<!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 Toolkit Release Notes"></meta>
<meta name="abstract" content="The Release Notes for the CUDA Toolkit."></meta>
<meta name="description" content="The Release Notes for the CUDA Toolkit."></meta>
<meta name="DC.Coverage" content="Release Notes"></meta>
<meta name="DC.subject" content="CUDA Toolkit, CUDA Toolkit 6.5 EA, CUDA Toolkit 6.5 EA libraries, CUDA Toolkit 6.5 EA release, CUDA Toolkit 6.5 EA installation, CUDA Toolkit issues, CUDA Toolkit core files, CUDA Toolkit resolved issues, CUDA Toolkit known issues, CUDA Toolkit documentation"></meta>
<meta name="keywords" content="CUDA Toolkit, CUDA Toolkit 6.5 EA, CUDA Toolkit 6.5 EA libraries, CUDA Toolkit 6.5 EA release, CUDA Toolkit 6.5 EA installation, CUDA Toolkit issues, CUDA Toolkit core files, CUDA Toolkit resolved issues, CUDA Toolkit known issues, CUDA Toolkit documentation"></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>Release Notes :: 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="../common/scripts/tynt/tynt.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-toolkit-release-notes/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
v6.5</a></div>
<div class="category"><a href="index.html" title="Release Notes">Release Notes</a></div>
<ul>
<li>
<div class="section-link"><a href="#overview">1. CUDA Toolkit Overview</a></div>
</li>
<li>
<div class="section-link"><a href="#new-features-title">2. New Features</a></div>
<ul>
<li>
<div class="section-link"><a href="#cuda-general-new-features">2.1. General CUDA</a></div>
</li>
<li>
<div class="section-link"><a href="#cuda-tools-title-new-features">2.2. CUDA Tools</a></div>
<ul>
<li>
<div class="section-link"><a href="#cuda-tools-general-new-features">2.2.1. General CUDA Tools</a></div>
</li>
<li>
<div class="section-link"><a href="#cuda-compiler-new-features">2.2.2. CUDA Compiler</a></div>
</li>
<li>
<div class="section-link"><a href="#cuda-occupancy-calculator-new-features">2.2.3. CUDA Occupancy Calculator</a></div>
</li>
<li>
<div class="section-link"><a href="#cupti-new-features">2.2.4. CUDA Profiling Tools Interface (CUPTI)</a></div>
</li>
<li>
<div class="section-link"><a href="#visual-profiler-new-features">2.2.5. NVIDIA Visual Profiler</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#cuda-libraries-title-new-features">2.3. CUDA Libraries</a></div>
<ul>
<li>
<div class="section-link"><a href="#cuda-libraries-general-new-features">2.3.1. General CUDA Libraries</a></div>
</li>
<li>
<div class="section-link"><a href="#cublas-new-features">2.3.2. cuBLAS Library</a></div>
</li>
<li>
<div class="section-link"><a href="#cufft-new-features">2.3.3. cuFFT Library</a></div>
</li>
<li>
<div class="section-link"><a href="#cusparse-new-features">2.3.4. cuSPARSE Library</a></div>
</li>
</ul>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#unsupported-features">3. Unsupported Features </a></div>
</li>
<li>
<div class="section-link"><a href="#deprecated-features">4. Deprecated Features </a></div>
</li>
<li>
<div class="section-link"><a href="#performance-improvements-title">5. Performance Improvements </a></div>
<ul>
<li>
<div class="section-link"><a href="#cuda-general-performance-improvements">5.1. General CUDA</a></div>
</li>
<li>
<div class="section-link"><a href="#cuda-libraries-title-performance-improvements">5.2. CUDA Libraries</a></div>
<ul>
<li>
<div class="section-link"><a href="#math-performance-improvements">5.2.1. CUDA Math Library</a></div>
</li>
</ul>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#resolved-issues-title">6. Resolved Issues </a></div>
<ul>
<li>
<div class="section-link"><a href="#cuda-general-resolved-issues">6.1. General CUDA</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#known-issues-title">7. Known Issues</a></div>
<ul>
<li>
<div class="section-link"><a href="#armv7-linux-general-cuda-known-issues">7.1. Linux on ARMv7 Specific Issues</a></div>
</li>
<li>
<div class="section-link"><a href="#cuda-general-known-issues">7.2. General CUDA</a></div>
</li>
<li>
<div class="section-link"><a href="#cuda-tools-title-known-issues">7.3. CUDA Tools</a></div>
<ul>
<li>
<div class="section-link"><a href="#cuda-compiler-known-issues">7.3.1. CUDA Compiler</a></div>
</li>
<li>
<div class="section-link"><a href="#cuda-gdb-known-issues">7.3.2. CUDA-GDB</a></div>
</li>
<li>
<div class="section-link"><a href="#nsight-ee-known-issues">7.3.3. Nsight Eclipse Edition</a></div>
</li>
<li>
<div class="section-link"><a href="#visual-profiler-known-issues">7.3.4. NVIDIA Visual Profiler</a></div>
</li>
</ul>
</li>
<li>
<div class="section-link"><a href="#cuda-libraries-title-known-issues">7.4. CUDA Libraries</a></div>
<ul>
<li>
<div class="section-link"><a href="#cufft-known-issues">7.4.1. cuFFT Library</a></div>
</li>
<li>
<div class="section-link"><a href="#thrust-known-issues">7.4.2. Thrust Library</a></div>
</li>
<li>
<div class="section-link"><a href="#cuda-samples-known-issues">7.4.3. CUDA Samples</a></div>
</li>
</ul>
</li>
</ul>
</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">Release Notes
(<a href="../../pdf/CUDA_Toolkit_Release_Notes.pdf">PDF</a>)
-
v6.5
(<a href="https://developer.nvidia.com/cuda-toolkit-archive">older</a>)
-
Last updated August 1, 2014
-
<a href="mailto:cudatools@nvidia.com?subject=CUDA Toolkit Documentation Feedback: Release Notes">Send Feedback</a>
-
<span class="st_facebook"></span><span class="st_twitter"></span><span class="st_linkedin"></span><span class="st_reddit"></span><span class="st_slashdot"></span><span class="st_tumblr"></span><span class="st_sharethis"></span></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 Toolkit Release Notes</a></h2>
<div class="body conbody"></div>
</div>
<div class="topic concept nested0" id="overview"><a name="overview" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#overview" name="overview" shape="rect">1. CUDA Toolkit Overview</a></h2>
<div class="body conbody">
<p class="p">This section provides an overview of the system requirements and major components of the
CUDA Toolkit and points to component locations after installation.
