208 lines
170 KiB
HTML
208 lines
170 KiB
HTML
|
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
|
||
|
|
<html xmlns="http://www.w3.org/1999/xhtml">
|
||
|
|
<head>
|
||
|
|
<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
|
||
|
|
<meta http-equiv="X-UA-Compatible" content="IE=9"/>
|
||
|
|
<meta name="generator" content="Doxygen 1.8.11"/>
|
||
|
|
<title>CUTLASS: gemm_splitk_parallel.h Source File</title>
|
||
|
|
<link href="tabs.css" rel="stylesheet" type="text/css"/>
|
||
|
|
<script type="text/javascript" src="jquery.js"></script>
|
||
|
|
<script type="text/javascript" src="dynsections.js"></script>
|
||
|
|
<link href="search/search.css" rel="stylesheet" type="text/css"/>
|
||
|
|
<script type="text/javascript" src="search/searchdata.js"></script>
|
||
|
|
<script type="text/javascript" src="search/search.js"></script>
|
||
|
|
<script type="text/javascript">
|
||
|
|
$(document).ready(function() { init_search(); });
|
||
|
|
</script>
|
||
|
|
<script type="text/x-mathjax-config">
|
||
|
|
MathJax.Hub.Config({
|
||
|
|
extensions: ["tex2jax.js"],
|
||
|
|
jax: ["input/TeX","output/HTML-CSS"],
|
||
|
|
});
|
||
|
|
</script><script type="text/javascript" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js"></script>
|
||
|
|
<link href="doxygen.css" rel="stylesheet" type="text/css" />
|
||
|
|
</head>
|
||
|
|
<body>
|
||
|
|
<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
|
||
|
|
<div id="titlearea">
|
||
|
|
<table cellspacing="0" cellpadding="0">
|
||
|
|
<tbody>
|
||
|
|
<tr style="height: 56px;">
|
||
|
|
<td id="projectlogo"><img alt="Logo" src="cutlass-logo-small.png"/></td>
|
||
|
|
<td id="projectalign" style="padding-left: 0.5em;">
|
||
|
|
<div id="projectname">CUTLASS
|
||
|
|
</div>
|
||
|
|
<div id="projectbrief">CUDA Templates for Linear Algebra Subroutines and Solvers</div>
|
||
|
|
</td>
|
||
|
|
</tr>
|
||
|
|
</tbody>
|
||
|
|
</table>
|
||
|
|
</div>
|
||
|
|
<!-- end header part -->
|
||
|
|
<!-- Generated by Doxygen 1.8.11 -->
|
||
|
|
<script type="text/javascript">
|
||
|
|
var searchBox = new SearchBox("searchBox", "search",false,'Search');
|
||
|
|
</script>
|
||
|
|
<div id="navrow1" class="tabs">
|
||
|
|
<ul class="tablist">
|
||
|
|
<li><a href="index.html"><span>Main Page</span></a></li>
|
||
|
|
<li><a href="modules.html"><span>Modules</span></a></li>
|
||
|
|
<li><a href="namespaces.html"><span>Namespaces</span></a></li>
|
||
|
|
<li><a href="annotated.html"><span>Classes</span></a></li>
|
||
|
|
<li class="current"><a href="files.html"><span>Files</span></a></li>
|
||
|
|
<li>
|
||
|
|
<div id="MSearchBox" class="MSearchBoxInactive">
|
||
|
|
<span class="left">
|
||
|
|
<img id="MSearchSelect" src="search/mag_sel.png"
|
||
|
|
onmouseover="return searchBox.OnSearchSelectShow()"
|
||
|
|
onmouseout="return searchBox.OnSearchSelectHide()"
|
||
|
|
alt=""/>
|
||
|
|
<input type="text" id="MSearchField" value="Search" accesskey="S"
|
||
|
|
onfocus="searchBox.OnSearchFieldFocus(true)"
|
||
|
|
onblur="searchBox.OnSearchFieldFocus(false)"
|
||
|
|
onkeyup="searchBox.OnSearchFieldChange(event)"/>
|
||
|
|
</span><span class="right">
|
||
|
|
<a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
|
||
|
|
</span>
|
||
|
|
</div>
|
||
|
|
</li>
|
||
|
|
</ul>
|
||
|
|
</div>
|
||
|
|
<div id="navrow2" class="tabs2">
|
||
|
|
<ul class="tablist">
|
||
|
|
<li><a href="files.html"><span>File List</span></a></li>
|
||
|
|
<li><a href="globals.html"><span>File Members</span></a></li>
|
||
|
|
</ul>
|
||
|
|
</div>
|
||
|
|
<!-- window showing the filter options -->
|
||
|
|
<div id="MSearchSelectWindow"
|
||
|
|
onmouseover="return searchBox.OnSearchSelectShow()"
|
||
|
|
onmouseout="return searchBox.OnSearchSelectHide()"
|
||
|
|
onkeydown="return searchBox.OnSearchSelectKey(event)">
|
||
|
|
</div>
|
||
|
|
|
||
|
|
<!-- iframe showing the search results (closed by default) -->
|
||
|
|
<div id="MSearchResultsWindow">
|
||
|
|
<iframe src="javascript:void(0)" frameborder="0"
|
||
|
|
name="MSearchResults" id="MSearchResults">
|
||
|
|
</iframe>
|
||
|
|
</div>
|
||
|
|
|
||
|
|
<div id="nav-path" class="navpath">
|
||
|
|
<ul>
|
||
|
|
<li class="navelem"><a class="el" href="dir_d44c64559bbebec7f509842c48db8b23.html">include</a></li><li class="navelem"><a class="el" href="dir_6baf2bb612a2f0daa69af3101ede80a1.html">cutlass</a></li><li class="navelem"><a class="el" href="dir_9aa36bd9cfad59a1f88859a38871c977.html">gemm</a></li><li class="navelem"><a class="el" href="dir_36528dc2736efa40b421028b7309c671.html">device</a></li> </ul>
|
||
|
|
</div>
|
||
|
|
</div><!-- top -->
|
||
|
|
<div class="header">
|
||
|
|
<div class="headertitle">
|
||
|
|
<div class="title">device/gemm_splitk_parallel.h</div> </div>
|
||
|
|
</div><!--header-->
|
||
|
|
<div class="contents">
|
||
|
|
<a href="device_2gemm__splitk__parallel_8h.html">Go to the documentation of this file.</a><div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno"> 1</span> <span class="comment">/***************************************************************************************************</span></div><div class="line"><a name="l00002"></a><span class="lineno"> 2</span> <span class="comment"> * Copyright (c) 2017-2019, NVIDIA CORPORATION. All rights reserved.</span></div><div class="line"><a name="l00003"></a><span class="lineno"> 3</span> <span class="comment"> *</span></div><div class="line"><a name="l00004"></a><span class="lineno"> 4</span> <span class="comment"> * Redistribution and use in source and binary forms, with or without modification, are permitted</span></div><div class="line"><a name="l00005"></a><span class="lineno"> 5</span> <span class="comment"> * provided that the following conditions are met:</span></div><div class="line"><a name="l00006"></a><span class="lineno"> 6</span> <span class="comment"> * * Redistributions of source code must retain the above copyright notice, this list of</span></div><div class="line"><a name="l00007"></a><span class="lineno"> 7</span> <span class="comment"> * conditions and the following disclaimer.