<!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" lang="en-us" xml:lang="en-us">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8"></meta>
<meta http-equiv="X-UA-Compatible" content="IE=edge"></meta>
<meta name="copyright" content="(C) Copyright 2005"></meta>
<meta name="DC.rights.owner" content="(C) Copyright 2005"></meta>
<meta name="DC.Type" content="concept"></meta>
<meta name="DC.Title" content="NVIDIA CUDA Installation Guide for Linux"></meta>
<meta name="abstract" content="The installation instructions for the CUDA Toolkit on Linux."></meta>
<meta name="description" content="The installation instructions for the CUDA Toolkit on Linux."></meta>
<meta name="DC.Coverage" content="Installation Guides"></meta>
<meta name="DC.subject" content="CUDA Linux, CUDA Linux requirements, CUDA Linux supported versions, CUDA Linux gcc, CUDA Linux install toolkit, CUDA Linux GPU, CUDA Linux download, CUDA Linux install verification, CUDA Linux examples, CUDA Linux binaries"></meta>
<meta name="keywords" content="CUDA Linux, CUDA Linux requirements, CUDA Linux supported versions, CUDA Linux gcc, CUDA Linux install toolkit, CUDA Linux GPU, CUDA Linux download, CUDA Linux install verification, CUDA Linux examples, CUDA Linux binaries"></meta>
<meta name="DC.Format" content="XHTML"></meta>
<meta name="DC.Identifier" content="abstract"></meta>
<link rel="stylesheet" type="text/css" href="../common/formatting/commonltr.css"></link>
<link rel="stylesheet" type="text/css" href="../common/formatting/site.css"></link>
<title>Installation Guide Linux :: CUDA Toolkit Documentation</title>
<!--[if lt IE 9]>
<script src="../common/formatting/html5shiv-printshiv.min.js"></script>
<![endif]-->
<script type="text/javascript" charset="utf-8" src="//assets.adobedtm.com/b92787824f2e0e9b68dc2e993f9bd995339fe417/satelliteLib-7ba51e58dc61bcb0e9311aadd02a0108ab24cc6c.js"></script>
<script type="text/javascript" charset="utf-8" src="../common/formatting/jquery.min.js"></script>
<script type="text/javascript" charset="utf-8" src="../common/formatting/jquery.ba-hashchange.min.js"></script>
<script type="text/javascript" charset="utf-8" src="../common/formatting/jquery.scrollintoview.min.js"></script>
<script type="text/javascript" src="../search/htmlFileList.js"></script>
<script type="text/javascript" src="../search/htmlFileInfoList.js"></script>
<script type="text/javascript" src="../search/nwSearchFnt.min.js"></script>
<script type="text/javascript" src="../search/stemmers/en_stemmer.min.js"></script>
<script type="text/javascript" src="../search/index-1.js"></script>
<script type="text/javascript" src="../search/index-2.js"></script>
<script type="text/javascript" src="../search/index-3.js"></script>
<link rel="canonical" href="http://docs.nvidia.com/cuda/cuda-installation-guide-linux/index.html"></link>
<link rel="stylesheet" type="text/css" href="../common/formatting/qwcode.highlight.css"></link>
</head>
<body>
<header id="header"><span id="company">NVIDIA</span><span id="site-title">CUDA Toolkit Documentation</span><form id="search" method="get" action="search">
<input type="text" name="search-text"></input><fieldset id="search-location">
<legend>Search In:</legend>
<label><input type="radio" name="search-type" value="site"></input>Entire Site</label>
<label><input type="radio" name="search-type" value="document"></input>Just This Document</label></fieldset>
<button type="reset">clear search</button>
<button id="submit" type="submit">search</button></form>
</header>
<div id="site-content">
<nav id="site-nav">
<div class="category closed"><a href="../index.html" title="The root of the site.">CUDA Toolkit
v10.1.168</a></div>
<div class="category"><a href="index.html" title="Installation Guide Linux">Installation Guide Linux</a></div>
<ul>
<li>
<div class="section-link"><a href="#introduction">1. Introduction</a></div>
<ul>
<li>
<div class="section-link"><a href="#system-requirements">1.1. System Requirements</a></div>
</li>
<li>
<div class="section-link"><a href="#about-this-document">1.2. About This Document</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#pre-installation-actions">2. Pre-installation Actions</a></div>
<ul>
<li>
<div class="section-link"><a href="#verify-you-have-cuda-enabled-system">2.1. Verify You Have a CUDA-Capable GPU</a></div>
</li>
<li>
<div class="section-link"><a href="#verify-you-have-supported-version-of-linux">2.2. Verify You Have a Supported Version of Linux</a></div>
</li>
<li>
<div class="section-link"><a href="#verify-that-gcc-is-installed">2.3. Verify the System Has gcc Installed</a></div>
</li>
<li>
<div class="section-link"><a href="#verify-kernel-packages">2.4. Verify the System has the Correct Kernel Headers and Development Packages Installed</a></div>
</li>
<li>
<div class="section-link"><a href="#choose-installation-method">2.5. Choose an Installation Method</a></div>
</li>
<li>
<div class="section-link"><a href="#download-nvidia-driver-and-cuda-software">2.6. Download the NVIDIA CUDA Toolkit</a></div>
</li>
<li>
<div class="section-link"><a href="#handle-uninstallation">2.7. Handle Conflicting Installation Methods</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#package-manager-installation">3. Package Manager Installation</a></div>
<ul>
<li>
<div class="section-link"><a href="#package-manager-overview">3.1. Overview</a></div>
</li>
<li>
<div class="section-link"><a href="#redhat-installation">3.2. Redhat/CentOS</a></div>
</li>
<li>
<div class="section-link"><a href="#fedora-installation">3.3. Fedora</a></div>
</li>
<li>
<div class="section-link"><a href="#sles-installation">3.4. SLES</a></div>
</li>
<li>
<div class="section-link"><a href="#suse-installation">3.5. OpenSUSE</a></div>
</li>
<li>
<div class="section-link"><a href="#ubuntu-installation">3.6. Ubuntu</a></div>
</li>
<li>
<div class="section-link"><a href="#package-manager-additional">3.7. Additional Package Manager Capabilities</a></div>
<ul>
<li>
<div class="section-link"><a href="#package-manager-packages">3.7.1. Available Packages</a></div>
</li>
<li>
<div class="section-link"><a href="#package-manager-upgrades">3.7.2. Package Upgrades</a></div>
</li>
<li>
<div class="section-link"><a href="#package-manager-metas">3.7.3. Meta Packages</a></div>
</li>
</ul>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#runfile">4. Runfile Installation</a></div>
<ul>
<li>
<div class="section-link"><a href="#runfile-overview">4.1. Overview</a></div>
</li>
<li>
<div class="section-link"><a href="#runfile-installation">4.2. Installation</a></div>
</li>
<li>
<div class="section-link"><a href="#runfile-installer-ui">4.3. Installer UI</a></div>
</li>
<li>
<div class="section-link"><a href="#runfile-nouveau">4.4. Disabling Nouveau</a></div>
<ul>
<li>
<div class="section-link"><a href="#runfile-nouveau-fedora">4.4.1. Fedora</a></div>
</li>
<li>
<div class="section-link"><a href="#runfile-nouveau-rhel">4.4.2. RHEL/CentOS</a></div>
</li>
<li>
<div class="section-link"><a href="#runfile-nouveau-suse">4.4.3. OpenSUSE</a></div>
</li>
<li>
<div class="section-link"><a href="#runfile-nouveau-sles">4.4.4. SLES</a></div>
</li>
<li>
<div class="section-link"><a href="#runfile-nouveau-ubuntu">4.4.5. Ubuntu</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#runfile-verifications">4.5. Device Node Verification</a></div>
</li>
<li>
<div class="section-link"><a href="#runfile-advanced">4.6. Advanced Options</a></div>
</li>
<li>
<div class="section-link"><a href="#runfile-uninstallation">4.7. Uninstallation</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#cluster">5. Cluster Management Packages</a></div>
<ul>
<li>
<div class="section-link"><a href="#cluster-overview">5.1. Overview</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#cross-platform">6. CUDA Cross-Platform Environment</a></div>
<ul>
<li>
<div class="section-link"><a href="#cross-installation">6.1. CUDA Cross-Platform Installation</a></div>
</li>
<li>
<div class="section-link"><a href="#cross-samples">6.2. CUDA Cross-Platform Samples</a></div>
</li>
<li>
<div class="section-link"><a href="#cross-nsight">6.3. Nsight Eclipse Edition</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#post-installation-actions">7. Post-installation Actions</a></div>
<ul>
<li>
<div class="section-link"><a href="#mandatory-post">7.1. Mandatory Actions</a></div>
<ul>
<li>
<div class="section-link"><a href="#environment-setup">7.1.1. Environment Setup</a></div>
</li>
<li>
<div class="section-link"><a href="#power9-setup">7.1.2. POWER9 Setup</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#recommended-post">7.2. Recommended Actions</a></div>
<ul>
<li>
<div class="section-link"><a href="#install-persistenced">7.2.1. Install Persistence Daemon</a></div>
</li>
<li>
<div class="section-link"><a href="#install-samples">7.2.2. Install Writable Samples</a></div>
</li>
<li>
<div class="section-link"><a href="#verify-installation">7.2.3. Verify the Installation</a></div>
<ul>
<li>
<div class="section-link"><a href="#verify-driver">7.2.3.1. Verify the Driver Version</a></div>
</li>
<li>
<div class="section-link"><a href="#compiling-examples">7.2.3.2. Compiling the Examples</a></div>
</li>
<li>
<div class="section-link"><a href="#running-binaries">7.2.3.3. Running the Binaries</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#install-nsight-plugins">7.2.4. Install Nsight Eclipse Plugins</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#optional-post">7.3. Optional Actions</a></div>
<ul>
<li>
<div class="section-link"><a href="#install-libraries">7.3.1. Install Third-party Libraries</a></div>
</li>
<li>
<div class="section-link"><a href="#install-source">7.3.2. Install the source code for cuda-gdb</a></div>
</li>
</ul>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#advanced-setup">8. Advanced Setup</a></div>
</li>
<li>
<div class="section-link"><a href="#faq">9. Frequently Asked Questions</a></div>
</li>
<li>
<div class="section-link"><a href="#additional-considerations">10. Additional Considerations</a></div>
</li>
<li>
<div class="section-link"><a href="#removing-cuda-tk-and-driver">11. Removing CUDA Toolkit and Driver</a></div>
</li>
</ul>
</nav>
<div id="resize-nav"></div>
<nav id="search-results">
<h2>Search Results</h2>
<ol></ol>
</nav>
<div id="contents-container">
<div id="breadcrumbs-container">
<div id="release-info">Installation Guide Linux
(<a href="../../pdf/CUDA_Installation_Guide_Linux.pdf">PDF</a>)
-
v10.1.168
(<a href="https://developer.nvidia.com/cuda-toolkit-archive">older</a>)
-
Last updated April 24, 2019
-
<a href="mailto:CUDAIssues@nvidia.com?subject=CUDA Toolkit Documentation Feedback: Installation Guide Linux">Send Feedback</a></div>
</div>
<article id="contents">
<div class="topic nested0" id="abstract"><a name="abstract" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#abstract" name="abstract" shape="rect">NVIDIA CUDA Installation Guide for Linux</a></h2>
<div class="body conbody">
<p class="shortdesc">The installation instructions for the CUDA Toolkit on Linux.</p>
</div>
</div>
<div class="topic concept nested0" id="introduction"><a name="introduction" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#introduction" name="introduction" shape="rect">1. Introduction</a></h2>
<div class="body conbody">
<p class="p">CUDA<sup>®</sup> is a parallel computing platform and programming model invented by NVIDIA. It enables dramatic increases in computing performance
by harnessing the power of the graphics processing unit (GPU).
</p>
<div class="p">CUDA was developed with several design goals in mind:
<ul class="ul">
<li class="li">Provide a small set of extensions to standard programming languages, like C, that enable a straightforward implementation
of parallel algorithms. With CUDA C/C++, programmers can focus on the task of parallelization of the algorithms rather than
spending time on their implementation.
</li>
<li class="li">Support heterogeneous computation where applications use both the CPU and GPU. Serial portions of applications are run on
the CPU, and parallel portions are offloaded to the GPU. As such, CUDA can be incrementally applied to existing applications.
The CPU and GPU are treated as separate devices that have their own memory spaces. This configuration also allows simultaneous
computation on the CPU and GPU without contention for memory resources.
</li>
</ul>
CUDA-capable GPUs have hundreds of cores that can collectively run thousands of computing threads. These cores have shared
resources including a register file and a shared memory. The on-chip shared memory allows parallel tasks running on these
cores to share data without sending it over the system memory bus.
</div>
<p class="p">This guide will show you how to install and check the correct operation of the CUDA development tools.</p>
</div>
<div class="topic concept nested1" id="system-requirements"><a name="system-requirements" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#system-requirements" name="system-requirements" shape="rect">1.1. System Requirements</a></h3>
<div class="body conbody">
<div class="p">To use CUDA on your system, you will need the following installed:
<ul class="ul">
<li class="li">CUDA-capable GPU</li>
<li class="li">A supported version of Linux with a gcc compiler and toolchain</li>
<li class="li">NVIDIA CUDA Toolkit (available at <a class="xref" href="http://developer.nvidia.com/cuda-downloads" target="_blank" shape="rect">http://developer.nvidia.com/cuda-downloads</a>)
</li>
</ul>
</div>
<p class="p">The CUDA development environment relies on tight integration with the host development
environment, including the host compiler and C runtime libraries, and is therefore only
supported on distribution versions that have been qualified for this CUDA Toolkit release.