</p>
<dl class="dl">
<dt class="dt dlterm">System Requirements</dt>
<dd class="dd">The CUDA Toolkit is supported for Linux, Mac OS X, and Microsoft Windows. Specific system
requirements are referenced below. <a name="overview__ul_e2j_msp_wn" shape="rect">
<!-- --></a><ul class="ul" id="overview__ul_e2j_msp_wn">
<li class="li">Linux: The latest information about support for the Linux platform can
be found online at <a class="xref" href="http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-linux/index.html" target="_blank" shape="rect">http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-linux/index.html</a>.
</li>
<li class="li">Mac OS: The latest information about support for Mac OS X can be found
online at <a class="xref" href="http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-mac-os-x/index.html" target="_blank" shape="rect">http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-mac-os-x/index.html</a>.
</li>
<li class="li">Windows: The latest information about support for Microsoft Windows can
be found online at <a class="xref" href="http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-microsoft-windows/index.html" target="_blank" shape="rect">http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-microsoft-windows/index.html</a>.
</li>
</ul>
</dd>
<dt class="dt dlterm">Compiler</dt>
<dd class="dd">The CUDA-C and CUDA-C++ compiler, <samp class="ph codeph">nvcc</samp>, is found in the
<samp class="ph codeph">bin/</samp> directory. It is built on top of the NVVM optimizer,
which is itself built on top of the LLVM compiler infrastructure. Developers who
want to target NVVM directly can do so using the Compiler SDK, which is
available in the <samp class="ph codeph">nvvm/</samp> directory.
</dd>
<dt class="dt dlterm">Tools</dt>
<dd class="dd">The following development tools are available in the <samp class="ph codeph">bin/</samp> directory (except
for NSight Visual Studio Edition (VSE) which is installed as a plug-in to
Microsoft Visual Studio).<a name="overview__ul_iyz_wy3_tm" shape="rect">
<!-- --></a><ul class="ul" id="overview__ul_iyz_wy3_tm">
<li class="li">IDEs: <samp class="ph codeph">nsight</samp> (Linux, Mac OS), NSight VSE (Windows)
</li>
<li class="li">Debuggers: <samp class="ph codeph">cuda-memcheck</samp>, <samp class="ph codeph">cuda-gdb</samp> (Linux, Mac OS),
NSight VSE (Windows)
</li>
<li class="li">Profilers: <samp class="ph codeph">nvprof</samp>, <samp class="ph codeph">nvvp</samp>, NSight VSE
(Windows)
</li>
<li class="li">Utilities: <samp class="ph codeph">cuobjdump</samp>, <samp class="ph codeph">nvdisasm</samp>,
<samp class="ph codeph">nvprune</samp></li>
</ul>
</dd>
<dt class="dt dlterm">Libraries</dt>
<dd class="dd">The scientific and utility libraries listed below are available in the <samp class="ph codeph">lib/</samp>
directory (DLLs on Windows are in <samp class="ph codeph">bin/</samp>), and their interfaces
are available in the <samp class="ph codeph">include/</samp> directory.<a name="overview__ul_ljm_jsj_tm" shape="rect">
<!-- --></a><ul class="ul" id="overview__ul_ljm_jsj_tm">
<li class="li"><samp class="ph codeph">cublas</samp> (BLAS)
</li>
<li class="li"><samp class="ph codeph">cublas_device</samp> (BLAS Kernel Interface)
</li>
<li class="li"><samp class="ph codeph">cuda_occupancy</samp> (Kernel Occupancy Calculation [header file implementation])
</li>
<li class="li"><samp class="ph codeph">cudadevrt</samp> (CUDA Device Runtime)
</li>
<li class="li"><samp class="ph codeph">cudart</samp> (CUDA Runtime)
</li>
<li class="li"><samp class="ph codeph">cufft</samp> (Fast Fourier Transform [FFT])
</li>
<li class="li"><samp class="ph codeph">cupti</samp> (Profiling Tools Interface)
</li>
<li class="li"><samp class="ph codeph">curand</samp> (Random Number Generation)
</li>
<li class="li"><samp class="ph codeph">cusparse</samp> (Sparse Matrix)
</li>
<li class="li"><samp class="ph codeph">npp</samp> (NVIDIA Performance Primitives [image and signal processing])
</li>
<li class="li"><samp class="ph codeph">nvblas</samp> ("Drop-in" BLAS)
</li>
<li class="li"><samp class="ph codeph">nvcuvid</samp> (CUDA Video Decoder [Windows, Linux])
</li>
<li class="li"><samp class="ph codeph">thrust</samp> (Parallel Algorithm Library [header file implementation])
</li>
</ul>
</dd>
<dt class="dt dlterm">CUDA Samples</dt>
<dd class="dd">Code samples that illustrate how to use various CUDA and library APIs are available in the
<samp class="ph codeph">samples/</samp> directory on Linux and Mac OS, and are installed
to <samp class="ph codeph">C:\ProgramData\NVIDIA Corporation\CUDA Samples</samp> on Windows.
On Linux and Mac OS, the <samp class="ph codeph">samples/</samp> directory is read-only and
the samples must be copied to another location if they are to be modified.
Further instructions can be found in the Getting Started Guides for Linux and
Mac OS.
</dd>
<dt class="dt dlterm">Documentation</dt>
<dd class="dd">The most current version of these release notes can be found online at <a class="xref" href="http://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html" target="_blank" shape="rect">http://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html</a>.
</dd>
<dd class="dd">Documentation, including Getting Started Guides, Programming Guides, API References, and
Tools Guides, can be found in PDF form in the <samp class="ph codeph">doc/pdf/</samp>
directory, or in HTML form at <samp class="ph codeph">doc/html/index.html</samp> and online at
<a class="xref" href="http://docs.nvidia.com/cuda/index.html" target="_blank" shape="rect">http://docs.nvidia.com/cuda/index.html</a>.
</dd>
<dt class="dt dlterm">Other</dt>
<dd class="dd">The Open64 source files are controlled under terms of the GPL license. Current
and previously released versions are located at <a class="xref" href="ftp://download.nvidia.com/CUDAOpen64/" target="_blank" shape="rect">ftp://download.nvidia.com/CUDAOpen64/</a>.
</dd>
<dd class="dd">The CUDA-GDB source files are controlled under terms of the GPL license.<a name="overview__ul_br5_hgn_lm" shape="rect">
<!-- --></a><ul class="ul" id="overview__ul_br5_hgn_lm">
<li class="li">The source code for CUDA-GDB that shipped with CUDA 5.5 and subsequent
versions is located at <a class="xref" href="https://github.com/NVIDIA/cuda-gdb" target="_blank" shape="rect">https://github.com/NVIDIA/cuda-gdb</a>.
</li>
<li class="li">The source code for CUDA-GDB that shipped with CUDA 5.0 and previous
versions is located at <a class="xref" href="ftp://download.nvidia.com/CUDAOpen64/" target="_blank" shape="rect">ftp://download.nvidia.com/CUDAOpen64/</a>.
