Conclusions

This book aimed to introduce to the reader how heterogeneous hardware acceleration is changing the programming landscape, while posing a number of challenges to existing programming languages. In particular, the focus of this book has been on dynamic programming languages executed on top of managed runtime environments. After having introduced heterogeneous hardware acceleration and explained why it has become prevalent in modern computing systems (Chap. 2 ), we elaborated on the programmings models that are currently being used to program such accelerators (Chap. 3 ). For simplicity, our discussions were focused mainly on a particular class of heterogeneous hardware accelerators; that of General Purpose GPUs. GPUs have become increasingly important due to their high performance and suitability for AI/ML workloads across different application domains. Nevertheless, both the discussion on heterogeneous programming models and the challenges they pose on MREs are transferable across different hardware accelerators and programming models.