</p>
<div class="tablenoborder">
<table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
<caption><span class="tablecap">Table 1. Native Linux Distribution Support in CUDA <span class="keyword">10.1</span></span></caption>
<thead class="thead" align="left">
<tr class="row">
<th class="entry" valign="top" width="NaN%" id="d54e155" rowspan="1" colspan="1">Distribution</th>
<th class="entry" valign="top" width="NaN%" id="d54e158" rowspan="1" colspan="1">Kernel*</th>
<th class="entry" valign="top" width="NaN%" id="d54e161" rowspan="1" colspan="1">GCC</th>
<th class="entry" valign="top" width="NaN%" id="d54e164" rowspan="1" colspan="1">GLIBC</th>
<th class="entry" valign="top" width="NaN%" id="d54e167" rowspan="1" colspan="1">ICC</th>
<th class="entry" valign="top" width="NaN%" id="d54e171" rowspan="1" colspan="1">PGI</th>
<th class="entry" valign="top" width="NaN%" id="d54e174" rowspan="1" colspan="1">XLC</th>
<th class="entry" valign="top" width="NaN%" id="d54e177" rowspan="1" colspan="1">CLANG</th>
</tr>
</thead>
<tbody class="tbody">
<tr class="row gray">
<td class="entry" colspan="8" align="center" valign="top" headers="d54e155 d54e158 d54e161 d54e164 d54e167 d54e171 d54e174 d54e177" rowspan="1">x86_64</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">RHEL 7.6</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">3.10</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">4.8.5</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.17</td>
<td class="entry" rowspan="11" valign="middle" width="NaN%" headers="d54e167" colspan="1">19.0</td>
<td class="entry" rowspan="11" valign="middle" width="NaN%" headers="d54e171" colspan="1">18.x, 19.x</td>
<td class="entry" rowspan="11" valign="middle" width="NaN%" headers="d54e174" colspan="1">NO</td>
<td class="entry" rowspan="11" valign="middle" width="NaN%" headers="d54e177" colspan="1">7.0.0</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">RHEL 6.10</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">2.6.32</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">4.4.7</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.12</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">CentOS 7.6</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">3.10</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">4.8.5</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.17</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">CentOS 6.10</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">2.6.32</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">4.4.7</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.12</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">Fedora 29</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.16</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">8.0.1</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.27</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">OpenSUSE Leap 15.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.15.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">7.3.1</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.26</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">SLES 15.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.12.14</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">7.2.1</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.26</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">SLES 12.4</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.12.14</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">4.8.5</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.22</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">Ubuntu 18.10</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.18.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">8.2.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.28</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">Ubuntu 18.04.1 (**)</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.15.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">7.3.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.27</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">Ubuntu 16.04.5 (**)</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.4</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">5.4.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.23</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">Ubuntu 14.04.5 (**)</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">3.13</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">4.8.4</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.19</td>
</tr>
<tr class="row gray">
<td class="entry" colspan="8" align="center" valign="top" headers="d54e155 d54e158 d54e161 d54e164 d54e167 d54e171 d54e174 d54e177" rowspan="1">POWER8(***)</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">RHEL 7.6</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">3.10</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">4.8.5</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.17</td>
<td class="entry" valign="top" width="NaN%" headers="d54e167" rowspan="1" colspan="1">NO</td>
<td class="entry" valign="top" width="NaN%" headers="d54e171" rowspan="1" colspan="1">18.x, 19.x</td>
<td class="entry" valign="top" width="NaN%" headers="d54e174" rowspan="1" colspan="1">13.1.x, 16.1.x</td>
<td class="entry" valign="top" width="NaN%" headers="d54e177" rowspan="1" colspan="1">7.0.0</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">Ubuntu 18.04.1</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.15.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">7.3.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.27</td>
<td class="entry" valign="top" width="NaN%" headers="d54e167" rowspan="1" colspan="1">NO</td>
<td class="entry" valign="top" width="NaN%" headers="d54e171" rowspan="1" colspan="1">18.x, 19.x</td>
<td class="entry" valign="top" width="NaN%" headers="d54e174" rowspan="1" colspan="1">13.1.x, 16.1.x</td>
<td class="entry" valign="top" width="NaN%" headers="d54e177" rowspan="1" colspan="1">7.0.0</td>
</tr>
<tr class="row gray">
<td class="entry" colspan="8" align="center" valign="top" headers="d54e155 d54e158 d54e161 d54e164 d54e167 d54e171 d54e174 d54e177" rowspan="1">POWER9(****)</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">Ubuntu 18.04.1</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.15.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">7.3.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.27</td>
<td class="entry" valign="top" width="NaN%" headers="d54e167" rowspan="1" colspan="1">NO</td>
<td class="entry" valign="top" width="NaN%" headers="d54e171" rowspan="1" colspan="1">18.x, 19.x</td>
<td class="entry" valign="top" width="NaN%" headers="d54e174" rowspan="1" colspan="1">13.1.x, 16.1.x</td>
<td class="entry" valign="top" width="NaN%" headers="d54e177" rowspan="1" colspan="1">7.0.0</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="NaN%" headers="d54e155" rowspan="1" colspan="1">RHEL 7.6 IBM Power LE</td>
<td class="entry" valign="top" width="NaN%" headers="d54e158" rowspan="1" colspan="1">4.14.0</td>
<td class="entry" valign="top" width="NaN%" headers="d54e161" rowspan="1" colspan="1">4.8.5</td>
<td class="entry" valign="top" width="NaN%" headers="d54e164" rowspan="1" colspan="1">2.17</td>
<td class="entry" valign="top" width="NaN%" headers="d54e167" rowspan="1" colspan="1">NO</td>
<td class="entry" valign="top" width="NaN%" headers="d54e171" rowspan="1" colspan="1">18.x, 19.x</td>
<td class="entry" valign="top" width="NaN%" headers="d54e174" rowspan="1" colspan="1">13.1.x, 16.1.x</td>
<td class="entry" valign="top" width="NaN%" headers="d54e177" rowspan="1" colspan="1">7.0.0</td>
</tr>
</tbody>
</table>
</div>
<p class="p">
(*) For specific kernel versions supported on Red Hat Enterprise Linux, visit <a class="xref" href="https://access.redhat.com/articles/3078" target="_blank" shape="rect">https://access.redhat.com/articles/3078</a>. For a list of kernel versions including the release dates for SUSE Linux Enterprise Server is available at <a class="xref" href="https://wiki.microfocus.com/index.php/SUSE/SLES/Kernel_versions" target="_blank" shape="rect">https://wiki.microfocus.com/index.php/SUSE/SLES/Kernel_versions</a>.
</p>
<p class="p">
(**) For Ubuntu LTS on x86-64, both the HWE kernel (e.g. 4.13.x for 16.04.4) and the server LTS kernel (e.g. 4.4.x for 16.04)
are supported in CUDA <span class="keyword">10.1</span>. Visit <a class="xref" href="https://wiki.ubuntu.com/Kernel/Support" target="_blank" shape="rect">https://wiki.ubuntu.com/Kernel/Support</a> for more information.
</p>
<p class="p">
(***) Only the Tesla GP100 GPU is supported for CUDA <span class="keyword">10.1</span> on POWER8.
</p>
<p class="p">
(****) Only the Tesla GV100 GPU is supported for CUDA <span class="keyword">10.1</span> on POWER9.
</p>
</div>
</div>
<div class="topic concept nested1" id="about-this-document"><a name="about-this-document" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#about-this-document" name="about-this-document" shape="rect">1.2. About This Document</a></h3>
<div class="body conbody">
<p class="p">This document is intended for readers familiar with the Linux environment and the compilation of C programs from the command
line. You do not need previous experience with CUDA or experience with parallel computation. Note: This guide covers installation
only on systems with X Windows installed.
</p>
<div class="note note"><span class="notetitle">Note:</span> Many commands in this document might require <em class="ph i">superuser</em> privileges. On most distributions of Linux, this will require you to log in as root. For systems that have enabled the sudo
package, use the sudo prefix for all necessary commands.
</div>
</div>
</div>
</div>
<div class="topic concept nested0" id="pre-installation-actions"><a name="pre-installation-actions" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#pre-installation-actions" name="pre-installation-actions" shape="rect">2. Pre-installation Actions</a></h2>
<div class="body conbody">
<div class="p">Some actions must be taken before the CUDA Toolkit and Driver can be installed on Linux:
<ul class="ul">
<li class="li">Verify the system has a CUDA-capable GPU.</li>
<li class="li">Verify the system is running a supported version of Linux.</li>
<li class="li">Verify the system has gcc installed.</li>
<li class="li">Verify the system has the correct kernel headers and development packages installed.</li>
<li class="li">Download the NVIDIA CUDA Toolkit.</li>
<li class="li">Handle conflicting installation methods.</li>
</ul>
</div>
<div class="note note"><span class="notetitle">Note:</span> You can override the install-time prerequisite checks by running the installer with the -override flag. Remember that the
prerequisites will still be required to use the NVIDIA CUDA Toolkit.
</div>
</div>
<div class="topic concept nested1" id="verify-you-have-cuda-enabled-system"><a name="verify-you-have-cuda-enabled-system" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#verify-you-have-cuda-enabled-system" name="verify-you-have-cuda-enabled-system" shape="rect">2.1. Verify You Have a CUDA-Capable GPU</a></h3>
<div class="body conbody">
<p class="p">To verify that your GPU is CUDA-capable, go to your distribution's equivalent of System Properties, or, from the command line,
enter:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> lspci | grep -i nvidia</pre><p class="p">If you do not see any settings, update the PCI hardware database that Linux maintains by entering <samp class="ph codeph">update-pciids</samp> (generally found in <samp class="ph codeph">/sbin</samp>) at the command line and rerun the previous <samp class="ph codeph">lspci</samp> command.
</p>
<p class="p">If your graphics card is from NVIDIA and it is listed in <a class="xref" href="http://developer.nvidia.com/cuda-gpus" target="_blank" shape="rect">http://developer.nvidia.com/cuda-gpus</a>, your GPU is CUDA-capable.
</p>
<p class="p">The Release Notes for the CUDA Toolkit also contain a list of supported products.</p>
</div>
</div>
<div class="topic concept nested1" id="verify-you-have-supported-version-of-linux"><a name="verify-you-have-supported-version-of-linux" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#verify-you-have-supported-version-of-linux" name="verify-you-have-supported-version-of-linux" shape="rect">2.2. Verify You Have a Supported Version of Linux</a></h3>
<div class="body conbody">
<p class="p">The CUDA Development Tools are only supported on some specific distributions of Linux. These are listed in the CUDA Toolkit
release notes.
</p>
<p class="p">To determine which distribution and release number you're running, type the following at the command line:</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> uname -m && cat /etc/*release</pre><p class="p">You should see output similar to the following, modified for your particular system:</p><pre class="pre screen" xml:space="preserve">x86_64
Red Hat Enterprise Linux Workstation release 6.0 (Santiago)</pre><p class="p">The <samp class="ph codeph">x86_64</samp> line indicates you are running on a 64-bit system. The remainder gives information about your distribution.
</p>
</div>
</div>
<div class="topic concept nested1" id="verify-that-gcc-is-installed"><a name="verify-that-gcc-is-installed" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#verify-that-gcc-is-installed" name="verify-that-gcc-is-installed" shape="rect">2.3. Verify the System Has gcc Installed</a></h3>
<div class="body conbody">
<p class="p">The <samp class="ph codeph">gcc</samp> compiler is required for development using the CUDA Toolkit. It is not required for running CUDA applications. It is generally
installed as part of the Linux installation, and in most cases the version of gcc installed with a supported version of Linux
will work correctly.
</p>
<p class="p">To verify the version of gcc installed on your system, type the following on the command line:</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> gcc --version</pre><p class="p">If an error message displays, you need to install the <dfn class="term">development tools</dfn> from your Linux distribution or obtain a version of <samp class="ph codeph">gcc</samp> and its accompanying toolchain from the Web.
</p>
</div>
</div>
<div class="topic concept nested1" id="verify-kernel-packages"><a name="verify-kernel-packages" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#verify-kernel-packages" name="verify-kernel-packages" shape="rect">2.4. Verify the System has the Correct Kernel Headers and Development Packages Installed</a></h3>
<div class="body conbody">
<p class="p">
The CUDA Driver requires that the kernel headers and development packages for the running version of the kernel be installed
at the time of the driver installation, as well whenever the driver is rebuilt. For example, if your system is running kernel
version 3.17.4-301, the 3.17.4-301 kernel headers and development packages must also be installed.
</p>
<p class="p">
While the Runfile installation performs no package validation, the RPM and Deb installations of the driver will make an attempt
to install the kernel header and development packages if no version of these packages is currently installed. However, it
will install the latest version of these packages, which may or may not match the version of the kernel your system is using.
<strong class="ph b">Therefore, it is best to manually ensure the correct version of the kernel headers and development packages are installed
prior to installing the CUDA Drivers, as well as whenever you change the kernel version.</strong></p>
<div class="p">
The version of the kernel your system is running can be found by running the following command:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> uname -r</pre>
This is the version of the kernel headers and development packages that must be installed prior to installing the CUDA Drivers.
This command will be used multiple times below to specify the version of the packages to install. Note that below are the
common-case scenarios for kernel usage. More advanced cases, such as custom kernel branches, should ensure that their kernel
headers and sources match the kernel build they are running.
</div>
<div class="note note"><span class="notetitle">Note:</span>
If you perform a system update which changes the version of the linux kernel being used, make sure to rerun the commands below
to ensure you have the correct kernel headers and kernel development packages installed. Otherwise, the CUDA Driver will fail
to work with the new kernel.
</div>
<div class="section">
<h3 class="title sectiontitle">RHEL/CentOS</h3>
<div class="p">
The kernel headers and development packages for the currently running kernel can be installed with:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum install kernel-devel-$(uname -r) kernel-headers-$(uname -r)</pre></div>
</div>
<div class="section">
<h3 class="title sectiontitle">Fedora</h3>
<div class="p">
The kernel headers and development packages for the currently running kernel can be installed with:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf install kernel-devel-$(uname -r) kernel-headers-$(uname -r)</pre></div>
</div>
<div class="section">
<h3 class="title sectiontitle">OpenSUSE/SLES</h3>
<div class="p">
Use the output of the <samp class="ph codeph">uname</samp> command to determine the running kernel's version and variant:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> uname -r
3.16.6-2-default</pre>
In this example, the version is <samp class="ph codeph">3.16.6-2</samp> and the variant is <samp class="ph codeph">default</samp>.
The kernel headers and development packages can then be installed with the following command, replacing <samp class="ph codeph"><variant></samp> and <samp class="ph codeph"><version></samp> with the variant and version discovered from the previous <samp class="ph codeph">uname</samp> command:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper install kernel-<variant>-devel=<version></pre></div>
</div>
<div class="section">
<h3 class="title sectiontitle">Ubuntu</h3>
<div class="p">
The kernel headers and development packages for the currently running kernel can be installed with:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-get install linux-headers-$(uname -r)</pre></div>
</div>
</div>
</div>
<div class="topic concept nested1" id="choose-installation-method"><a name="choose-installation-method" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#choose-installation-method" name="choose-installation-method" shape="rect">2.5. Choose an Installation Method</a></h3>
<div class="body conbody">
<p class="p">
The CUDA Toolkit can be installed using either of two different installation mechanisms: distribution-specific packages (RPM
and Deb packages), or a distribution-independent package (runfile packages). The distribution-independent package has the
advantage of working across a wider set of Linux distributions, but does not update the distribution's native package management
system. The distribution-specific packages interface with the distribution's native package management system. It is recommended
to use the distribution-specific packages, where possible.
</p>
<div class="note note"><span class="notetitle">Note:</span>
Standalone installers are not provided for architectures other than the x86_64 release. For both native as well as cross development,
the toolkit must be installed using the distribution-specific installer. See the <a class="xref" href="index.html#cross-installation" shape="rect">CUDA Cross-Platform Installation</a> section for more details.
</div>
</div>
</div>
<div class="topic concept nested1" id="download-nvidia-driver-and-cuda-software"><a name="download-nvidia-driver-and-cuda-software" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#download-nvidia-driver-and-cuda-software" name="download-nvidia-driver-and-cuda-software" shape="rect">2.6. Download the NVIDIA CUDA Toolkit</a></h3>
<div class="body conbody">
<p class="p">The NVIDIA CUDA Toolkit is available at <a class="xref" href="http://developer.nvidia.com/cuda-downloads" target="_blank" shape="rect">http://developer.nvidia.com/cuda-downloads</a>.
</p>
<p class="p">Choose the platform you are using and download the NVIDIA CUDA Toolkit</p>
<p class="p">The CUDA Toolkit contains the CUDA driver and tools needed to create, build and run a CUDA application as well as libraries,
header files, CUDA samples source code, and other resources.
</p>
<div class="section">
<h3 class="title sectiontitle">Download Verification</h3>
<p class="p">The download can be verified by comparing the MD5 checksum posted at <a class="xref" href="http://developer.nvidia.com/cuda-downloads/checksums" target="_blank" shape="rect">http://developer.nvidia.com/cuda-downloads/checksums</a> with that of the downloaded file. If either of the checksums differ, the downloaded file is corrupt and needs to be downloaded
again.
</p>
<div class="p">To calculate the MD5 checksum of the downloaded file, run the following:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> md5sum <file></pre></div>
</div>
</div>
</div>
<div class="topic concept nested1" id="handle-uninstallation"><a name="handle-uninstallation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#handle-uninstallation" name="handle-uninstallation" shape="rect">2.7. Handle Conflicting Installation Methods</a></h3>
<div class="body conbody">
<p class="p">
Before installing CUDA, any previously installations that could conflict should be uninstalled. This will not affect systems
which have not had CUDA installed previously, or systems where the installation method has been preserved (RPM/Deb vs. Runfile).
See the following charts for specifics.