</li>
</ul>
</dd>
</dl>
</div>
</div>
<div class="topic concept nested0" id="new-features-title"><a name="new-features-title" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#new-features-title" name="new-features-title" shape="rect">2. New
Features</a></h2>
<div class="body conbody">
<p class="p"></p>
</div>
<div class="topic concept nested1" id="cuda-general-new-features"><a name="cuda-general-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-general-new-features" name="cuda-general-new-features" shape="rect">2.1. General CUDA</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">Added support for using <samp class="ph codeph">_shfl</samp> intrinsics with
all first class types. User source code already implementing this feature should be
guarded with <samp class="ph codeph">(CUDA_VERSION <= 6000)</samp> in order to compile against
CUDA 6.5.
</li>
<li class="li">On Linux, Xid 13 dmesg error reporting has been improved
to provide more detail and also to indicate which of the various potential causes of
the Xid 13 error was to blame.
</li>
<li class="li">The Linux <samp class="ph codeph">.run</samp> installation now comes with an
uninstallation script, <samp class="ph codeph">uninstall_cuda_6.5.pl</samp>, to help with
uninstalling the toolkit during conversions to Debian/RPM installations.
</li>
<li class="li">On Linux, stubs that applications can link against at build
time have been added for each library. This removes the need to have the full
library installed when building an application. In addition to the CUDA Toolkit
libraries, a stub has been provided for the CUDA Driver library
(<samp class="ph codeph">libcuda.so</samp>). See the <cite class="cite">NVIDIA CUDA Getting Started Guide
for LINUX</cite> for details on how to use these stubs.
</li>
</ul>
</div>
</div>
<div class="topic concept nested1" id="cuda-tools-title-new-features"><a name="cuda-tools-title-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-tools-title-new-features" name="cuda-tools-title-new-features" shape="rect">2.2. CUDA Tools</a></h3>
<div class="topic concept nested2" id="cuda-tools-general-new-features"><a name="cuda-tools-general-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-tools-general-new-features" name="cuda-tools-general-new-features" shape="rect">2.2.1. General CUDA Tools</a></h3>
<div class="body conbody"><a name="cuda-tools-general-new-features__ul_oyy_jwr_zl" shape="rect">
<!-- --></a><ul class="ul" id="cuda-tools-general-new-features__ul_oyy_jwr_zl">
<li class="li">Improved support for CUDA FORTRAN in the command-line
debugging and profiling tools in the CUDA Toolkit, including new debugging support
for FORTRAN arrays (in Linux only), improved source-to-assembly code correlation,
and improved documentation. This improved support is available with PGI compiler
version 14.4 and higher. CUDA FORTRAN support is a beta feature in the CUDA 6.5
release.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="cuda-compiler-new-features"><a name="cuda-compiler-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-compiler-new-features" name="cuda-compiler-new-features" shape="rect">2.2.2. CUDA Compiler</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">(Windows) Support has been added for the C++ compiler (VC 12)
in Microsoft Visual Studio 2013 for Windows.
</li>
<li class="li">The default target GPU architecture (<samp class="ph codeph">-arch</samp>)
for <samp class="ph codeph">nvcc</samp> has changed from sm_10 in previous releases to sm_20 in
this release. Note that sm_20 is not the minimum target architecture supported by
<samp class="ph codeph">nvcc</samp>, since sm_11, sm_12, and sm_13 are still valid target GPU
architectures if specified explicitly.
</li>
<li class="li">A new tool in CUDA 6.5, <samp class="ph codeph">nvprune</samp>, prunes an
object to only contain the compiled code for the specified architectures (for
example, selects only the sm_35 code for <samp class="ph codeph">libcublas_static.a</samp>). See
the <cite class="cite">CUDA Binary Utilities</cite> document for more information.
</li>
<li class="li">(Linux) The <samp class="ph codeph">cuobjdump</samp> utility for examining
CUDA binaries is now available on Linux distributions running natively on the ARM
architecture; this includes Android OS.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="cuda-occupancy-calculator-new-features"><a name="cuda-occupancy-calculator-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-occupancy-calculator-new-features" name="cuda-occupancy-calculator-new-features" shape="rect">2.2.3. CUDA Occupancy Calculator</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">Added CUDA occupancy calculator and occupancy-based launch
configuration API interfaces. These functions help set up execution configurations with
reasonable occupancy.
<p class="p">The stand-alone programmatic occupancy calculator implementation,
<samp class="ph codeph">cuda_occupancy.h</samp>, is rewritten and out of beta. Note that the API has
changed significantly from the beta version included with CUDA 6.0. This file includes
stand-alone implementations of both the occupancy calculator and the occupancy-based
launch configuration functions, so applications can use them without depending on the
entire CUDA software stack.
</p>
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="cupti-new-features"><a name="cupti-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cupti-new-features" name="cupti-new-features" shape="rect">2.2.4. CUDA Profiling Tools Interface (CUPTI)</a></h3>
<div class="body conbody"><a name="cupti-new-features__ul_jy3_4dc_yl" shape="rect">
<!-- --></a><ul class="ul" id="cupti-new-features__ul_jy3_4dc_yl">
<li class="li">Instruction classification is done for the source-correlated
instruction execution activity <samp class="ph codeph">CUpti_ActivityInstructionExecution</samp>.
See <samp class="ph codeph">CUpti_ActivityInstructionClass</samp> for the instruction classes.
</li>
<li class="li">
Two new device attributes were added to the activity
<samp class="ph codeph">CUpti_DeviceAttribute</samp>: <a name="cupti-new-features__ul_lbn_wvf_yn" shape="rect">
<!-- --></a><ul class="ul" id="cupti-new-features__ul_lbn_wvf_yn">
<li class="li"><samp class="ph codeph">CUPTI_DEVICE_ATTR_FLOP_SP_PER_CYCLE</samp>. Peak single-precision
floating-point operations that can be performed in one cycle by the
device.
</li>
<li class="li"><samp class="ph codeph">CUPTI_DEVICE_ATTR_FLOP_DP_PER_CYCLE</samp>. Peak double-precision
floating-point operations that can be performed in one cycle by the
device.
</li>
</ul>
</li>
<li class="li">
Two new metric device properties were added: <a name="cupti-new-features__ul_gxl_ctm_14" shape="rect">
<!-- --></a><ul class="ul" id="cupti-new-features__ul_gxl_ctm_14">
<li class="li"><samp class="ph codeph">CUPTI_METRIC_PROPERTY_FLOP_SP</samp>. Peak single-precision
floating point operations that can be performed in one cycle by the
device.
</li>
<li class="li"><samp class="ph codeph">CUPTI_METRIC_PROPERTY_FLOP_DP</samp>. Peak double-precision
floating-point operations that can be performed in one cycle by the
device.
