Orchard: Heterogeneous Parallelism and Fine-grained Fusion for Complex Tree Traversals
Published in ACM TACO, Vol. 21, No. 2, Article 41, 2024
Orchard automatically fuses and parallelizes tree traversals that are difficult to optimize by hand, combining heterogeneous parallelism with fine-grained fusion. Measured speedups range from 1x to 5x over baseline traversals.
Recommended citation: Vidush Singhal, Laith Sakka, Kirshanthan Sundararajah, Ryan Newton, Milind Kulkarni. (2024). "Orchard: Heterogeneous Parallelism and Fine-grained Fusion for Complex Tree Traversals." ACM Transactions on Architecture and Code Optimization, 21(2), Article 41.
Download Paper