</span></div><div class="line"><a name="l00008"></a><span class="lineno"> 8</span> <span class="comment"> * * Redistributions in binary form must reproduce the above copyright notice, this list of</span></div><div class="line"><a name="l00009"></a><span class="lineno"> 9</span> <span class="comment"> * conditions and the following disclaimer in the documentation and/or other materials</span></div><div class="line"><a name="l00010"></a><span class="lineno"> 10</span> <span class="comment"> * provided with the distribution.</span></div><div class="line"><a name="l00011"></a><span class="lineno"> 11</span> <span class="comment"> * * Neither the name of the NVIDIA CORPORATION nor the names of its contributors may be used</span></div><div class="line"><a name="l00012"></a><span class="lineno"> 12</span> <span class="comment"> * to endorse or promote products derived from this software without specific prior written</span></div><div class="line"><a name="l00013"></a><span class="lineno"> 13</span> <span class="comment"> * permission.</span></div><div class="line"><a name="l00014"></a><span class="lineno"> 14</span> <span class="comment"> *</span></div><div class="line"><a name="l00015"></a><span class="lineno"> 15</span> <span class="comment"> * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND ANY EXPRESS OR</span></div><div class="line"><a name="l00016"></a><span class="lineno"> 16</span> <span class="comment"> * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND</span></div><div class="line"><a name="l00017"></a><span class="lineno"> 17</span> <span class="comment"> * FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL NVIDIA CORPORATION BE LIABLE</span></div><div class="line"><a name="l00018"></a><span class="lineno"> 18</span> <span class="comment"> * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,</span></div><div class="line"><a name="l00019"></a><span class="lineno"> 19</span> <span class="comment"> * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;</span></div><div class="line"><a name="l00020"></a><span class="lineno"> 20</span> <span class="comment"> * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,</span></div><div class="line"><a name="l00021"></a><span class="lineno"> 21</span> <span class="comment"> * STRICT LIABILITY, OR TOR (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE</span></div><div class="line
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a36698ab97ffbc16808b60f041a1e583a"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a36698ab97ffbc16808b60f041a1e583a">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::WarpShape</a></div><div class="ttdeci">WarpShape WarpShape</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:136</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a949dbf8f84e6350649a171bf3b45478a"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a949dbf8f84e6350649a171bf3b45478a">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::GemmKernel</a></div><div class="ttdeci">typename UnderlyingOperator::GemmKernel GemmKernel</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:499</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_a850da307d8741296e515add0f716eaf9"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#a850da307d8741296e515add0f716eaf9">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::ref_D</a></div><div class="ttdeci">TensorRef< ElementC, LayoutC > ref_D</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:513</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a08550ee41dd02322c514364035f5d098"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a08550ee41dd02322c514364035f5d098">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::LayoutB</a></div><div class="ttdeci">typename layout::LayoutTranspose< LayoutA >::type LayoutB</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:129</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_a2aabb13f196a087b77245c67c8664b7b"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#a2aabb13f196a087b77245c67c8664b7b">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::ref_C</a></div><div class="ttdeci">TensorRef< ElementC const, LayoutC > ref_C</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:512</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_ae436b25ceca72104f23d09442de78f73"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#ae436b25ceca72104f23d09442de78f73">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::Operator</a></div><div class="ttdeci">Operator Operator</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:142</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1MatrixShape_html"><div class="ttname"><a href="structcutlass_1_1MatrixShape.html">cutlass::MatrixShape</a></div><div class="ttdoc">Describes the size of a matrix tile. </div><div class="ttdef"><b>Definition:</b> matrix_shape.h:42</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_aa69c465611c07990cdc79605c16b04ff"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#aa69c465611c07990cdc79605c16b04ff">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::ReductionKernel</a></div><div class="ttdeci">typename UnderlyingOperator::ReductionKernel ReductionKernel</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:500</div></div>
|
||
|
|