</li>
</ul>
</li>
<li class="li">Activity record <samp class="ph codeph">CUpti_ActivityGlobalAccess2</samp>
for source-level global access information replaces
<samp class="ph codeph">CUpti_ActivityGlobalAccess</samp>, which has been deprecated. The new
record adds information needed to map SASS assembly instructions to CUDA C source
code; it also provides ideal L2 transaction counts based on access patterns.
</li>
<li class="li">Activity record <samp class="ph codeph">CUpti_ActivityBranch2</samp> for
source-level branch information replaces <samp class="ph codeph">CUpti_ActivityBranch</samp>,
which has been deprecated. The new record adds information needed to map SASS
assembly instructions to CUDA C source code.
</li>
<li class="li">Added a new sample to show how to map SASS assembly
instructions to CUDA C source lines.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="visual-profiler-new-features"><a name="visual-profiler-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#visual-profiler-new-features" name="visual-profiler-new-features" shape="rect">2.2.5. NVIDIA Visual Profiler</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">Visual Profiler now displays peak single-precision flops and
peak double-precision flops for a GPU under <span class="ph uicontrol">Device</span>
properties.
</li>
<li class="li">The Visual Profiler <span class="ph uicontrol">Kernel</span> profile
analysis view has been updated with several enhancements. <a name="visual-profiler-new-features__ul_qxh_cl1_wn" shape="rect">
<!-- --></a><ul class="ul" id="visual-profiler-new-features__ul_qxh_cl1_wn">
<li class="li">Initially, the instruction with the maximum execution count is
highlighted.
</li>
<li class="li">A bar is shown in the background of the counter value for the
<span class="ph uicontrol">Exec Count</span> column to make it easier to identify
instructions with high execution counts.
</li>
<li class="li">The current assembly instruction block is highlighted using two horizontal
lines around the block. Also, <span class="ph uicontrol">next</span> and
<span class="ph uicontrol">previous</span> buttons have been added to move to the
next or previous block of assembly instructions.
</li>
<li class="li">Syntax highlighting is done for the CUDA C source.</li>
<li class="li">A tooltip describing each column has been added.</li>
</ul>
</li>
<li class="li">The Visual Profiler <span class="ph uicontrol">Kernel</span> memory
analysis view has been updated with several enhancements. <a name="visual-profiler-new-features__ul_gn5_gn1_wn" shape="rect">
<!-- --></a><ul class="ul" id="visual-profiler-new-features__ul_gn5_gn1_wn">
<li class="li">Added ECC overhead, which provides a count of memory transactions required
for ECC.
</li>
<li class="li">For L2 cache, added a split of transactions for L1 reads, L1 writes, texture
reads, atomic reads, and noncoherent reads.
</li>
<li class="li">For L1 cache, added a count of atomic transactions.</li>
</ul>
</li>
<li class="li">Visual Profiler and <samp class="ph codeph">nvprof</samp> now support a new
application replay mode for collecting multiple events and metrics. In this mode,
the application is run multiple times instead of using kernel replay. This is useful
for cases when the kernel uses a large amount of device memory and the use of kernel
replay is slow due to the high overhead of saving and restoring device memory for
each kernel replay run. In Visual Profiler, this new application replay mode is
enabled in the <span class="ph uicontrol">New Session</span> dialog.
</li>
</ul>
</div>
</div>
</div>
<div class="topic concept nested1" id="cuda-libraries-title-new-features"><a name="cuda-libraries-title-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-libraries-title-new-features" name="cuda-libraries-title-new-features" shape="rect">2.3. CUDA Libraries</a></h3>
<div class="topic concept nested2" id="cuda-libraries-general-new-features"><a name="cuda-libraries-general-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-libraries-general-new-features" name="cuda-libraries-general-new-features" shape="rect">2.3.1. General CUDA Libraries</a></h3>
<div class="body conbody"><a name="cuda-libraries-general-new-features__ul_oyy_jwr_zl" shape="rect">
<!-- --></a><ul class="ul" id="cuda-libraries-general-new-features__ul_oyy_jwr_zl">
<li class="li">Starting with the CUDA 6.5 release on Linux and Mac OS, the
cuBLAS, cuSPARSE, cuFFT, cuRAND, and NPP libraries are provided as static libraries
in addition to being provided as shared libraries. These new static libraries depend
on a common thread abstraction layer library cuLIBOS (<samp class="ph codeph">libculibos.a</samp>)
that is now distributed as part of the toolkit. Consequently, cuLIBOS must be
provided to the linker when at least one of these static libraries is being linked
against. For example, on Linux, to compile an application using cuBLAS and cuRAND
against the static versions of these libraries, the following command should be
used:<pre xml:space="preserve">gcc myApp.c libcublas_static.a libcurand_static.a libculibos.a -o myApp</pre>Note
that <samp class="ph codeph">libculibos.a</samp> is <strong class="ph b">not</strong> needed when the shared version of
these libraries is used.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="cublas-new-features"><a name="cublas-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cublas-new-features" name="cublas-new-features" shape="rect">2.3.2. cuBLAS Library</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">The <samp class="ph codeph">cublas<T>trsmBatched()</samp> routine no
longer limits the <samp class="ph codeph">m</samp> and <samp class="ph codeph">n</samp> dimensions to 32.
However, the routine is still intended to be used for matrices of relatively small
size, for which the performance of calling <samp class="ph codeph">cublas<T>trsm()</samp>
multiple times would be limited by kernel launch overhead. Performance has also been
significantly improved for <samp class="ph codeph">n</samp> > 1.
</li>
<li class="li">The cuBLAS Library now offers the batched routines
<samp class="ph codeph">cublas<T>geqrfBatched()</samp> and
<samp class="ph codeph">cublas<T>gelsBatched()</samp>, which are respectively a batched QR
factorization and a batched least-squares solver for over-determined systems.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="cufft-new-features"><a name="cufft-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cufft-new-features" name="cufft-new-features" shape="rect">2.3.3. cuFFT Library</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">User-specified routines can now operate directly on cuFFT input
or output data. The new <samp class="ph codeph">cufftXt*Callback()</samp> APIs are used to specify
which user-defined routines will be called when each data point is loaded or stored
by the cuFFT kernels, potentially reducing the overall number of accesses to device
memory.
</li>
<li class="li">Starting with cuFFT in CUDA 6.5, single 2D or 3D FFTs on
multiple GPUs can be performed without the need for transposing data between
successive FFTs. In prior releases it was necessary to transpose the data before
performing a second FFT on multiple GPUs.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="cusparse-new-features"><a name="cusparse-new-features" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cusparse-new-features" name="cusparse-new-features" shape="rect">2.3.4. cuSPARSE Library</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">The cuSPARSE Library added two new routines that support the
BSR format, <samp class="ph codeph">cusparse<T>bsrmm()</samp> and
<samp class="ph codeph">cusparse<T>bsrsm()</samp>, which are respectively the
multiplication of a matrix in BSR format by a dense matrix, and the solve of a
triangular matrix in BSR format against multiple right-hand sides.