</p>
<div class="tablenoborder">
<table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
<caption><span class="tablecap">Table 2. CUDA Toolkit Installation Compatibility Matrix</span></caption>
<tbody class="tbody">
<tr class="row">
<td class="entry" rowspan="2" colspan="2" align="center" valign="top"> </td>
<td class="entry" colspan="2" align="center" valign="top" rowspan="1">Installed Toolkit Version == X.Y</td>
<td class="entry" colspan="2" align="center" valign="top" rowspan="1">Installed Toolkit Version != X.Y</td>
</tr>
<tr class="row">
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">RPM/Deb</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">run</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">RPM/Deb</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">run</td>
</tr>
<tr class="row">
<td class="entry" rowspan="2" align="center" valign="top" colspan="1">Installing Toolkit Version X.Y</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">RPM/Deb</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">No Action</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">Uninstall Run</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">No Action</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">No Action</td>
</tr>
<tr class="row">
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">run</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">Uninstall RPM/Deb</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">Uninstall Run</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">No Action</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">No Action</td>
</tr>
</tbody>
</table>
</div>
<div class="tablenoborder">
<table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
<caption><span class="tablecap">Table 3. NVIDIA Driver Installation Compatibility Matrix</span></caption>
<tbody class="tbody">
<tr class="row">
<td class="entry" rowspan="2" colspan="2" valign="top"> </td>
<td class="entry" colspan="2" valign="top" rowspan="1">Installed Driver Version == X.Y</td>
<td class="entry" colspan="2" valign="top" rowspan="1">Installed Driver Version != X.Y</td>
</tr>
<tr class="row">
<td class="entry" valign="top" rowspan="1" colspan="1">RPM/Deb</td>
<td class="entry" valign="top" rowspan="1" colspan="1">run</td>
<td class="entry" valign="top" rowspan="1" colspan="1">RPM/Deb</td>
<td class="entry" valign="top" rowspan="1" colspan="1">run</td>
</tr>
<tr class="row">
<td class="entry" rowspan="2" valign="top" colspan="1">Installing Driver Version X.Y</td>
<td class="entry" valign="top" rowspan="1" colspan="1">RPM/Deb</td>
<td class="entry" valign="top" rowspan="1" colspan="1">No Action</td>
<td class="entry" valign="top" rowspan="1" colspan="1">Uninstall Run</td>
<td class="entry" valign="top" rowspan="1" colspan="1">No Action</td>
<td class="entry" valign="top" rowspan="1" colspan="1">Uninstall Run</td>
</tr>
<tr class="row">
<td class="entry" valign="top" rowspan="1" colspan="1">run</td>
<td class="entry" valign="top" rowspan="1" colspan="1">Uninstall RPM/Deb</td>
<td class="entry" valign="top" rowspan="1" colspan="1">No Action</td>
<td class="entry" valign="top" rowspan="1" colspan="1">Uninstall RPM/Deb</td>
<td class="entry" valign="top" rowspan="1" colspan="1">No Action</td>
</tr>
</tbody>
</table>
</div>
<div class="p">
Use the following command to uninstall a Toolkit runfile installation:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo /usr/local/cuda-X.Y/bin/uninstall_cuda_X.Y.pl</pre>
Use the following command to uninstall a Driver runfile installation:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo /usr/bin/nvidia-uninstall</pre>
Use the following commands to uninstall a RPM/Deb installation:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum remove <package_name> # Redhat/CentOS
<strong class="ph b">$</strong> sudo dnf remove <package_name> # Fedora
<strong class="ph b">$</strong> sudo zypper remove <package_name> # OpenSUSE/SLES
<strong class="ph b">$</strong> sudo apt-get --purge remove <package_name> # Ubuntu</pre></div>
</div>
</div>
</div>
<div class="topic concept nested0" id="package-manager-installation"><a name="package-manager-installation" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#package-manager-installation" name="package-manager-installation" shape="rect">3. Package Manager Installation</a></h2>
<div class="body conbody">
<p class="p">
Basic instructions can be found in the <a class="xref" href="http://docs.nvidia.com/cuda/cuda-quick-start-guide/index.html#linux" target="_blank" shape="rect">Quick Start Guide</a>. Read on for more detailed instructions.
</p>
</div>
<div class="topic concept nested1" id="package-manager-overview"><a name="package-manager-overview" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#package-manager-overview" name="package-manager-overview" shape="rect">3.1. Overview</a></h3>
<div class="body conbody">
<p class="p">
The Package Manager installation interfaces with your system's package management system.
When using RPM or Deb, the downloaded package is a repository package.
Such a package only informs the package manager where to find the actual installation packages, but will not install them.
</p>
<p class="p">
If those packages are available in an online repository, they will be automatically downloaded in a later step. Otherwise,
the repository package also installs a local repository containing the installation packages on the system.
Whether the repository is available online or installed locally, the installation procedure is identical and made of several
steps.
</p>
<div class="p">
Distribution-specific instructions detail how to install CUDA:
<ul class="ul">
<li class="li"><a class="xref" href="index.html#redhat-installation" shape="rect">Redhat/CentOS</a></li>
<li class="li"><a class="xref" href="index.html#fedora-installation" shape="rect">Fedora</a></li>
<li class="li"><a class="xref" href="index.html#sles-installation" shape="rect">SLES</a></li>
<li class="li"><a class="xref" href="index.html#suse-installation" shape="rect">OpenSUSE</a></li>
<li class="li"><a class="xref" href="index.html#ubuntu-installation" shape="rect">Ubuntu</a></li>
</ul>
</div>
<p class="p">
Finally, some helpful <a class="xref" href="index.html#package-manager-additional" shape="rect">package manager capabilities</a> are detailed.
</p>
<p class="p">
These instructions are for native development only. For cross-platform development, see the <a class="xref" href="index.html#cross-platform" shape="rect">CUDA Cross-Platform Environment</a> section.
</p>
<div class="note note"><span class="notetitle">Note:</span>
The package "cuda-core" has been deprecated in CUDA 9.1. Please use "cuda-compiler" instead.
</div>
</div>
</div>
<div class="topic concept nested1" id="redhat-installation"><a name="redhat-installation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#redhat-installation" name="redhat-installation" shape="rect">3.2. Redhat/CentOS</a></h3>
<div class="body conbody">
<ol class="ol">
<li class="li">
Perform the <a class="xref" href="index.html#pre-installation-actions" shape="rect"> pre-installation actions.</a></li>
<li class="li"><strong class="ph b">Satisfy third-party package dependency</strong><div class="p"><a name="redhat-installation__ul_gzl_qbv_bgb" shape="rect">
<!-- --></a><ul class="ul" id="redhat-installation__ul_gzl_qbv_bgb">
<li class="li"><strong class="ph b">Satisfy DKMS dependency</strong>: The NVIDIA driver RPM packages depend on other
external packages, such as DKMS and <samp class="ph codeph">libvdpau</samp>. Those packages are only
available on third-party repositories, such as <a class="xref" href="http://fedoraproject.org/wiki/EPEL" target="_blank" shape="rect">EPEL</a>. Any such third-party repositories must be added to the package manager
repository database before installing the NVIDIA driver RPM packages, or missing
dependencies will prevent the installation from proceeding.
</li>
<li class="li"><strong class="ph b">Enable optional repos</strong>:
<p class="p">On <strong class="ph b">RHEL 7 Linux</strong> only, execute the following steps
to enable optional repositories.
</p><a name="redhat-installation__ul_r2y_xbv_bgb" shape="rect">
<!-- --></a><ul class="ul" id="redhat-installation__ul_r2y_xbv_bgb">
<li class="li"><strong class="ph b">On x86_64 workstation:</strong><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> subscription-manager repos --enable=rhel-7-workstation-optional-rpms</pre></li>
<li class="li"><strong class="ph b">On POWER9 system:</strong><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> subscription-manager repos --enable=rhel-7-for-power-9-optional-rpms</pre></li>
<li class="li"><strong class="ph b">On x86_64 server:
</strong><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> subscription-manager repos --enable=rhel-7-server-optional-rpms </pre></li>
</ul>
</li>
</ul>
</div>
</li>
<li class="li"><strong class="ph b">Address custom xorg.conf, if applicable</strong><p class="p">The driver relies on an automatically generated xorg.conf file at /etc/X11/xorg.conf. If
a custom-built xorg.conf file is present, this functionality will be disabled and the
driver may not work. You can try removing the existing xorg.conf file, or adding the
contents of /etc/X11/xorg.conf.d/00-nvidia.conf to the xorg.conf file. The xorg.conf file
will most likely need manual tweaking for systems with a non-trivial GPU configuration.
</p>
</li>
<li class="li"><strong class="ph b">Install repository meta-data</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo rpm --install cuda-repo-<distro>-<version>.<architecture>.rpm</pre></div>
</li>
<li class="li"><strong class="ph b">Clean Yum repository cache</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum clean expire-cache</pre></div>
</li>
<li class="li"><strong class="ph b">Install CUDA</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum install cuda</pre></div>
<div class="p">
If the i686 libvdpau package dependency fails to install, try using the following steps to fix the issue:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> yumdownloader libvdpau.i686
<strong class="ph b">$</strong> sudo rpm -U --oldpackage libvdpau*.rpm</pre></div>
</li>
<li class="li"><strong class="ph b">Add libcuda.so symbolic link, if necessary</strong><p class="p">
The libcuda.so library is installed in the /usr/lib{,64}/nvidia directory. For pre-existing projects which use libcuda.so,
it may be useful to add a symbolic link from libcuda.so in the /usr/lib{,64} directory.
</p>
</li>
<li class="li">
Perform the <a class="xref" href="index.html#post-installation-actions" shape="rect"> post-installation actions.</a></li>
</ol>
</div>
</div>
<div class="topic concept nested1" id="fedora-installation"><a name="fedora-installation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#fedora-installation" name="fedora-installation" shape="rect">3.3. Fedora</a></h3>
<div class="body conbody">
<ol class="ol">
<li class="li">
Perform the <a class="xref" href="index.html#pre-installation-actions" shape="rect"> pre-installation actions.</a></li>
<li class="li"><strong class="ph b">Address custom xorg.conf, if applicable</strong><p class="p">
The driver relies on an automatically generated xorg.conf file at /etc/X11/xorg.conf. If a custom-built xorg.conf file is
present, this functionality will be disabled and the driver may not work.
You can try removing the existing xorg.conf file, or adding the contents of /etc/X11/xorg.conf.d/00-nvidia.conf to the xorg.conf
file. The xorg.conf file will most likely need manual tweaking for systems with a non-trivial GPU configuration.
</p>
</li>
<li class="li"><strong class="ph b">Satisfy Akmods dependency</strong><p class="p">
The NVIDIA driver RPM packages depend on the Akmods framework which is provided by the <a class="xref" href="http://rpmfusion.org/Configuration" target="_blank" shape="rect">RPMFusion free repository</a>.
The RPMFusion free repository must be added to the package manager repository database before installing the NVIDIA driver
RPM packages, or missing dependencies will prevent the installation from proceeding.
</p>
</li>
<li class="li"><strong class="ph b">Install repository meta-data</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo rpm --install cuda-repo-<distro>-<version>.<architecture>.rpm</pre></div>
</li>
<li class="li"><strong class="ph b">Clean DNF repository cache</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf clean expire-cache</pre></div>
</li>
<li class="li"><strong class="ph b">Install CUDA</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf install cuda</pre>
The CUDA driver installation may fail if
the RPMFusion non-free repository is enabled. In this case, CUDA installations should
temporarily disable the RPMFusion non-free repository:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf --disablerepo="rpmfusion-nonfree*" install cuda</pre></div>
<div class="p"> If a system has installed both packages with the same instance of <samp class="ph codeph">dnf</samp>,
some driver components may be missing. Such an installation can be corrected by running:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf install cuda-drivers</pre></div>
<div class="p"> If the i686 libvdpau package dependency fails to install, try using the following
steps to fix the issue:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> dnf download libvdpau.i686
<strong class="ph b">$</strong> sudo rpm -U --oldpackage libvdpau*.rpm</pre></div>
It may be necessary to rebuild the grub configuration files, particularly if you use a
non-default partition scheme. If so, then run this below command, and reboot the system:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo grub2-mkconfig -o /boot/grub2/grub.cfg</pre>
Remember to reboot the
system. </li>
<li class="li"><strong class="ph b">Add libcuda.so symbolic link, if necessary</strong><p class="p">
The libcuda.so library is installed in the /usr/lib{,64}/nvidia directory. For pre-existing projects which use libcuda.so,
it may be useful to add a symbolic link from libcuda.so in the /usr/lib{,64} directory.
</p>
</li>
<li class="li">
Perform the <a class="xref" href="index.html#post-installation-actions" shape="rect"> post-installation actions.</a></li>
</ol>
</div>
</div>
<div class="topic concept nested1" id="sles-installation"><a name="sles-installation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#sles-installation" name="sles-installation" shape="rect">3.4. SLES</a></h3>
<div class="body conbody">
<ol class="ol">
<li class="li liexpand">
Perform the <a class="xref" href="index.html#pre-installation-actions" shape="rect"> pre-installation actions.</a></li>
<li class="li liexpand">On SLES12 SP4, install the Mesa-libgl-devel Linux packages before proceeding. See <a class="xref" href="https://pkgs.org/download/Mesa-libGL-devel" target="_blank" shape="rect">Mesa-libGL-devel.</a></li>
<li class="li liexpand"><strong class="ph b">Install repository meta-data</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo rpm --install cuda-repo-<distro>-<version>.<architecture>.rpm</pre></div>
</li>
<li class="li liexpand"><strong class="ph b">Refresh Zypper repository cache</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper refresh</pre></div>
</li>
<li class="li liexpand"><strong class="ph b">Install CUDA</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper install cuda</pre></div>
</li>
<li class="li liexpand"><strong class="ph b">Add the user to the video group</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo usermod -a -G video <username></pre></div>
</li>
<li class="li liexpand"><strong class="ph b">Install CUDA Samples GL dependencies</strong><p class="p">
The CUDA Samples package on SLES does not include dependencies on GL and X11 libraries as these are provided in the SLES SDK.
These packages must be installed separately, depending on which samples you want to use.
</p>
</li>
<li class="li liexpand">
Perform the <a class="xref" href="index.html#post-installation-actions" shape="rect"> post-installation actions.</a></li>
</ol>
</div>
</div>
<div class="topic concept nested1" id="suse-installation"><a name="suse-installation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#suse-installation" name="suse-installation" shape="rect">3.5. OpenSUSE</a></h3>
<div class="body conbody">
<ol class="ol">
<li class="li">
Perform the <a class="xref" href="index.html#pre-installation-actions" shape="rect"> pre-installation actions.</a></li>
<li class="li"><strong class="ph b">Install repository meta-data</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo rpm --install cuda-repo-<distro>-<version>.<architecture>.rpm</pre></div>
</li>
<li class="li"><strong class="ph b">Refresh Zypper repository cache</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper refresh</pre></div>
</li>
<li class="li"><strong class="ph b">Install CUDA</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper install cuda</pre></div>
</li>
<li class="li"><strong class="ph b">Add the user to the video group</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo usermod -a -G video <username></pre></div>
</li>
<li class="li">
Perform the <a class="xref" href="index.html#post-installation-actions" shape="rect"> post-installation actions.</a></li>
</ol>
</div>
</div>
<div class="topic concept nested1" id="ubuntu-installation"><a name="ubuntu-installation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#ubuntu-installation" name="ubuntu-installation" shape="rect">3.6. Ubuntu</a></h3>
<div class="body conbody">
<ol class="ol">
<li class="li">
Perform the <a class="xref" href="index.html#pre-installation-actions" shape="rect"> pre-installation actions.</a></li>
<li class="li"><strong class="ph b">Install repository meta-data</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dpkg -i cuda-repo-<distro>_<version>_<architecture>.deb</pre></div>
</li>
<li class="li"><strong class="ph b">Installing the CUDA public GPG key</strong><p class="p">When installing using the local repo:</p>
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-key add /var/cuda-repo-<version>/7fa2af80.pub</pre></div>
<p class="p">When installing using network repo on Ubuntu 18.04/18.10:</p>
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/<distro>/<architecture>/7fa2af80.pub</pre></div>
<p class="p">When installing using network repo on Ubuntu 16.04:</p>
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/<distro>/<architecture>/7fa2af80.pub</pre></div>
</li>
<li class="li"><strong class="ph b">Update the Apt repository cache</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-get update</pre></div>
</li>
<li class="li"><strong class="ph b">Install CUDA</strong><div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-get install cuda</pre></div>
</li>
<li class="li">
Perform the <a class="xref" href="index.html#post-installation-actions" shape="rect"> post-installation actions.</a></li>
</ol>
</div>
</div>
<div class="topic concept nested1" id="package-manager-additional"><a name="package-manager-additional" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#package-manager-additional" name="package-manager-additional" shape="rect">3.7. Additional Package Manager Capabilities</a></h3>
<div class="body conbody">
<p class="p">Below are some additional capabilities of the package manager that users can take advantage of.</p>
</div>
<div class="topic concept nested2" id="package-manager-packages"><a name="package-manager-packages" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#package-manager-packages" name="package-manager-packages" shape="rect">3.7.1. Available Packages</a></h3>
<div class="body conbody">
<p class="p">
The recommended installation package is the <samp class="ph codeph">cuda</samp> package.
This package will install the full set of other CUDA packages required for native development and should cover most scenarios.
</p>
<p class="p">
The <samp class="ph codeph">cuda</samp> package installs all the available packages for native developments.
That includes the compiler, the debugger, the profiler, the math libraries,...
For x86_64 patforms, this also include NSight Eclipse Edition and the visual profiler
It also includes the NVIDIA driver package.
