squall/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity
06eb90cc0d
cutlass/docs/tensor__ref_8h_source.html

171 lines
88 KiB
HTML
Raw Normal View History

CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
<!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"/>
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<meta name="generator" content="Doxygen 1.8.11"/>
<title>CUTLASS: tensor_ref.h Source File</title>
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
<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>
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<script type="text/javascript">
$(document).ready(function() { init_search(); });
</script>
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
<script type="text/x-mathjax-config">
MathJax.Hub.Config({
extensions: ["tex2jax.js"],
jax: ["input/TeX","output/HTML-CSS"],
});
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
</script><script type="text/javascript" src="http://cdn.mathjax.org/mathjax/latest/MathJax.js"></script>
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
<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;">
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<td id="projectlogo"><img alt="Logo" src="cutlass-logo-small.png"/></td>
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
<td id="projectalign" style="padding-left: 0.5em;">
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<div id="projectname">CUTLASS
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
</div>
<div id="projectbrief">CUDA Templates for Linear Algebra Subroutines and Solvers</div>
</td>
</tr>
</tbody>
</table>
</div>
<!-- end header part -->
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<!-- Generated by Doxygen 1.8.11 -->
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
<script type="text/javascript">
var searchBox = new SearchBox("searchBox", "search",false,'Search');
</script>
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<div id="navrow1" class="tabs">
<ul class="tablist">
<li><a href="index.html"><span>Main&#160;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&#160;List</span></a></li>
<li><a href="globals.html"><span>File&#160;Members</span></a></li>
</ul>
</div>
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
<!-- 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>
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<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> </ul>
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
</div>
</div><!-- top -->
<div class="header">
<div class="headertitle">
<div class="title">tensor_ref.h</div> </div>
</div><!--header-->
<div class="contents">
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<a href="tensor__ref_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>&#160;<span class="comment">/***************************************************************************************************</span></div><div class="line"><a name="l00002"></a><span class="lineno"> 2</span>&#160;<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>&#160;<span class="comment"> *</span></div><div class="line"><a name="l00004"></a><span class="lineno"> 4</span>&#160;<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>&#160;<span class="comment"> * provided that the following conditions are met:</span></div><div class="line"><a name="l00006"></a><span class="lineno"> 6</span>&#160;<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>&#160;<span class="comment"> * conditions and the following disclaimer.</span></div><div class="line"><a name="l00008"></a><span class="lineno"> 8</span>&#160;<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>&#160;<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>&#160;<span class="comment"> * provided with the distribution.</span></div><div class="line"><a name="l00011"></a><span class="lineno"> 11</span>&#160;<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>&#160;<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>&#160;<span class="comment"> * permission.</span></div><div class="line"><a name="l00014"></a><span class="lineno"> 14</span>&#160;<span class="comment"> *</span></div><div class="line"><a name="l00015"></a><span class="lineno"> 15</span>&#160;<span class="comment"> * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS &quot;AS IS&quot; AND ANY EXPRESS OR</span></div><div class="line"><a name="l00016"></a><span class="lineno"> 16</span>&#160;<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>&#160;<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>&#160;<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>&#160;<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>&#160;<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>&#160;<span class="comment"> * STRICT LIABILITY, OR TOR (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE</span></div><div class="line"><a name="l00022">