</li>
</ul>
</div>
</div>
</div>
</div>
<div class="topic concept nested0" id="unsupported-features"><a name="unsupported-features" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#unsupported-features" name="unsupported-features" shape="rect">3. Unsupported Features
</a></h2>
<div class="body conbody">
<p class="p"> The following features are officially unsupported in the current release. Developers
must employ alternative solutions to these features in their software.
</p>
<dl class="dl">
<dt class="dt dlterm">Windows XP 64-bit Edition Support</dt>
<dd class="dd">With this release, CUDA no longer supports the 64-bit version of the Windows XP
operating system, although CUDA on the 32-bit version of Windows XP is still
supported. We recommend that developers and users of the 64-bit version of
Windows XP migrate to Windows 7 or Windows 8.1, which are supported in the
current and future CUDA releases.
</dd>
<dt class="dt dlterm">Windows Vista Support</dt>
<dd class="dd">This CUDA release no longer supports the Windows Vista operating system. We
recommend that users and developers migrate to Windows 7 or Windows 8.1, which
are supported in the current and future releases.
</dd>
<dt class="dt dlterm">Windows Server 2012 Support</dt>
<dd class="dd">CUDA on the Windows Server 2012 operating system is not supported in this CUDA
release. We recommend that users and developers migrate to Windows Server 2012
R2, which is supported in the current and future releases.
</dd>
<dt class="dt dlterm">(Linux) Support for 32-bit Applications on
x86-based Linux Distributions
</dt>
<dd class="dd">Several portions of the CUDA Toolkit are no longer available for developing
32-bit applications on x86-based Linux distributions:<a name="unsupported-features__ul_jth_pjb_d4" shape="rect">
<!-- --></a><ul class="ul" id="unsupported-features__ul_jth_pjb_d4">
<li class="li">.deb installer packages for 32-bit CUDA Toolkit components </li>
<li class="li">CUDA Toolkit scientific libraries, including cuBLAS, cuSPARSE, cuFFT,
cuRAND, and NPP
</li>
<li class="li">Thrust </li>
<li class="li">Quadro and Tesla products</li>
<li class="li">Tesla (sm_1x) and Fermi (sm_2x) architectures</li>
<li class="li">CUDA Samples</li>
</ul>
</dd>
<dd class="dd">The above list also applies to 32-bit components and 32-bit rpm/deb packages on
64-bit x86-based Linux distributions. The 64-bit components are unaffected by
these changes.
</dd>
<dt class="dt dlterm">(Mac OS X) Support for 32-bit CUDA and OpenCL
Applications on Mac OS X
</dt>
<dd class="dd">Developing and running 32-bit CUDA and OpenCL applications on Mac OS X platforms
is no longer supported in the CUDA Toolkit and in the CUDA Driver. Legacy 32-bit
CUDA and OpenCL applications will not run on this version of the CUDA Driver on
Mac OS X platforms.
</dd>
<dt class="dt dlterm">Targeting sm_10 (G80) for
CUDA Applications
</dt>
<dd class="dd">The CUDA Toolkit no longer supports the sm_10 target architecture (the G80
architecture) for CUDA and OpenCL applications.
</dd>
<dt class="dt dlterm">CUDA Video Encoder (NVCUVENC)</dt>
<dd class="dd">Building applications with the CUDA Video Encoder interface is no longer
supported; however, the driver will continue to run applications built against
this interface. We recommend using the NVIDIA Encoder API (NVENC), a newer video
encoding interface that is available at <a class="xref" href="https://developer.nvidia.com/nvidia-video-codec-sdk" target="_blank" shape="rect">https://developer.nvidia.com/nvidia-video-codec-sdk</a>.
</dd>
</dl>
</div>
</div>
<div class="topic concept nested0" id="deprecated-features"><a name="deprecated-features" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#deprecated-features" name="deprecated-features" shape="rect">4. Deprecated Features
</a></h2>
<div class="body conbody">
<p class="p">The following features are deprecated in the current release of the CUDA software. The
features still work in the current release, but their documentation may have been
removed, and they will become officially unsupported in a future release. We recommend
that developers employ alternative solutions to these features in their software.
</p>
<dl class="dl">
<dt class="dt dlterm">Tesla and Quadro Products and CUDA Toolkit on
32-bit Windows Platforms
</dt>
<dd class="dd">Support for the CUDA Toolkit on 32-bit Windows platforms is deprecated, as is
support for Tesla and Quadro products for the CUDA driver on 32-bit Windows
platforms. Additionally, on 64-bit Windows platforms, support for the following
features for 32-bit CUDA and OpenCL applications is deprecated from the CUDA
driver and CUDA toolkit, as appropriate: <a name="deprecated-features__ul_wgw_hkp_yn" shape="rect">
<!-- --></a><ul class="ul" id="deprecated-features__ul_wgw_hkp_yn">
<li class="li">Tesla and Quadro products</li>
<li class="li">CUDA Toolkit scientific libraries, including cuBLAS, cuSPARSE, cuFFT,
cuRAND, and NPP
</li>
<li class="li">Thrust</li>
<li class="li">CUDA samples</li>
</ul>
</dd>
<dd class="dd">This deprecation notice doesn't impact any 64-bit components. </dd>
</dl>
<dl class="dl">
<dt class="dt dlterm">Interop with IDirect3D9 objects on Microsoft
Windows 7 and Later
</dt>
<dd class="dd">This release deprecates support for interop with <samp class="ph codeph">IDirect3D9</samp>
objects on Windows 7 and later Microsoft operating systems. This applies to the
<samp class="ph codeph">cuD3D9*()</samp> and
<samp class="ph codeph">cuGraphicsD3D9RegisterResource()</samp> routines in the Driver
API, as well as the corresponding <samp class="ph codeph">cudaD3D9*()</samp> and
<samp class="ph codeph">cudaGraphicsD3D9RegisterResource()</samp> routines in the Runtime
API. We recommend using <samp class="ph codeph">IDirect3D9ex</samp> objects, which will work
with these same routines, instead.
</dd>
</dl>
<dl class="dl">
<dt class="dt dlterm">Linux RHEL 5 and CentOS 5 Support</dt>
<dd class="dd">Support for CUDA on the RHEL 5 and CentOS 5 Linux distributions is deprecated in
this CUDA release and will be dropped in a future release. We recommend that
users and developers migrate to RHEL 6, which is supported in the current and
future releases.
