Project Study: Speeding-up of Linear Programs Solving on GPU

Linear Programs (LPs) appear in a large number of applications and efficiently solving them is essential. While solution algorithms for LPs cannot easily be parallelized, recent studies suggest that GPUs still can offer a significant performance boost, especially in cases where multiple (up to millions) small-scale LPs have to be solved efficiently and in parallel. As compared to parallelizing the inner workings of solution algorithm, this topic is till now under-researched.