<div class="ttc" id="namespacecutlass_html"><div class="ttname"><a href="namespacecutlass.html">cutlass</a></div><div class="ttdef"><b>Definition:</b> aligned_buffer.h:35</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_adee4f1a66aa6b6cb0400f6159ec52eb9"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#adee4f1a66aa6b6cb0400f6159ec52eb9">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::problem_size</a></div><div class="ttdeci">GemmCoord problem_size</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:509</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1kernel_1_1DefaultGemmSplitKParallel_html"><div class="ttname"><a href="structcutlass_1_1gemm_1_1kernel_1_1DefaultGemmSplitKParallel.html">cutlass::gemm::kernel::DefaultGemmSplitKParallel</a></div><div class="ttdef"><b>Definition:</b> default_gemm_splitk_parallel.h:88</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_af324f9dd7abc0e3ae6007e6dcf186190"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#af324f9dd7abc0e3ae6007e6dcf186190">cutlass::gemm::device::GemmSplitKParallel::operator()</a></div><div class="ttdeci">Status operator()(Arguments const &args, void *workspace=nullptr, cudaStream_t stream=nullptr)</div><div class="ttdoc">Runs the kernel using initialized state. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:395</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_af1f647942f7734bbf01e473118f2512c"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#af1f647942f7734bbf01e473118f2512c">cutlass::gemm::device::GemmSplitKParallel::kStages</a></div><div class="ttdeci">static int const kStages</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:143</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1reduction_1_1kernel_1_1ReduceSplitK_html_af788ae48c72021b8ce49da15dfa72be3"><div class="ttname"><a href="classcutlass_1_1reduction_1_1kernel_1_1ReduceSplitK.html#af788ae48c72021b8ce49da15dfa72be3">cutlass::reduction::kernel::ReduceSplitK::block_shape</a></div><div class="ttdeci">static CUTLASS_HOST_DEVICE dim3 block_shape()</div><div class="ttdoc">Determines the threadblock shape. </div><div class="ttdef"><b>Definition:</b> reduce_split_k.h:138</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_adbe8a410fe634ab05b8cf69356b79b26"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#adbe8a410fe634ab05b8cf69356b79b26">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::ElementC</a></div><div class="ttdeci">ElementC_ ElementC</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:460</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_a8b10e75e5d6cd348dacc085f5264ee95"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#a8b10e75e5d6cd348dacc085f5264ee95">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::ref_A</a></div><div class="ttdeci">TensorRef< ElementA const, LayoutA > ref_A</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:510</div></div>
|
||
|
|
<div class="ttc" id="reduction__operators_8h_html"><div class="ttname"><a href="reduction__operators_8h.html">reduction_operators.h</a></div><div class="ttdoc">Kernel performing a reduction over densely packed tensors in global memory. </div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1GemmCoord_html"><div class="ttname"><a href="structcutlass_1_1gemm_1_1GemmCoord.html">cutlass::gemm::GemmCoord</a></div><div class="ttdef"><b>Definition:</b> include/cutlass/gemm/gemm.h:94</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_a04818e67f94c5440ac6c367798e17fc2"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#a04818e67f94c5440ac6c367798e17fc2">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::epilogue</a></div><div class="ttdeci">EpilogueOutputOp::Params epilogue</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:514</div></div>
|
||
|
|
<div class="ttc" id="conversion__op_8h_html"><div class="ttname"><a href="conversion__op_8h.html">conversion_op.h</a></div><div class="ttdoc">Functor performing conversion operations used by epilogues. </div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_a3f1ad816609ea90bb3aed753dad546ca"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#a3f1ad816609ea90bb3aed753dad546ca">cutlass::gemm::device::GemmSplitKParallel::Arguments::split_k_slices</a></div><div class="ttdeci">int split_k_slices</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:191</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_af6008a17370eec923078a23564972383"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#af6008a17370eec923078a23564972383">cutlass::gemm::device::GemmSplitKParallel::Arguments::reduction</a></div><div class="ttdeci">ReductionOp::Params reduction</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:193</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1reduction_1_1thread_1_1ReduceAdd_html"><div class="ttname"><a href="structcutlass_1_1reduction_1_1thread_1_1ReduceAdd.html">cutlass::reduction::thread::ReduceAdd</a></div><div class="ttdoc">Mixed-precision reduction. </div><div class="ttdef"><b>Definition:</b> reduction_operators.h:50</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a1de7cf5d8bad27b3ff6c803dbc572077"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a1de7cf5d8bad27b3ff6c803dbc572077">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::run</a></div><div class="ttdeci">Status run(cudaStream_t stream=nullptr)</div><div class="ttdoc">Runs the kernel using initialized state. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:604</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a44facae3996ed3da5fdb4398e469b773"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a44facae3996ed3da5fdb4398e469b773">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::update</a></div><div class="ttdeci">Status update(Arguments const &args, void *workspace=nullptr)</div><div class="ttdoc">Lightweight update given a subset of arguments. