Acronym of CUDA

( Compute Unified Device Architecture ) K'veda Ch (CUDA) is a platform for parallel processing and the programming model is that by now, Nvidia is in units of the graphics processing, the company is supported.K'veda CH to the developer Gan software allows a GPU that features CUDA-enabled IS for the purpose of processing. an approach that GPGPU known. K'veda CH to the developer Gan possibility of direct access to memory and set of instructions in a graphics processing unit.

platform kueda to work with programming languages like C, C and Fortran, is designed.This access makes for specialists to use the resources of the GPU easier to be contrary to the way of working, another API because DIRECT3D and OpenGL, which requires the ability to professional in graphics programming.Such kovda of the framework such as OpenACC and OpenCL support.

kueda by the library, equipped with a perm, etc. compiler, such as OpenACC and, likewise, the development of the industry standard of the languages included C, C and Fortran developers is accessible.Programmers C /C of the ' CUDA C/C' s they use that compiled with the "NDTV Good Times".NDTV Good Times is a compiler C /C based on LLVM, the company Nvidia is.Programmers Fortran can also be 'CUDA Fortran' and to use that compiled with the PGI CUDA Fortran Complier company The Portland Group. In addition to the library. , the compiler and the CUDA C/C ++ and CUDA Fortran, etc. Soko K'veda CH from other interfaces, computational includes support.

OpenCL group Khronos
DirectCompute, Microsoft
shadow calculations, OpenGL
C-AMP
Also, the wrappers of the third-person (Third party wrappers for languages like Perl (the Perl), but Python (Python), the AR (R) and. Fortran (FORTRAN), etc., Java (Java) in. Ruby (Ruby), the Haskell (Haskell), a Matlab (Matlab), for iPhone LCD (IDL), etc. (Lua) And also by default (Mathematica) are available.

In Industry, computer games, etc. GPU only for graphics rendering but also in physical calculations (physical effects like smoke, fire, discharge, and debris) are also.Example, such as PhysX and bullet include this case.K'veda CH as well as for applications, the acceleration rate in computational biology, etc. encryption and other spheres are also used.

Perm a API low level, and an API provides high-level.SDK basic Perm on 15 February 2007 for Windows, Micro Soft and Linux to public release.Support in Mac OS, in the second version was added, which replace the version of the test, on February 14, 2008.Kueda with all GPU of the series, G8x to the next includes a force majeure, etc. and Tesla(graphic) works.Kueda with most operating systems, standard works.Nvidia says the programs that for the series, G8x have been developed also without change on future generations of graphics cards, depending on the compatibility of binary .

kueda several excellence in calculations, the traditional on the GPU (in the total order GPGPU) that of graphical interfaces..

reading sparse, i.e., the code can address arbitrary memory read.
virtual memory integrated (kovda version 4.0 onwards)
the integrated memory (kovda version 6.0 onwards)
shared memory perm area is a memory shared. the show, which can be among the yarns to be shared. This memory can be used as a cache managed under the user access and bandwidth, the more we have the ability to make use of the search tissue.
download faster and re-reading (reread)
full support for integers and the operations of the bit, including the search for tissue correct

---

Back