<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="classcutlass_1_1IdentityTensorLayout_html_a471e797514da29a5cad6c61fffe9eb5c"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html#a471e797514da29a5cad6c61fffe9eb5c">cutlass::IdentityTensorLayout::operator()</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE LongIndex operator()(Coord&lt; Rank &gt; const &amp;coord) const </div><div class="ttdoc">Returns the offset of a coordinate in linear memory. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:85</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a300283c640d4e6aadc9c695befa26fec"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a300283c640d4e6aadc9c695befa26fec">cutlass::TensorRef&lt; Element, Layout &gt;::stride</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Index stride(int dim) const</div><div class="ttdoc">Returns the layout object&amp;#39;s stride in a given physical dimension. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:289</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a7fb24829405f63b9fa9bbcb110141d9e"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a7fb24829405f63b9fa9bbcb110141d9e">cutlass::TensorRef&lt; Element, Layout &gt;&lt; Element, Layout &gt;::Stride</a></div><div class="ttdeci">typename Layout::Stride Stride</div><div class="ttdoc">Layout&amp;#39;s stride vector. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:174</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a425adc5418cc80c9929579046d3111ef"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a425adc5418cc80c9929579046d3111ef">cutlass::TensorRef&lt; Element, Layout &gt;::stride</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Index &amp; stride(int dim)</div><div class="ttdoc">Returns the layout object&amp;#39;s stride in a given physical dimension. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:295</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a5fbf003d2e2321f816be60b0cbd0cfb7"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a5fbf003d2e2321f816be60b0cbd0cfb7">cutlass::TensorRef&lt; Element, Layout &gt;::stride</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Stride &amp; stride()</div><div class="ttdoc">Returns the layout object&amp;#39;s stride vector. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:283</div></div>
<div class="ttc" id="structcutlass_1_1platform_1_1remove__const_html_ac3662947fa50251daf58240a9c798085"><div class="ttname"><a href="structcutlass_1_1platform_1_1remove__const.html#ac3662947fa50251daf58240a9c798085">cutlass::platform::remove_const::type</a></div><div class="ttdeci">T type</div><div class="ttdef"><b>Definition:</b> platform.h:351</div></div>
<div class="ttc" id="classcutlass_1_1IdentityTensorLayout_html"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html">cutlass::IdentityTensorLayout</a></div><div class="ttdef"><b>Definition:</b> tensor_ref.h:45</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_ac7db3ca62ab1dfe0d3ea08bcadbc9352"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#ac7db3ca62ab1dfe0d3ea08bcadbc9352">cutlass::TensorRef&lt; Element, Layout &gt;::data</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Element * data() const</div><div class="ttdoc">Returns the pointer to referenced data. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:254</div></div>
<div class="ttc" id="classcutlass_1_1IdentityTensorLayout_html_a5e1b58137ca0996e3fa0f727a1d85761"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html#a5e1b58137ca0996e3fa0f727a1d85761">cutlass::IdentityTensorLayout::Stride</a></div><div class="ttdeci">Coord&lt; kStrideRank, Index &gt; Stride</div><div class="ttdoc">Stride vector. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:63</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a5c5c66a59e9759f11f832fb71f4234c2"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a5c5c66a59e9759f11f832fb71f4234c2">cutlass::TensorRef&lt; Element, Layout &gt;::const_ref</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE ConstTensorRef const_ref() const</div><div class="ttdoc">Returns a reference to constant-valued tensor. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:224</div></div>
<div class="ttc" id="classcutlass_1_1IdentityTensorLayout_html_ae64fd86033500257785f6923da2558c0"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html#ae64fd86033500257785f6923da2558c0">cutlass::IdentityTensorLayout::Index</a></div><div class="ttdeci">int32_t Index</div><div class="ttdoc">Index type used for coordinates. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:54</div></div>
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
<div class="ttc" id="coord_8h_html"><div class="ttname"><a href="coord_8h.html">coord.h</a></div><div class="ttdoc">A Coord is a coordinate of arbitrary rank into a tensor or matrix. </div></div>
