, size)</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00045">softmax_layer.cl:45</a></div></div>
-<div class="ttc" id="fixed__point_8h_xhtml_a36f754c05b6fddf6df0d8d0a74f8159f"><div class="ttname"><a href="fixed__point_8h.xhtml#a36f754c05b6fddf6df0d8d0a74f8159f">VEC_DATA_TYPE</a></div><div class="ttdeci">#define VEC_DATA_TYPE(type, size)</div><div class="ttdef"><b>Definition:</b> <a href="fixed__point_8h_source.xhtml#l00093">fixed_point.h:93</a></div></div>
-<div class="ttc" id="softmax__layer_8cl_xhtml_a0712735973f172ac9efc7d48a31e47ad"><div class="ttname"><a href="softmax__layer_8cl.xhtml#a0712735973f172ac9efc7d48a31e47ad">idx16</a></div><div class="ttdeci">__constant uint16 idx16</div><div class="ttdef"><b>Definition:</b> <a href="softmax__layer_8cl_source.xhtml#l00087">softmax_layer.cl:87</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="namespacearm__compute_1_1test_1_1fixed__point__arithmetic_1_1detail_xhtml_ad91bb73431b4de1f4946ed949d444849"><div class="ttname"><a href="namespacearm__compute_1_1test_1_1fixed__point__arithmetic_1_1detail.xhtml#ad91bb73431b4de1f4946ed949d444849">arm_compute::test::fixed_point_arithmetic::detail::max</a></div><div class="ttdeci">fixed_point< T > max(fixed_point< T > x, fixed_point< T > y)</div><div class="ttdef"><b>Definition:</b> <a href="tests_2validation_2_fixed_point_8h_source.xhtml#l00889">FixedPoint.h:889</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 & src</div><div class="ttdef"><b>Definition:</b> <a href="_c_l_2_convolution_8cpp_source.xhtml#l00133">Convolution.cpp:133</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 & 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>
@@ -1743,8 +1294,6 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00087">87</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00114">softmax_layer_max()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00196">softmax_layer_shift_exp_sum()</a>.</p>
-
</div>
</div>
<a class="anchor" id="a4884a666a1e93fbf8c27bd7d2da3c8bb"></a>
@@ -1759,7 +1308,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00088">88</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">softmax_layer_max_shift_exp_sum_parallel()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">softmax_layer_max_shift_exp_sum_parallel()</a>.</p>
</div>
</div>
@@ -1775,7 +1324,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00082">82</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
@@ -1791,8 +1340,6 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00086">86</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00114">softmax_layer_max()</a>.</p>
-
</div>
</div>
<a class="anchor" id="af7a08044d0e491a0ee1520a24a107a2b"></a>
@@ -1807,7 +1354,7 @@
<p>Definition at line <a class="el" href="softmax__layer_8cl_source.xhtml#l00081">81</a> of file <a class="el" href="softmax__layer_8cl_source.xhtml">softmax_layer.cl</a>.</p>
-<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00500">softmax_layer_max_shift_exp_sum_parallel()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00350">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
+<p>Referenced by <a class="el" href="softmax__layer_8cl_source.xhtml#l00328">softmax_layer_max_shift_exp_sum_parallel()</a>, and <a class="el" href="softmax__layer_8cl_source.xhtml#l00178">softmax_layer_max_shift_exp_sum_serial()</a>.</p>
</div>
</div>
@@ -1817,7 +1364,7 @@
<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="softmax__layer_8cl.xhtml">softmax_layer.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>