Nvidia has announced that it is in its latest Cuda 4.1 Toolkit provides support for the LLVM compiler. The compiler supplies according to Nvidia speed lwa efficient code for applications that use the gpu use and is in addition mulitplatform.
Until recently, it had developers who code for Cuda to write use of various tools from Nvidia. With the release of version 4.1 of the Cuda Toolkit, should it come to an end by the link with LLVM, a modular compiler.
The LLVM compiler would have the advantage that it is faster Cuda code can convert. Also, the final executable to perform better than the current compile-tools from Nvidia. In addition, LLVM is a modular compiler, making this not only can handle different programming languages, but also on a variety of platforms it can run. An additional advantage is that LLVM in addition to the x86 architecture code can be compiled for ARM processors.
Nvidia hopes with the use of the LLVM compiler more developers to attract for writing compute-intensive applications that use gpu’s. Nonetheless, the LLVM compiler Nvidia a partly closed project; the company wants the code to share with partners.