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<div class="ttc" id="classcutlass_1_1TensorRef_html_ae0e372b28e665820e6a2d17fc9f68d2b"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#ae0e372b28e665820e6a2d17fc9f68d2b">cutlass::TensorRef&lt; Element, Layout &gt;&lt; Element, Layout &gt;::Layout</a></div><div class="ttdeci">Layout Layout</div><div class="ttdoc">Mapping function from logical coordinate to linear memory. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:152</div></div>
<div class="ttc" id="classcutlass_1_1IdentityTensorLayout_html_a041b285984b1940d70a2b7768416e996"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html#a041b285984b1940d70a2b7768416e996">cutlass::IdentityTensorLayout::stride</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Stride &amp; stride()</div><div class="ttdoc">Returns the stride of the layout. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:97</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a1c539841446205c482fe18cc0a99d913"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a1c539841446205c482fe18cc0a99d913">cutlass::TensorRef&lt; Element, Layout &gt;::reset</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE void reset(Element *ptr, Layout const &amp;layout)</div><div class="ttdoc">Updates the pointer and layout object. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:241</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_ac5f24cda8149ddfa8f19467ff2629f49"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#ac5f24cda8149ddfa8f19467ff2629f49">cutlass::TensorRef&lt; Element, Layout &gt;::operator-=</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE TensorRef &amp; operator-=(TensorCoord const &amp;b)</div><div class="ttdoc">Returns a TensorRef offset by a given amount. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:356</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a10e36681cae687d9a45ec01bad2626b7"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a10e36681cae687d9a45ec01bad2626b7">cutlass::TensorRef&lt; Element, Layout &gt;::operator[]</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Reference operator[](TensorCoord const &amp;coord) const</div><div class="ttdoc">Returns a reference to the element at a given Coord. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:313</div></div>
<div class="ttc" id="platform_8h_html"><div class="ttname"><a href="platform_8h.html">platform.h</a></div><div class="ttdoc">C++ features that may be otherwise unimplemented for CUDA device functions. </div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a26c607b811a8f828f5ec5b00f3b874ea"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a26c607b811a8f828f5ec5b00f3b874ea">cutlass::TensorRef&lt; Element, Layout &gt;::TensorRef</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE TensorRef(Element *ptr=nullptr, Layout const &amp;layout=Layout())</div><div class="ttdoc">Constructs a TensorRef with a pointer and layout object. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:207</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_aca7d373c988016ee946f85ea958b21a7"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#aca7d373c988016ee946f85ea958b21a7">cutlass::TensorRef&lt; Element, Layout &gt;::layout</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Layout layout() const</div><div class="ttdoc">Returns the layout object. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:271</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a4bed879c428963070de8ffbdc5d6e4f9"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a4bed879c428963070de8ffbdc5d6e4f9">cutlass::TensorRef&lt; Element, Layout &gt;::add_coord_offset</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE TensorRef &amp; add_coord_offset(TensorCoord const &amp;coord)</div><div class="ttdoc">Adds an offset to each pointer. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:326</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a6c1be7001a3fb05f8ac2d1fe7ea94c68"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a6c1be7001a3fb05f8ac2d1fe7ea94c68">cutlass::TensorRef&lt; Element, Layout &gt;&lt; Element, Layout &gt;::Element</a></div><div class="ttdeci">Element Element</div><div class="ttdoc">Data type of individual access. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:149</div></div>
<div class="ttc" id="cutlass_8h_html_a4b1c9f25ab6eaa25e1f2258dd63e6ce4"><div class="ttname"><a href="cutlass_8h.html#a4b1c9f25ab6eaa25e1f2258dd63e6ce4">CUTLASS_PRAGMA_UNROLL</a></div><div class="ttdeci">#define CUTLASS_PRAGMA_UNROLL</div><div class="ttdef"><b>Definition:</b> cutlass.h:110</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a965e8b3b7f92dc51d4d3821ea6a25012"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a965e8b3b7f92dc51d4d3821ea6a25012">cutlass::TensorRef&lt; Element, Layout &gt;::data</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Reference data(LongIndex idx) const</div><div class="ttdoc">Returns a reference to the element at a given linear index. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:258</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_aa275fd660406665884f25c6f693a8892"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#aa275fd660406665884f25c6f693a8892">cutlass::TensorRef&lt; Element, Layout &gt;::operator-</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE TensorRef operator-(TensorCoord const &amp;b) const</div><div class="ttdoc">Returns a TensorRef offset by a given amount. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:348</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a2b79f0d7c0c70774199ba4226bae0fe0"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a2b79f0d7c0c70774199ba4226bae0fe0">cutlass::TensorRef&lt; Element, Layout &gt;::TensorRef</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE TensorRef(NonConstTensorRef const &amp;ref)</div><div class="ttdoc">Converting constructor from TensorRef to non-constant data. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:217</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a191e88bc0fb310be655d700e937ab97c"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a191e88bc0fb310be655d700e937ab97c">cutlass::TensorRef&lt; Element, Layout &gt;::stride</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Stride stride() const</div><div class="ttdoc">Returns the layout object&amp;#39;s stride vector. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:277</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_ace218cdb46555a46bd71dbdfc2c317c1"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#ace218cdb46555a46bd71dbdfc2c317c1">cutlass::TensorRef&lt; Element, Layout &gt;&lt; Element, Layout &gt;::TensorCoord</a></div><div class="ttdeci">typename Layout::TensorCoord TensorCoord</div><div class="ttdoc">Coordinate in logical tensor space. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:171</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_ac968ea9cb34fa99d29c64608c53bd4d4"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#ac968ea9cb34fa99d29c64608c53bd4d4">cutlass::TensorRef&lt; Element, Layout &gt;::good</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE bool good() const</div><div class="ttdoc">Returns true if the TensorRef is non-null. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:248</div></div>
<div class="ttc" id="structcutlass_1_1sizeof__bits_html"><div class="ttname"><a href="structcutlass_1_1sizeof__bits.html">cutlass::sizeof_bits</a></div><div class="ttdoc">Defines the size of an element in bits. </div><div class="ttdef"><b>Definition:</b> numeric_types.h:42</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a9c2149162016bc19c7735b824d57eb9e"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a9c2149162016bc19c7735b824d57eb9e">cutlass::TensorRef&lt; Element, Layout &gt;::reset</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE void reset(Element *ptr=nullptr)</div><div class="ttdoc">Updates only the pointer. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:235</div></div>
<div class="ttc" id="structcutlass_1_1ReferenceFactory_html"><div class="ttname"><a href="structcutlass_1_1ReferenceFactory.html">cutlass::ReferenceFactory</a></div><div class="ttdef"><b>Definition:</b> subbyte_reference.h:557</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html"><div class="ttname"><a href="classcutlass_1_1TensorRef.html">cutlass::TensorRef</a></div><div class="ttdef"><b>Definition:</b> tensor_ref.h:146</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a0aba29ebf8715d217817a49356f95d3f"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a0aba29ebf8715d217817a49356f95d3f">cutlass::TensorRef&lt; Element, Layout &gt;&lt; Element, Layout &gt;::Reference</a></div><div class="ttdeci">typename platform::conditional&lt; sizeof_bits&lt; Element &gt;::value &gt;=8, Element &amp;, SubbyteReference&lt; Element &gt; &gt;::type Reference</div><div class="ttdoc">Reference type to an element. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:159</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="classcutlass_1_1TensorRef_html_a4166ac2a0754574ac21d5d57d74f34e5"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a4166ac2a0754574ac21d5d57d74f34e5">cutlass::TensorRef&lt; Element, Layout &gt;::offset</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE LongIndex offset(TensorCoord const &amp;coord) const</div><div class="ttdoc">Computes the offset of an index from the origin of the tensor. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:301</div></div>
<div class="ttc" id="structcutlass_1_1platform_1_1conditional_html"><div class="ttname"><a href="structcutlass_1_1platform_1_1conditional.html">cutlass::platform::conditional</a></div><div class="ttdoc">std::conditional (true specialization) </div><div class="ttdef"><b>Definition:</b> platform.h:325</div></div>
Updated doxygen
2018-09-20 05:02:08 +08:00
