The CUDA Toolkit has all the development tools, libraries, and documentation you need to create applications for the CUDA architecture.

<p><b>Highlights:</b></p>
<ul>
<li>Easier Application Porting<ul>
<li>Share GPUs across multiple threads</li>
<li>Use all GPUs in the system concurrently from a single host thread</li>
<li>No-copy pinning of system memory, a faster alternative to
cudaMallocHost()</li>
<li>C++ new/delete and support for virtual functions</li>
<li>Support for inline PTX assembly</li>
<li>Thrust library of templated performance primitives such as sort,
reduce, etc.</li>
<li>Nvidia Performance Primitives (NPP) library for image/video
processing</li>
<li>Layered Textures for working with same size/format textures at
larger sizes and higher performance</li>
</ul>
</li>
<li>Faster Multi-GPU Programming<ul>
<li>Unified Virtual Addressing</li>
<li>GPUDirect v2.0 support for Peer-to-Peer Communication</li>
</ul>
</li>
<li>New & Improved Developer Tools<ul>
<li>Automated Performance Analysis in Visual Profiler</li>
<li>C++ debugging in CUDA-GDB for Linux and MacOS</li>
<li>GPU binary disassembler for Fermi architecture (cuobjdump)</li>
<li><a href="http://developer.nvidia.com/nvidia-parallel-nsight">
Parallel Nsight 2.0</a> now available for Windows developers with new
debugging and profiling features.</li>
</ul>
</li>
</ul>
<p><b>Features:</b></p>
<ul>
<li>C/C++ compiler</li>
<li>Visual Profiler</li>
<li>GPU-accelerated BLAS library</li>
<li>GPU-accelerated FFT library</li>
<li>GPU-accelerated Sparse Matrix library</li>
<li>GPU-accelerated RNG library</li>
<li>Additional tools and documentation</li>
</ul>
<p><b>What's New:</b></p>
<p>This release of the CUDA Toolkit version 4.1 features a new LLVM-based CUDA
compiler, 1000+ new image processing functions, and a redesigned Visual Profiler
with automated performance analysis and integrated expert guidance.</p>