</p>
<p class="p">
On supported platforms, the <samp class="ph codeph">cuda-cross-armhf</samp>, <samp class="ph codeph">cuda-cross-aarch64</samp>, and <samp class="ph codeph">cuda-cross-ppc64el</samp>
packages install all the packages required for cross-platform development to ARMv7, ARMv8, and POWER8, respectively. The libraries
and header
files of the target architecture's display driver package are also installed to enable the cross compilation of driver applications.
The <samp class="ph codeph">cuda-cross-<arch></samp> packages do not install the native display driver.
</p>
<p class="p">
The packages installed by the packages above can also be installed individually by specifying their names explicitly.
The list of available packages be can obtained with:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> yum --disablerepo="*" --enablerepo="cuda*" list available # RedHat
<strong class="ph b">$</strong> dnf --disablerepo="*" --enablerepo="cuda*" list available # Fedora
<strong class="ph b">$</strong> zypper packages -r cuda # OpenSUSE & SLES
<strong class="ph b">$</strong> cat /var/lib/apt/lists/*cuda*Packages | grep "Package:" # Ubuntu</pre></div>
</div>
<div class="topic concept nested2" id="package-manager-upgrades"><a name="package-manager-upgrades" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#package-manager-upgrades" name="package-manager-upgrades" shape="rect">3.7.2. Package Upgrades</a></h3>
<div class="body conbody">
<p class="p">
The <samp class="ph codeph">cuda</samp> package points to the latest stable release of the CUDA Toolkit.
When a new version is available, use the following commands to upgrade the toolkit and driver:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum install cuda # RedHat
<strong class="ph b">$</strong> sudo dnf install cuda # Fedora
<strong class="ph b">$</strong> sudo zypper install cuda # OpenSUSE & SLES
<strong class="ph b">$</strong> sudo apt-get install cuda # Ubuntu</pre><p class="p">
The <samp class="ph codeph">cuda-cross-<arch></samp> packages can also be upgraded in the same manner.
</p>
<p class="p">
The cuda-drivers package points to the latest driver release available in the CUDA repository.
When a new version is available, use the following commands to upgrade the driver:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum install cuda-drivers # RedHat
<strong class="ph b">$</strong> sudo dnf install cuda-drivers # Fedora
<strong class="ph b">$</strong> sudo zypper install cuda-drivers \
nvidia-gfxG04-kmp-default # OpenSUSE & SLES
<strong class="ph b">$</strong> sudo apt-get install cuda-drivers # Ubuntu</pre><p class="p">
Some desktop environments, such as GNOME or KDE, will display an notification alert when new packages are available.
</p>
<p class="p">
To avoid any automatic upgrade, and lock down the toolkit installation to the X.Y release, install the <samp class="ph codeph">cuda-X-Y</samp> or <samp class="ph codeph">cuda-cross-<arch>-X-Y</samp> package.
</p>
<p class="p">
Side-by-side installations are supported.
For instance, to install both the X.Y CUDA Toolkit and the X.Y+1 CUDA Toolkit, install the <samp class="ph codeph">cuda-X.Y</samp> and <samp class="ph codeph">cuda-X.Y+1</samp> packages.
</p>
</div>
</div>
<div class="topic concept nested2" id="package-manager-metas"><a name="package-manager-metas" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#package-manager-metas" name="package-manager-metas" shape="rect">3.7.3. Meta Packages</a></h3>
<div class="body conbody">
<p class="p">
Meta packages are RPM/Deb packages which contain no (or few) files but have multiple dependencies. They are used to install
many CUDA packages when you may not know the details of the packages you want. Below is the list of meta packages.
</p>
<div class="tablenoborder">
<table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
<caption><span class="tablecap">Table 4. Meta Packages Available for CUDA <span class="keyword">10.1</span></span></caption>
<thead class="thead" align="left">
<tr class="row">
<th class="entry" align="center" valign="top" id="d54e1849" rowspan="1" colspan="1">Meta Package</th>
<th class="entry" align="center" valign="top" id="d54e1852" rowspan="1" colspan="1">Purpose</th>
</tr>
</thead>
<tbody class="tbody">
<tr class="row">
<td class="entry" valign="top" headers="d54e1849" rowspan="1" colspan="1">cuda</td>
<td class="entry" valign="top" headers="d54e1852" rowspan="1" colspan="1">Installs all CUDA Toolkit and Driver packages. Handles upgrading to the next version of the <samp class="ph codeph">cuda</samp> package when it's released.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" headers="d54e1849" rowspan="1" colspan="1">cuda-<span class="keyword">10</span>-<span class="keyword">1</span></td>
<td class="entry" valign="top" headers="d54e1852" rowspan="1" colspan="1">Installs all CUDA Toolkit and Driver packages. Remains at version <span class="keyword">10.1</span> until an additional version of CUDA is installed.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" headers="d54e1849" rowspan="1" colspan="1">cuda-toolkit-<span class="keyword">10</span>-<span class="keyword">1</span></td>
<td class="entry" valign="top" headers="d54e1852" rowspan="1" colspan="1">Installs all CUDA Toolkit packages required to develop CUDA applications. Does not include the driver.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" headers="d54e1849" rowspan="1" colspan="1">cuda-tools-<span class="keyword">10</span>-<span class="keyword">1</span></td>
<td class="entry" valign="top" headers="d54e1852" rowspan="1" colspan="1">Installs all CUDA command line and visual tools.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" headers="d54e1849" rowspan="1" colspan="1">cuda-runtime-<span class="keyword">10</span>-<span class="keyword">1</span></td>
<td class="entry" valign="top" headers="d54e1852" rowspan="1" colspan="1">Installs all CUDA Toolkit packages required to run CUDA applications, as well as the Driver packages.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" headers="d54e1849" rowspan="1" colspan="1">cuda-compiler-<span class="keyword">10</span>-<span class="keyword">1</span></td>
<td class="entry" valign="top" headers="d54e1852" rowspan="1" colspan="1">Installs all CUDA compiler packages.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" headers="d54e1849" rowspan="1" colspan="1">cuda-libraries-<span class="keyword">10</span>-<span class="keyword">1</span></td>
<td class="entry" valign="top" headers="d54e1852" rowspan="1" colspan="1">Installs all runtime CUDA Library packages.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" headers="d54e1849" rowspan="1" colspan="1">cuda-libraries-dev-<span class="keyword">10</span>-<span class="keyword">1</span></td>
<td class="entry" valign="top" headers="d54e1852" rowspan="1" colspan="1">Installs all development CUDA Library packages.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" headers="d54e1849" rowspan="1" colspan="1">cuda-drivers</td>
<td class="entry" valign="top" headers="d54e1852" rowspan="1" colspan="1">Installs all Driver packages. Handles upgrading to the next version of the Driver packages when they're released.</td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
</div>
</div>
<div class="topic concept nested0" id="runfile"><a name="runfile" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#runfile" name="runfile" shape="rect">4. Runfile Installation</a></h2>
<div class="body conbody">
<p class="p">
Basic instructions can be found in the <a class="xref" href="http://docs.nvidia.com/cuda/cuda-quick-start-guide/index.html#linux" target="_blank" shape="rect">Quick Start Guide</a>. Read on for more detailed instructions.
</p>
<p class="p">
This section describes the installation and configuration of CUDA when using the standalone installer. The standalone installer
is a ".run" file and is completely self-contained.
</p>
</div>
<div class="topic concept nested1" id="runfile-overview"><a name="runfile-overview" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-overview" name="runfile-overview" shape="rect">4.1. Overview</a></h3>
<div class="body conbody">
<p class="p">The Runfile installation installs the NVIDIA Driver, the CUDA Toolkit, and CUDA Samples, via an
interactive ncurses-based interface.
</p>
<p class="p">The <a class="xref" href="index.html#runfile-installation" shape="rect">installation steps</a> are listed below.
Distribution-specific instructions for <a class="xref" href="index.html#runfile-nouveau" shape="rect">disabling the
Nouveau drivers</a>, and the steps for <a class="xref" href="index.html#runfile-verifications" shape="rect">verifying device node creation</a>, are also provided.
</p>
<p class="p">Finally, the <a class="xref" href="index.html#runfile-advanced" shape="rect">advanced options</a> for the installer and the
<a class="xref" href="index.html#runfile-uninstallation" shape="rect">uninstallation steps</a> are detailed below.
</p>
<p class="p">The Runfile installation does not include support for cross-platform development. For
cross-platform development, see the <a class="xref" href="index.html#cross-platform" shape="rect">CUDA Cross-Platform
Environment</a> section.
</p>
</div>
</div>
<div class="topic concept nested1" id="runfile-installation"><a name="runfile-installation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-installation" name="runfile-installation" shape="rect">4.2. Installation</a></h3>
<div class="body conbody">
<ol class="ol">
<li class="li liexpand">
<p class="p">
Perform the <a class="xref" href="index.html#pre-installation-actions" shape="rect">pre-installation actions</a>.
</p>
</li>
<li class="li liexpand">
<p class="p"><a class="xref" href="index.html#runfile-nouveau" shape="rect">Disable the Nouveau drivers</a>.
</p>
</li>
<li class="li liexpand">
<p class="p">
Reboot into text mode (runlevel 3).
</p>
<p class="p">
This can usually be accomplished by adding the number "3" to the end of the system's kernel boot parameters.
</p>
<p class="p">
Since the NVIDIA drivers are not yet installed, the text terminals may not display correctly. Temporarily adding "nomodeset"
to the system's kernel boot parameters may fix this issue.
</p>
<p class="p">
Consult your system's bootloader documentation for information on how to make the above boot parameter changes.
</p>
<p class="p">
The reboot is required to completely unload the Nouveau drivers and prevent the graphical interface from loading. The CUDA
driver cannot be installed while the Nouveau drivers are loaded or while the graphical interface is active.
</p>
</li>
<li class="li liexpand">
<p class="p">
Verify that the Nouveau drivers are not loaded.
If the Nouveau drivers are still loaded, consult your distribution's documentation to see if further steps are needed to disable
Nouveau.
</p>
</li>
<li class="li liexpand">
<div class="p"> Run the installer and follow the on-screen prompts:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo sh cuda_<version>_linux.run</pre></div>
<p class="p">See <a class="xref" href="index.html#runfile-installer-ui" shape="rect">Installer UI</a> for navigating the ncurses-based installer
UI.
</p>
<p class="p">As of CUDA 10.1 some libraries will be installed in the system standard locations
rather than in the Toolkit installation directory. Depending on your distribution
these installed locations can be either:<samp class="ph codeph"> /usr/lib/x84_64-linux-gnu</samp>,
or <samp class="ph codeph">/usr/lib64</samp>, or <samp class="ph codeph">/usr/lib</samp>. See the <a class="xref" href="index.html#runfile-advanced" shape="rect">Advanced Options</a> for how to change this location.
</p>
<div class="p"> The default installation locations for the toolkit and samples are:
<div class="tablenoborder"><a name="runfile-installation__table_q5k_kxh_bhb" shape="rect">
<!-- --></a><table cellpadding="4" cellspacing="0" summary="" id="runfile-installation__table_q5k_kxh_bhb" class="table" frame="border" border="1" rules="all">
<thead class="thead" align="left">
<tr class="row">
<th class="entry" valign="top" id="d54e2139" rowspan="1" colspan="1">Component</th>
<th class="entry" valign="top" id="d54e2142" rowspan="1" colspan="1">Default Installation Directory</th>
</tr>
</thead>
<tbody class="tbody">
<tr class="row">
<td class="entry" valign="top" headers="d54e2139" rowspan="1" colspan="1">CUDA Toolkit</td>
<td class="entry" valign="top" headers="d54e2142" rowspan="1" colspan="1">/usr/local/cuda-<span class="keyword">10.1</span></td>
</tr>
<tr class="row">
<td class="entry" valign="top" headers="d54e2139" rowspan="1" colspan="1">CUDA Samples</td>
<td class="entry" valign="top" headers="d54e2142" rowspan="1" colspan="1">$(HOME)/NVIDIA_CUDA-<span class="keyword">10.1</span>_Samples
</td>
</tr>
</tbody>
</table>
</div>
</div>
<p class="p">
The <samp class="ph codeph">/usr/local/cuda</samp> symbolic link points to the location where the CUDA Toolkit was installed. This link allows projects to use the latest CUDA
Toolkit without any configuration file update.
</p>
<div class="p"> The installer must be executed with sufficient privileges to perform some actions.
When the current privileges are insufficient to perform an action, the installer will
ask for the user's password to attempt to install with root privileges. Actions that
cause the installer to attempt to install with root privileges are: <a name="runfile-installation__ul_owm_5xh_bhb" shape="rect">
<!-- --></a><ul class="ul" id="runfile-installation__ul_owm_5xh_bhb">
<li class="li liexpand">installing the CUDA Driver</li>
<li class="li liexpand">installing the CUDA Toolkit to a location the user does not have permission to
write to
</li>
<li class="li liexpand">installing the CUDA Samples to a location the user does not have permission to
write to
</li>
<li class="li liexpand">creating the <samp class="ph codeph">/usr/local/cuda</samp> symbolic link
</li>
</ul>
</div>
<p class="p">Running the installer with <strong class="ph b">sudo</strong>, as shown above, will give permission to
install to directories that require root permissions. Directories and files created
while running the installer with <strong class="ph b">sudo</strong> will have root ownership.
</p>
<p class="p"> If installing the driver, the installer will also ask if the openGL libraries should
be installed. If the GPU used for display is not an NVIDIA GPU, the NVIDIA openGL
libraries should not be installed. Otherwise, the openGL libraries used by the
graphics driver of the non-NVIDIA GPU will be overwritten and the GUI will not work.
If performing a silent installation, the <samp class="ph codeph">--no-opengl-libs</samp> option
should be used to prevent the openGL libraries from being installed. See the <a class="xref" href="index.html#runfile-advanced" shape="rect">Advanced Options</a> section for more details.
</p>
<p class="p">
If the GPU used for display is an NVIDIA GPU, the X server configuration file, <samp class="ph codeph">/etc/X11/xorg.conf</samp>, may need to be modified. In some cases, <samp class="ph codeph">nvidia-xconfig</samp> can be used to automatically generate a <samp class="ph codeph">xorg.conf</samp> file that works for the system. For non-standard systems, such as those with more than one GPU, it is recommended to manually
edit the <samp class="ph codeph">xorg.conf</samp> file. Consult the <samp class="ph codeph">xorg.conf</samp> documentation for more information.
</p>
<div class="note note"><span class="notetitle">Note:</span>
Installing Mesa may overwrite the <samp class="ph codeph">/usr/lib/libGL.</samp>so that was previously installed by the NVIDIA driver, so a reinstallation of the NVIDIA driver might be required after installing
these libraries.
</div>
</li>
<li class="li liexpand">
<p class="p">
Reboot the system to reload the graphical interface.
</p>
</li>
<li class="li liexpand">
<p class="p">
Verify the <a class="xref" href="index.html#runfile-verifications" shape="rect">device nodes</a> are created properly.
</p>
</li>
<li class="li liexpand">
<p class="p">
Perform the <a class="xref" href="index.html#post-installation-actions" shape="rect">post-installation actions</a>.
</p>
</li>
</ol>
</div>
</div>
<div class="topic concept nested1" id="runfile-installer-ui"><a name="runfile-installer-ui" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-installer-ui" name="runfile-installer-ui" shape="rect">4.3. Installer UI</a></h3>
<div class="body conbody">
<p class="p">The installer UI has three main states:</p>
<ol class="ol">
<li class="li liexpand"><strong class="ph b">EULA Acceptance.</strong><a name="runfile-installer-ui__ol_bpj_t5h_bhb" shape="rect">
<!-- --></a><ol class="ol" type="a" id="runfile-installer-ui__ol_bpj_t5h_bhb">
<li class="li">Scroll through the EULA using the arrow keys, the page up/down keys, or a scroll
wheel.
</li>
</ol>
</li>
<li class="li liexpand"><strong class="ph b">Component Selection.</strong><a name="runfile-installer-ui__ol_tff_x5h_bhb" shape="rect">
<!-- --></a><ol class="ol" type="a" id="runfile-installer-ui__ol_tff_x5h_bhb">
<li class="li">Navigate the menu using the arrow keys. The left/right keys will expand/collapse
sub-elements.
</li>
<li class="li">Select or deselect items to install by pressing the spacebar or enter key with
the cursor on that item.
</li>
<li class="li">With the cursor over an item with advanced options available, press 'A' to see
that options menu. This is currently available for CUDA Toolkit and CUDA Samples
items only.
