Archive notice: This article was originally published on November 19, 2009. Links and embedded videos are preserved as part of the historical record.
Together with its partners, NVIDIA says it will release numerous new GPU-computing software products over the coming months. The updates cover a broad spectrum of GPU-computing programming languages, tools and libraries, including updates to NVIDIA’s CUDA C compiler adding support for C++ and the next generation of the CUDA GPU architecture, codenamed “Fermi”. NVIDIA is also releasing its R195 driver, containing extensions to the OpenCL 1.0-compatible driver and toolkit. It is joined by a beta version of NVIDIA “Nexus”, the first development environment for massively parallel computing to integrate with Microsoft Visual Studio.
The NVIDIA updates are complemented by partner products such as the Portland Group’s CUDA Fortran compiler, Allinea’s Distributed Debugging Tool and the TotalView debugger.
All updates to the parallel-computing development tools from NVIDIA and its partners at a glance:
- CUDA Toolkit 3.0 Beta (1): The beta version of CUDA Toolkit 3.0 allows developers to create applications for the Fermi architecture today. The beta release includes features such as ECC Reporting, Dual DMA Engine, Concurrent Kernel Execution and support for NVIDIA Fermi hardware debugging in cuda-gdb. Performance profiling is available for CUDA Visual Profiler and OpenCL Visual Profiler, together with support for the unified API for Direct3D and OpenGL, including Direct3D 11.
- OpenCL 1.0 Extensions (2): As the only manufacturer, NVIDIA supports OpenCL features beyond the lowest common denominator, such as Double Precision, OpenGL interoperability and OpenCL Installable Client Device (ICD), in addition to NVIDIA-only support for 2D Image, 32-Bit Atomics and Byte Addressable Stores.
- NVIDIA “Nexus” (3): This is the first integrated CPU/GPU development environment for Microsoft Visual Studio. Nexus consists of three components: Debugger, Performance Analyzer and Graphics Inspector. NVIDIA Nexus will significantly increase productivity because developers of GPU-based applications can continue to use the familiar Microsoft Visual Studio tools.
- Portland Group CUDA Fortran Compiler (4): Using the CUDA Fortran compiler, Fortran developers can use the massively parallel computing power of NVIDIA graphics processors to create HPC applications for scientific calculations. Applications include numerical modelling of oceans and weather forecasts, as well as many areas of geophysics and bioinformatics.
- HPC Debugging Solutions from Allinea (5) and TotalView (6): These tools contain CUDA GPU functions that complement existing Linux tools using MPI, OpenMPI and pthreads. Developers can use them to debug applications running on hybrid clusters consisting of x86-64-bit CPU and Tesla GPU servers.
- Numerical Analysis Packages: Familiar numerical-analysis and mathematical-modelling solutions such as MATLAB from MathWorks (7), Mathematica (8) from Wolfram Research and LabVIEW (9) from National Instruments are also making significant progress in using CUDA-capable GPUs more effectively.
- CUDA Libraries: Developers can also use an extensive collection of CUDA-capable libraries from NVIDIA and its partners, including BLAS, FFT, LAPACK (EM Photonics CULA), MAGMA (Innovative Computing Laboratory at the University of Tennessee, Knoxville), NVIDIA Performance Primitives (NPP), CUDA Vision Workbench (CVWB), and video- and image-processing libraries.
To take advantage of all the innovations and extensions in CUDA-optimised tools and applications, a range of CUDA consultants (10), including Acceleware Corp., ANEO, CAPS, Elegant Mathematics, EM Photonics, Fixstars, GASS Ltd., HPC Project, Infosys, SagivTech, Stone Ridge Technology and Tech-X Corp., offer training and continuing-education courses. With the breadth of programming languages, APIs, libraries and tools supported by NVIDIA GPUs based on the CUDA architecture, developers receive one of the most comprehensive parallel-computing toolsets.
1) http://www.nvidia.com/object/cuda_home.html#
2) http://www.nvidia.com/object/cuda_opencl.html
3) http://developer.nvidia.com/object/nexus.html
4) http://www.pgroup.com/resources/cudafortran.htm
5) http://www.allinea.com/?page=48
6) http://www.totalviewtech.com/products/totalview.html?via=midbox
8) http://www.wolfram.com/products/mathematica/index.html
10) http://www.nvidia.com/object/cuda_consultants.html
Further links:
- More information
More information
More information
More information
More information
More information
More information
More information
More information
More information