</dd>
</dl>
<dl class="dl">
<dt class="dt dlterm">Support for sm_10, sm_11, sm_12, and sm_13
Architectures
</dt>
<dd class="dd">The sm_10 architecture is deprecated within the CUDA Driver, and the sm_11,
sm_12, and sm_13 architectures are deprecated within the CUDA Toolkit and the
CUDA Driver. Support for these architectures will be removed in the next major
version of the CUDA Toolkit and Driver. Note that support for the sm_10
architecture has already been removed from the CUDA Toolkit.
</dd>
</dl>
<dl class="dl">
<dt class="dt dlterm">Developing and Running 32-bit CUDA and OpenCL
Applications on x86 Linux Platforms
</dt>
<dd class="dd">Support for developing and running 32-bit CUDA and OpenCL applications on x86
Linux platforms is deprecated. This implies the following:<a name="deprecated-features__ul_xcb_nhp_yn" shape="rect">
<!-- --></a><ul class="ul" id="deprecated-features__ul_xcb_nhp_yn">
<li class="li">Support is currently still available in the toolkit and driver.</li>
<li class="li">Support may be dropped from the toolkit in a future release, and
similarly from the driver.
</li>
<li class="li">New features may not have support for 32-bit x86 Linux
applications.
</li>
<li class="li">This notice applies to running applications on a 32-bit Linux kernel,
and also to running 32-bit applications on a 64-bit Linux kernel.
</li>
<li class="li">This notice applies to x86 architectures only; 32-bit Linux applications
are still officially supported and are not deprecated on the ARM
architecture.
</li>
</ul>
</dd>
</dl>
<dl class="dl">
<dt class="dt dlterm">CUPTI Activity Records</dt>
<dd class="dd">Activity record <samp class="ph codeph">CUpti_ActivityGlobalAccess</samp> for source-level
global access information has been deprecated and replaced by the new activity
record <samp class="ph codeph">CUpti_ActivityGlobalAccess2</samp>. Activity record
<samp class="ph codeph">CUpti_ActivityBranch</samp> for source-level branch information
has been deprecated and replaced by the new activity record
<samp class="ph codeph">CUpti_ActivityBranch2</samp>.
</dd>
</dl>
</div>
</div>
<div class="topic concept nested0" id="performance-improvements-title"><a name="performance-improvements-title" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#performance-improvements-title" name="performance-improvements-title" shape="rect">5. Performance Improvements
</a></h2>
<div class="topic concept nested1" id="cuda-general-performance-improvements"><a name="cuda-general-performance-improvements" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-general-performance-improvements" name="cuda-general-performance-improvements" shape="rect">5.1. General CUDA</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">MPS performance has been improved: launch performance has been
improved from 7 to 5 microseconds; launch and synchronize performance has been
improved from 35 to 15 microseconds.
</li>
</ul>
</div>
</div>
<div class="topic concept nested1" id="cuda-libraries-title-performance-improvements"><a name="cuda-libraries-title-performance-improvements" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-libraries-title-performance-improvements" name="cuda-libraries-title-performance-improvements" shape="rect">5.2. CUDA Libraries</a></h3>
<div class="topic concept nested2" id="math-performance-improvements"><a name="math-performance-improvements" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#math-performance-improvements" name="math-performance-improvements" shape="rect">5.2.1. CUDA Math Library</a></h3>
<div class="body conbody"><a name="math-performance-improvements__ul_fq4_jfq_wn" shape="rect">
<!-- --></a><ul class="ul" id="math-performance-improvements__ul_fq4_jfq_wn">
<li class="li">Performance has been increased for these single-precision
functions in CUDA 6.5: <samp class="ph codeph">acoshf()</samp>, <samp class="ph codeph">asinhf()</samp>,
<samp class="ph codeph">atanf()</samp>, <samp class="ph codeph">atan2f()</samp>, <samp class="ph codeph">atanhf()</samp>,
<samp class="ph codeph">cyl_bessel_i0f()</samp>, <samp class="ph codeph">cyl_bessel_i1f()</samp>,
<samp class="ph codeph">cbrtf()</samp>, <samp class="ph codeph">coshf()</samp>, <samp class="ph codeph">erfcf()</samp>,
<samp class="ph codeph">erfcinvf()</samp>, <samp class="ph codeph">erfcxf()</samp>,
<samp class="ph codeph">erfinvf()</samp>, <samp class="ph codeph">expf()</samp>, <samp class="ph codeph">exp10f()</samp>,
<samp class="ph codeph">expm1f()</samp>, <samp class="ph codeph">fdiv_rd()</samp>,
<samp class="ph codeph">fdiv_rn()</samp>, <samp class="ph codeph">fdiv_ru()</samp>,
<samp class="ph codeph">fdiv_rz()</samp>, <samp class="ph codeph">fmodf()</samp>,
<samp class="ph codeph">frcp_rd()</samp>, <samp class="ph codeph">frcp_rn()</samp>, <samp class="ph codeph">frcp_ru()</samp>,
<samp class="ph codeph">frcp_rz()</samp>, <samp class="ph codeph">frsqrt_rn()</samp>,
<samp class="ph codeph">hypotf()</samp>, <samp class="ph codeph">logf()</samp>, <samp class="ph codeph">log10f()</samp>,
<samp class="ph codeph">log1pf()</samp>, <samp class="ph codeph">log2f()</samp>,
<samp class="ph codeph">normcdff()</samp>, <samp class="ph codeph">normcdfinvf()</samp>,
<samp class="ph codeph">powf()</samp>, <samp class="ph codeph">remainderf()</samp>, <samp class="ph codeph">remquof()</samp>,
<samp class="ph codeph">rhypotf()</samp>, <samp class="ph codeph">sincospif()</samp>,
<samp class="ph codeph">sinhf()</samp>, <samp class="ph codeph">sinpif()</samp>, and
<samp class="ph codeph">tanhf()</samp>. Of these, <samp class="ph codeph">atanf()</samp>,
<samp class="ph codeph">expf()</samp>, <samp class="ph codeph">exp10f()</samp>, <samp class="ph codeph">expm1f()</samp>,
<samp class="ph codeph">hypotf()</samp>, and <samp class="ph codeph">rhypotf()</samp> show especially marked
improvement.