</li>
</ol>
</li>
<li class="li liexpand"><strong class="ph b">Advanced Options.</strong><a name="runfile-installer-ui__ol_x21_gvh_bhb" shape="rect">
<!-- --></a><ol class="ol" type="a" id="runfile-installer-ui__ol_x21_gvh_bhb">
<li class="li">Options such as setting the install path for a specific component are available
here.
</li>
</ol>
</li>
</ol>
</div>
</div>
<div class="topic concept nested1" id="runfile-nouveau"><a name="runfile-nouveau" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-nouveau" name="runfile-nouveau" shape="rect">4.4. Disabling Nouveau</a></h3>
<div class="body conbody">
<p class="p">
To install the Display Driver, the Nouveau drivers must first be disabled. Each distribution of Linux has a different method
for disabling Nouveau.
</p>
<div class="p">
The Nouveau drivers are loaded if the following command prints anything:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> lsmod | grep nouveau</pre></div>
</div>
<div class="topic concept nested2" id="runfile-nouveau-fedora"><a name="runfile-nouveau-fedora" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-nouveau-fedora" name="runfile-nouveau-fedora" shape="rect">4.4.1. Fedora</a></h3>
<div class="body conbody">
<div class="p">
<ol class="ol">
<li class="li">
Create a file at <samp class="ph codeph">/usr/lib/modprobe.d/blacklist-nouveau.conf</samp> with the following contents:
<pre class="pre screen" xml:space="preserve">blacklist nouveau
options nouveau modeset=0</pre></li>
<li class="li">
Regenerate the kernel initramfs:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dracut --force</pre></li>
<li class="li">
Run the below command:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo grub2-mkconfig -o /boot/grub2/grub.cfg</pre></li>
<li class="li">
Reboot the system.
</li>
</ol>
</div>
</div>
</div>
<div class="topic concept nested2" id="runfile-nouveau-rhel"><a name="runfile-nouveau-rhel" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-nouveau-rhel" name="runfile-nouveau-rhel" shape="rect">4.4.2. RHEL/CentOS</a></h3>
<div class="body conbody">
<div class="p">
<ol class="ol">
<li class="li">
Create a file at <samp class="ph codeph">/etc/modprobe.d/blacklist-nouveau.conf</samp> with the following contents:
<pre class="pre screen" xml:space="preserve">blacklist nouveau
options nouveau modeset=0</pre></li>
<li class="li">
Regenerate the kernel initramfs:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dracut --force</pre></li>
</ol>
</div>
</div>
</div>
<div class="topic concept nested2" id="runfile-nouveau-suse"><a name="runfile-nouveau-suse" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-nouveau-suse" name="runfile-nouveau-suse" shape="rect">4.4.3. OpenSUSE</a></h3>
<div class="body conbody">
<div class="p">
<ol class="ol">
<li class="li">
Create a file at <samp class="ph codeph">/etc/modprobe.d/blacklist-nouveau.conf</samp> with the following contents:
<pre class="pre screen" xml:space="preserve">blacklist nouveau
options nouveau modeset=0</pre></li>
<li class="li">
Regenerate the kernel initrd:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo /sbin/mkinitrd</pre></li>
</ol>
</div>
</div>
</div>
<div class="topic concept nested2" id="runfile-nouveau-sles"><a name="runfile-nouveau-sles" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-nouveau-sles" name="runfile-nouveau-sles" shape="rect">4.4.4. SLES</a></h3>
<div class="body conbody">
<p class="p">
No actions to disable Nouveau are required as Nouveau is not installed on SLES.
</p>
</div>
</div>
<div class="topic concept nested2" id="runfile-nouveau-ubuntu"><a name="runfile-nouveau-ubuntu" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-nouveau-ubuntu" name="runfile-nouveau-ubuntu" shape="rect">4.4.5. Ubuntu</a></h3>
<div class="body conbody">
<div class="p">
<ol class="ol">
<li class="li">
Create a file at <samp class="ph codeph">/etc/modprobe.d/blacklist-nouveau.conf</samp> with the following contents:
<pre class="pre screen" xml:space="preserve">blacklist nouveau
options nouveau modeset=0</pre></li>
<li class="li">
Regenerate the kernel initramfs:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo update-initramfs -u</pre></li>
</ol>
</div>
</div>
</div>
</div>
<div class="topic concept nested1" id="runfile-verifications"><a name="runfile-verifications" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-verifications" name="runfile-verifications" shape="rect">4.5. Device Node Verification</a></h3>
<div class="body conbody">
<p class="p">
Check that the device files<samp class="ph codeph">/dev/nvidia*</samp> exist and have the correct (0666) file permissions.
These files are used by the CUDA Driver to communicate with the kernel-mode portion of the NVIDIA Driver.
Applications that use the NVIDIA driver, such as a CUDA application or the X server (if any), will normally automatically
create these files
if they are missing using the <dfn class="term">setuid</dfn><samp class="ph codeph">nvidia-modprobe</samp> tool that is bundled with the NVIDIA Driver.
However, some systems disallow setuid binaries, so if these files do not exist, you can create them manually
by using a startup script such as the one below:
</p><pre xml:space="preserve">#!/bin/bash
/sbin/modprobe nvidia
if [ "$?" -eq 0 ]; then
# Count the number of NVIDIA controllers found.
NVDEVS=`lspci | grep -i NVIDIA`
N3D=`echo "$NVDEVS" | grep "3D controller" | wc -l`
NVGA=`echo "$NVDEVS" | grep "VGA compatible controller" | wc -l`
N=`expr $N3D + $NVGA - 1`
for i in `seq 0 $N`; do
mknod -m 666 /dev/nvidia$i c 195 $i
done
mknod -m 666 /dev/nvidiactl c 195 255
else
exit 1
fi
/sbin/modprobe nvidia-uvm
if [ "$?" -eq 0 ]; then
# Find out the major device number used by the nvidia-uvm driver
D=`grep nvidia-uvm /proc/devices | awk '{print $1}'`
mknod -m 666 /dev/nvidia-uvm c $D 0
else
exit 1
fi</pre></div>
</div>
<div class="topic concept nested1" id="runfile-advanced"><a name="runfile-advanced" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-advanced" name="runfile-advanced" shape="rect">4.6. Advanced Options</a></h3>
<div class="body conbody">
<div class="tablenoborder">
<table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
<thead class="thead" align="left">
<tr class="row">
<th class="entry" valign="top" width="18.181818181818183%" id="d54e2540" rowspan="1" colspan="1">Action</th>
<th class="entry" valign="top" width="27.27272727272727%" id="d54e2543" rowspan="1" colspan="1">Options Used</th>
<th class="entry" valign="top" width="54.54545454545454%" id="d54e2546" rowspan="1" colspan="1">Explanation</th>
</tr>
</thead>
<tbody class="tbody">
<tr class="row">
<td class="entry" rowspan="7" valign="top" width="18.181818181818183%" headers="d54e2540" colspan="1">Silent Installation</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--silent</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Required for any silent installation. Performs an installation with no
further user-input and minimal command-line output based on the options
provided below. Silent installations are useful for scripting the installation
of CUDA. Using this option implies acceptance of the EULA. The following flags
can be used to customize the actions taken during installation. At least one of
--driver, --uninstall, --toolkit, and --samples must be passed if running with
non-root permissions.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--driver</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Install the CUDA Driver.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--toolkit</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Install the CUDA Toolkit.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--toolkitpath=<path></td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Install the CUDA Toolkit to the <path> directory. If not provided, the
default path of <samp class="ph codeph">/usr/local/cuda-<span class="keyword">10.1</span></samp> is used.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--samples</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Install the CUDA Samples.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--samplespath=<path></td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Install the CUDA Samples to the <path> directory. If not provided, the
default path of <samp class="ph codeph">$(HOME)/NVIDIA_CUDA-<span class="keyword">10.1</span>_Samples</samp> is used.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--defaultroot=<path></td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Install libraries to the <path> directory. If the <path> is not
provided, then the default path of your distribution is used. <em class="ph i">This only
applies to the libraries installed outside of the CUDA Toolkit
path.</em></td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="18.181818181818183%" headers="d54e2540" rowspan="1" colspan="1">Extraction</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--extract=<path></td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">
<p class="p"> Extracts to the <path> the following: the driver runfile, the raw files
of the toolkit and samples to <path>.
</p>
<p class="p"> This is especially useful when one wants to install the driver using one or
more of the command-line options provided by the driver installer which are
not exposed in this installer.
</p>
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="18.181818181818183%" headers="d54e2540" rowspan="1" colspan="1">Overriding Installation Checks</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--override</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Ignores compiler, third-party library, and toolkit detection checks which
would prevent the CUDA Toolkit and CUDA Samples from installing.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="18.181818181818183%" headers="d54e2540" rowspan="1" colspan="1">No OpenGL Libraries</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--no-opengl-libs</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Prevents the driver installation from installing NVIDIA's GL libraries.
Useful for systems where the display is driven by a non-NVIDIA GPU. In such
systems, NVIDIA's GL libraries could prevent X from loading properly.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="18.181818181818183%" headers="d54e2540" rowspan="1" colspan="1">No man pages</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--no-man-page</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Do not install the man pages under /usr/share/man.</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="18.181818181818183%" headers="d54e2540" rowspan="1" colspan="1">Overriding Kernel Source</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--kernel-source-path=<path></td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Tells the driver installation to use <path> as the kernel source
directory when building the NVIDIA kernel module. Required for systems where
the kernel source is installed to a non-standard location.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="18.181818181818183%" headers="d54e2540" rowspan="1" colspan="1">Running nvidia-xconfig</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--run-nvidia-xconfig</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Tells the driver installation to run nvidia-xconfig to update the system X
configuration file so that the NVIDIA X driver is used. The pre-existing X
configuration file will be backed up.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="18.181818181818183%" headers="d54e2540" rowspan="1" colspan="1">No nvidia-drm kernel module</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--no-drm</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Do not install the nvidia-drm kernel module. This option should only be
used to work around failures to build or install the nvidia-drm kernel module
on systems that do not need the provided features.
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="18.181818181818183%" headers="d54e2540" rowspan="1" colspan="1">Custom Temporary Directory Selection</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--tmpdir=<path></td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Performs any temporary actions within <path> instead of /tmp. Useful in
cases where /tmp cannot be used (doesn't exist, is full, is mounted with
'noexec', etc.).
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="18.181818181818183%" headers="d54e2540" rowspan="1" colspan="1">Show Installer Options</td>
<td class="entry" valign="top" width="27.27272727272727%" headers="d54e2543" rowspan="1" colspan="1">--help</td>
<td class="entry" valign="top" width="54.54545454545454%" headers="d54e2546" rowspan="1" colspan="1">Prints the list of command-line options to stdout.</td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
<div class="topic concept nested1" id="runfile-uninstallation"><a name="runfile-uninstallation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#runfile-uninstallation" name="runfile-uninstallation" shape="rect">4.7. Uninstallation</a></h3>
<div class="body conbody">
<div class="p">To uninstall the CUDA Toolkit, run the uninstallation script provided in the bin directory of the
toolkit. By default, it is located in <samp class="ph codeph">/usr/local/cuda-<span class="keyword">10.1</span>/bin</samp>:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo /usr/local/cuda-<span class="keyword">10.1</span>/bin/cuda-uninstaller</pre></div>
<div class="p">To uninstall the NVIDIA Driver, run <samp class="ph codeph">nvidia-uninstall</samp>:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo /usr/bin/nvidia-uninstall</pre>
To enable the Nouveau drivers, remove the blacklist file created in the <a class="xref" href="index.html#runfile-nouveau" shape="rect">Disabling Nouveau</a> section, and regenerate the kernel initramfs/initrd again as described in that section.
</div>
</div>
</div>
</div>
<div class="topic concept nested0" id="cluster"><a name="cluster" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#cluster" name="cluster" shape="rect">5. Cluster Management Packages</a></h2>
<div class="body conbody"></div>
<div class="topic concept nested1" id="cluster-overview"><a name="cluster-overview" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cluster-overview" name="cluster-overview" shape="rect">5.1. Overview</a></h3>
<div class="body conbody">
<p class="p">
Cluster management packages are provided as an alternative set of RPM and Deb packages intended to be used by deployment management
tools as standalone packages. These packages are available for RHEL 6, RHEL 7, Ubuntu 14.04, and Ubuntu 16.04 on the x86_64
architecture. There are three parts to the cluster management packages: the CUDA toolkit packages, the NVIDIA driver packages,
and the README.
</p>
<p class="p">
The cluster management toolkit packages are split into a runtime package, <samp class="ph codeph">cuda-cluster-runtime-<span class="keyword">10</span>-<span class="keyword">1</span></samp>, and a development package, <samp class="ph codeph">cuda-cluster-devel-<span class="keyword">10</span>-<span class="keyword">1</span></samp>. The development package depends on the runtime package. The driver packages are the same as what is provided in the general
RPM and Deb solution used in the <a class="xref" href="index.html#package-manager-installation" shape="rect">Package Management</a> section.
</p>
<p class="p">
The README describes the package and kernel source dependencies of the cluster management packages. The README also describes
the order of installation of the standalone cluster management packages.
</p>
</div>
</div>
</div>
<div class="topic concept nested0" id="cross-platform"><a name="cross-platform" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#cross-platform" name="cross-platform" shape="rect">6. CUDA Cross-Platform Environment</a></h2>
<div class="body conbody">
<p class="p">
Cross-platform development is only supported on Ubuntu systems, and is only provided via the Package Manager installation
process.
</p>
<p class="p">
We recommend selecting Ubuntu 14.04 as your cross-platform development environment. This selection helps prevent host/target
incompatibilities, such as GCC or GLIBC version mismatches.
</p>
</div>
<div class="topic concept nested1" id="cross-installation"><a name="cross-installation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cross-installation" name="cross-installation" shape="rect">6.1. CUDA Cross-Platform Installation</a></h3>
<div class="body conbody">
<p class="p">
Some of the following steps may have already been performed as part of the <a class="xref" href="index.html#ubuntu-installation" shape="rect">native Ubuntu installation</a>. Such steps can safely be skipped.
</p>
<p class="p">
These steps should be performed on the x86_64 host system, rather than the target system. To install the native CUDA Toolkit
on the target system, refer to the native <a class="xref" href="index.html#ubuntu-installation" shape="rect">Ubuntu</a> installation section.
</p>
<ol class="ol">
<li class="li liexpand">
Perform the <a class="xref" href="index.html#pre-installation-actions" shape="rect"> pre-installation actions.</a></li>
<li class="li liexpand">Install repository meta-data package with:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dpkg -i cuda-repo-cross-<identifier>_all.deb </pre><p class="p">where
<samp class="ph codeph"><identifier></samp> indicates the operating system, architecture,
and/or the version of the package.
</p>
</li>
<li class="li liexpand">Update the Apt repository cache:
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-get update</pre></div>
</li>
<li class="li liexpand">Install the appropriate cross-platform CUDA Toolkit: <a name="cross-installation__ol_xhf_tyl_fhb" shape="rect">
<!-- --></a><ol class="ol" type="a" id="cross-installation__ol_xhf_tyl_fhb">
<li class="li">For aarch64:
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-get install cuda-cross-aarch64</pre></div>
</li>
<li class="li">For QNX:<pre class="pre screen" xml:space="preserve">$ sudo apt-get install cuda-cross-qnx</pre></li>
</ol>
</li>
<li class="li liexpand">
Perform the <a class="xref" href="index.html#post-installation-actions" shape="rect"> post-installation actions.</a></li>
</ol>
</div>
</div>
<div class="topic concept nested1" id="cross-samples"><a name="cross-samples" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cross-samples" name="cross-samples" shape="rect">6.2. CUDA Cross-Platform Samples</a></h3>
<div class="body conbody">
<p class="p">
This section describes the options used to build cross-platform samples. <samp class="ph codeph">TARGET_ARCH=<arch></samp> and <samp class="ph codeph">TARGET_OS=<os></samp> should be chosen based on the supported targets shown below. <samp class="ph codeph">TARGET_FS=<path></samp> can be used to point nvcc to libraries and headers used by the sample.