<div class="ttc" id="platform_8h_html_adde4c9ea91b753491851361a4198c009"><div class="ttname"><a href="platform_8h.html#adde4c9ea91b753491851361a4198c009">static_assert</a></div><div class="ttdeci">#define static_assert(__e, __m)</div><div class="ttdef"><b>Definition:</b> platform.h:153</div></div>
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
<div class="ttc" id="classcutlass_1_1IdentityTensorLayout_html_a813d116ff0e45679a2a7960a7c10fd1b"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html#a813d116ff0e45679a2a7960a7c10fd1b">cutlass::IdentityTensorLayout::kRank</a></div><div class="ttdeci">static int const kRank</div><div class="ttdoc">Logical rank of tensor. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:48</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a66a9bab939e2d57c130bb76a5527f482"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a66a9bab939e2d57c130bb76a5527f482">cutlass::TensorRef&lt; Element, Layout &gt;::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="structcutlass_1_1Coord_html"><div class="ttname"><a href="structcutlass_1_1Coord.html">cutlass::Coord</a></div><div class="ttdoc">Statically-sized array specifying Coords within a tensor. </div><div class="ttdef"><b>Definition:</b> coord.h:43</div></div>
<div class="ttc" id="classcutlass_1_1IdentityTensorLayout_html_aea0b83b611144c3f5860712967234ab4"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html#aea0b83b611144c3f5860712967234ab4">cutlass::IdentityTensorLayout::LongIndex</a></div><div class="ttdeci">int64_t LongIndex</div><div class="ttdoc">Long index type used for offsets. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:57</div></div>
<div class="ttc" id="namespacecutlass_html_accfe64331a1403e14cc312d1b4c844e1"><div class="ttname"><a href="namespacecutlass.html#accfe64331a1403e14cc312d1b4c844e1">cutlass::make_TensorRef</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE TensorRef&lt; Element, Layout &gt; make_TensorRef(Element *ptr, Layout const &amp;layout)</div><div class="ttdoc">Constructs a TensorRef, deducing types from arguments. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:368</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a11ec4b07a2132e647ca2ebe5112ce5ec"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a11ec4b07a2132e647ca2ebe5112ce5ec">cutlass::TensorRef&lt; Element, Layout &gt;&lt; Element, Layout &gt;::Index</a></div><div class="ttdeci">typename Layout::Index Index</div><div class="ttdoc">Index type. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:165</div></div>
<div class="ttc" id="classcutlass_1_1IdentityTensorLayout_html_a2aff6124a25853f227cdd7f3b36733c4"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html#a2aff6124a25853f227cdd7f3b36733c4">cutlass::IdentityTensorLayout::IdentityTensorLayout</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE IdentityTensorLayout(Stride const &amp;stride=Stride())</div><div class="ttdef"><b>Definition:</b> tensor_ref.h:81</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a8758907a1c9b1fcd00e7ece626d03b76"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a8758907a1c9b1fcd00e7ece626d03b76">cutlass::TensorRef&lt; Element, Layout &gt;::at</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Reference at(TensorCoord const &amp;coord) const</div><div class="ttdoc">Returns a reference to the element at a given Coord. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:307</div></div>
<div class="ttc" id="classcutlass_1_1SubbyteReference_html"><div class="ttname"><a href="classcutlass_1_1SubbyteReference.html">cutlass::SubbyteReference</a></div><div class="ttdef"><b>Definition:</b> subbyte_reference.h:294</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a7771fcd932e36c51c15572305f5b5520"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a7771fcd932e36c51c15572305f5b5520">cutlass::TensorRef&lt; Element, Layout &gt;::layout</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE Layout &amp; layout()</div><div class="ttdoc">Returns the layout object. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:265</div></div>
<div class="ttc" id="namespacecutlass_html_aa43b0a7d59635cb2d9ac96a077c988c3"><div class="ttname"><a href="namespacecutlass.html#aa43b0a7d59635cb2d9ac96a077c988c3">cutlass::TensorRef_aligned</a></div><div class="ttdeci">bool TensorRef_aligned(TensorRef&lt; Element, Layout &gt; const &amp;ref, int alignment)</div><div class="ttdef"><b>Definition:</b> tensor_ref.h:382</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_ad4e1a9d4912a18547cd5391d63e8e7ac"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#ad4e1a9d4912a18547cd5391d63e8e7ac">cutlass::TensorRef&lt; Element, Layout &gt;::operator+</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE TensorRef operator+(TensorCoord const &amp;b) const</div><div class="ttdoc">Returns a TensorRef offset by a given amount. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:333</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_ae82b8ba3bbf7ebb28fe063932db3dc6b"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#ae82b8ba3bbf7ebb28fe063932db3dc6b">cutlass::TensorRef&lt; Element, Layout &gt;::operator+=</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE TensorRef &amp; operator+=(TensorCoord const &amp;b)</div><div class="ttdoc">Returns a TensorRef offset by a given amount. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:341</div></div>
