INTELL (Integrated code generation and paral lelization)

Abstract

The advancement in GPU technology can be attributed mostly due to heavy investment in gaming industry. This has led to a need to use GPUs for high performance computing, for example GPUs have become an irreplaceable component of the world’s most powerful super computers. The use of NVIDIA’s GP-GPU by commercial products such as Adobe Photoshop CS6 adds to the validity of this argument. One important issue using GPU technology is that the advancement in architecture is progressing very rapidly which results in a wide variety of different architectures. To utilize these GPUs to their optimal potential, the software must be optimized according to these varying architectures. This can be a tedious and time consuming process. INTELL is a benchmarking tool that provides users with the ability to test and evaluate new and expensive hardware without having to purchase it. With the rapid development in GPU technology it has become increasingly difficult for people to evaluate the true capabilities of the new systems based on specifications data alone because actual performance depends on a number of different variables. INTELL allows developers and researchers to utilize the capabilities of GPU’s through a web interface. It provides users with the ability to test different kernels on different architectures by actually performing the tests. These results are then provided to the user in the form of graphs for evaluation. This allows researchers and developers to reduce the cost of their projects.