</li>
<li class="li">Performance has been increased for these
double-precision functions in CUDA 6.5: <samp class="ph codeph">acosh()</samp>,
<samp class="ph codeph">asin()</samp>, <samp class="ph codeph">asinh()</samp>, <samp class="ph codeph">atan()</samp>,
<samp class="ph codeph">atanh()</samp>, cyl_bessel_i0, <samp class="ph codeph">cyl_bessel_i1()</samp>,
<samp class="ph codeph">cbrt()</samp>, <samp class="ph codeph">cospi()</samp>, <samp class="ph codeph">div()</samp>,
<samp class="ph codeph">erfc()</samp>, <samp class="ph codeph">erfcx()</samp>, <samp class="ph codeph">erfinv()</samp>,
<samp class="ph codeph">exp2()</samp>, <samp class="ph codeph">fmod()</samp>, <samp class="ph codeph">hypot()</samp>,
<samp class="ph codeph">log()</samp>, <samp class="ph codeph">log10()</samp>, <samp class="ph codeph">log1p()</samp>,
<samp class="ph codeph">log2()</samp>, <samp class="ph codeph">normcdf()</samp>, pow(),
<samp class="ph codeph">rcbrt()</samp>, <samp class="ph codeph">remainder()</samp>,
<samp class="ph codeph">remquo()</samp>, <samp class="ph codeph">rhypot()</samp>,
<samp class="ph codeph">sincospi()</samp>, <samp class="ph codeph">sinpi()</samp>, and
<samp class="ph codeph">tan()</samp>. Of these, <samp class="ph codeph">acosh()</samp>,
<samp class="ph codeph">atan()</samp>, <samp class="ph codeph">cbrt()</samp>, <samp class="ph codeph">hypot()</samp>, and
<samp class="ph codeph">rhypot()</samp> show especially marked improvement.
</li>
<li class="li">Performance of the double-precision square root function,
<samp class="ph codeph">sqrt()</samp>, was significantly improved for GPUs with compute
capability 2.0 and above.
</li>
<li class="li">Performance of the double-precision reciprocal square root
function, <samp class="ph codeph">rsqrt()</samp>, was significantly improved for GPUs with compute
capability 2.0 and above.
</li>
</ul>
</div>
</div>
</div>
</div>
<div class="topic concept nested0" id="resolved-issues-title"><a name="resolved-issues-title" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#resolved-issues-title" name="resolved-issues-title" shape="rect">6. Resolved Issues
</a></h2>
<div class="topic concept nested1" id="cuda-general-resolved-issues"><a name="cuda-general-resolved-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-general-resolved-issues" name="cuda-general-resolved-issues" shape="rect">6.1. General CUDA</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">(Linux) A driver packaging issue that forced users on Redhat
and Fedora to ensure that the <samp class="ph codeph">xorg-x11-drv-nvidia-devel</samp> package was
installed has been resovled.
</li>
<li class="li">The device memory heap size, set using
<samp class="ph codeph">cudaDeviceSetLimit(cudaLimitMallocHeapSize, *)</samp> or
<samp class="ph codeph">cuCtxSetLimit(CU_LIMIT_MALLOC_HEAP_SIZE, *)</samp>, is no longer
limited to a size of 4,294,967,296 (4 GB).
</li>
</ul>
</div>
</div>
</div>
<div class="topic concept nested0" id="known-issues-title"><a name="known-issues-title" shape="rect">
<!-- --></a><h2 class="title topictitle1"><a href="#known-issues-title" name="known-issues-title" shape="rect">7. Known Issues</a></h2>
<div class="topic concept nested1" id="armv7-linux-general-cuda-known-issues"><a name="armv7-linux-general-cuda-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#armv7-linux-general-cuda-known-issues" name="armv7-linux-general-cuda-known-issues" shape="rect">7.1. Linux on ARMv7 Specific Issues</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">Mapping host memory allocated outside
of CUDA to device memory is not allowed on ARM; because of this,
<samp class="ph codeph">cudaHostRegister()</samp> is not supported by the CUDA driver on ARM
platforms. If required, <samp class="ph codeph">cudaHostAlloc()</samp> with the flag
<samp class="ph codeph">cudaHostAllocMapped</samp> can be used to allocate device-mapped
host-accessible memory.
</li>
</ul>
</div>
</div>
<div class="topic concept nested1" id="cuda-general-known-issues"><a name="cuda-general-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-general-known-issues" name="cuda-general-known-issues" shape="rect">7.2. General CUDA</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">The <samp class="ph codeph">cuda</samp> and
<samp class="ph codeph">gpu-deployment-kit</samp> packages must be installed by separate
executions of <samp class="ph codeph">yum</samp>. See the <cite class="cite">Linux Getting Started Guide</cite>
for more details.
</li>
<li class="li">On openSUSE and SLES, X will fail to load if the CUDA Toolkit
RPM packages are installed using relocation immediately following an installation of
the <samp class="ph codeph">cuda-drivers</samp> package (and its dependencies). Users should
reboot in between the driver and toolkit installations. Executing
<samp class="ph codeph">nvidia-xconfig</samp> may rescue a system where X has failed to load
in this situation.
</li>
<li class="li">The CUDA drivers may fail to install if the RPMFusion
repository is enabled at the same time as the CUDA repository. When installing CUDA,
the <samp class="ph codeph">--disablerepo="rpmfusion-nonfree*"</samp> option should be used. For
example, to install the <samp class="ph codeph">cuda</samp> package: <samp class="ph codeph">yum
--disablerepo="rpmfusion-nonfree*" install cuda</samp>.
</li>
<li class="li">(Mac OS) When CUDA applications are run on 2012 MacBook Pro
models, allowing or forcing the system to go to sleep causes a system crash (kernel
panic). To prevent the computer from automatically going to sleep, set the
<span class="ph uicontrol">Computer Sleep</span> option slider to
<span class="ph uicontrol">Never</span> in the <span class="ph uicontrol">Energy Saver</span> pane of
the <span class="ph uicontrol">System Preferences</span>.
</li>
<li class="li">The CUDA reference manual incorrectly describes the
type of <samp class="ph codeph">CUdeviceptr</samp> as an <samp class="ph codeph">unsigned int</samp> on all
platforms. On 64-bit platforms, a <samp class="ph codeph">CUdeviceptr</samp> is an
<samp class="ph codeph">unsigned long long</samp>, not an <samp class="ph codeph">unsigned int</samp>.
</li>
<li class="li">Peer access is disabled between two devices if either
of them is in SLI mode.
</li>
<li class="li">On multi-GPU configurations without P2P support
between any pair of devices that support Unified Memory, managed memory allocations
are placed in zero-copy memory. When data is migrated, this results in lower
performance than the default managed memory behavior. In certain cases, the
environment variable <samp class="ph codeph">CUDA_MANAGED_FORCE_DEVICE_ALLOC</samp> can be set to
force managed allocations to be in device memory and to enable migration on these
hardware configurations. Normally, using the environment variable
<samp class="ph codeph">CUDA_VISIBLE_DEVICES</samp> is recommended to restrict CUDA to only
use those GPUs that have P2P support. Please refer to the environment variables
section in the <cite class="cite">CUDA C Programming Guide</cite> for further details.