</p>
<div class="tablenoborder">
<table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
<caption><span class="tablecap">Table 5. Supported Target Arch/OS Combinations</span></caption>
<tbody class="tbody">
<tr class="row">
<td class="entry" rowspan="2" colspan="2" align="center" valign="top"> </td>
<td class="entry" colspan="4" align="center" valign="top" rowspan="1">TARGET OS</td>
</tr>
<tr class="row">
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">linux</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">darwin</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">android</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">qnx</td>
</tr>
<tr class="row">
<td class="entry" rowspan="2" align="center" valign="middle" colspan="1">TARGET ARCH</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">x86_64</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">YES</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">YES</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">NO</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">NO</td>
</tr>
<tr class="row">
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">aarch64</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">YES</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">NO</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">YES</td>
<td class="entry" align="center" valign="top" rowspan="1" colspan="1">YES</td>
</tr>
</tbody>
</table>
</div>
</div>
<div class="topic concept nested2" id="targetarch"><a name="targetarch" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#targetarch" name="targetarch" shape="rect">TARGET_ARCH</a></h3>
<div class="body conbody">
<div class="p">
The target architecture must be specified when cross-compiling applications. If not specified, it defaults to the host architecture.
Allowed architectures are:
<ul class="ul">
<li class="li"><samp class="ph codeph">x86_64</samp> - 64-bit x86 CPU architecture
</li>
<li class="li"><samp class="ph codeph">aarch64</samp> - 64-bit ARM CPU architecture, like that found on Jetson TX1 onwards
</li>
</ul>
</div>
</div>
</div>
<div class="topic concept nested2" id="targetos"><a name="targetos" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#targetos" name="targetos" shape="rect">TARGET_OS</a></h3>
<div class="body conbody">
<div class="p">
The target OS must be specified when cross-compiling applications. If not specified, it defaults to the host OS. Allowed OSes
are:
<ul class="ul">
<li class="li"><samp class="ph codeph">linux</samp> - for any Linux distributions
</li>
<li class="li"><samp class="ph codeph">darwin</samp> - for Mac OS X
</li>
<li class="li"><samp class="ph codeph">android</samp> - for any supported device running Android
</li>
<li class="li"><samp class="ph codeph">qnx</samp> - for any supported device running QNX
</li>
</ul>
</div>
</div>
</div>
<div class="topic concept nested2" id="targetfs"><a name="targetfs" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#targetfs" name="targetfs" shape="rect">TARGET_FS</a></h3>
<div class="body conbody">
<div class="p">
The most reliable method to cross-compile the CUDA Samples is to use the TARGET_FS variable. To do so, mount the target's
filesystem on the host, say at <samp class="ph codeph">/mnt/target</samp>. This is typically done using <samp class="ph codeph">exportfs</samp>. In cases where <samp class="ph codeph">exportfs</samp> is unavailable, it is sufficient to copy the target's filesystem to <samp class="ph codeph">/mnt/target</samp>. To cross-compile a sample, execute:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> make TARGET_ARCH=<arch> TARGET_OS=<os> TARGET_FS=/mnt/target</pre></div>
</div>
<div class="topic concept nested3" id="cross-compile-to-arm-architectures"><a name="cross-compile-to-arm-architectures" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cross-compile-to-arm-architectures" name="cross-compile-to-arm-architectures" shape="rect">Cross Compiling to ARM architectures</a></h3>
<div class="body conbody">
<div class="p">
While cross compiling the samples from x86_64 installation to ARM architectures, that is, <samp class="ph codeph">aarch64</samp> or <samp class="ph codeph">armv7l</samp>, if you intend to run the executable on tegra GPU then <samp class="ph codeph">SMS</samp> variable need to override SM architectures to the tegra GPU through <samp class="ph codeph">SMS=<TEGRA_GPU_SM_ARCH></samp>, where <samp class="ph codeph"><TEGRA_GPU_SM_ARCH></samp> is the SM architecture of tegra GPU on which you want the generated binary to run on. For instance it can be <samp class="ph codeph">SMS="32 53 62"</samp>. Note you can also add SM arch of discrete GPU to this list <TEGRA_GPU_SM_ARCH> if you intend to run on embedded board having
discrete GPU as well. To cross compile a sample, execute:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> make TARGET_ARCH=<arch> TARGET_OS=<os> SMS=<TEGRA_GPU_SM_ARCHS> TARGET_FS=/mnt/target</pre></div>
</div>
</div>
<div class="topic concept nested3" id="copy-libs"><a name="copy-libs" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#copy-libs" name="copy-libs" shape="rect">Copying Libraries</a></h3>
<div class="body conbody">
<div class="p">
If the TARGET_FS option is not available, the libraries used should be copied from the target system to the host system, say
at <samp class="ph codeph">/opt/target/libs</samp>. If the sample uses GL, the GL headers must also be copied, say at <samp class="ph codeph">/opt/target/include</samp>. The linker must then be told where the libraries are with the <samp class="ph codeph">-rpath-link</samp> and/or <samp class="ph codeph">-L</samp> options. To ignore unresolved symbols from some libraries, use the <samp class="ph codeph">--unresolved-symbols</samp> option as shown below. <samp class="ph codeph">SAMPLE_ENABLED</samp> should be used to force the sample to build. For example, to cross-compile a sample which uses such libraries, execute:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> make TARGET_ARCH=<arch> TARGET_OS=<os> \
EXTRA_LDFLAGS="-rpath-link=/opt/target/libs -L/opt/target/libs --unresolved-symbols=ignore-in-shared-libs" \
EXTRA_CCFLAGS="-I /opt/target/include" \
SAMPLE_ENABLED=1</pre></div>
</div>
</div>
</div>
</div>
<div class="topic concept nested1" id="cross-nsight"><a name="cross-nsight" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cross-nsight" name="cross-nsight" shape="rect">6.3. Nsight Eclipse Edition</a></h3>
<div class="body conbody">
<p class="p">
Nsight Eclipse Edition supports cross-platform development. See the <a class="xref" href="http://docs.nvidia.com/cuda/nsight-eclipse-edition-getting-started-guide/index.html#remote-development" target="_blank" shape="rect">Nsight Eclipse Edition Getting Started Guide</a> for more details.
</p>
</div>
</div>
</div>
<div class="topic concept nested0" id="post-installation-actions"><a name="post-installation-actions" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#post-installation-actions" name="post-installation-actions" shape="rect">7. Post-installation Actions</a></h2>
<div class="body conbody">
<p class="p">
The post-installation actions must be manually performed. These actions are split into mandatory, recommended, and optional
sections.
</p>
</div>
<div class="topic concept nested1" id="mandatory-post"><a name="mandatory-post" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#mandatory-post" name="mandatory-post" shape="rect">7.1. Mandatory Actions</a></h3>
<div class="body conbody">
<p class="p">
Some actions must be taken after the installation before the CUDA Toolkit and Driver can be used.
</p>
</div>
<div class="topic concept nested2" id="environment-setup"><a name="environment-setup" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#environment-setup" name="environment-setup" shape="rect">7.1.1. Environment Setup</a></h3>
<div class="body conbody">
<p class="p">The <samp class="ph codeph">PATH</samp> variable needs to include <samp class="ph codeph">/usr/local/cuda-<span class="keyword">10.1</span>/bin</samp>
and <samp class="ph codeph">/usr/local/cuda-<span class="keyword">10.1</span>/NsightCompute-<tool-version></samp>.
<samp class="ph codeph"><tool-version></samp> refers to the version of Nsight Compute that ships with the CUDA toolkit, e.g. <samp class="ph codeph">2019.1</samp>.
</p>
<p class="p">To add this path to the <samp class="ph codeph">PATH</samp> variable:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> export PATH=/usr/local/cuda-<span class="keyword">10.1</span>/bin:/usr/local/cuda-<span class="keyword">10.1</span>/NsightCompute-2019.1${PATH:+:${PATH}}</pre><p class="p">In addition, when using the runfile installation method, the <samp class="ph codeph">LD_LIBRARY_PATH</samp> variable needs to contain <samp class="ph codeph">/usr/local/cuda-<span class="keyword">10.1</span>/lib64</samp> on a 64-bit system, or <samp class="ph codeph">/usr/local/cuda-<span class="keyword">10.1</span>/lib</samp> on a 32-bit system
</p>
<ul class="ul">
<li class="li">
<p class="p">To change the environment variables for 64-bit operating systems:</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> export LD_LIBRARY_PATH=/usr/local/cuda-<span class="keyword">10.1</span>/lib64\
${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}</pre></li>
<li class="li">
<p class="p">To change the environment variables for 32-bit operating systems:</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> export LD_LIBRARY_PATH=/usr/local/cuda-<span class="keyword">10.1</span>/lib\
${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}</pre></li>
</ul>
<p class="p">Note that the above paths change when using a custom install path with the runfile installation method.</p>
</div>
</div>
<div class="topic concept nested2" id="power9-setup"><a name="power9-setup" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#power9-setup" name="power9-setup" shape="rect">7.1.2. POWER9 Setup</a></h3>
<div class="body conbody">
<p class="p">
Because of the addition of new features specific to the NVIDIA POWER9 CUDA driver, there are some additional setup requirements
in order for the driver to function properly. These additional steps are not handled by the installation of CUDA packages,
and failure to ensure these extra requirements are met will result in a non-functional CUDA driver installation.
</p>
<div class="p">
There are two changes that need to be made manually after installing the NVIDIA CUDA driver to ensure proper operation:
<ol class="ol">
<li class="li">
<div class="p">
The NVIDIA Persistence Daemon should be automatically started for POWER9 installations. Check that it is running with the
following command:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> systemctl status nvidia-persistenced</pre>
If it is not active, run the following command:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo systemctl enable nvidia-persistenced</pre></div>
</li>
<li class="li">
<p class="p">
Disable a udev rule installed by default in some Linux distributions that cause hot-pluggable memory to be automatically onlined
when it is physically probed. This behavior prevents NVIDIA software from bringing NVIDIA device memory online with non-default
settings.
This udev rule must be disabled in order for the NVIDIA CUDA driver to function properly on POWER9 systems.
</p>
<div class="p">
On RedHat Enterprise Linux 7, this rule can be found in:
<pre class="pre screen" xml:space="preserve">/lib/udev/rules.d/40-redhat.rules</pre></div>
<div class="p"> On Ubuntu 17.04, this rule can be found in:
<pre class="pre screen" xml:space="preserve">/lib/udev/rules.d/40-vm-hotadd.rules</pre></div>
<div class="p">The rule generally takes a form where it detects the addition of a memory
block and changes the 'state' attribute to online. For example, in RHEL7, the rule
looks like this:
<pre class="pre screen" xml:space="preserve">SUBSYSTEM=="memory", ACTION=="add", PROGRAM="/bin/uname -p", RESULT!="s390*", ATTR{state}=="offline", ATTR{state}="online"</pre>
This
rule must be disabled by copying the file to /etc/udev/rules.d and commenting out,
removing, or changing the hot-pluggable memory rule in the /etc copy so that it
does not apply to POWER9 NVIDIA systems. For example, on RHEL 7.5 or earlier
versions:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo cp /lib/udev/rules.d/40-redhat.rules /etc/udev/rules.d
<strong class="ph b">$</strong> sudo sed -i '/SUBSYSTEM=="memory", ACTION=="add"/d' /etc/udev/rules.d/40-redhat.rules</pre>
On RHEL 7.6 and later versions:
<pre class="pre screen" xml:space="preserve">$ sudo cp /lib/udev/rules.d/40-redhat.rules /etc/udev/rules.d
<strong class="ph b">$</strong> sudo sed -i 's/SUBSYSTEM!="memory", ACTION!="add", GOTO="memory_hotplug_end"/SUBSYSTEM=="memory", ACTION=="add", GOTO="memory_hotplug_end"/' /etc/udev/rules.d/40-redhat.rules</pre></div>
</li>
</ol>
</div>
<p class="p">
You will need to reboot the system to initialize the above changes.
</p>
<div class="note note"><span class="notetitle">Note:</span> For NUMA best practices on IBM Newell POWER9, see <a class="xref" href="../cuda-c-best-practices-guide/index.html#numa-best-practices" target="_blank" shape="rect">NUMA Best Practices.</a></div>
</div>
</div>
</div>
<div class="topic concept nested1" id="recommended-post"><a name="recommended-post" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#recommended-post" name="recommended-post" shape="rect">7.2. Recommended Actions</a></h3>
<div class="body conbody">
<p class="p">
Other actions are recommended to verify the integrity of the installation.
</p>
</div>
<div class="topic concept nested2" id="install-persistenced"><a name="install-persistenced" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#install-persistenced" name="install-persistenced" shape="rect">7.2.1. Install Persistence Daemon</a></h3>
<div class="body conbody">
<p class="p">
NVIDIA is providing a user-space daemon on Linux to support persistence of driver state across CUDA job runs. The daemon approach
provides a more elegant and robust solution to this problem than persistence mode. For more details on the NVIDIA Persistence
Daemon, see the documentation <a class="xref" href="http://docs.nvidia.com/deploy/driver-persistence/index.html#persistence-daemon" target="_blank" shape="rect">here</a>.
</p>
<div class="p">
The NVIDIA Persistence Daemon can be started as the root user by running:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> /usr/bin/nvidia-persistenced --verbose</pre>
This command should be run on boot. Consult your Linux distribution's init documentation for details on how to automate this.
</div>
</div>
</div>
<div class="topic concept nested2" id="install-samples"><a name="install-samples" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#install-samples" name="install-samples" shape="rect">7.2.2. Install Writable Samples</a></h3>
<div class="body conbody">
<div class="p">
In order to modify, compile, and run the samples, the samples must be installed with write permissions. A convenience installation
script is provided:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> cuda-install-samples-<span class="keyword">10.1</span>.sh <dir></pre>
This script is installed with the cuda-samples-<span class="keyword">10</span>-<span class="keyword">1</span> package. The cuda-samples-<span class="keyword">10</span>-<span class="keyword">1</span> package installs only a read-only copy in /usr/local/cuda-<span class="keyword">10.1</span>/samples.
</div>
</div>
</div>
<div class="topic concept nested2" id="verify-installation"><a name="verify-installation" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#verify-installation" name="verify-installation" shape="rect">7.2.3. Verify the Installation</a></h3>
<div class="body conbody">
<p class="p">Before continuing, it is important to verify that the CUDA toolkit can find and communicate correctly with the CUDA-capable
hardware. To do this, you need to compile and run some of the included sample programs.
</p>
<div class="note note"><span class="notetitle">Note:</span> Ensure the PATH and, if using the runfile installation method, LD_LIBRARY_PATH variables are <a class="xref" href="index.html#environment-setup" shape="rect">set correctly</a>.
</div>
</div>
<div class="topic concept nested3" id="verify-driver"><a name="verify-driver" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#verify-driver" name="verify-driver" shape="rect">7.2.3.1. Verify the Driver Version</a></h3>
<div class="body conbody">
<p class="p">If you installed the driver, verify that the correct version of it is loaded. If you did not install the driver, or are using
an operating system where the driver is not loaded via a kernel module, such as L4T, skip this step.
</p>
<div class="p">When the driver is loaded, the driver version can be found by executing the command
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> cat /proc/driver/nvidia/version</pre>
Note that this command will not work on an iGPU/dGPU system.
</div>
</div>
</div>
<div class="topic concept nested3" id="compiling-examples"><a name="compiling-examples" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#compiling-examples" name="compiling-examples" shape="rect">7.2.3.2. Compiling the Examples</a></h3>
<div class="body conbody">
<p class="p">The version of the CUDA Toolkit can be checked by running <samp class="ph codeph">nvcc -V</samp> in a terminal window. The <samp class="ph codeph">nvcc</samp> command runs the compiler driver that compiles CUDA programs. It calls the <samp class="ph codeph">gcc</samp> compiler for C code and the NVIDIA PTX compiler for the CUDA code.
</p>
<p class="p">The NVIDIA CUDA Toolkit includes sample programs in source form. You should compile them by changing to <samp class="ph codeph">~/NVIDIA_CUDA-<span class="keyword">10.1</span>_Samples</samp> and typing <samp class="ph codeph">make</samp>. The resulting binaries will be placed under <samp class="ph codeph">~/NVIDIA_CUDA-<span class="keyword">10.1</span>_Samples/bin</samp>.
</p>
</div>
</div>
<div class="topic concept nested3" id="running-binaries"><a name="running-binaries" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#running-binaries" name="running-binaries" shape="rect">7.2.3.3. Running the Binaries</a></h3>
<div class="body conbody">
<p class="p">After compilation, find and run <samp class="ph codeph">deviceQuery</samp> under <samp class="ph codeph">~/NVIDIA_CUDA-<span class="keyword">10.1</span>_Samples</samp>. If the CUDA software is installed and configured correctly, the output for <samp class="ph codeph">deviceQuery</samp> should look similar to that shown in <a class="xref" href="index.html#running-binaries__valid-results-from-sample-cuda-devicequery-program" shape="rect">Figure 1</a>.