<div class="ttc" id="classcutlass_1_1TensorRef_html_a6bbcd0e512915565cabfeccdb1b6417d"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#a6bbcd0e512915565cabfeccdb1b6417d">cutlass::TensorRef&lt; Element, Layout &gt;::add_pointer_offset</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE TensorRef &amp; add_pointer_offset(LongIndex offset_)</div><div class="ttdoc">Adds an offset to each pointer. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:319</div></div>
<div class="ttc" id="subbyte__reference_8h_html"><div class="ttname"><a href="subbyte__reference_8h.html">subbyte_reference.h</a></div><div class="ttdoc">Provides a mechanism for packing and unpacking elements smaller than one byte. </div></div>
<div class="ttc" id="classcutlass_1_1IdentityTensorLayout_html_ad0ed9dc11a6284d25ea588d6933d2965"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html#ad0ed9dc11a6284d25ea588d6933d2965">cutlass::IdentityTensorLayout::kStrideRank</a></div><div class="ttdeci">static int const kStrideRank</div><div class="ttdoc">Rank of stride vector. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:51</div></div>
<div class="ttc" id="structcutlass_1_1Coord_html_a057a417a4d4a6e2f69e0b55a6f7ee902"><div class="ttname"><a href="structcutlass_1_1Coord.html#a057a417a4d4a6e2f69e0b55a6f7ee902">cutlass::Coord::dot</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE LongIndex dot(Coord const &amp;b, LongIndex sum=LongIndex(0)) const </div><div class="ttdoc">Computes the dot product with anotherCoord object. </div><div class="ttdef"><b>Definition:</b> coord.h:246</div></div>
<div class="ttc" id="structcutlass_1_1Coord_html_abe58b7c8f153a6029c2adc173f340fe0"><div class="ttname"><a href="structcutlass_1_1Coord.html#abe58b7c8f153a6029c2adc173f340fe0">cutlass::Coord::max_dim_index</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE int max_dim_index() const </div><div class="ttdoc">Returns the index of the dimension with greatest value. </div><div class="ttdef"><b>Definition:</b> coord.h:130</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="classcutlass_1_1TensorRef_html_adeada5e33b231f125a4aaeaf963bd3a3"><div class="ttname"><a href="classcutlass_1_1TensorRef.html#adeada5e33b231f125a4aaeaf963bd3a3">cutlass::TensorRef&lt; Element, Layout &gt;&lt; Element, Layout &gt;::LongIndex</a></div><div class="ttdeci">typename Layout::LongIndex LongIndex</div><div class="ttdoc">Long index used for pointer offsets. </div><div class="ttdef"><b>Definition:</b> tensor_ref.h:168</div></div>
<div class="ttc" id="classcutlass_1_1IdentityTensorLayout_html_a3dc530520b5eb35bc57c75de7954b59f"><div class="ttname"><a href="classcutlass_1_1IdentityTensorLayout.html#a3dc530520b5eb35bc57c75de7954b59f">cutlass::IdentityTensorLayout::capacity</a></div><div class="ttdeci">CUTLASS_HOST_DEVICE LongIndex capacity(TensorCoord const &amp;size) const </div><div class="ttdoc">Compute the number of contiguous elements needed to store a tensor with the given size...</div><div class="ttdef"><b>Definition:</b> tensor_ref.h:103</div></div>
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
</div><!-- fragment --></div><!-- contents -->
<!-- start footer part -->
<hr class="footer"/><address class="footer"><small>
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
Generated by &#160;<a href="http://www.doxygen.org/index.html">
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
<img class="footer" src="doxygen.png" alt="doxygen"/>
CUTLASS 2.0 (#62) CUTLASS 2.0 Substantially refactored for - Better performance, particularly for native Turing Tensor Cores - Robust and durable templates spanning the design space - Encapsulated functionality embodying modern C++11 programming techniques - Optimized containers and data types for efficient, generic, portable device code Updates to: - Quick start guide - Documentation - Utilities - CUTLASS Profiler Native Turing Tensor Cores - Efficient GEMM kernels targeting Turing Tensor Cores - Mixed-precision floating point, 8-bit integer, 4-bit integer, and binarized operands Coverage of existing CUTLASS functionality: - GEMM kernels targeting CUDA and Tensor Cores in NVIDIA GPUs - Volta Tensor Cores through native mma.sync and through WMMA API - Optimizations such as parallel reductions, threadblock rasterization, and intra-threadblock reductions - Batched GEMM operations - Complex-valued GEMMs Note: this commit and all that follow require a host compiler supporting C++11 or greater.
2019-11-20 08:55:34 +08:00
</a> 1.8.11
CUTLASS v1.0 release
2018-05-17 02:44:56 +08:00
</small></address>
</body>
</html>
Reference in New Issue Copy Permalink