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:598</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1reduction_1_1kernel_1_1ReduceSplitK_1_1Params_html"><div class="ttname"><a href="structcutlass_1_1reduction_1_1kernel_1_1ReduceSplitK_1_1Params.html">cutlass::reduction::kernel::ReduceSplitK::Params</a></div><div class="ttdoc">Params structure. </div><div class="ttdef"><b>Definition:</b> reduce_split_k.h:80</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_ab2e2468a859f14502fd18013859ec9e6"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#ab2e2468a859f14502fd18013859ec9e6">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::InstructionShape</a></div><div class="ttdeci">InstructionShape InstructionShape</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:137</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_ac5d77730905dfaaa58056ff6d78bf876"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#ac5d77730905dfaaa58056ff6d78bf876">cutlass::gemm::device::GemmSplitKParallel::Arguments::Arguments</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Arguments()</div><div class="ttdoc">Default ctor. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:201</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1GemmCoord_html_a18835ec84cbb6250143327e93697c7e9"><div class="ttname"><a href="structcutlass_1_1gemm_1_1GemmCoord.html#a18835ec84cbb6250143327e93697c7e9">cutlass::gemm::GemmCoord::k</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Index const & k() const </div><div class="ttdoc">Returns the GEMM K coordinate. </div><div class="ttdef"><b>Definition:</b> include/cutlass/gemm/gemm.h:145</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_aabbedd03888a6090f049f53f53bf4e45"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#aabbedd03888a6090f049f53f53bf4e45">cutlass::gemm::device::GemmSplitKParallel::Arguments::convert</a></div><div class="ttdeci">ConvertScaledOp::Params convert</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:192</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1layout_1_1ColumnMajor_html"><div class="ttname"><a href="classcutlass_1_1layout_1_1ColumnMajor.html">cutlass::layout::ColumnMajor</a></div><div class="ttdoc">Mapping function for column-major matrices. </div><div class="ttdef"><b>Definition:</b> layout/matrix.h:142</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_ad0c614a548bcade989eb25633b45bb0f"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#ad0c614a548bcade989eb25633b45bb0f">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::GemmSplitKParallel</a></div><div class="ttdeci">GemmSplitKParallel()</div><div class="ttdoc">Constructs the GEMM. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:562</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a4026bdedc7cf880b58acbdfa716f7f28"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a4026bdedc7cf880b58acbdfa716f7f28">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::ElementC</a></div><div class="ttdeci">ElementC ElementC</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:130</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_ad174d9818e8a4857b65e5da6da7a45b3"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#ad174d9818e8a4857b65e5da6da7a45b3">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::ThreadblockShape</a></div><div class="ttdeci">ThreadblockShape ThreadblockShape</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:135</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_a9a22df7c4d515a48e03fd6f16e074217"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#a9a22df7c4d515a48e03fd6f16e074217">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::ref_B</a></div><div class="ttdeci">TensorRef< ElementB const, LayoutB > ref_B</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:511</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a2b8223379b65f5117d4d8dfd4f6cbc13"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a2b8223379b65f5117d4d8dfd4f6cbc13">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::EpilogueOutputOp</a></div><div class="ttdeci">EpilogueOutputOp EpilogueOutputOp</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:139</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a4d9c70e23eef0a15d849b5b0ebadfcdd"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a4d9c70e23eef0a15d849b5b0ebadfcdd">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::get_workspace_size</a></div><div class="ttdeci">static size_t get_workspace_size(Arguments const &args)</div><div class="ttdoc">Gets the workspace size. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:586</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a2a48eb6e51e2315e945882d5e70ebb2f"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a2a48eb6e51e2315e945882d5e70ebb2f">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::ArchTag</a></div><div class="ttdeci">ArchTag ArchTag</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:134</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_ad6d811ca346ce6467a291497edc85623"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#ad6d811ca346ce6467a291497edc85623">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::operator()</a></div><div class="ttdeci">Status operator()(Arguments const &args, void *workspace=nullptr, cudaStream_t stream=nullptr)</div><div class="ttdoc">Runs the kernel using initialized state. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:615</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a72f5de19ad97e08241157d5106f2f66a"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a72f5de19ad97e08241157d5106f2f66a">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::operator()</a></div><div class="ttdeci">Status operator()(cudaStream_t stream=nullptr)</div><div class="ttdoc">Runs the kernel using initialized state. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:610</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a114b122602b425909f9be0df461353a4"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a114b122602b425909f9be0df461353a4">cutlass::gemm::device::GemmSplitKParallel::can_implement</a></div><div class="ttdeci">static Status can_implement(Arguments const &args)</div><div class="ttdoc">Determines whether the GEMM can execute the given problem. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:244</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_a1550cfad6be8e22c9dc13b05f5601845"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#a1550cfad6be8e22c9dc13b05f5601845">cutlass::gemm::device::GemmSplitKParallel::Arguments::ref_D</a></div><div class="ttdeci">TensorRef< ElementC, LayoutC > ref_D</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:189</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_abfb1166a9c55270ff8f1b265516a418c"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#abfb1166a9c55270ff8f1b265516a418c">cutlass::gemm::device::GemmSplitKParallel::GemmSplitKParallel</a></div><div class="ttdeci">GemmSplitKParallel()</div><div class="ttdoc">Constructs the GEMM. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:241</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1layout_1_1LayoutTranspose_html"><div class="ttname"><a href="structcutlass_1_1layout_1_1LayoutTranspose.html">cutlass::layout::LayoutTranspose</a></div><div class="ttdoc">Defines transposes of matrix layouts. </div><div class="ttdef"><b>Definition:</b> layout/matrix.h:921</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_a24cbcd47c87175248ed1c55a9a0e5426"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#a24cbcd47c87175248ed1c55a9a0e5426">cutlass::gemm::device::GemmSplitKParallel::Arguments::problem_size</a></div><div class="ttdeci">GemmCoord problem_size</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:185</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1TensorRef_html"><div class="ttname"><a href="classcutlass_1_1TensorRef.html">cutlass::TensorRef< ElementA const, LayoutA ></a></div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a2d8d3a504dd8807ed09e25f37a658783"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a2d8d3a504dd8807ed09e25f37a658783">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::ReductionOp</a></div><div class="ttdeci">ReductionOp_ ReductionOp</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:470</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a4a100b8064c285bdb06fe3e523a07d1c"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a4a100b8064c285bdb06fe3e523a07d1c">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::GemmKernel</a></div><div class="ttdeci">typename kernel::DefaultGemmSplitKParallel< ElementA, LayoutA, kAlignmentA, ElementB, LayoutB, kAlignmentB, ElementAccumulator, LayoutC, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, ConvertScaledOp, ThreadblockSwizzle, kStages, Operator >::GemmKernel GemmKernel</div><div class="ttdoc">GEMM kernel. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:165</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_a63048fa3419753d96a60eaee28f6cfe4"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#a63048fa3419753d96a60eaee28f6cfe4">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::reduction</a></div><div class="ttdeci">ReductionOp::Params reduction</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:517</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_ae59bd456dc03a51e64a655367382535d"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#ae59bd456dc03a51e64a655367382535d">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::LayoutA</a></div><div class="ttdeci">typename layout::LayoutTranspose< LayoutB >::type LayoutA</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:127</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a5d934b66a5ead9df63bea323abf8f83f"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a5d934b66a5ead9df63bea323abf8f83f">cutlass::gemm::device::GemmSplitKParallel::operator()</a></div><div class="ttdeci">Status operator()(cudaStream_t stream=nullptr)</div><div class="ttdoc">Runs the kernel using initialized state. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:390</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_ac82ba3da12b03bc91586a3947ce99fc5"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#ac82ba3da12b03bc91586a3947ce99fc5">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::ReductionOp</a></div><div class="ttdeci">ReductionOp ReductionOp</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:140</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a5a1c15eb1246208cf00bd811b8df85ba"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a5a1c15eb1246208cf00bd811b8df85ba">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::ElementB</a></div><div class="ttdeci">ElementA ElementB</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:128</div></div>
|
||
|
|
<div class="ttc" id="namespacecutlass_html_ac5a88c5840a28a9e0206b9cc7812a18da4d3b5847a0b17037c6b69faf8b1d4d71"><div class="ttname"><a href="namespacecutlass.html#ac5a88c5840a28a9e0206b9cc7812a18da4d3b5847a0b17037c6b69faf8b1d4d71">cutlass::Status::kErrorInternal</a></div><div class="ttdoc">An error within CUTLASS occurred. </div></div>
|
||
|
|
<div class="ttc" id="device__kernel_8h_html"><div class="ttname"><a href="device__kernel_8h.html">device_kernel.h</a></div><div class="ttdoc">Template for generic CUTLASS kernel. </div></div>