</p>
<div class="fig fignone" id="running-binaries__valid-results-from-sample-cuda-devicequery-program"><a name="running-binaries__valid-results-from-sample-cuda-devicequery-program" shape="rect">
<!-- --></a><span class="figcap">Figure 1. Valid Results from deviceQuery CUDA Sample</span><br clear="none"></br><div class="imagecenter"><img class="image imagecenter" src="graphics/valid-results-from-sample-cuda-devicequery-program.png" width="550" alt="Valid Results from deviceQuery CUDA Sample."></img></div><br clear="none"></br></div>
<p class="p">The exact appearance and the output lines might be different on your system. The important outcomes are that a device was
found (the first highlighted line), that the device matches the one on your system (the second highlighted line), and that
the test passed (the final highlighted line).
</p>
<p class="p">If a CUDA-capable device and the CUDA Driver are installed but <samp class="ph codeph">deviceQuery</samp> reports that no CUDA-capable devices are present, this likely means that the <samp class="ph codeph">/dev/nvidia*</samp> files are missing or have the wrong permissions.
</p>
<div class="p">On systems where <samp class="ph codeph">SELinux</samp> is enabled, you might need to temporarily disable this security feature to run <samp class="ph codeph">deviceQuery</samp>. To do this, type:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> setenforce 0</pre>
from the command line as the <dfn class="term">superuser</dfn>.
</div>
<p class="p">Running the <samp class="ph codeph">bandwidthTest</samp> program ensures that the system and the CUDA-capable device are able to communicate correctly. Its output is shown in <a class="xref" href="index.html#running-binaries__valid-results-from-sample-cuda-bandwidthtest-program" shape="rect">Figure 2</a>.
</p>
<div class="fig fignone" id="running-binaries__valid-results-from-sample-cuda-bandwidthtest-program"><a name="running-binaries__valid-results-from-sample-cuda-bandwidthtest-program" shape="rect">
<!-- --></a><span class="figcap">Figure 2. Valid Results from bandwidthTest CUDA Sample</span><br clear="none"></br><div class="imagecenter"><img class="image imagecenter" src="graphics/valid-results-from-sample-cuda-bandwidthtest-program.png" width="550" alt="Valid Results from bandwidthTest CUDA Sample."></img></div><br clear="none"></br></div>
<p class="p">Note that the measurements for your CUDA-capable device description will vary from system to system. The important point is
that you obtain measurements, and that the second-to-last line (in <a class="xref" href="index.html#running-binaries__valid-results-from-sample-cuda-bandwidthtest-program" shape="rect">Figure 2</a>) confirms that all necessary tests passed.
</p>
<p class="p">Should the tests not pass, make sure you have a CUDA-capable NVIDIA GPU on your system and make sure it is properly installed.
</p>
<p class="p">If you run into difficulties with the link step (such as libraries not being found), consult the <cite class="cite">Linux Release Notes</cite> found in the <samp class="ph codeph">doc</samp> folder in the CUDA Samples directory.
</p>
</div>
</div>
</div>
<div class="topic concept nested2" id="install-nsight-plugins"><a name="install-nsight-plugins" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#install-nsight-plugins" name="install-nsight-plugins" shape="rect">7.2.4. Install Nsight Eclipse Plugins</a></h3>
<div class="body conbody">
<div class="p">
To install Nsight Eclipse plugins, an installation script is provided:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> /usr/local/cuda-<span class="keyword">10.1</span>/bin/nsight_ee_plugins_manage.sh install <eclipse-dir></pre>
Refer to <a class="xref" href="http://docs.nvidia.com/cuda/nsightee-plugins-install-guide/index.html" target="_blank" shape="rect">Nsight Eclipse Plugins Installation Guide</a> for more details.
</div>
</div>
</div>
</div>
<div class="topic concept nested1" id="optional-post"><a name="optional-post" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#optional-post" name="optional-post" shape="rect">7.3. Optional Actions</a></h3>
<div class="body conbody">
<p class="p">
Other options are not necessary to use the CUDA Toolkit, but are available to provide additional features.
</p>
</div>
<div class="topic concept nested2" id="install-libraries"><a name="install-libraries" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#install-libraries" name="install-libraries" shape="rect">7.3.1. Install Third-party Libraries</a></h3>
<div class="body conbody">
<p class="p">
Some CUDA samples use third-party libraries which may not be installed by default on your system. These samples attempt to
detect any required libraries when building. If a library is not detected, it waives itself and warns you which library is
missing. To build and run these samples, you must install the missing libraries. These dependencies may be installed if the
RPM or Deb <samp class="ph codeph">cuda-samples-<span class="keyword">10</span>-<span class="keyword">1</span></samp> package is used. In cases where these dependencies are not installed, follow the instructions below.
</p>
<div class="section">
<h4 class="title sectiontitle">RHEL/CentOS</h4>
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum install freeglut-devel libX11-devel libXi-devel libXmu-devel \
make mesa-libGLU-devel</pre></div>
</div>
<div class="section">
<h4 class="title sectiontitle">Fedora</h4>
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf install freeglut-devel libX11-devel libXi-devel libXmu-devel \
make mesa-libGLU-devel</pre></div>
</div>
<div class="section">
<h4 class="title sectiontitle">SLES</h4>
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper install libglut3 libX11 libXi6 libXmu6 libGLU1 make</pre></div>
</div>
<div class="section">
<h4 class="title sectiontitle">OpenSUSE</h4>
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper install freeglut-devel libX11-devel libXi-devel libXmu-devel \
make Mesa-libGL-devel</pre></div>
</div>
<div class="section">
<h4 class="title sectiontitle">Ubuntu</h4>
<div class="p"><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-get install g++ freeglut3-dev build-essential libx11-dev \
libxmu-dev libxi-dev libglu1-mesa libglu1-mesa-dev</pre></div>
</div>
</div>
</div>
<div class="topic concept nested2" id="install-source"><a name="install-source" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#install-source" name="install-source" shape="rect">7.3.2. Install the source code for cuda-gdb</a></h3>
<div class="body conbody">
<p class="p"> The <samp class="ph codeph">cuda-gdb</samp> source must be explicitly selected for installation with the
runfile installation method. During the installation, in the component selection page, expand
the component "CUDA Tools 10.1" and select the <samp class="ph codeph">cuda-gdb-src</samp> for installation.
It is unchecked by default.
</p>
<p class="p">
To obtain a copy of the source code for cuda-gdb using the RPM and Debian installation methods, the <strong class="ph b">cuda-gdb-src</strong> package must be installed.
</p>
<p class="p">
The source code is installed as a tarball in the /usr/local/cuda-<span class="keyword">10.1</span>/extras directory.
</p>
</div>
</div>
</div>
</div>
<div class="topic concept nested0" id="advanced-setup"><a name="advanced-setup" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#advanced-setup" name="advanced-setup" shape="rect">8. Advanced Setup</a></h2>
<div class="body conbody">
<p class="p">
Below is information on some advanced setup scenarios which are not covered in the basic instructions above.
</p>
<div class="tablenoborder">
<table cellpadding="4" cellspacing="0" summary="" class="table" frame="border" border="1" rules="all">
<caption><span class="tablecap">Table 6. Advanced Setup Scenarios when Installing CUDA</span></caption>
<thead class="thead" align="left">
<tr class="row">
<th class="entry" align="center" valign="top" width="25%" id="d54e3856" rowspan="1" colspan="1">Scenario</th>
<th class="entry" align="center" valign="top" width="75%" id="d54e3859" rowspan="1" colspan="1">Instructions</th>
</tr>
</thead>
<tbody class="tbody">
<tr class="row">
<td class="entry" valign="top" width="25%" headers="d54e3856" rowspan="1" colspan="1">
Install CUDA using the Package Manager installation method without installing the NVIDIA GL libraries.
</td>
<td class="entry" valign="top" width="75%" headers="d54e3859" rowspan="1" colspan="1"><strong class="ph b">RHEL 7/CentOS 7</strong><p class="p">
Install CUDA using the following command:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum install cuda-toolkit-<span class="keyword">10</span>-<span class="keyword">1</span> \
nvidia-driver-cuda nvidia-kmod</pre><p class="p">
If the system is using a non-NVIDIA GPU to render the display, remove the files at <samp class="ph codeph">/etc/X11/xorg.conf</samp> and <samp class="ph codeph">/usr/lib64/nvidia/xorg/libglx.so</samp>, and remove the <samp class="ph codeph">nomodeset</samp> kernel parameter from the boot command-line.
</p>
<p class="p">
Follow the instructions <a class="xref" href="index.html#runfile-nouveau-rhel" shape="rect">here</a> to ensure that Nouveau is disabled.
</p><strong class="ph b">RHEL 6/CentOS 6</strong><p class="p">
Install CUDA using the following command:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum install cuda-toolkit-<span class="keyword">10</span>-<span class="keyword">1</span> \
xorg-x11-drv-nvidia-libs nvidia-kmod</pre><p class="p">
If the system is using a non-NVIDIA GPU to render the display, remove the file at <samp class="ph codeph">/etc/X11/xorg.conf</samp>.
</p>
<p class="p">
Follow the instructions <a class="xref" href="index.html#runfile-nouveau-rhel" shape="rect">here</a> to ensure that Nouveau is disabled.
</p><strong class="ph b">Fedora</strong><p class="p">
Install CUDA using the following command:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf install cuda-toolkit-<span class="keyword">10</span>-<span class="keyword">1</span> \
nvidia-driver-cuda akmod-nvidia</pre><p class="p">
Follow the instructions <a class="xref" href="index.html#runfile-nouveau-fedora" shape="rect">here</a> to ensure that Nouveau is disabled.
</p>
<p class="p">
If performing an upgrade over a previous installation, the NVIDIA kernel module may need to be rebuilt by following the instructions
<a class="xref" href="index.html#faq8" shape="rect">here</a>.
</p><strong class="ph b">OpenSUSE/SLES</strong><p class="p">
On some system configurations the NVIDIA GL libraries may need to be locked before installation using:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper addlock nvidia-glG04</pre><p class="p">
Install CUDA using the following command:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper install --no-recommends cuda-toolkit-<span class="keyword">10</span>-<span class="keyword">1</span> \
nvidia-computeG04 \
nvidia-gfxG04-kmp-default</pre><p class="p">
Follow the instructions <a class="xref" href="index.html#runfile-nouveau-suse" shape="rect">here</a> to ensure that Nouveau is disabled.
</p><strong class="ph b">Ubuntu</strong><p class="p">
This functionality isn't supported on Ubuntu. Instead, the driver packages integrate with the Bumblebee framework to provide
a solution for users who wish to control what applications the NVIDIA drivers are used for. See Ubuntu's <a class="xref" href="https://wiki.ubuntu.com/Bumblebee" target="_blank" shape="rect">Bumblebee wiki</a> for more information.
</p>
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="25%" headers="d54e3856" rowspan="1" colspan="1">
Upgrade from a RPM/Deb driver installation which includes the diagnostic driver packages to a driver installation which does
not include the diagnostic driver packages.
</td>
<td class="entry" valign="top" width="75%" headers="d54e3859" rowspan="1" colspan="1"><strong class="ph b">RHEL/CentOS</strong><p class="p"> Remove diagnostic packages using the following command: </p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum remove cuda-drivers-diagnostic \
xorg-x11-drv-nvidia-diagnostic</pre><p class="p"> Follow the instructions <a class="xref" href="index.html#package-manager-upgrades" shape="rect">here</a> to continue installation as normal.
</p><strong class="ph b">Fedora</strong><p class="p"> Remove diagnostic packages using the following command: </p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf remove cuda-drivers-diagnostic \
xorg-x11-drv-nvidia-diagnostic</pre><p class="p"> Follow the instructions <a class="xref" href="index.html#package-manager-upgrades" shape="rect">here</a> to continue installation as normal.
</p><strong class="ph b">OpenSUSE/SLES</strong><p class="p"> Remove diagnostic packages using the following command: </p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper remove cuda-drivers-diagnostic \
nvidia-diagnosticG04</pre><p class="p"> Follow the instructions <a class="xref" href="index.html#package-manager-upgrades" shape="rect">here</a> to continue installation as normal.
</p><strong class="ph b">Ubuntu</strong><p class="p"> Remove diagnostic packages using the following command: </p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-get --purge remove cuda-drivers-diagnostic \
nvidia-384-diagnostic</pre><p class="p"> Follow the instructions <a class="xref" href="index.html#package-manager-upgrades" shape="rect">here</a> to continue installation as normal.
</p>
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="25%" headers="d54e3856" rowspan="1" colspan="1">
Use a specific GPU for rendering the display.
</td>
<td class="entry" valign="top" width="75%" headers="d54e3859" rowspan="1" colspan="1">
Add or replace a <strong class="ph b">Device</strong> entry in your xorg.conf file, located at <samp class="ph codeph">/etc/X11/xorg.conf</samp>. The <strong class="ph b">Device</strong> entry should resemble the following:
<pre class="pre screen" xml:space="preserve">Section "Device"
Identifier "Device0"
Driver "driver_name"
VendorName "vendor_name"
BusID "bus_id"
EndSection
</pre>
The details will you will need to add differ on a case-by-case basis. For example, if you have two NVIDIA GPUs and you want
the first GPU to be used for display, you would replace "driver_name" with "nvidia", "vendor_name" with "NVIDIA Corporation"
and "bus_id" with the Bus ID of the GPU.
<p class="p">
The Bus ID will resemble "PCI:00:02.0" and can be found by running <samp class="ph codeph">lspci</samp>.
</p>
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="25%" headers="d54e3856" rowspan="1" colspan="1">
Install CUDA to a specific directory using the Package Manager installation method.
</td>
<td class="entry" valign="top" width="75%" headers="d54e3859" rowspan="1" colspan="1"><strong class="ph b">RPM</strong><p class="p">
The RPM packages don't support custom install locations through the package managers (Yum and Zypper), but it is possible
to install the RPM packages to a custom location using rpm's <samp class="ph codeph">--relocate</samp> parameter:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo rpm --install --relocate /usr/local/cuda-<span class="keyword">10.1</span>=/new/toolkit package.rpm</pre><p class="p">
You will need to install the packages in the correct dependency order; this task is normally taken care of by the package
managers. For example, if package "foo" has a dependency on package "bar", you should install package "bar" first, and package
"foo" second. You can check the dependencies of a RPM package as follows:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> rpm -qRp package.rpm</pre><p class="p">
Note that the driver packages cannot be relocated.
</p><strong class="ph b">Deb</strong><p class="p">
The Deb packages do not support custom install locations. It is however possible to extract the contents of the Deb packages
and move the files to the desired install location. See the next scenario for more details one xtracting Deb packages.
</p>
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="25%" headers="d54e3856" rowspan="1" colspan="1">
Extract the contents of the installers.
</td>
<td class="entry" valign="top" width="75%" headers="d54e3859" rowspan="1" colspan="1"><strong class="ph b">Runfile</strong><p class="p">
The Runfile can be extracted into the standalone Toolkit, Samples and Driver Runfiles by using the <samp class="ph codeph">--extract</samp> parameter. The Toolkit and Samples standalone Runfiles can be further extracted by running:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> ./runfile.run --tar mxvf</pre><p class="p">
The Driver Runfile can be extracted by running:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> ./runfile.run -x</pre><strong class="ph b">RPM</strong><p class="p">
The RPM packages can be extracted by running:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> rpm2cpio package.rpm | cpio -idmv</pre><strong class="ph b">Deb</strong><p class="p">
The Deb packages can be extracted by running:
</p><pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> dpkg-deb -x package.deb output_dir</pre></td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="25%" headers="d54e3856" rowspan="1" colspan="1">
<p class="p">
Modify Ubuntu's apt package manager to query specific architectures for specific repositories.
</p>
<p class="p">
This is useful when a foreign architecture has been added, causing "404 Not Found" errors to appear when the repository meta-data
is updated.