</li>
</ul>
</div>
</div>
<div class="topic concept nested1" id="cuda-tools-title-known-issues"><a name="cuda-tools-title-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-tools-title-known-issues" name="cuda-tools-title-known-issues" shape="rect">7.3. CUDA Tools</a></h3>
<div class="topic concept nested2" id="cuda-compiler-known-issues"><a name="cuda-compiler-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-compiler-known-issues" name="cuda-compiler-known-issues" shape="rect">7.3.1. CUDA Compiler</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">
(Mac OS) When Clang is used as the host compiler,
32-bit target compilation on OS X is not supported. This is because the Clang
compiler doesn't support the <samp class="ph codeph">-malign-double</samp> switch that the NVCC
compiler needs to properly align double-precision structure fields when compiling
for a 32-bit target (GCC does support this switch). Note that GCC is the default
host compiler used by NVCC on OS X 10.8 and Clang is the default on OS X 10.9.
</li>
<li class="li">
The NVCC compiler doesn't accept Unicode
characters in any filename or path provided as a command-line parameter.
</li>
<li class="li">
A CUDA program may not compile correctly if
a <samp class="ph codeph">type</samp> or <samp class="ph codeph">typedef</samp><samp class="ph codeph">T</samp> is private to a class or a structure, and at least one of the
following is satisfied: <a name="cuda-compiler-known-issues__ul_e4m_jgt_5n" shape="rect">
<!-- --></a><ul class="ul" id="cuda-compiler-known-issues__ul_e4m_jgt_5n">
<li class="li"><samp class="ph codeph">T</samp> is a parameter type for a <samp class="ph codeph">__global__</samp>
function.
</li>
<li class="li"><samp class="ph codeph">T</samp> is an argument type for a template instantiation of a
<samp class="ph codeph">__global__</samp> function.
</li>
</ul>
This restriction will be fixed in a future release.
</li>
<li class="li">
(Mac OS) The documentation surrounding the use of the
flag <samp class="ph codeph">-malign-double</samp> suggests it be used to make the struct size the
same between host and device code. We know now that this flag causes problems with
other host libraries. The CUDA documentation will be updated to reflect this. The
workaround for this issue is to manually add padding so that the structs between the
host compiler and CUDA are consistent.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="cuda-gdb-known-issues"><a name="cuda-gdb-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-gdb-known-issues" name="cuda-gdb-known-issues" shape="rect">7.3.2. CUDA-GDB</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">There can be a significant performance degradation for large
routines when the debugger steps over inlined routines. This happens because inlined
code blocks may have multiple exit points under the hood, and the debugger steps
every single instruction until an exit point is reached.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="nsight-ee-known-issues"><a name="nsight-ee-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#nsight-ee-known-issues" name="nsight-ee-known-issues" shape="rect">7.3.3. Nsight Eclipse Edition</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">On Linux, the NVIDIA Visual Profiler
(<samp class="ph codeph">nvvp</samp>) and the Nsight IDE (<samp class="ph codeph">nsight</samp>) do not run
properly when the oxygen-gtk theme is used. If you experience such crashes, please
uninstall the oxygen-gtk theme. The command to do this on OpenSUSE is <samp class="ph codeph">sudo
zypper rm gtk2-theme-oxygen</samp> and on Ubuntu is <samp class="ph codeph">sudo apt-get
remove gtk2-engines-oxygen</samp>.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="visual-profiler-known-issues"><a name="visual-profiler-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#visual-profiler-known-issues" name="visual-profiler-known-issues" shape="rect">7.3.4. NVIDIA Visual Profiler</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">(Windows) Using the mouse wheel button to scroll does
not work within the Visual Profiler on Windows.
</li>
<li class="li">(Mac OS) Visual Profiler events and
metrics do not work correctly on Mac OS X 10.9.3. Mac OS X 10.9.2 can be used as a
workaround.
</li>
</ul>
</div>
</div>
</div>
<div class="topic concept nested1" id="cuda-libraries-title-known-issues"><a name="cuda-libraries-title-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-libraries-title-known-issues" name="cuda-libraries-title-known-issues" shape="rect">7.4. CUDA Libraries</a></h3>
<div class="topic concept nested2" id="cufft-known-issues"><a name="cufft-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cufft-known-issues" name="cufft-known-issues" shape="rect">7.4.1. cuFFT Library</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">In the CUDA 6.5 Early Access release, there are some
limitations in the cuFFT callback implementation. <a name="cufft-known-issues__ul_wd4_lt3_yn" shape="rect">
<!-- --></a><ul class="ul" id="cufft-known-issues__ul_wd4_lt3_yn">
<li class="li">The static version of the cuFFT library is not supported on 32-bit Windows
systems; consequently, the callback feature is not supported there
either.
</li>
<li class="li">If the size of any dimension cannot be factored into a combination of powers
of 2, 3, 5, and 7 (that is, the size has a prime factor of 11 or greater),
the callback routine cannot safely call
<samp class="ph codeph">__syncthreads()</samp>.
</li>
<li class="li">For 2D and 3D transforms, if the size of any dimension has a prime factor of
131 or greater, <samp class="ph codeph">cufftUnsetCallback()</samp> does not function
correctly.
</li>
<li class="li">For 2D and 3D C2C transforms, if any dimension has a prime factor of 131 or
greater, the <samp class="ph codeph">store()</samp> callback does not function
correctly.
</li>
<li class="li">For multi-GPU C2R and R2C plans, callbacks are not supported if the batch
size is greater than one and any dimension has a prime factor of 131 or
greater.
</li>
</ul>
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="thrust-known-issues"><a name="thrust-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#thrust-known-issues" name="thrust-known-issues" shape="rect">7.4.2. Thrust Library</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">(Linux) There is a known issue that causes the
<samp class="ph codeph">TestGetTemporaryBufferDispatchExplicit</samp> and
<samp class="ph codeph">TestGetTemporaryBufferDispatchImplicit</samp> unit tests provided with
the Thrust library to fail on the SLES 11 Linux distribution.
</li>
<li class="li">(Linux) There is a known issue that causes the
<samp class="ph codeph">segmentationTreeThrust</samp> CUDA sample in the
<samp class="ph codeph">6_Advanced</samp> directory to fail on the SLES 11 Linux
distribution.
</li>
</ul>
</div>
</div>
<div class="topic concept nested2" id="cuda-samples-known-issues"><a name="cuda-samples-known-issues" shape="rect">
<!-- --></a><h3 class="title topictitle2"><a href="#cuda-samples-known-issues" name="cuda-samples-known-issues" shape="rect">7.4.3. CUDA Samples</a></h3>
<div class="body conbody">
<ul class="ul">
<li class="li">On 32-bit Windows systems, certain samples may fail to compile
due to the compiler exhausting available memory, especially if compiling is done in
Debug mode or if the sample is using Dynamic Parallelism.
</li>
</ul>
</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">2007</span>-<span class="ph">2014</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></body>
</html>
distrib
>
Mageia
>
5
>
x86_64
>
media
>
nonfree-release
>
by-pkgid
>
d44b02ea46d82d6a48df31bbd1a088f3
>
files
>
1210