|
||
|
|
<div class="ttc" id="reduce__split__k_8h_html"><div class="ttname"><a href="reduce__split__k_8h.html">reduce_split_k.h</a></div><div class="ttdoc">Kernel performing a reduction over densely packed tensors in global memory. </div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_aaaf935dfdc267b089de7b304be979d4d"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#aaaf935dfdc267b089de7b304be979d4d">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::UnderlyingArguments</a></div><div class="ttdeci">typename UnderlyingOperator::Arguments UnderlyingArguments</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:498</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a696cd49441ddb490d32a374135731c68"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a696cd49441ddb490d32a374135731c68">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Operator</a></div><div class="ttdeci">Operator_ Operator</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:472</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_a48ced96adaf371f03c1c9a50db9f50f2"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#a48ced96adaf371f03c1c9a50db9f50f2">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::convert</a></div><div class="ttdeci">ConvertScaledOp::Params convert</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:516</div></div>
|
||
|
|
<div class="ttc" id="cutlass_8h_html_a28c2443a142676d3d71effdae1a986b1"><div class="ttname"><a href="cutlass_8h.html#a28c2443a142676d3d71effdae1a986b1">CUTLASS_HOST_DEVICE</a></div><div class="ttdeci">#define CUTLASS_HOST_DEVICE</div><div class="ttdef"><b>Definition:</b> cutlass.h:89</div></div>
|
||
|
|
<div class="ttc" id="numeric__types_8h_html"><div class="ttname"><a href="numeric__types_8h.html">numeric_types.h</a></div><div class="ttdoc">Top-level include for all CUTLASS numeric types. </div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1reduction_1_1kernel_1_1ReduceSplitK_html"><div class="ttname"><a href="classcutlass_1_1reduction_1_1kernel_1_1ReduceSplitK.html">cutlass::reduction::kernel::ReduceSplitK</a></div><div class="ttdef"><b>Definition:</b> reduce_split_k.h:55</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1reduction_1_1kernel_1_1ReduceSplitK_html_a696cecdb049ddb78b3d40530abbba1fb"><div class="ttname"><a href="classcutlass_1_1reduction_1_1kernel_1_1ReduceSplitK.html#a696cecdb049ddb78b3d40530abbba1fb">cutlass::reduction::kernel::ReduceSplitK::grid_shape</a></div><div class="ttdeci">static CUTLASS_HOST_DEVICE dim3 grid_shape(cutlass::MatrixCoord problem_size)</div><div class="ttdoc">Computes the grid size given a chosen threadblock shape. </div><div class="ttdef"><b>Definition:</b> reduce_split_k.h:128</div></div>
|
||
|
|
<div class="ttc" id="default__gemm__configuration_8h_html"><div class="ttname"><a href="default__gemm__configuration_8h.html">default_gemm_configuration.h</a></div><div class="ttdoc">Definitions for GEMM structures. </div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1TensorRef_html_a66a9bab939e2d57c130bb76a5527f482"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a66a9bab939e2d57c130bb76a5527f482">cutlass::TensorRef::non_const_ref</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE NonConstTensorRef non_const_ref() const </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:229</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_ae8fb82c40078cf84c211f10f726caaf5"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#ae8fb82c40078cf84c211f10f726caaf5">cutlass::gemm::device::GemmSplitKParallel::get_workspace_size</a></div><div class="ttdeci">static size_t get_workspace_size(Arguments const &args)</div><div class="ttdoc">Gets the workspace size. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:252</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html">cutlass::gemm::device::GemmSplitKParallel</a></div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:123</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1layout_1_1RowMajor_html"><div class="ttname"><a href="classcutlass_1_1layout_1_1RowMajor.html">cutlass::layout::RowMajor</a></div><div class="ttdoc">Mapping function for row-major matrices. </div><div class="ttdef"><b>Definition:</b> layout/matrix.h:50</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_aa111536496495825c5d2ac2ea7360998"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#aa111536496495825c5d2ac2ea7360998">cutlass::gemm::device::GemmSplitKParallel::Arguments::ref_C</a></div><div class="ttdeci">TensorRef< ElementC const, LayoutC > ref_C</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:188</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a4d9f086305f76d7f885bf032f3d2c7c9"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a4d9f086305f76d7f885bf032f3d2c7c9">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::initialize</a></div><div class="ttdeci">Status initialize(Arguments const &args, void *workspace)</div><div class="ttdoc">Initializes GEMM state from arguments. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:592</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_affb7a5c96c9e8b04eb94a464e5fdc48b"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#affb7a5c96c9e8b04eb94a464e5fdc48b">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::ConvertScaledOp</a></div><div class="ttdeci">ConvertScaledOp ConvertScaledOp</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:138</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a465591fbfde2a9aa6330d9adcbf82bd6"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a465591fbfde2a9aa6330d9adcbf82bd6">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::can_implement</a></div><div class="ttdeci">static Status can_implement(Arguments const &args)</div><div class="ttdoc">Determines whether the GEMM can execute the given problem. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:580</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_a0224c595b49468f336b40b865b04e3df"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#a0224c595b49468f336b40b865b04e3df">cutlass::gemm::device::GemmSplitKParallel::Arguments::Arguments</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Arguments(GemmCoord problem_size_, TensorRef< ElementA const, LayoutA > ref_A_, TensorRef< ElementB const, LayoutB > ref_B_, TensorRef< ElementC const, LayoutC > ref_C_, TensorRef< ElementC, LayoutC > ref_D_, typename EpilogueOutputOp::Params epilogue_=typename EpilogueOutputOp::Params(), int split_k_slices=1, typename ConvertScaledOp::Params convert_=typename ConvertScaledOp::Params(), typename ReductionOp::Params reduction_=typename ReductionOp::Params())</div><div class="ttdoc">Constructs an Arguments structure. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:205</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a56649fb86263ffd12aae9f7ae2e6506e"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a56649fb86263ffd12aae9f7ae2e6506e">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::ElementAccumulator</a></div><div class="ttdeci">ElementAccumulator ElementAccumulator</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:132</div></div>
|
||
|
|
<div class="ttc" id="namespacecutlass_html_ac5a88c5840a28a9e0206b9cc7812a18da6467125354474c5cff2f3c920abb7d3f"><div class="ttname"><a href="namespacecutlass.html#ac5a88c5840a28a9e0206b9cc7812a18da6467125354474c5cff2f3c920abb7d3f">cutlass::Status::kErrorWorkspaceNull</a></div><div class="ttdoc">The given workspace is null when it is required to be non-null. </div></div>
|
||
|
|
<div class="ttc" id="namespacecutlass_html_ac5a88c5840a28a9e0206b9cc7812a18da8c632159fa131f09d04f94e3cbcd8782"><div class="ttname"><a href="namespacecutlass.html#ac5a88c5840a28a9e0206b9cc7812a18da8c632159fa131f09d04f94e3cbcd8782">cutlass::Status::kSuccess</a></div><div class="ttdoc">Operation was successful. </div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_ac37c811b2f58c7756ef720fdfc0d072b"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#ac37c811b2f58c7756ef720fdfc0d072b">cutlass::gemm::device::GemmSplitKParallel::Arguments::ref_B</a></div><div class="ttdeci">TensorRef< ElementB const, LayoutB > ref_B</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:187</div></div>
|
||
|
|
<div class="ttc" id="gemm_2threadblock_2threadblock__swizzle_8h_html"><div class="ttname"><a href="gemm_2threadblock_2threadblock__swizzle_8h.html">threadblock_swizzle.h</a></div><div class="ttdoc">Implements several possible threadblock-swizzling functions mapping blockIdx to GEMM problems...</div></div>
|
||
|
|
<div class="ttc" id="arch_8h_html"><div class="ttname"><a href="arch_8h.html">arch.h</a></div><div class="ttdoc">Defines tags for architecture-specific configurations. </div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_aacc6cfe6ebe0b33ec7577c654303f70d"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#aacc6cfe6ebe0b33ec7577c654303f70d">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::ThreadblockSwizzle</a></div><div class="ttdeci">ThreadblockSwizzle ThreadblockSwizzle</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:141</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_aff78ac3c99bb15cf8a7d7a1ece736cd1"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#aff78ac3c99bb15cf8a7d7a1ece736cd1">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::split_k_slices</a></div><div class="ttdeci">int split_k_slices</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:515</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a07ad725857d7eb191cbfc135df22b781"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a07ad725857d7eb191cbfc135df22b781">cutlass::gemm::device::GemmSplitKParallel::update</a></div><div class="ttdeci">Status update(Arguments const &args, void *workspace=nullptr)</div><div class="ttdoc">Lightweight update given a subset of arguments. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:312</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_aeadc76210d06ec22776aca4a58de9930"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#aeadc76210d06ec22776aca4a58de9930">cutlass::gemm::device::GemmSplitKParallel::run</a></div><div class="ttdeci">Status run(cudaStream_t stream=nullptr)</div><div class="ttdoc">Runs the kernel using initialized state. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:329</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_a92fb61dbfc8ca622f852cc1673fa1066"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#a92fb61dbfc8ca622f852cc1673fa1066">cutlass::gemm::device::GemmSplitKParallel::Arguments::epilogue</a></div><div class="ttdeci">EpilogueOutputOp::Params epilogue</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:190</div></div>
|
||
|
|