</p>
</td>
<td class="entry" valign="top" width="75%" headers="d54e3859" rowspan="1" colspan="1">
<p class="p">
Each repository you wish to restrict to specific architectures must have its <samp class="ph codeph">sources.list</samp> entry modified. This is done by modifying the <samp class="ph codeph">/etc/apt/sources.list</samp> file and any files containing repositories you wish to restrict under the <samp class="ph codeph">/etc/apt/sources.list.d/</samp> directory. Normally, it is sufficient to modify only the entries in <samp class="ph codeph">/etc/apt/sources.list</samp></p>
<div class="p">
An architecture-restricted repository entry looks like:
<pre class="pre screen" xml:space="preserve">deb [arch=<arch1>,<arch2>] <url></pre>
For example, if you wanted to restrict a repository to only the amd64 and i386 architectures, it would look like:
<pre class="pre screen" xml:space="preserve">deb [arch=amd64,i386] <url></pre>
It is not necessary to restrict the <samp class="ph codeph">deb-src</samp> repositories, as these repositories don't provide architecture-specific packages.
</div>
<p class="p">
For more details, see the <samp class="ph codeph">sources.list</samp> manpage.
</p>
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="25%" headers="d54e3856" rowspan="1" colspan="1">
<p class="p">
The nvidia.ko kernel module fails to load, saying some symbols are unknown.
</p>
<div class="p">
For example:
<pre class="pre screen" xml:space="preserve">nvidia: Unknown symbol drm_open (err 0)</pre></div>
</td>
<td class="entry" valign="top" width="75%" headers="d54e3859" rowspan="1" colspan="1">
<p class="p">
Check to see if there are any optionally installable modules that might provide these symbols which are not currently installed.
</p>
<p class="p">
For the example of the <samp class="ph codeph">drm_open</samp> symbol, check to see if there are any packages which provide drm_open and are not already installed. For instance, on Ubuntu
14.04, the <samp class="ph codeph">linux-image-extra</samp> package provides the DRM kernel module (which provides drm_open). This package is optional even though the kernel headers
reflect the availability of DRM regardless of whether this package is installed or not.
</p>
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="25%" headers="d54e3856" rowspan="1" colspan="1">
<p class="p">
The runfile installer fails to extract due to limited space in the TMP directory.
</p>
</td>
<td class="entry" valign="top" width="75%" headers="d54e3859" rowspan="1" colspan="1">
<p class="p">
This can occur on systems with limited storage in the TMP directory (usually /tmp), or on systems which use a tmpfs in memory
to handle temporary storage. In this case, the <samp class="ph codeph">--tmpdir</samp> command-line option should be used to instruct the runfile to use a directory with sufficient space to extract into. More
information on this option can be found <a class="xref" href="index.html#runfile-advanced" shape="rect">here</a>.
</p>
</td>
</tr>
<tr class="row">
<td class="entry" valign="top" width="25%" headers="d54e3856" rowspan="1" colspan="1">
<p class="p">
Re-enable Wayland after installing the RPM driver on Fedora.
</p>
</td>
<td class="entry" valign="top" width="75%" headers="d54e3859" rowspan="1" colspan="1">
<div class="p">
Wayland is disabled during installation of the Fedora driver RPM due to compatability issues. To re-enable wayland, comment
out this line in /etc/gdm/custom.conf:
<pre class="pre screen" xml:space="preserve">WaylandEnable=false</pre></div>
</td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
<div class="topic concept nested0" id="faq"><a name="faq" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#faq" name="faq" shape="rect">9. Frequently Asked Questions</a></h2>
<div class="topic concept nested1" id="faq1"><a name="faq1" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq1" name="faq1" shape="rect">How do I install the Toolkit in a different location?</a></h3>
<div class="body conbody">
<div class="p">
The Runfile installation asks where you wish to install the Toolkit and the Samples during an interactive install. If installing
using a non-interactive install, you can use the --toolkitpath and --samplespath parameters to change the install location:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> ./runfile.run --silent \
--toolkit --toolkitpath=/my/new/toolkit \
--samples --samplespath=/my/new/samples</pre></div>
<p class="p">
The RPM and Deb packages cannot be installed to a custom install location directly using the package managers. See the "Install
CUDA to a specific directory using the Package Manager installation method" scenario in the <a class="xref" href="index.html#advanced-setup" shape="rect">Advanced Setup</a> section for more information.
</p>
</div>
</div>
<div class="topic concept nested1" id="faq2"><a name="faq2" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq2" name="faq2" shape="rect">Why do I see "nvcc: No such file or directory" when I try to build a CUDA application?</a></h3>
<div class="body conbody">
<div class="p">
Your PATH environment variable is not set up correctly. Ensure that your PATH includes the bin directory where you installed
the Toolkit, usually /usr/local/cuda-<span class="keyword">10.1</span>/bin.
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> export PATH=/usr/local/cuda-<span class="keyword">10.1</span>/bin${PATH:+:${PATH}}</pre></div>
</div>
</div>
<div class="topic concept nested1" id="faq3"><a name="faq3" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq3" name="faq3" shape="rect">Why do I see "error while loading shared libraries: <lib name>: cannot open shared object file: No such file or directory"
when I try to run a CUDA application that uses a CUDA library?</a></h3>
<div class="body conbody">
<div class="p">
Your LD_LIBRARY_PATH environment variable is not set up correctly. Ensure that your LD_LIBRARY_PATH includes the lib and/or
lib64 directory where you installed the Toolkit, usually /usr/local/cuda-<span class="keyword">10.1</span>/lib{,64}:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> export LD_LIBRARY_PATH=/usr/local/cuda-<span class="keyword">10.1</span>/lib\
${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}</pre></div>
</div>
</div>
<div class="topic concept nested1" id="faq4"><a name="faq4" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq4" name="faq4" shape="rect">Why do I see multiple "404 Not Found" errors when updating my repository meta-data on Ubuntu?</a></h3>
<div class="body conbody">
<p class="p">
These errors occur after adding a foreign architecture because apt is attempting to query for each architecture within each
repository listed in the system's sources.list file. Repositories that do not host packages for the newly added architecture
will present this error. While noisy, the error itself does no harm. Please see the <a class="xref" href="index.html#advanced-setup" shape="rect">Advanced Setup</a> section for details on how to modify your sources.list file to prevent these errors.
</p>
</div>
</div>
<div class="topic concept nested1" id="faq5"><a name="faq5" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq5" name="faq5" shape="rect">How can I tell X to ignore a GPU for compute-only use?</a></h3>
<div class="body conbody">
<p class="p">
To make sure X doesn't use a certain GPU for display, you need to specify which <strong class="ph b">other</strong> GPU to use for display. For more information, please refer to the "Use a specific GPU for rendering the display" scenario
in the <a class="xref" href="index.html#advanced-setup" shape="rect">Advanced Setup</a> section.
</p>
</div>
</div>
<div class="topic concept nested1" id="faq6"><a name="faq6" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq6" name="faq6" shape="rect">Why doesn't the cuda-repo package install the CUDA Toolkit and Drivers?</a></h3>
<div class="body conbody">
<p class="p">
When using RPM or Deb, the downloaded package is a repository package.
Such a package only informs the package manager where to find the actual installation packages, but will not install them.
</p>
<p class="p">
See the <a class="xref" href="index.html#package-manager-overview" shape="rect">Package Manager Installation</a> section for more details.
</p>
</div>
</div>
<div class="topic concept nested1" id="faq7"><a name="faq7" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq7" name="faq7" shape="rect">How do I get CUDA to work on a laptop with an iGPU and a dGPU running Ubuntu14.04?</a></h3>
<div class="body conbody">
<div class="p">
After installing CUDA, set the driver value for the intel device in /etc/X11/xorg.conf to 'modesetting' as shown below:
<pre class="pre screen" xml:space="preserve">Section "Device"
Identifier "intel"
Driver <strong class="ph b">"modesetting"</strong>
...
EndSection</pre>
To prevent Ubuntu from reverting the change in xorg.conf, edit /etc/default/grub to add "nogpumanager" to GRUB_CMDLINE_LINUX_DEFAULT.
</div>
<div class="p">Run the following command to update grub before rebooting:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo update-grub</pre></div>
</div>
</div>
<div class="topic concept nested1" id="faq8"><a name="faq8" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq8" name="faq8" shape="rect">What do I do if the display does not load, or CUDA does not work, after performing a system update?</a></h3>
<div class="body conbody">
<p class="p">
System updates may include an updated Linux kernel. In many cases, a new Linux kernel will be installed without properly updating
the required Linux kernel headers and development packages. To ensure the CUDA driver continues to work when performing a
system update, rerun the commands in the <a class="xref" href="index.html#verify-kernel-packages" shape="rect">Kernel Headers and Development Packages</a> section.
</p>
<div class="p">
Additionally, on Fedora, the Akmods framework will sometimes fail to correctly rebuild the NVIDIA kernel module packages when
a new Linux kernel is installed. When this happens, it is usually sufficient to invoke Akmods manually and regenerate the
module mapping files by running the following commands in a virtual console, and then rebooting:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo akmods --force
<strong class="ph b">$</strong> sudo depmod</pre>
You can reach a virtual console by hitting <samp class="ph codeph">ctrl+alt+f2</samp> at the same time.
</div>
</div>
</div>
<div class="topic concept nested1" id="faq9"><a name="faq9" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq9" name="faq9" shape="rect">How do I install a CUDA driver with a version less than 367 using a network repo?</a></h3>
<div class="body conbody">
<p class="p">
To install a CUDA driver at a version earlier than 367 using a network repo, the required packages will need to be explicitly
installed at the desired version. For example, to install 352.99, instead of installing the cuda-drivers metapackage at version
352.99, you will need to install all required packages of cuda-drivers at version 352.99.
</p>
</div>
</div>
<div class="topic concept nested1" id="faq10"><a name="faq10" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#faq10" name="faq10" shape="rect">How do I install an older CUDA version using a network repo?</a></h3>
<div class="body conbody">
<p class="p">
Depending on your system configuration, you may not be able to install old versions of CUDA using the cuda metapackage. In
order to install a specific version of CUDA, you may need to specify all of the packages that would normally be installed
by the cuda metapackage at the version you want to install.
</p>
<p class="p">
If you are using yum to install certain packages at an older version, the dependencies may not resolve as expected. In this
case you may need to pass "--setopt=obsoletes=0" to yum to allow an install of packages which are obsoleted at a later version
than you are trying to install.
</p>
</div>
</div>
</div>
<div class="topic concept nested0" id="additional-considerations"><a name="additional-considerations" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#additional-considerations" name="additional-considerations" shape="rect">10. Additional Considerations</a></h2>
<div class="body conbody">
<p class="p">Now that you have CUDA-capable hardware and the NVIDIA CUDA Toolkit installed, you can examine and enjoy the numerous included
programs. To begin using CUDA to accelerate the performance of your own applications, consult the <cite class="cite">CUDA C Programming Guide</cite>, located in <samp class="ph codeph">/usr/local/cuda-<span class="keyword">10.1</span>/doc</samp>.
</p>
<p class="p">A number of helpful development tools are included in the CUDA Toolkit to assist you as you develop your CUDA programs, such
as NVIDIA<sup>®</sup> Nsight™ Eclipse Edition, NVIDIA Visual Profiler, cuda-gdb, and cuda-memcheck.
</p>
<p class="p">For technical support on programming questions, consult and participate in the developer forums at <a class="xref" href="http://developer.nvidia.com/cuda/" target="_blank" shape="rect">http://developer.nvidia.com/cuda/</a>.
</p>
</div>
</div>
<div class="topic concept nested0" id="removing-cuda-tk-and-driver"><a name="removing-cuda-tk-and-driver" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#removing-cuda-tk-and-driver" name="removing-cuda-tk-and-driver" shape="rect">11. Removing CUDA Toolkit and Driver</a></h2>
<div class="body conbody">
<p class="p">Follow the below steps to properly uninstall the CUDA Toolkit and NVIDIA Drivers from
your system. These steps will ensure that the uninstallation will be clean.
</p>
<div class="section">
<h2 class="title sectiontitle">RHEL/CentOS</h2>
<div class="p">To remove CUDA Toolkit: <pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum remove "*cublas*" "cuda*"</pre>
To remove NVIDIA Drivers: <pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo yum remove "*nvidia*" </pre></div>
</div>
<div class="section">
<h2 class="title sectiontitle">Fedora</h2>
<div class="p"> To remove CUDA Toolkit: <pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf remove "*cublas*" "cuda*"</pre>
To remove NVIDIA Drivers: <pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo dnf remove "*nvidia*" </pre></div>
</div>
<div class="section">
<h2 class="title sectiontitle">OpenSUSE/SLES</h2>
<div class="p">To remove CUDA Toolkit:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper remove "*cublas*" "cuda*"</pre>
To remove NVIDIA
Drivers: <pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo zypper remove "*nvidia*" </pre></div>
</div>
<div class="section">
<h2 class="title sectiontitle">Ubuntu</h2>
<div class="p"> To remove CUDA Toolkit:
<pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-get --purge remove "*cublas*" "cuda*"</pre>
To remove
NVIDIA Drivers: <pre class="pre screen" xml:space="preserve"><strong class="ph b">$</strong> sudo apt-get --purge remove "*nvidia*"</pre></div>
</div>
</div>
</div>
<div class="topic concept nested0" id="notices-header"><a name="notices-header" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#notices-header" name="notices-header" shape="rect">Notices</a></h2>
<div class="topic reference nested1" id="notice"><a name="notice" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#notice" name="notice" shape="rect"></a></h3>
<div class="body refbody">
<div class="section">
<h3 class="title sectiontitle">Notice</h3>
<p class="p">ALL NVIDIA DESIGN SPECIFICATIONS, REFERENCE BOARDS, FILES, DRAWINGS, DIAGNOSTICS, LISTS, AND OTHER DOCUMENTS (TOGETHER AND
SEPARATELY, "MATERIALS") ARE BEING PROVIDED "AS IS." NVIDIA MAKES NO WARRANTIES, EXPRESSED, IMPLIED, STATUTORY, OR OTHERWISE
WITH RESPECT TO THE MATERIALS, AND EXPRESSLY DISCLAIMS ALL IMPLIED WARRANTIES OF NONINFRINGEMENT, MERCHANTABILITY, AND FITNESS
FOR A PARTICULAR PURPOSE.
</p>
<p class="p">Information furnished is believed to be accurate and reliable. However, NVIDIA Corporation assumes no responsibility for the
consequences of use of such information or for any infringement of patents or other rights of third parties that may result
from its use. No license is granted by implication of otherwise under any patent rights of NVIDIA Corporation. Specifications
mentioned in this publication are subject to change without notice. This publication supersedes and replaces all other information
previously supplied. NVIDIA Corporation products are not authorized as critical components in life support devices or systems
without express written approval of NVIDIA Corporation.
</p>
</div>
</div>
</div>
<div class="topic reference nested1" id="trademarks"><a name="trademarks" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#trademarks" name="trademarks" shape="rect"></a></h3>
<div class="body refbody">
<div class="section">
<h3 class="title sectiontitle">Trademarks</h3>
<p class="p">NVIDIA and the NVIDIA logo are trademarks or registered trademarks of NVIDIA Corporation
in the U.S. and other countries. Other company and product names may be trademarks of
the respective companies with which they are associated.
</p>
</div>
</div>
</div>
<div class="topic reference nested1" id="copyright-past-to-present"><a name="copyright-past-to-present" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#copyright-past-to-present" name="copyright-past-to-present" shape="rect"></a></h3>
<div class="body refbody">
<div class="section">
<h3 class="title sectiontitle">Copyright</h3>
<p class="p">© <span class="ph">2009</span>-<span class="ph">2019</span> NVIDIA
Corporation. All rights reserved.
</p>
<p class="p">This product includes software developed by the Syncro Soft SRL (http://www.sync.ro/).</p>
</div>
</div>
</div>
</div>
<hr id="contents-end"></hr>
</article>
</div>
</div>
<script language="JavaScript" type="text/javascript" charset="utf-8" src="../common/formatting/common.min.js"></script>
<script language="JavaScript" type="text/javascript" charset="utf-8" src="../common/scripts/google-analytics/google-analytics-write.js"></script>
<script language="JavaScript" type="text/javascript" charset="utf-8" src="../common/scripts/google-analytics/google-analytics-tracker.js"></script>
<script type="text/javascript">var switchTo5x=true;</script><script type="text/javascript" src="http://w.sharethis.com/button/buttons.js"></script><script type="text/javascript">stLight.options({publisher: "998dc202-a267-4d8e-bce9-14debadb8d92", doNotHash: false, doNotCopy: false, hashAddressBar: false});</script><script type="text/javascript">_satellite.pageBottom();</script></body>
</html>
distrib
>
Mageia
>
7
>
x86_64
>
media
>
nonfree-updates
>
by-pkgid
>
b86a85131cc739c1c53d0b55840a4328
>
files
>
1444
