Polly can now automatically offload suitable kernels to GPUs in the form of Polly-ACC, saving a lot of developer time and effort. Julia, a modern language built just for scientific computing from scratch, has much to gain from the productivity offered by Polly’s GPU offload capabilities which makes it appealing to seasoned researchers and novice programmers who’d want to leverage a GPU’s computational power with the least amount of effort. This project aims to integrate Polly-ACC into Julia and ensure it is able to accelerate compute-intensive parts of Julia programs by focusing on a set of representative benchmarks. It also aims to better optimise code using cues from run-time parameters, leveraging the opportunities that JIT compilation has to offer. This could be a stepping stone into making an LLVM powered language a preferred platform to program heterogeneous systems.

Student

Singapuram Sanjay S.V.

Mentors

  • Tobias Grosser-1
close

2017