<div class="ttc" id="default__gemm__splitk__parallel_8h_html"><div class="ttname"><a href="default__gemm__splitk__parallel_8h.html">default_gemm_splitk_parallel.h</a></div><div class="ttdoc">Default kernel-level GEMM definitions combine threadblock-scoped matrix multiply-add with the appropr...</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html">cutlass::gemm::device::GemmSplitKParallel::Arguments</a></div><div class="ttdoc">Argument structure. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:179</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_ab2572b47bae5f0ae0f06647cfe86ad4e"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#ab2572b47bae5f0ae0f06647cfe86ad4e">cutlass::gemm::device::GemmSplitKParallel::LayoutC</a></div><div class="ttdeci">LayoutC_ LayoutC</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:131</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_a37c45d8dc800de6a631b8a096704559a"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#a37c45d8dc800de6a631b8a096704559a">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::Arguments</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Arguments(GemmCoord problem_size_, TensorRef< ElementA const, LayoutA > ref_A_, TensorRef< ElementB const, LayoutB > ref_B_, TensorRef< ElementC const, LayoutC > ref_C_, TensorRef< ElementC, LayoutC > ref_D_, typename EpilogueOutputOp::Params epilogue_=typename EpilogueOutputOp::Params(), int split_k_slices=1, typename ConvertScaledOp::Params convert_=typename ConvertScaledOp::Params(), typename ReductionOp::Params reduction_=typename ReductionOp::Params())</div><div class="ttdoc">Constructs an Arguments structure. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:529</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a7085b7cf85bc1bcd202ea6928656d966"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a7085b7cf85bc1bcd202ea6928656d966">cutlass::gemm::device::GemmSplitKParallel::initialize</a></div><div class="ttdeci">Status initialize(Arguments const &args, void *workspace)</div><div class="ttdoc">Initializes GEMM state from arguments. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:266</div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments_html_a3f4d6497cd54e624b3e410996a8b7d10"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_1_1Arguments.html#a3f4d6497cd54e624b3e410996a8b7d10">cutlass::gemm::device::GemmSplitKParallel::Arguments::ref_A</a></div><div class="ttdeci">TensorRef< ElementA const, LayoutA > ref_A</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:186</div></div>
|
||
|
|
<div class="ttc" id="cutlass_8h_html"><div class="ttname"><a href="cutlass_8h.html">cutlass.h</a></div><div class="ttdoc">Basic include for CUTLASS. </div></div>
|
||
|
|
<div class="ttc" id="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b_html_acf6c5b216c0c82f0c7797627d651743f"><div class="ttname"><a href="structcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01Elementafcb1aeaf2035a7ac769d7acc233423b.html#acf6c5b216c0c82f0c7797627d651743f">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::Arguments::Arguments</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Arguments()</div><div class="ttdoc">Default ctor. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:525</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457_html_a535863339cab9879474e31f2fd543804"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_3_01ElementA___00_01LayoutA___00_01ElementBbe7c1f7154ad5b5bf9d4d28301e2b457.html#a535863339cab9879474e31f2fd543804">cutlass::gemm::device::GemmSplitKParallel< ElementA_, LayoutA_, ElementB_, LayoutB_, ElementC_, layout::ColumnMajor, ElementAccumulator_, OperatorClass_, ArchTag_, ThreadblockShape_, WarpShape_, InstructionShape_, EpilogueOutputOp_, ConvertScaledOp_, ReductionOp_, ThreadblockSwizzle_, Stages, kAlignmentA, kAlignmentB, Operator_ >::to_underlying_arguments</a></div><div class="ttdeci">static UnderlyingArguments to_underlying_arguments(Arguments const &args)</div><div class="ttdoc">Helper to construct a transposed equivalent for the underying GEMM operator. </div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:565</div></div>
|
||
|
|
<div class="ttc" id="namespacecutlass_html_ac5a88c5840a28a9e0206b9cc7812a18d"><div class="ttname"><a href="namespacecutlass.html#ac5a88c5840a28a9e0206b9cc7812a18d">cutlass::Status</a></div><div class="ttdeci">Status</div><div class="ttdoc">Status code returned by CUTLASS operations. </div><div class="ttdef"><b>Definition:</b> cutlass.h:39</div></div>
|
||
|
|
<div class="ttc" id="include_2cutlass_2gemm_2kernel_2gemm_8h_html"><div class="ttname"><a href="include_2cutlass_2gemm_2kernel_2gemm_8h.html">gemm.h</a></div><div class="ttdoc">Template for a pipelined GEMM kernel. Does not compute batching or support split-K. </div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a372091a0c8d52bb9aab75c274039ceb7"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a372091a0c8d52bb9aab75c274039ceb7">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::ElementA</a></div><div class="ttdeci">ElementB ElementA</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:126</div></div>
|
||
|
|
<div class="ttc" id="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel_html_a1bc1686f83acac48e8a1695b71518b16"><div class="ttname"><a href="classcutlass_1_1gemm_1_1device_1_1GemmSplitKParallel.html#a1bc1686f83acac48e8a1695b71518b16">cutlass::gemm::device::GemmSplitKParallel< ElementB, typename layout::LayoutTranspose< LayoutB >::type, ElementA, typename layout::LayoutTranspose< LayoutA >::type, ElementC, layout::RowMajor, ElementAccumulator, OperatorClass, ArchTag, ThreadblockShape, WarpShape, InstructionShape, EpilogueOutputOp, ConvertScaledOp, ReductionOp, ThreadblockSwizzle, Stages, kAlignmentA, kAlignmentB, Operator >::OperatorClass</a></div><div class="ttdeci">OperatorClass OperatorClass</div><div class="ttdef"><b>Definition:</b> device/gemm_splitk_parallel.h:133</div></div>
|
||
|
|
</div><!-- fragment --></div><!-- contents -->
|
||
|
|
<!-- start footer part -->
|
||
|
|
<hr class="footer"/><address class="footer"><small>
|
||
|
|
Generated by  <a href="http://www.doxygen.org/index.html">
|
||
|
|
<img class="footer" src="doxygen.png" alt="doxygen"/>
|
||
|
|
</a> 1.8.11
|
||
|
|
</small></address>
|
||
|
|
</body>
|
||
|
|
</html>
|